Identification of conserved antigens for early serodiagnosis of relapsing fever Borrelia

Borrelia hermsii is a blood-borne pathogen transmitted by the argasid tick Ornithodoros hermsi. Since spirochaete clearance in mice is associated with an IgM-mediated response, an immunoproteomic analysis was used to identify proteins reactive with IgM. We report that IgM from both mice and human patients infected with B. hermsii not only reacted with the previously identified variable membrane proteins but also identified candidate antigens including heat-shock proteins, an adhesin protein, ABC transporter proteins, flagellar proteins, housekeeping proteins, an immune evasion protein, and proteins with unknown function. Furthermore, IgM reactivity to recombinant glycerophosphodiester phosphodiesterase was detected during early spirochaete infection and prior to a detectable IgG response. Lastly, a conserved hypothetical protein was produced in Escherichia coli and tested with immune serum against B. hermsii and Borrelia recurrentis. These results identify a much larger set of immunoreactive proteins, and could help in the early serodiagnosis of this tick-borne infection.

Relapsing fever Borrelia can reach upwards of 10 7 spirochaetes per ml of infected blood (Bryceson et al., 1970;Coffey & Eveland, 1967), and can be detected by microscopy. However, detection by microscopy lacks sensitivity between spirochaetaemic episodes. Glycerophosphodiester phosphodiesterase (GlpQ) has been used as a diagnostic antigen for infection with relapsing fever spirochaetes in North America and Africa (Nordstrand et al., 2007;Porcella et al., 2000;Schwan et al., 1996). However, in one report 79 % of Ethiopian patients infected with B. recurrentis did not have detectable IgG responses to recombinant GlpQ (rGlpQ) during early spirochaete infection as determined by an ELISA (Porcella et al., 2000).
With variable membrane proteins facilitating immune evasion, and the robust IgM-mediated response during early infection, understanding the broader antigenic repertoire against conserved proteins is important. However, little is known about proteins reactive with IgM in relapsing fever spirochaetes (Anda et al., 1996;Colombo & Alugupalli, 2008;Connolly et al., 2004). In this study, protein candidates reactive with IgM were identified using serum from mice and human patients infected with Borrelia hermsii. The amino acid identity of Borrelia orthologues was compared to B. hermsii, and early IgM responses to rGlpQ were determined using serum samples from Ethiopian patients that had previously tested negative to rGlpQ during early spirochaete infection (Porcella et al., 2000). Furthermore, a hypothetical protein from B. hermsii was produced and the immunogenicity of this recombinant protein was tested as a possible diagnostic antigen.

METHODS
Animal inoculation and immune sera collection. Low-passage B. hermsii DAH isolate was grown in BSK medium containing 12 % rabbit serum (Barbour, 1984;Battisti et al., 2008) to 1.0610 8 spirochaetes ml 21 , and used to needle inoculate three mice intraperitoneally with 1.0610 5 spirochaetes per mouse. Prior to, and for the first 4 days and the seventh day after inoculation, blood was obtained, and the spirochaetaemias were quantified by brightfield microscopy (Schwan et al., 2003). Immune serum collected from the mice was pooled. All animal handling was in compliance with the Rocky Mountain Laboratories Animal Care and Use Committee.
Human serum was available from two patients who were infected with B. hermsii. Serum samples from both patients were collected within 1 month of the initial spirochaetaemic episode. Serum samples from patients infected with B. recurrentis were available from a hospital in Addis Ababa, Ethiopia, as previously described (Porcella et al., 2000). The acute-phase serum samples were collected shortly after hospitalization; the convalescent-phase serum samples were collected 8-13 days later. Acute-phase serum samples used in this study had tested IgG negative to rGlpQ while convalescent-phase samples were positive (Porcella et al., 2000).
Soluble and insoluble fractions were resuspended in 200 ml 2D Rehydration Sample Buffer 1 containing 2 mM TBP and 0.2 % ampholytes, and used to rehydrate 11 cm pH 4-7, pH 5-8, or pH 7-10 ReadyStrip IPG Strips (Bio-Rad). Rehydration was performed under active conditions, and IPG strips were focused for a total of 25 000 V h using the PROTEAN IEF (Bio-Rad).
IPG strips were equilibrated for 20 min as specified by the manufacturer (Bio-Rad) with equilibration buffer 1 containing 5 mM TBP. IPG strips were transferred onto 10.5-14 % Criterion Precast Gels (Bio-Rad), electrophoresed, and stained with the colloidal Coomassie SimplyBlue SafeStain (Invitrogen) or transferred onto a nitrocellulose membrane using the iBlot Dry Blotting System (Invitrogen).
Mouse serum samples collected prior to and 7 days after infection, and immune serum from human patients infected with B. hermsii, were used to probe 2D immunoblots at a 1 : 100 dilution in I-Block blocking reagent (TROPIX). Secondary mAbs were goat anti-mouse IgM-HRP (Invitrogen) at 1 : 4000 or goat-anti-human IgM-peroxidase (Sigma-Aldrich) at 1 : 20 000. Membranes were developed by chemiluminescence with the ECL Western Blotting Detection Reagent (GE Healthcare), following the manufacturer's instructions. Images from 2D gels and 2D immunoblots were acquired with the GelDoc (Bio-Rad), and overlaid in order to select IgM-reactive protein spots.
2D gel protein sample preparation for mass spectrometry.
Protein samples of interest were excised from colloidal Coomassiestained (Invitrogen) gels using the 1.0, 2.0, or 3.0 mm Harris UniCore sampling punch tool (Sigma-Aldrich) and placed into a new Ultra Low Attachment 96-well Costar Plate (Corning Inc.). Gel plugs were destained, reduced, alkylated, and then rehydrated in 25 ml of 10 ng ml 21 sequencing grade trypsin solution (Promega).
After an overnight incubation at 37 uC, supernatants containing eluted trypsinized peptides were removed and placed into a new Ultra Low Attachment 96-well Costar Plate. Peptides were further extracted by adding 5 % formic acid (Sigma-Aldrich) and 50 % acetonitrile (ACN) to the gel plugs for 30 min at 37 uC. The solution was removed and pooled with the supernatants from the overnight trypsinization and the extraction repeated. Trypsinized peptides were dried under vacuum (Jouan) and dissolved in 30 ml of 0.1 % trifluoroacetic acid (TFA) (JT Baker).
Peptide samples were desalted with the C18 ZipTips (Millipore), following the manufacturer's instructions with minor modifications. Briefly, ZipTips were conditioned with 90 % methanol and 0.1 % TFA, equilibrated with 0.1 % TFA, and peptides bound by pipetting. Peptides were washed with 0.5 % acetic acid (Sigma-Aldrich), eluted into a Thermo-Fast 96-well skirted plate (ABgene House) with 70 % ACN and 1 % acetic acid, and dried under vacuum centrifugation for mass spectrometric analysis.
Mass spectrometry, B. hermsii database search, and basic local alignment search tool (BLAST) analysis. Identification of proteins from 2D gels was performed by mass spectrometry, and this was repeated four separate times. Tryptic digests were analysed by coupling the Nanomate (Advion BioSciences), an automated chipbased nano-electrospray interface source, to a quadrupole-time of flight mass spectrometer, QStarXL MS/MS System (Applied Biosystems/ Sciex). Peptide sequence information was provided by MS/MS. AnalystQS software (Applied Biosystems/Sciex) was used for data acquisition. Data processing and database searching were performed with the MASCOT software (Matrix Science). A protein database was generated from the genome sequence of B. hermsii DAH and submitted to MASCOT as a separate database for searching. Generally, MASCOT ion scores greater than 24 correspond to a probability of 95 % or greater that the peptide match is not a random event.
To determine the amino acid identity of non-variable membrane proteins, protein BLAST analysis was performed at http://blast. ncbi.nlm.nih.gov/Blast.cgi. Also, individual variable small proteins (Vsps) and variable large protein (Vlp) 5 were named according to BLAST analysis. A BLAST threshold E-value of 1.0610 24 was used to identify borrelia proteins.
bh0238 was expressed using the pBAD/TOPO ThioFusion expression system following the manufacturer's instructions (Invitrogen). Forward and reverse primers used to amplify bh0238 from B. hermsii DAH genomic DNA were 59-ATGACTAGATTTTTAGTGGAGG-TTAGCATGAG-39 and 59-TTTTATTGAAAAGAGTACCCATTTA-TCATCC-39, respectively. PCR amplification consisted of 34 cycles with a denaturing temperature of 94 uC for 30 s, annealing temperature of 55 uC for 30 s, and an extension temperature of 72 uC for 2.5 min. After confirming the correct orientation and nucleotide sequence within the pBAD/TOPO ThioFusion vector using the Vector NTI software package (Invitrogen), bh0238 was induced with a final concentration of 0.2 % arabinose and purified using a His-Bind Quick Column (Novagen) following the manufacturer's instructions.
1D SDS-PAGE and 1D immunoblotting. For detecting anti-GlpQ Abs during early mouse infection, whole-cell lysates of B. hermsii DAH, 6 mg His-tagged rGlpQ (Schwan et al., 2003), and Escherichia coli containing the empty shuttle vector were separated by 1D SDS-PAGE as previously described (Schwan et al., 1989) and transferred to a nitrocellulose membrane as stated above. Murine immune sera collected for the first 4 days post-inoculation were used at a dilution Serum samples from three mice infected with B. hermsii were collected, pooled, and used to probe immunoblots (A). Serum from patient 1 (B) and patient 2 (C) was collected after B. hermsii infection. Proteins identified by mass spectrometry from 2D gels are numbered on 2D immunoblots (A-C). 1D immunoblots confirmed that murine pre-immune sera (E) and human serum from an uninfected donor were negative to B. hermsii whole-cell lysate (F). Molecular mass standards are shown on the left. of 1 : 100 to probe immunoblots. To differentiate between IgM and IgG responses, goat-anti-mouse IgM-HRP or HRP-recombinant-Protein A (ZYMED) were used at a 1 : 4000 dilution.
rBH0238 produced in the pBAD/TOPO ThioFusion system (Invitrogen) was examined as follows. E. coli harvested prior to and 2 h after induction or B. recurrentis lysate and 2 mg of purified rBH0238 were separated by SDS-PAGE (Schwan et al., 1989), and stained with the colloidal Coomassie SimplyBlue SafeStain (Invitrogen) or transferred onto a nitrocellulose membrane as stated above. Immunoblots were probed with a monoclonal anti-polyhistidine peroxidase conjugate (Sigma-Aldrich) at a 1 : 2000 dilution, immune serum from mice infected with B. hermsii at a 1 : 100 dilution, or acute and convalescent serum from Ethiopian patients infected with B. recurrentis at a 1 : 500 dilution. The secondary molecule used was HRP-recombinant-Protein A (ZYMED). In addition, convalescent-phase serum samples were used to determine seroconversion rBH0238. ELISA with rGlpQ. IgM reactivity to rGlpQ was determined by ELISA as previously described, with minor modifications (Porcella et al., 2000). Sera from Ethiopian patients infected with B. recurrentis were available from their acute phase of infection. Immune and negative control sera were diluted 1 : 100 in diluent (16 PBS, 5 % horse serum, 0.05 % Tween 20, 0.001 % dextran sulfate) and tested in triplicate. IgM binding to rGlpQ was determined using goat-antihuman IgM-peroxidase (Sigma-Aldrich) at 1 : 20 000 followed by the addition of a substrate consisting of 50 % 2,29-azino-di-(3-ethylbenzthiazoline sulfonate). After 30 min incubation, the A 405 was measured with a Labsystems Multiskan Plus microtitre plate reader (Fisher Scientific). Samples were considered positive if their mean absorbance was more than three standard deviations above the mean of negative control serum samples.

Evaluation of mouse and patient serum samples infected with B. hermsii
All mice inoculated with B. hermsii DAH became spirochaetaemic within 3 days after injection (data not shown). IgM from these mice and patients infected with B. hermsii recognized antigens ranging from 20 to 100 kDa ( Fig. 1A-C), while murine preimmune sera or sera from uninfected human patients had little reactivity to B. hermsii whole-cell lysate ( Fig. 1E and F).

Protein spot selection for mass spectrometry
Image overlays from 2D electrophoresis gels and 2D immunoblots probed with immune serum from mice or humans identified 114 protein candidates for mass spectrometry analysis within the isoelectric focusing range of pH 5-8 ( Fig. 1A-D). Vsps, Vlp 5, and the nucleoside-binding proteins were the only IgM-reactive proteins identified in pH 4-7 and pH 7-10 2D gels (data not shown), and of these proteins, Vlp C54silD and the nucleoside-binding protein were the only additional proteins not identified in the pH range 5-8 (Table 1). Also, there were no apparent differences between protein fractions that were tested by 2D electrophoresis and immunoblotting.

B. hermsii database analysis and amino acid identity of proteins reactive with immune IgM
Proteins identified by mass spectrometry were grouped into eight categories where, except for Vsps, protein plugs containing an individual protein are shown (Table 1). Given that several Vsps were identified within a protein plug, mapping peptides identified by mass spectrometry to their respective Vsp confirmed that unique peptides were identified from each Vsp (data not shown).
While 114 protein samples were analysed by mass spectrometry, identifying the same protein in several spots was common. This could be due to post-translational modifications including lipidation of the identified variable membrane proteins and several of the ABC transporter proteins (Bono et al., 1998;Burman et al., 1990;Kornacki & Oliver, 1998;Shang et al., 1998). Also, protein binding to residual DNA and RNA could cause heterogeneous protein migration.
Amino acid alignments of non-variable membrane proteins indicated the conserved nature of the identified proteins when compared to B. hermsii (Table 2). Similar to GlpQ and the factor H/plasminogen-binding protein (FhbA) (Hovis et al., 2006;Rossmann et al., 2007;Schwan et al., 1996), no orthologue of hypothetical protein FJ446703 was identified in Borrelia burgdorferi (Table 2). Additionally, BLAST analysis identified P66, FhbA, and hypothetical proteins FJ446703 and BH0238 as unique proteins within the genus Borrelia.
Comparisons of IgM-reactive protein candidates between human patients and mice infected with B. hermsii Immunoblot comparisons confirmed similar IgM profiles in human patients and mice infected with B. hermsii, where 74 % of the protein spots were recognized by both human and mouse serum samples (Fig. 2). In addition, lack of IgM reactivity to Vlp 5 from human serum samples was likely due to the expression of other vlps during their infection.
IgM reactivity to rGlpQ during early B. hermsii and B. recurrentis infection Murine and human IgM bound to rGlpQ during early spirochaete infection (Figs 3 and 4). Murine IgM bound to rGlpQ at the expected molecular mass by the fourth day after spirochaete infection (Fig. 3A), and prior to detection of IgG antibodies (Fig. 3B), while preimmune serum was negative to the antigens tested (Fig. 3C).  IgM binding to rGlpQ was detected in acute-phase serum samples from eight of 15 Ethiopian patients infected with B. recurrentis (Fig. 4). In addition, these serum samples were negative when previously tested for IgG responses against rGlpQ (Porcella et al., 2000).

Expression and immunogenicity of rBH0238
Given the conserved nature of BH0238, its apparent immunogenicity in B. hermsii, and preliminary data suggesting that its orthologue is immunogenic in B. recurrentis (data not shown), this gene was expressed as a potential diagnostic antigen. Inducing transformed E. coli with a final concentration of 0.2 % arabinose produced a protein band at the expected molecular mass of 48 kDa (Fig. 5A). Immunoblotting with an anti-polyhistidine mAb confirmed reactivity with a protein at 48 kDa and 35 kDa, suggesting degradation of rBH0238 (Fig. 5B). In addition, there was no mAb reactivity in uninduced E. coli (Fig. 5B).
An additional objective of our study was to determine early IgM reactivity to rGlpQ (Porcella et al., 2000). In mice, IgM reactivity to rGlpQ was detected within 4 days of spirochaete infection. Furthermore, IgM responses to rGlpQ were detected from eight additional patients infected with B. recurrentis. While promising, these results also suggest the importance of identifying additional immunogens for early detection of relapsing fever spirochaetes, especially for infections where the organisms are below the level of detection.  Three diagnostic antigens of interest include FhbA (Hovis et al., 2006), and hypothetical proteins FJ446703 and BH0238. The factor H-binding protein and hypothetical protein FJ446703 could be diagnostic antigens in North America because they discriminate between relapsing fever spirochaetes and Lyme-disease-causing spirochaetes. While rBH0238 was recognized with serum from patients infected with B. recurrentis, studies indicate that patients with Lyme disease also seroconvert to this antigen (data not shown), limiting rBH0238 as an antigen that could discriminate between the two species of Borrelia. However, given this protein's immunogenicity and BLAST analysis indicating that it is a Borrelia-specific protein, combining rBH0238 with additional immunogens may aid in diagnosing infection of relapsing fever spirochaetes.
Two factors complicating the treatment of relapsing fever patients include high spirochaete counts associated with a poorer prognosis (Melkert, 1991) and the reported misdiagnosis of relapsing fever Borrelia (Nordstrand et al., 2007). Collectively, the data reported here identify proteins whose immunogenic properties can be further investigated and may be useful for early serodiagnosis. In addition, these findings may aid in managing relapsing fever Borrelia in endemic regions and have the potential to alleviate healthcare costs in developing countries.