Delineating Surface Epitopes of Lyme Disease Pathogen Targeted by Highly Protective Antibodies of New Zealand White Rabbits

Lyme disease (LD), the most prevalent vector-borne illness in the United States and Europe, is caused by Borreliella burgdorferi. No vaccine is available for humans. Dogmatically, B. burgdorferi can establish a persistent infection in the mammalian host (e.g., mice) due to a surface antigen, VlsE. This antigenically variable protein allows the spirochete to continually evade borreliacidal antibodies.

ABSTRACT

Lyme disease (LD), the most prevalent vector-borne illness in the United States and Europe, is caused by Borreliella burgdorferi. No vaccine is available for humans. Dogmatically, B. burgdorferi can establish a persistent infection in the mammalian host (e.g., mice) due to a surface antigen, VlsE. This antigenically variable protein allows the spirochete to continually evade borreliacidal antibodies. However, our recent study has shown that the B. burgdorferi spirochete is effectively cleared by anti-B. burgdorferi antibodies of New Zealand White rabbits, despite the surface expression of VlsE. Besides homologous protection, the rabbit antibodies also cross-protect against heterologous B. burgdorferi spirochetes and significantly reduce the pathology of LD arthritis in persistently infected mice. Thus, this finding that NZW rabbits develop a unique repertoire of very potent antibodies targeting the protective surface epitopes, despite abundant VlsE, prompted us to identify the specificities of the protective rabbit antibodies and their respective targets. By applying subtractive reverse vaccinology, which involved the use of random peptide phage display libraries coupled with next-generation sequencing and our computational algorithms, repertoires of nonprotective (early) and protective (late) rabbit antibodies were identified and directly compared. Consequently, putative surface epitopes that are unique to the protective rabbit sera were mapped. Importantly, the relevance of newly identified protection-associated epitopes for their surface exposure has been strongly supported by prior empirical studies. This study is significant because it now allows us to systematically test the putative epitopes for their protective efficacy with an ultimate goal of selecting the most efficacious targets for development of a long-awaited LD vaccine.

INTRODUCTION

The Borreliella burgdorferi sensu lato complex comprises 21 genospecies, of which Borreliella burgdorferi sensu stricto (B. burgdorferi), Borreliella afzelii, and Borreliella garinii are the main tick-borne pathogens responsible for Lyme disease (LD; also known as Lyme borreliosis) in humans, dogs, and other animal species (1–6). Recently, a rise in the incidence rates of human LD has been steadily recorded in the Northern Hemisphere (7–9). In the United States, B. burgdorferi alone causes approximately 300,000 annual LD cases, and in Europe, the LD genospecies are collectively responsible for 65,500 to 85,000 LD records each year (2–5).

The debilitating nature of LD is primarily due to the ability of B. burgdorferi to establish a life-long persistence in humans. When early, usually nondescript symptoms are missed, a chronic stage of LD with a variety of clinical signs (e.g., arthritis, carditis, encephalitis) can develop (10). Unfortunately, treatment of this chronic LD stage is very challenging. Paradoxically, in contrast to dogs, no vaccine against this medically important but preventable disease is currently available for humans (11–16).

The years-long persistence of B. burgdorferi is partially attributable to the variable major protein (VMP)-like sequence (vls) locus, which is housed on the 28-kb linear plasmid lp28-1 (17). This vls locus, composed of the vlsE expression site and an array of nonexpressible vls cassettes, allows continuous vlsE recombination to occur. The unprogrammed vls recombination events are inducible in vivo only when, shortly after the tick-mediated transmission, LD spirochetes have adapted in the mammalian host (e.g., mice and humans) (17–24). The final product of the vls system is expression of an antigenically varying surface protein, VlsE. This highly immunogenic VlsE is absolutely required for B. burgdorferi to evade, during its long-term persistence, otherwise protective anti-B. burgdorferi antibodies (25–35).

Recently, however, it has been shown that, as opposed to humans (36–41) and all other known LD animal models (42–57), VlsE-competent B. burgdorferi is effectively cleared by the humoral immune response of New Zealand White (NZW) rabbits (58). Host-adapted spirochetes, which are considered highly immune evasive in mice due to their abundant VlsE expression (34, 59), are efficiently killed by rabbit anti-B. burgdorferi antibodies (the rabbit antibodies). The mice that had received the rabbit antibodies consistently abrogated an establishment of infection by VlsE-expressing B. burgdorferi (58). In addition to being protective against homologous infection, the rabbit antibodies are also cross-protective against in vitro-grown heterologous B. burgdorferi. Furthermore, the rabbit antibodies have the capacity to significantly reduce the pathology of LD arthritis in actively infected mice, despite the fact that spirochetes have already disseminated to avascular collagenous tissues and, hence, are protected from the host immune response (60). Thus, in contrast to infected LD patients or mice, NZW rabbits develop a unique repertoire of very potent antibodies which have the capacity to specifically target protective epitopes. Importantly, these protective targets seem to be freely accessible to the rabbit antibodies, despite the copious amounts of VlsE molecules on the spirochetal surface (58). This recent finding (58) is significant because it has opened up an opportunity to comprehensively examine the specificities of the rabbit antibodies responsible for the sterilizing immunity against LD pathogens by defining their respective targets.

In the present study, in order to delineate the surface epitopes associated with the observed protection, we compared the repertoires of nonprotective (early) and protective (late) antibodies of NZW rabbits. The approach involved the use of random peptide phage display libraries coupled with next-generation sequencing (RPPDL/NGS), followed by our previously developed computational algorithms (59). As a result, putative surface epitopes that are unique to the protective sera of NZW rabbits have been identified. The newly identified protection-associated epitopes were distinct from those previously mapped via the mouse model (59). Importantly, whenever published data are available, the relevance of these novel epitopes for their surface exposure has been verified by a number of empirical studies. Also, some of the putative epitopes belong to the proteins of B. burgdorferi that have consistently been considered to be strong vaccine candidates. Together, the present study now allows us to systematically test the newly and previously (59) identified epitopes for their protective efficacy with an ultimate goal of selecting the most promising targets for designing a much-needed LD vaccine.

RESULTS

Our recent (58) and present studies have confirmed some of the prior findings (58, 61, 62) that, by week 4 to 5 postinfection (p.i.), NZW rabbits clear the wild-type B. burgdorferi spirochete from the inoculated skin. In contrast to the findings at days 14 and 21 p.i., when rabbit skin biopsy specimens were consistently culture positive, the day 28 skin tissues of some NZW rabbits were already negative by culture. Importantly, as shown by quantitative PCR, the NZW rabbits whose biopsy specimens from day 28 were still culture positive consistently harbored very few spirochetes, suggesting the final stage of skin infection clearance at about this time point (58). Moreover, only when they were passively immunized with day 28 rabbit sera did immunocompetent mice gain resistance to the host-adapted B. burgdorferi B31-A3 (B31) strain (58). Thus, by day 28 p.i., NZW rabbits develop a repertoire of highly protective antibodies, which are responsible for the clearance of VlsE-expressing spirochetes (58). In order to identify the specificities of the protective antibodies and, consequently, map their targets, we analyzed the global antibody repertoires of immune sera collected from B31-infected rabbits at days 14 and 28 p.i., the time points at which the rabbit antibodies are not yet protective or are already protective, respectively.

Identification of antibody repertoires in nonprotective and protective sera of NZW rabbits.

To generate day 14 and day 28 anti-B. burgdorferi sera, three NZW rabbits were intradermally inoculated with in vitro-grown B. burgdorferi B31 as described previously (62). The progress of infection was monitored by culturing skin biopsy specimens sampled around each inoculation site at days 7, 14, 21, 28, 35, 42, 49, and 56 p.i. Dark-field microscopy demonstrated that skin tissues from all the NZW rabbits were consistently culture positive on days 7, 14, and 21, indicating a successful establishment of infection with wild-type B. burgdorferi (Table 1). At day 28 p.i., however, only two out of three animals remained culture positive for spirochetes. Skin biopsy specimens from all the inoculation sites of rabbit L tested culture negative. From day 35 p.i. and onwards, however, all the skin tissues from the three NZW rabbits were consistently culture negative (Table 1), which indicated that the three animals had cleared the skin infection (61).

TABLE 1.

Culture results for skin tissues taken weekly from B. burgdorferi B31-infected NZW rabbits

Rabbit identifier	Culture result on the following day postinfectiona :
	7	14	21	28	35	42	49	56
L	+	+	+	−	−	−	−	−
M	+	+	+	+	−	−	−	−
N	+	+	+	+	−	−	−	−

^a+ and −, culture-positive and culture-negative skin biopsy specimens, respectively.

Three individual serum samples from day 14 and three individual serum samples from day 28 were analyzed via RPPDL/NGS, the approach that was successfully applied to identify antibody repertoires in B. burgdorferi-infected mice (59, 63). As a background control, preimmune sera collected from the three NZW rabbits prior to their challenge were pooled in equal volumes and analyzed as a single sample. A permutation test was used to detect any significant difference between the day 14 and day 28 peptide profiles. As a result, 309 nonredundant peptides which were missing from all the day 14 and preimmune sera were identified in each of the three day 28 samples (P < 0.05). Thus, by day 28 p.i., infected NZW rabbits developed an additional antibody repertoire, which may well account for the observed protection against B. burgdorferi.

Identification of protection-associated surface epitopes of B. burgdorferi.

To examine whether antibodies of the day 28 repertoire represented by the 309 peptides were specific to B31 surface proteins, the previously developed in silico algorithm was applied (59). Specifically, the 309 peptides were first mapped to the entire proteome of B. burgdorferi B31 (a total of 1,391 proteins) (64). This mapping resulted in the identification of 1,427 BLASTP hits with at least 4 amino acid matches to 661 proteins (Fig. 1). To increase the stringency, any hits with gaps were excluded and only the matches with 5 or more amino acids were considered. As a result, 501 gapless hits led to the identification of 321 proteins, some of which contained multiple peptide hits and/or the same peptide hit aligned to multiple proteins. Of these 501 hits, we focused on peptides with hits to a single protein only in order to eliminate any potentially cross-reactive peptides. This filtering process identified 72 proteins (some proteins contained multiple nonoverlapping peptide hits) and their respective 94 hits (peptides) (Fig. 1). Out of these 72 proteins, 7 surface-exposed proteins were mapped by seven peptides, which most likely represented epitopes that were targeted by the protective rabbit (day 28) antibodies (Table 2). Together, the relevance of seven protection-associated (PA) peptides for B. burgdorferi surface antigens was validated by the fact that each 7-mer aligned to the B31 surface protein with no fewer than 5 amino acid matches. The newly identified PA peptides were then divided into three categories, categories I to III (Table 3). Category I included a peptide that had a 5-amino-acid match to its respective surface protein and no other 5- or 4-amino-acid matches to any other B31 proteins. Category I peptide pA206B57 was mapped to a surface-exposed protein, BB0337 (Table 3). Category II consisted of peptides that had 5-amino-acid matches to their respective surface proteins and no other 5-amino-acid matches to any of the nonsurface proteins of the B31 strain. Category II peptides pA58, pA152, pA259, and pA20B4 were aligned with their 5-amino-acid matches to the following surface proteins: BB0089, BBO39, BB0210, and BB0603, respectively. The four peptides of this category also aligned to some nonsurface proteins with 4-amino-acid matches (Table 3). Category III peptides pA202B71 and pA234 had 5-amino-acid matches to their respective surface proteins and additional single 5-amino-acid and multiple 4-amino-acid matches to other nonsurface proteins. Interestingly, some 4-amino-acid matches of the pA202B71 and pA234 peptides were also aligned to additional surface proteins of the B31 strain (Table 3). Together, the category I to III peptides represented seven protection-associated putative linear epitopes (PA epitopes) of seven surface proteins of B. burgdorferi B31.

FIG 1.

Identification of surface antibody-binding epitopes and their respective proteins associated with protection of New Zealand White rabbits against B. burgdorferi. Three NZW rabbits were infected with the wild-type B. burgdorferi B31 strain. At days 0 (prior inoculation), 14, and 28 postinfection, serum samples were individually collected from the three B31-infected animals. Day 14 sera (three samples) and day 28 sera (three samples) were considered nonprotective and protective, respectively. To identify peptide (antibody) repertoires of protective (repertoire A) and nonprotective (repertoire B) sera, the six serum samples and control preimmune sera were analyzed by use of a random peptide phage display library. Repertoires A and B and the preimmune repertoire were then compared via the permutation test. Consequently, 309 peptides (repertoire C) were consistently identified in each of the protective serum samples and absent in all three nonprotective and pooled preimmune serum samples. The 309 peptides were mapped, via BLASTP analysis, to the amino acid sequences of 1,391 proteins from the complete genome of B. burgdorferi B31 (64). Only matches that contained at least 4 amino acids were initially considered, for a total of 1,427 hits. To increase the stringency, 501 gapless hits with at least 5 contiguous identical amino acids were then selected. Because some proteins contained multiple peptide hits and/or the same peptide hit aligned to different proteins, these hits encompassed 321 B. burgdorferi proteins, of which 7 proteins were known to be surface exposed.

TABLE 2.

Surface proteins of B. burgdorferi B31 identified in silico through epitope mapping

GenBank accession no.	ORFa (plasmid)	Protein name	Protein length (no. of amino acids)	Reference(s)b
NP_051372.1	BBO39 (cp32-7)	ErpL	229	102, 103
NP_212223.1	BB0089	Outer surface protein BB0089 (predicted)	324	99
NP_212471.1	BB0337	Enolase	433	87, 89, 90, 131
NP_212599.1	BB0465	Outer surface protein BB0465 (predicted)	231	99
NP_212737.1	BB0603	Integral outer membrane protein P66	618	132–135
NP_212972.1	BB0838	Outer surface protein BB0838	1,146	99
YP_008686569.1	BB0210	Surface-located membrane protein 1 (Lmp1)	1,119	85

^aORF, open reading frame.

^aListed are citations for studies that demonstrated or predicted the surface localization of B. burgdorferi proteins.

TABLE 3.

Protection-associated peptides with amino acid gapless matches to surface and nonsurface proteins of B. burgdorferi B31

Category and peptide identifier	Protein name	Alignment length (no. of amino acids)	No. of identical amino acid matches	No. of amino acid mismatches	No. of 4-amino-acid matches to nonsurface proteins	No. of 5- amino-acid matches to nonsurface proteins	No. of 4-amino-acid matches to other surface proteins [protein(s)]
Category I, pA206B57	BB0337 (enolase)	7	5	2	0	0	0
Category II
pA58	BB0089	6	5	1	2	0	0
pA152	BBO39 (ErpL)	6	5	1	2	0	0
pA259	BB0210 (Lmp1)	6	5	1	3	0	0
pA20B4	BB0603 (P66)	6	5	1	1	0	1 (BBA36)
Category III
pA202B71	BB0838	5	5	0	2	1	2 (BB0161, BB0405)
pA234	BB0465	6	5	1	3	1	2 (BBB07, BB0603)

To examine how conserved the newly identified PA epitopes are among the three main pathogenic genospecies, B. burgdorferi, B. afzelii, and B. garinii, the seven surface proteins of B. burgdorferi B31 mapped by category I to III peptides were aligned, via BLASTP analysis, to their respective counterparts of B. garinii and B. afzelii. Moreover, the proteins of B. burgdorferi, B. afzelii, and B. garinii were also individually aligned to the seven PA peptides. The results demonstrated that most PA epitopes exhibited a very high degree of conservation between the main pathogenic genospecies. Specifically, the PA epitopes pA206B57, pA58, pA20B4, pA202B71, and pA234 showed very high conservation ratios (the number of identical amino acid matches per alignment length), 5/7, 5/6, 5/6, 5/5, and 5/6, respectively, across the three genospecies (Table 4). Interestingly, despite the fact that the two proteins of B. garinii and B. afzelii had an overall low degree of amino acid sequence identity (78 to 82%) relative to BB0210 of B. burgdorferi, their respective PA epitopes mapped by pA259 also exhibited a high conservation ratio, 4/6. Finally, the pA152-mapped epitope represented the least conserved protein, BBO39 (61 to 62%), and, as expected, had the lowest conservation ratio, 3/6 (Table 4). Together, the results demonstrate that most newly identified PA epitopes were highly conserved between the three LD genospecies.

TABLE 4.

Conservation of linear epitopes between the main LD pathogens relative to protection-associated peptides

Peptide category and peptide	Protein name	No. of identical amino acid matches/no. of amino acids in alignment (protein GenBank accession no./% identity with the respective B. burgdorferi B31 protein)
		B. burgdorferi	B. afzelii	B. garinii
Category I, pA206B57	BB0337 (enolase)	5/7 (NP_212471.1/NAa )	5/7 (WP_011600962.1/97)	5/7 (WP_015026804.1/97)
Category II
pA58	BB0089	5/6 (NP_212223.1/NA)	5/6 (WP_014486340.1/91)	5/6 (WP_038628417.1/93)
pA152	BBO39 (ErpL)	5/6 (NP_051372.1/NA)	3/6 (WP_014486142.1/62)	3/6 (NP_051372.1/61)
pA259	BB0210 (Lmp1)	5/6 (YP_008686569.1/NA)	4/6 (WP_048830530.1/78)	4/6 (WP_004793842.1/82)
pA20B4	BB0603 (P66)	5/6 (NP_212737.1/NA)	5/6 (WP_073999082.1/93)	5/6 (WP_029362299.1/91)
Category III
pA202B71	BB0838	5/5 (NP_212972.1/NA)	5/5 (WP_004789632.1/91)	5/5 (WP_004792444.1/91)
pA234	BB0465	5/6 (NP_212599.1/NA)	5/6 (WP_004789852.1/93)	5/6 (WP_031505460.1/91)

^aNA, not applicable.

Comparison of antibody repertoires developed in NZW rabbits and C3H mice upon B. burgdorferi infection.

In contrast to NZW rabbits, which efficiently clear the LD pathogen, immunocompetent mice become persistently infected with B. burgdorferi, despite their strong anti-Borreliella antibody responses (17, 26, 30, 33, 35, 53, 54). To examine whether the novel PA epitopes targeted by the day 28 rabbit antibodies are also recognized by mouse anti-B31 antibodies, sera sampled from three B31-infected C3H/HeJ (C3H) mice at day 28 p.i. were analyzed via RPPDL/NGS. To identify any antibody repertoire that would be unique to the day 28 rabbit sera, the global mimotope profiles derived from the day 28 mouse sera and day 14 and pooled preimmune rabbit sera were compared with the day 28 rabbit mimotope profile. Consequently, 87 nonredundant 7-mers were identified in each of the three day 28 rabbit samples. These 87 peptides were not recognized by any of the three day 14 (n = 3) or preimmune rabbit serum samples (P < 0.05, permutation test). All the 87 peptides were part of the 309 peptides which had earlier been identified to be specific to the day 28 rabbit sera compared to the day 14 rabbit sera (Fig. 1). Interestingly, out of the seven PA peptides that mapped the seven surface-exposed proteins, pA152 and pA259 were detected in all three day 28 mouse samples. Furthermore, the other two PA peptides, pA58 and pA234, were identified in sera from mouse 2 and mouse 3, respectively. The remaining three peptides, pA206B57, pA20B4, and pA202B71, were not detected in any of the day 28 mouse sera and thus were unique to the protective rabbit sera.

Comparison of protection-associated sequences with previously identified conformational surface epitopes of B. burgdorferi.

In addition to linear epitopes, we attempted to map conformational surface epitopes associated with the PA peptides uniquely recognized by the rabbit day 28 sera vis-à-vis the PA peptides that were previously identified in mice (59). The analysis involved a total of five conformational epitopes within outer surface proteins A and B of B. burgdorferi B31 (OspA and OspB, respectively) (Fig. 2). Similar to the mouse PA peptides, the day 28 rabbit-recognized peptides overlapped only a fraction of the epitope residues. Among the four OspA epitopes (A1 to A4) and the sole OspB epitope (B1), rabbit-recognized peptide sequences overlapped only two OspA epitopes (A2 and A3). Peptide q141 (GMLEKTT) aligned with a sequence of A2 (¹⁶⁰TTLVVKE¹⁶⁶), while peptides q198 (ASSPSTK) and q204 (TIKTGSW) aligned with a sequence of A3 (²⁰⁶SSAATKKTAA²¹⁵).

FIG 2.

Conformational epitopes (labeled A1 through A4 and B1) of B. burgdorferi B31 outer surface proteins A (OspA) and B (OspB) with aligned with 7-mer phage-displayed peptides. Each epitope label is followed by its cognate monoclonal antibody (MAb) name. A1 (136), A2 (137), A3 (109), and A4 (138) are on OspA, while B1 (139) is on OspB. Epitope residues are underlined in the protein sequences, with representative sequence positions being numbered. Peptide residues are rendered in uppercase if they are part of a BLASTP hit alignment and lowercase otherwise. Peptide identifiers (in parentheses) are prefixed with either “p” for protection-associated peptides in mice from previous work (59) or “q” for peptides uniquely recognized by day 28 rabbit sera. Residues of A3 marked with asterisks are critical for LA-2 binding and complement-dependent bactericidal activity (140). Residues common to both A3 and A4 are marked with carets. A3 is topologically analogous to B1, with B1 being centered around OspB residue K253 (marked with a plus sign), which is essential for binding of OspB by H6831 (139). The other MAbs, MAb B3G11 and MAb N5G5, bind 10-mer peptide analogs (sequences highlighted with a magenta background) of OspA and OspB, respectively (141).

As before (59), aligned peptides were also analyzed in relation to three protective sequences (Fig. 3), from OspA, OspB, and OspC, which had previously been shown to elicit antipeptide antibodies with complement-dependent (anti-OspA and anti-OspC) or complement-independent (anti-OspB) bactericidal activity (65–67). Unlike the mouse PA peptides, only one rabbit-recognized peptide (q204) overlapped a protective sequence (OspB ²³⁸KWEDSTSTLTISADSKKTKD²⁵⁷). Moreover, the six rabbit-recognized peptides that yielded BLASTP hits on OspC (q009, q029, q080, q128, q131, and q156) all aligned with ¹¹⁹GLKEK¹²³ and shared the N-terminal consensus sequence GLLEK. In contrast, mouse PA peptides yielded a larger number of BLASTP hits that were more evenly distributed along the entire OspC sequence (Fig. 4). The GLLEK consensus sequence itself aligned with the N-terminal end of the 7-mer sequence GLLQKPL derived from the most frequently observed residues in the entire set of rabbit-recognized peptides (as summarized in the sequence logo presented in Fig. 5).

FIG 3.

Protective sequences of B. burgdorferi B31 outer surface proteins A (OspA), B (OspB), and C (OspC) with aligned 7-mer phage-displayed peptides. Underlined sequences (with the N- and C-terminal residue positions being numbered) of OspA, OspB, and OspC elicit antipeptide antibodies with complement-dependent (anti-OspA and anti-OspC) and complement-independent (anti-OspB) bactericidal activity (65–67). Peptide identifiers (on the lines of their respective sequences) are prefixed with either “p” for protection-associated peptides in mice from previous work (59) or “q” for peptides uniquely recognized by day 28 rabbit sera. Peptide residues are rendered in uppercase if they are part of a BLASTP hit alignment and lowercase otherwise. BLASTP analysis E values and bit scores are in parentheses. OspA and OspB sequences are aligned to emphasize their homology, with OspB residue K253 being marked with a plus sign, as in Fig. 2 (within conformational epitope B1). A subsequence exactly matching OspC residues 131 to 149 (highlighted with a yellow background) forms part of a chimeric vaccinogen that elicits antipeptide antibodies having complement-dependent bactericidal activity (142).

FIG 4.

B. burgdorferi B31 outer surface protein C (OspC) sequence with aligned epitope sequences (labeled E1 through E6) and 7-mer phage-displayed peptides. Epitope sequences (with the N- and C-terminal residue positions being numbered) were inferred from 12-mer phage-displayed peptide sequences of B. burgdorferi SKT-2 OspC (143). Homologous sequences in B. burgdorferi B31 OspC are underlined. Epitope sequence differences between the two B. burgdorferi strains are highlighted with a cyan background. Peptide residues are rendered in uppercase if they are part of a BLASTP hit alignment and lowercase otherwise. Peptide identifiers (in parentheses) are prefixed with either “p” for protection-associated peptides in mice from previous work (59) or “q” for peptides uniquely recognized by 28-day rabbit sera. Angled brackets delineate residues 130 to 150, which constitute the protective sequence that contains a vaccinogen component subsequence (highlighted with a yellow background). The 15-mer peptide whose sequence exactly matches residues 133 to 147 (highlighted with a magenta overhead band) is bound by bactericidal complement-independent monoclonal antibody 16.22 (144).

FIG 5.

Sequence logo generated from the 7-mer phage-displayed peptides uniquely recognized by 28-day rabbit sera using the WebLogo (version 2.8.2) sequence logo generator (125).

Considering that the BLASTP E values are rather large for the GLLEK consensus peptides aligned with OspC ¹¹⁹GLKEK¹²³ (Fig. 3) while these peptides more closely match nucleoside 2-deoxyribosyltransferase superfamily protein ⁴⁴GLLEK⁴⁸ (GenBank accession number NP_212560.2) and even lipoprotein ¹⁰²GLVEK¹⁰⁶ (GenBank accession number NP_045706.1) (noting that GLLEK and GLVEK each occur only once in the B31 proteome), the alignments with OspC ¹¹⁹GLKEK¹²³ likely represent only weak sequence similarity rather than antigenic cross-reactivity, especially in view of the nonconservative substitution of K (cationic) for L (nonpolar). The multiplicity of GLLEK consensus peptides might be at least partly due to the simultaneous antibody-mediated selection of various related sequences, for example, sequences comprising the subsequence GLLE or LLEK (which occur in the proteome 9 and 49 more times, respectively, outside GLLEK) or, more generally, bearing similarity to GLLQKPL (e.g., with E replacing Q).

GLLQKPL itself may represent more than a single epitope sequence, as it is absent from the B31 proteome (64). Subsequences of GLLQKPL occur within the B31 proteome, the longest of which are GLLQK and LQKPL (matching hypothetical protein BB_0509 ²¹²GLLQK²¹⁶ [GenBank accession number NP_212643.1] and hypothetical protein BB_0748 ¹⁰⁵LQKPL¹⁰⁹ [GenBank accession number NP_212882.1], respectively). Additionally, LLQK occurs seven more times (outside GLLQK), while LQKP and QKPL each occur only once more (outside LQKPL, matching hypothetical protein BB_0024 ⁸⁸LQKP⁹¹ [GenBank accession number NP_212158.1] and hypothetical protein BB_0531 ³³⁹QKPL³⁴² [GenBank accession number YP_008853942.1], respectively).

Generalizing LQKPL to a consensus sequence, XQKPX, where X is a nonpolar residue (i.e., with a side chain incapable of hydrogen bonding) other than P, only three exact matches were found in the B31 proteome, namely, LQKPL itself, LQKPI (matching hypothetical protein BB_0024 ⁸⁸LQKPI⁹² [GenBank accession number NP_212158.1]), and VQKPV (matching lipoprotein ¹⁷⁶VQKPV¹⁸⁰ [GenBank accession number NP_045709.2], encoded by open reading frame BBA36 on plasmid lp54). VQKPV thus maps to a lipoprotein (BBA36) previously identified to be a potential surface-exposed protein target for protective mouse antibodies (see Tables 5 and 6 of reference 59) and yields multiple BLAST hits with the rabbit-recognized peptides (Fig. 6). Using the entire lipoprotein BBA36 sequence, the CPHmodels web server identified a template (PDB accession number 2K0N, chain A, for the nuclear magnetic resonance ensemble solution structure of the Saccharomyces cerevisiae yeast Gal11p kix domain as the template) for remote homology modeling (after an initial failed attempt to identify a template via BLAST analysis of the sequences in the Protein Data Bank [PDB]). This approach assigned atomic coordinates only to residues 47 to 126 (of 212 residues). However, when the C-terminal sequence that lacked assigned atomic coordinates (residues 127 to 212) was used as the input, the CPHmodels web server identified an alternative template (PDB accession number 1CMB, chain A, for the Escherichia coli strain K-12 Met repressor) for another round of remote homology modeling (again, after an initial failed attempt to identify a template via BLAST analysis of the sequences in PDB). Atomic coordinates were thus also assigned to residues 147 to 199, placing ¹⁷⁶VQKPV¹⁸⁰ within a region of putative disordered (i.e., nonhelical, non-β-sheet) secondary structure (residues 175 to 183 between a 3₁₀-helix and an α-helix), with ¹⁷⁶VQKPV¹⁸⁰ comprising residues of a β-turn and non-hydrogen-bonded bends (Fig. 6). This region largely overlaps a sequence with a putative highly mobile (i.e., high-B-factor) disordered secondary structure predicted by the DisEMBL tool and entirely encompassing ¹⁷⁶VQKPV¹⁸⁰. The latter sequence itself comprises residues (¹⁷⁶VQKP¹⁷⁹) predicted to be natively disordered by the PrDOS server and abuts residues (¹⁷⁴ES¹⁷⁵) likewise predicted by IUPred2 to be disordered. Such disorder among protein epitopes may favor their binding by paratopes through conformational adjustments, for example, to enable binding of native protein targets by antipeptide antibodies that have been produced in response to peptide-based immunogens (e.g., vaccine candidates), wherein peptide epitopes are themselves typically disordered (68). Hence, BBA36 ¹⁷⁶VQKPV¹⁸⁰ may constitute or at least form part of an epitope that could be targeted by antibodies for effecting protective immunity, considering that BBA36 is upregulated by cultivation in mammalian hosts, while specific antibodies against BBA36 are bactericidal (69).

FIG 6.

The B. burgdorferi B31 (GenBank accession number NP_045709.2) lipoprotein BBA36 C-terminal sequence with aligned 7-mer phage-displayed peptides. Representative sequence positions are numbered. Peptide residues are rendered in uppercase if they are part of a BLASTP hit alignment and lowercase otherwise. BLASTP analysis E values and bit scores are in parentheses. The line above the protein sequence displays the DSSP secondary-structure assignment codes (H, α-helix; G, 3₁₀-helix; T, hydrogen-bonded turn; S, non-hydrogen-bonded bend) for residues with atomic coordinates generated by CPHmodels (or an asterisk for unassigned coordinates). Protein residues predicted to be conformationally disordered are highlighted with colored overhead bands: blue for residues predicted by DisEMBL (version 1.5) to be in loops/coils (i.e., nonhelical and non-β-strand regions) with high B factors (i.e., α-carbon Debye-Waller atomic temperature factors), green for residues predicted by PrDOS to be natively disordered, and red for residues predicted by IUPred2 to be in long disordered regions.

DISCUSSION

Subtractive RV to identify potential targets for LD vaccine.

Novel approaches to vaccine candidate selection and design have been based on recent advances in genomics and various innovative strategies, such as reverse vaccinology (RV) (70). The first RV application resulted in the development of a successful vaccine against meningococcal serogroup B (71). Subtractive RV was later developed to identify antigens that are present and absent in pathogenic and commensal microorganisms, respectively (72, 73). Since then, the RV approach has been widely and successfully applied to a wide range of bacterial pathogens (70, 74–78). Recently, subtractive RV was applied to identify putative protective epitopes on various surface antigens of B. burgdorferi (59). In that study, a repeated immunization assay involved an attenuated mutant of B. burgdorferi, B. burgdorferi ΔVlsE, and immunocompetent mice. The ΔVlsE clone lacked VlsE, the antigenically variable surface protein that is thought to physically shield protective epitopes of LD spirochetes from host antibodies (30, 34, 79–81). It is hypothesized that, during natural infection, VlsE molecules mainly mask dominant protective epitopes, whereas subdominant ones remain exposed, as they do not present any risk to the spirochete due to their inability to induce a strong immune response (59). By directly (due to the lack of VlsE) exposing subdominant (and dominant) surface epitopes through a series of immunizations with the ΔVlsE mutant, 50% of mice became immune to challenge with the highly immune-evasive VlsE-expressing wild type (59). These results indicate that, when abundantly expressed on the spirochetal surface, VlsE molecules do not physically block every single surface epitope and, importantly, that some of these exposed (not shielded by VlsE) epitopes are targeted by protective mouse antibodies. By analyzing and directly comparing individual serum samples from protected and nonprotected mice via RPPDL/NGS, an antibody repertoire unique to the protective mouse sera was thus identified. A downstream application of computational algorithms resulted in mapping of PA epitopes on several surface proteins of B. burgdorferi. Importantly, whenever prior data were available (69, 82–88), the mapped surface proteins were shown to be immunogenic, capable of inducing protection against wild-type challenge, and critical for B. burgdorferi pathogenesis or growth, all of which are desired characteristics of a strong vaccine candidate (59). Similarly, whenever available, published studies empirically proved the surface exposure of the PA epitope-containing domains of these proteins (59).

Similar to our prior work, in this study, we applied our previously developed approach to identify surface epitopes of B. burgdorferi that are specifically targeted by the protective rabbit antibodies. In contrast to the 50% protection rate achieved for mice via repeated immunization with the ΔVlsE mutant (59), the rabbit antibodies consistently prevented 100% of mice from host-adapted, VlsE-expressing spirochetes. Moreover, the rabbit antibodies significantly reduced the pathology of LD arthritis in persistently infected mice (58). Thus, in the present work, via subtractive RV (58), we have compared the antibody repertoires of nonprotective (day 14) and protective (day 28) rabbit sera with the goal of identifying the specificities of protective antibodies and, therefore, their respective surface targets. As a result, novel PA epitopes mapped to several surface proteins of B. burgdorferi have been identified. Interestingly, the newly identified epitopes are distinct from those recently delineated via the mouse repeated immunization assay (58). Importantly, when available, the published data cited below have empirically demonstrated that newly identified PA epitopes are surface exposed and that their respective proteins are considered to be strong LD vaccine candidates.

Protection-associated antibody targets identified via the NZW rabbit model.

The current study has identified a total of seven PA epitopes that were missing and present, respectively, in the nonprotective (day 14) and protective (day 28) sera of NZW rabbits. These PA epitopes were mapped to seven surface proteins of B. burgdorferi and, overall, demonstrated a very high degree of conservation among the main pathogenic genospecies, B. burgdorferi, B. afzelii, and B. garinii. Moreover, three out of seven PA epitopes, mapped by pA206B57, pA20B4, and pA202B71, were not recognized by any of the nonprotective day 28 mouse sera and thus were unique to the protective rabbit sera. The other PA epitopes were recognized by either a single mouse serum sample (pA58 and pA234) or all three mouse serum samples (pA152 and pA259). It is possible that these shared epitopes induce protective antibodies only in the NZW rabbit and not the mouse.

The chromosomally encoded enolase of B. burgdorferi (BB0337) mapped by pA206B57 is a plasminogen receptor which is highly conserved across all the genospecies of the B. burgdorferi sensu lato complex (87, 89). Enolase was shown to be immunogenic both in LD patients and in experimentally infected mice (90). Although antienolase antibodies do not significantly reduce the spirochetal numbers in immunized animals, the antibodies do interfere with LD pathogen persistence in the tick vector (87). Importantly, as shown by the prior study (59), the enolase also contains two additional PA epitopes, defined by p222 and p358. Thus, these three PA epitopes should be tested for their ability to induce antibodies that would interfere with LD spirochetal survival in Ixodes ticks.

Surface-located membrane protein 1 (Lmp1 or BB0210) is a chromosomally encoded adhesin (128-kDa) which is required for B. burgdorferi to persist in murine tissues (85, 86). Despite being immunogenic in mice and humans (86), Lmp1 is important for LD spirochetes to resist adaptive humoral responses (86). Lmp1 deletion impairs the ability of B. burgdorferi mutants to establish persistent infection in immunocompetent mice but not in immunodeficient mice. Out of three domains, specifically, Lmp-N, Lmp-M, and Lmp-C, the amino terminus is immunogenic and membrane embedded, whereas the carboxy region of Lmp1 is surface exposed (86). The fact that the PA epitope mapped by pA259 is localized within this surface-exposed Lmp-C domain reinforces the relevance of this novel putative epitope for its surface exposure. Interestingly, the Lmp-C domain also contains another PA epitope previously identified by use of the mouse model (59).

BB0603 (P66) is a chromosomally encoded integral membrane porin (91) and, simultaneously, critical adhesin responsible for binding to mammalian integrins (92, 93). P66 is composed of N- and C-terminal intramembranous domains (∼53 kDa) with a short surface-exposed loop (∼4 kDa). P66 is upregulated in fed ticks and the mammalian host (92, 94, 95). Previous epitope mapping demonstrated that P66 contains numerous linear epitopes that are recognized by the sera of LD patients (96). The immunogenicity and virulence-associated adhesion function of P66 have led to an intense interest in this protein as a potential LD vaccine candidate (97). Importantly, a prior study demonstrated that immunization with the P66 protein results in the protection of four out of six mice against host-adapted, VlsE-expressing B. burgdorferi B31 (98). The PA epitope mapped by pA20B4 is localized within this small surface-exposed (the only-exposed) domain of P66, strongly supporting the relevance of the pA20B4-mapped PA epitope for its surface exposure by P66.

In addition to Lmp1, P66, and enolase, two other chromosomally encoded proteins (outer membrane proteins), BB0838 and BB0465, were also mapped by PA peptides. BB0838, mapped by p202B71, is a 120-kDa outer membrane protein which was experimentally demonstrated to be surface exposed (99). The fact that no transposon was inserted in the bb0838 gene upon global transposon mutagenesis (100, 101) strongly suggests that BB0838 plays an essential role in the life cycle of B. burgdorferi. BB0465, mapped by pA234, is consistently predicted to be surface exposed by numerous algorithms and the presence of a signal sequence (99).

Finally, plasmid-encoded BBO39 (ErpL), defined by pA152, is a surface protein expressed by B. burgdorferi in fed ticks and in the mammalian host (102, 103). This immunogenic protein was shown to be recognized by sera from tick-infected laboratory mice and LD patients (102, 104). This protein is thought to function as a factor H-binding protein that protects LD spirochetes from the mammalian complement.

Challenges in mapping conformational epitopes.

Identification of conformational epitopes using the phage-displayed peptide sequence data presented herein was challenging for reasons related to the complexity of host immune responses to B. burgdorferi infection, as detailed in our previous work (59). Briefly, epitope mapping via phage display, which is applicable to purified monoclonal antibodies (105), is apparently difficult to scale with increasing complexity in the setting of polyclonal antibodies elicited in response to diverse pathogen-expressed antigens, as exemplified by the outer surface proteins of B. burgdorferi.

Nevertheless, the representative sequence GLLQKPL constructed from the most frequently observed residues among 7-mer peptides uniquely recognized by day 28 rabbit antibodies in the present work comprises subsequences with similarity to protein segments that are identified as BLAST hits on the B. burgdorferi B31 proteome when the peptides are used as query sequences. The protein segments may thus comprise residues of epitopes recognized by the antibodies. In particular, BBA36 ¹⁷⁶VQKPV¹⁸⁰ (aligning with the C-terminal end of GLLQKPL via the consensus sequence XQKPX, where X denotes a nonpolar residue) may comprise residues of an epitope that could be targeted by antibodies for protective immunity against B. burgdorferi.

A future subunit vaccine against Lyme disease.

To date, with increasing incidences of LD worldwide (2–5, 7–9), there is an urgent need to develop an efficacious LD vaccine. An ideal vaccine should be safe, potent, and stable at storage and should provide life-long protection, preferably following only a single immunization (106, 107). Subunit vaccines have a number of advantages over conventional live-attenuated or whole killed counterparts (108). In addition to being safer, subunit vaccines can be manufactured in a well-controlled process and can be freeze-dried, which allows better lot-to-lot consistency and nonrefrigerated storage or transport, respectively (107). Importantly, the feasibility of generating an efficacious LD subunit vaccine using surface epitopes of B. burgdorferi is strongly supported by the protective efficacy of the OspA LA-2 epitope (109), whose monoclonal antibodies protected SCID mice from clinical signs of LD (110).

Overall, similar to a new formulation of a canine vaccine (Vanguard crLyme; Zoetis), the long-term goal for the development of a subunit LD vaccine for humans is to engineer a single protein that would contain multiple cross-protective targets. In contrast to the available canine vaccine, which is solely based on two plasmid-encoded proteins, OspA and OspC (7 variants) (111), the ideal vaccine formulation would contain protective epitopes from multiple surface antigens of B. burgdorferi sensu lato. In order to ensure target retention by LD spirochetes, it is desirable that targeted proteins be chromosomally encoded. Moreover, the vaccine should ideally disrupt all the aspects of B. burgdorferi infection cycle: survival in ticks, tick-mediated transmission to the natural (e.g., mice) and incidental (e.g., humans) host, and mammalian infection. The last aspect requires that a good vaccine also be efficacious at the LD spirochetemic phase, when LD spirochetes have not yet invaded collagen-rich tissues, the host environment that facilitates the ability of B. burgdorferi to hide from host antibodies (60). However, at this early stage, OspC already becomes permanently replaced by the abundant VlsE (112–114), the highly variable surface antigen that ensures, during the natural infection, a continually successful escape of B. burgdorferi from potent antibodies. Thus, the immediate goal toward the development of a subunit LD vaccine designed to be efficacious against the VlsE-expressing spirochete is identification of protective surface epitopes that, despite the abundant VlsE presence, would still remain accessible to protective antibodies. Thus, toward this goal, the very first step—of identifying putative surface epitopes that, despite VlsE, are reachable for the protective antibodies—was taken by our recent (59) and present studies. The next step, which is under way, is to systematically test these putative epitopes for their protective efficacy using mouse and/or rabbit infectivity models. If they are proven to be protective, these epitopes can then be used for the development of a long-awaited, efficacious subunit LD vaccine for humans.

In summary, our approach has involved an unbiased search for putative protective epitopes using a top-down strategy: from the delineation of the antibody repertoire that is unique to the protective rabbit sera to the mapping of newly identified PA epitopes to surface proteins of B. burgdorferi. Consequently, identification of these novel PA epitopes now allows us to experimentally test them as potential subunit vaccine candidates.

MATERIALS AND METHODS

Ethics statement.

Animals were maintained at Texas A&M University in an animal facility accredited by the Association for the Assessment and Accreditation of Laboratory Animal Care International (AAALAC). The experimental practices were approved by the Institutional Animal Care and Use Committee of Texas A&M University (IACUC 2017-0390) and were carried out in accordance with Public Health Service Policy on Humane Care and Use of Laboratory Animals (115), Guide for the Care and Use of Agricultural Animals in Research and Teaching (116), and Guide for the Care and Use of Laboratory Animals (117).

Bacterial strains and animal infection.

The wild-type B. burgdorferi B31-A3 strain (B31) (118) was used in the present study (Table 1). Spirochetes were grown in liquid Barbour-Stoenner-Kelly medium with 6% rabbit serum (referred to here as BSK-II; Gemini Bio-Products, USA) and incubated at 35°C under 2% CO₂. B. burgdorferi spirochetes used for challenge were passaged in vitro no more than three times.

Three female New Zealand White (NZW) rabbits of 12 to 14 weeks of age (Charles River Laboratories, USA) were first bled via the marginal ear vein to obtain preimmune serum. Then, each animal was inoculated intradermally at six sites along the spine with B31 (118) at 10⁶ cells per site as previously described (62). The three B31-inoculated NZW rabbits were bled via the marginal ear vein, and skin biopsy specimens (diameter, 3 mm; Integra Miltex, USA) were sampled around each inoculation site at weeks 1, 2, 3, 4, 5, 6, 7, and 8 postinfection (p.i.). Blood samples were immediately placed at 4°C and on the next day were centrifuged at 5,000 × g for 10 min to collect the serum. The obtained sera were stored at −80°C until used. Skin tissues that represented the six inoculation sites of each NZW rabbit were cultured in 5 ml of BSK-II with 0.02 mg ml⁻¹ phosphomycin, 0.05 mg ml⁻¹ rifampin, and 2.5 mg ml⁻¹ amphotericin B (the antibiotic cocktail) at 35°C under 2% CO₂. The skin tissue cultures were examined on a weekly basis by dark-field microscopy for up to 6 weeks for the presence of viable spirochetes. The three NZW rabbits were humanely sacrificed at week 8 p.i.

Three male C3H/HeJ (C3H) mice of 4 to 6 weeks of age were obtained from The Jackson Laboratory (ME, USA). The mice were infected via subcutaneous inoculation of 1 × 10⁴ total B31 spirochetes in the scapular region. The infection was confirmed by culturing 50 μl of blood drawn at day 7 p.i. via maxillary bleed and ear, heart, bladder, and tibiotarsal joint tissues harvested at day 28 p.i. in BSK-II with the antibiotic cocktail. Blood and heart tissues were transferred into 8-ml polystyrene tubes (Becton, Dickinson Labware, NJ, USA) that contained 3 ml of BSK-II. Bladder, tibiotarsal joint, or ear skin tissues were cultured in 1.7-ml polypropylene microcentrifuge tubes (Denville Scientific Inc., MA, USA) with 1.0 ml of BSK-II. All the tissues were incubated at 35°C under 2.5% CO₂ for up to 4 weeks. The presence of live spirochetes was confirmed by dark-field microscopy.

Phage display library.

First, 20 μl or 50 μl of each NZW rabbit or C3H mouse serum sample, respectively, and 10 μl of random phage display peptide library Ph.D.-7 (New England BioLabs, MA, USA) were first diluted in 200 μl of Tris-buffered saline buffer with 0.1% Tween 20 (TBST) and 1% bovine serum albumin (BSA) and then incubated at 25°C for 18 h (119). Antibody-bound phages were isolated by addition of 20 μl of protein G-agarose beads (Santa Cruz Biotechnology, Inc., TX, USA) to the phage-antibody mixture for 1 h. To remove unbound phages, the bead mixture was transferred to a 96-well MultiScreen-Mesh filter plate (EMD Millipore, MA, USA) using a 20-μm-pore-size nylon mesh. In order to remove unbound phages, vacuum was applied to the exterior of the nylon mesh. Then, the beads were washed four times with 100 μl of TBST buffer per each well. Antibody-bound phages were eluted with 100 μl of 100 mM Tris-glycine buffer (pH 2.2) and subsequently neutralized by adding 20 μl of 1 M Tris buffer (pH 9.1). The solution was then used for amplification of eluted phages by infecting bacteria per the manufacturer's instructions. Amplified phages were subjected to two rounds of biopanning, after which antibody-bound phages were isolated by the use of protein G-agarose beads. DNA was isolated by phenol-chloroform extraction and ethanol precipitation. Lastly, 21-nucleotide-long DNA fragments that encoded random peptides were amplified by PCR utilizing the following forward and reverse primers, respectively: 5′-AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGCTCTTCCGATCT(INDEX)TGGTACCTTTCTATTCTCACTCT-3′ and 5′-CAAGCAGAAGAGGGCATACGAGCTCTTCCGATCTAACAGTTTCGGCCGAACCTCCACC-3′. The INDEX of each forward primer was defined by the unique 6-nucleotide-long sequence. Thus, for each serum sample, a forward primer with the unique INDEX was used (119). After purification, the generated PCR-amplified DNA library was sequenced by use of an Illumina HiSeq 2500 platform at the University of Buffalo Genomics and Bioinformatics Core, New York State Center of Excellence Bioinformatics & Life Sciences, Buffalo, NY.

DNA read analysis.

The sequencing resulted in approximately 1.79 × 10⁸ DNA reads. The reads were demultiplexed based on the unique bar codes. Each read was tagged by the unique INDEX and contained a 21-nucleotide-long sequence which encoded the respective peptide. The 21-nucleotide sequences were extracted between positions 30 and 50 and translated into 7-mers in the first frame. The mean number of peptides that had no stop codon per serum sample was 5.9 × 10⁶, of which approximately 2.1 × 10⁵ peptides per sample were nonredundant. All the data were analyzed by the use of Python (https://www.python.org).

Comparison of peptide profiles of protective and nonprotective sera.

The strength of association between a peptide and a serum sample was measured as follows: a peptide, P, was associated with the day 28 serum sample if X(P), the lowest frequency of P among the day 28 serum samples, was higher than Y(P), the highest frequency of P among the day 14 rabbit serum samples or among the day 14 and day 28 rabbit serum samples. The strength of association was then measured by the size of the gap: [X(P) − Y(P) + 1]/[Y(P) + 1].

Mapping of protection-associated peptides onto linear amino acid sequences of B. burgdorferi surface proteins.

Nonredundant peptides were analyzed across all the serum samples to identify those that were unique to the day 28 sera of the three NZW rabbits. The resultant 309 peptides were then mapped to the amino acid sequences of 1,391 B. burgdorferi B31 proteins (64) (GenBank accession number GCF_000008685.2; https://www.ncbi.nlm.nih.gov/genome/738?genome_assembly_id=168382) via BLASTP analysis (120, 121). Only hits that showed at least 4 exact amino acid matches (identity threshold) were considered. To select for longer matches, hits of 5 or more amino acids in length without gaps were examined, and this resulted in 501 hits to 321 proteins. Furthermore, those peptides in this subset that were shared by multiple proteins were filtered out. Thus, the analysis resulted in the selection of only those peptides that had at least 5 identical amino acids in an aligned portion which exclusively aligned to a single protein. In this way, a possibility of the cross-matching of the same peptide to multiple proteins was avoided.

Alignment of peptides recognized uniquely by day 28 rabbit sera with conformational epitopes of outer surface proteins A, B, and C.

The sequences of the peptides recognized uniquely by day 28 rabbit sera were compared with the protein sequences of the reference genome of B. burgdorferi B31 via BLASTP analysis (121), according to a previously described approach (59). BLASTP parameters were adjusted as appropriate for short (7-mer) peptide sequences, with an E value threshold of 2,000, substitution matrix PAM30, and word size 2, without applying composition-based statistics (122); and the formation of gaps in the sequence alignments was suppressed by setting the BLASTP gap-opening and -extension penalties to their maximum possible value (32,767), as gaps correspond to insertions or deletions that are likely to disrupt antibody-antigen binding if they occur within epitope sequences. Only BLASTP hits with peptide-protein sequence alignments at least 4 residues in length were considered for further analysis. No cutoff values were set for minimum sequence identity. The latter allowed for the potential inclusion of epitope structures (e.g., helices) with residues that are at least partly oriented away from the antibody surface. In this way, the residues were more likely to be represented at phage-displayed sequence positions of high residue variability due to the diversity of side chains compatible with particular main chain conformations.

The peptides were aligned with protein antigen sequences that contained known epitopes and pertinent epitope-related sequences of B. burgdorferi, identified mainly via the Immune Epitope Database (IEDB; version 3.0) (123). IEDB searches were conducted from 26 to 28 July 2018 using the B Cell Assay Details interface (http://www.iedb.org/bcelldetails_v3.php) with “epitope source organism” set to “Borreliella burgdorferi (Lyme disease spirochete)” (including B. burgdorferi B31 and other strains). Conformational epitopes were identified by setting “epitope structure type” to “discontinuous epitopes.” Additionally, sequences that elicit antipeptide antibodies with biological activity against B. burgdorferi were identified by setting “epitope structure type” to “linear epitopes,” “1st immunogen epitope relation” to “epitope,” and “assay” to “biological activity.” Likewise, phage-displayed sequences bound by antibodies to B. burgdorferi were identified by setting “epitope structure type” to “linear epitopes” and “assay” to “phage display.” Phage-displayed sequences of protection-associated peptides in mice from previous work (59) were also included for comparison. As before, any other epitope-related sequences that have been identified only via enzyme-linked immunosorbent assay and Western blot analysis were excluded from consideration. These binding assays may potentially yield false-positive results due to artifactually unfolded protein epitopes and may also yield negative results due to the masking of epitopes (124).

A sequence logo was generated from the 7-mer phage-displayed peptides uniquely recognized by day 28 rabbit sera using the WebLogo (version 2.8.2) sequence logo generator (125). The most frequently occurring residues in each sequence position were combined to produce a single representative 7-mer sequence, which was compared with the BLASTP hits of the peptides to establish a consensus sequence that in turn was used to identify unique sequences in the B. burgdorferi B31 proteome yielding multiple BLAST hits with the peptides. If these hits mapped to a previously identified potential surface-exposed protein target for protective mouse antibodies (59), the protein sequence context was then further characterized by homology modeling with CPHmodels (version 3.2; http://www.cbs.dtu.dk/services/CPHmodels/) (126) vis-à-vis disorder prediction with the DisEMBL (version 1.5; http://dis.embl.de/) (127), PrDOS (http://prdos.hgc.jp/cgi-bin/top.cgi) (128), and IUPred2 (https://iupred2a.elte.hu/) (129) web interfaces. Default online server settings were used for all homology modeling and disorder prediction tools, with the PrDOS false-positive rate set to 5.0% and the IUPred2 prediction mode set for long disordered regions (e.g., disordered domains). For residues with atomic coordinates obtained via homology modeling, the DSSP (version 2.2.1) program (130) was used to assign secondary structure codes.

Statistics.

The statistical significance of the difference between the number of peptides associated with day 28 rabbit sera and the number of peptides associated with day 14 rabbit (and day 28 mouse) sera was measured using the permutation test. The permutation test was used due to the low numbers of samples (three serum samples from day 14, three serum samples from day 28, and one sample of pooled preimmune serum from the three NZW rabbits). For each possible permutation, the difference between the numbers of associated peptides was found and compared with the actual difference using the permutation test. A P value of <0.05 was considered significantly different.