INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Bartonella bacilliformis is the etiologic agent of bartonellosis (Carrion's disease), a unique biphasic disease that is prevalent among inhabitants of the western slopes of the Andes Mountains in Columbia, Ecuador, and Peru. The primary phase of the disease is known as Oroya fever and is characterized by a very severe hemolytic anemia that was fatal in approximately 40% of cases in the preantibiotic era. The cause of death is primarily the severe anemia, in which nearly 100% of the erythrocytes are parasitized by bartonellae. Bartonellosis also induces transient immunosuppression that results in the onset of potentially life-threatening opportunistic infections such as salmonellosis, shigellosis, and tuberculosis. The secondary phase of bartonellosis, known as verruga peruana, manifests itself 4 to 8 weeks after the onset of Oroya fever. This phase is rarely fatal and is characterized by nodular eruptions involving the face, neck, and extremities (3, 7, 24). Recently, variants of classical Peruvian bartonellosis in which only the verruga phase of the disease was present were observed in the lowland province of Manabi in Ecuador (2). This has led to suggestions that the milder form of bartonellosis may be caused by less-virulent strains of B. bacilliformis (2).

In the valleys of the Andes where bartonellosis is endemic, approximately 60% of the population are seropositive for the bacterium and 5 to 10% of the population are active carriers of the disease (14). Outbreaks of bartonellosis can reach epidemic proportions in these areas, such as the outbreak of 1870 in Oroya, Peru (after which the disease was named), in which more than 7,000 railroad workers died of the disease. More recently, delayed diagnosis resulted in the death of 14 people (88% case fatality) in an epidemic in Peru in 1987 (9). Bartonellosis thus remains a significant health problem in regions where it is endemic and requires research attention for the development of rapid tests for diagnosis and treatment of the disease. Humans are the only known natural reservoir for B. bacilliformis, which suggests that eradication of the disease is achievable by vaccinating the population in the regions where the disease is endemic.

The skin lesions of the verruga phase of Carrion's disease are very similar to the lesions that are associated with bacillary angiomatosis (BA), a vascular proliferative disease that is mostly seen among immunocompromised individuals. B. henselae, one of the recently included members of the genus Bartonella, was identified as a causative agent of BA (21). B. henselae has also been implicated in the etiology of cat scratch disease (CSD) and a number of other disease syndromes. Based on the phylogenetic similarities between B. bacilliformis and B. henselae, it is conceivable that factors common between these two organisms may be responsible for the pathological similarities between verruga peruana and BA. Identification and characterization of such factors could lead to a better understanding of the mechanisms of pathogenesis employed by these organisms.

The present study was initiated to identify and characterize immunogenic proteins of B. bacilliformis that are expressed during the infectious process. We screened a genomic DNA lambda library with serum from a patient who had the chronic verruga phase of bartonellosis and were able to isolate several immunoreactive clones expressing bartonella-specific proteins (18). In this paper we describe the cloning and characterization of an immunoreactive 43-kDa lipoprotein of B. bacilliformis.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains, growth conditions, and plasmids. B. bacilliformis strains KC584 and KC583 were obtained from the American Type Culture Collections (ATCC), Manassas, Va. Both strains were grown on heart infusion agar plates supplemented with 5% defibrinated rabbit blood (BBL-Becton Dickinson, Cockeysville, Md.) at 28°C for 7 to 14 days under humid conditions. B. henselae Houston-1 (ATCC 49822) strain was grown on the same plates at 32°C in the presence of 5% CO₂ for 5 to 7 days. Bacteria were harvested and resuspended in phosphate-buffered saline (PBS). All E. coli strains were grown at 37°C in media supplemented with appropriate antibiotics.

Human sera. The anti-B. bacilliformis human sera used in this study had indirect fluorescent-antibody assay (IFA) titers ranging from 512 to 1,024. These sera were from clinical cases of bartonellosis from Peru. The sera and their titers were generously provided by Judith Chamberlain of the Department of Preventive Medicine and Biometrics, Uniformed Health Services University, Bethesda, Md. The anti-B. henselae human sera used in this study were from suspected CSD cases and were submitted to the Centers for Disease Control and Prevention for confirmative diagnosis. These sera had high titers (2,048) for B. henselae as determined by IFA. Sera with negative IFA titers (32) to Bartonella spp. were used as controls.

Preadsorption of sera with E. coli antigens. All of the sera used in this study were preadsorbed with E. coli antigens to remove cross-reacting antibodies prior to their use in Western blotting. Then, 1.5-ml aliquots of fresh overnight cultures of E. coli strains XL1-Blue MRF' and JM105 were pelleted by centrifugation. The supernatant was discarded, and the pellet was resuspended in 200 µl of protoplasting buffer (15 mM Tris-HCl, pH 8.0; 0.45 M sucrose; 8 mM EDTA). Next, 5 µl of lysozyme (50 mg/ml) was added, and the cells were incubated at room temperature for 15 min, followed by a 2-min incubation at 37°C. The sera were diluted to a volume of 500 µl and added to the lysed cells. The mixture was incubated at room temperature for 1 h, with periodic mixing. This was followed by centrifugation for 10 min at 10,000 rpm in a microcentrifuge to remove the cellular debris. The supernatant was carefully collected and used for immunoassays.

DNA sequencing and analysis. DNA sequencing was done using a model 377 automated nucleic acid sequencer (Applied Biosystems, Foster City, Calif.). DNA and protein analyses were performed with the Wisconsin software package (version 8) of the Genetics Computer Group (Madison, Wis.) and DNASTAR (Lasergene, Inc.).

DNA manipulations. The primers that were used for subcloning ORF-401 were as follows: forward primer (5'-TGA GCA GAA TCC AAT GAG AAG ATT CAT GTA-3') and reverse primer (5'-ACC TAC CTG CAG TAA ACT GAT ATC ATA GCG-3'). The underlined sequences indicate sites for the restriction enzymes EcoRI and PstI for the forward and reverse primers, respectively.

PCR-directed mutagenesis. The technique of overlap extension by PCR was used to mutagenize the polypeptide encoded by ORF-401 (9). Two complementary 30-bp oligonucleotides with the sequences 5'-AGGTTCTAGATCTGGCACACAGCGTTTTTT-3' (oligonucleotide A) and 5'-AAAAAACGCTGTGTGCCAGATCTAGAACCT-3' (oligonucleotide B) were synthesized for the mutagenesis. The underlined residues indicate the positions at which these oligonucleotides differ from the wild-type sequence to produce a CysSer change at position 33 of the polypeptide. The nucleotides in boldface denote restriction enzyme recognition sequences. These oligonucleotides were also designed to introduce an XbaI restriction site into the resultant PCR product. In the first round of PCR, oligonucleotides A and B were used in separate reactions along with two more oligonucleotides, 5'-TGAGCAGAATTCAATGAGAAGATTCATGTA-3' (oligonucleotide C) and 5'-ACCTACCTGCAGTAAACTGATATCATAGCG-3' (oligonucleotide D), to generate two overlapping PCR products. These PCR products were purified and used as templates in the second round of PCR with oligonucleotides C and D to generate a mutated PCR product of 1,300 bp. The mutation was confirmed by sequencing and restriction digestion with XbaI, and the PCR product that contained the mutation was cloned into the vector pKK223-3 to generate the mutant clone pKMUT-9. Expression of the mutated polypeptide by pKMUT-9 was confirmed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and Western blot analysis.

Labeling of lipoproteins with [³H]palmitate. The E. coli strain JM105 carrying pKIP-7, pKMUT-9, or the vector pKK223-3 was grown in Luria broth supplemented with ampicillin (50 µg/ml). When the optical density at 550 nm reached 0.3 to 0.5, 1 mM isopropyl--D-thiogalactopyranoside (IPTG) and 100 µCi of [9,10(n)-³H]palmitate were added, and the incubation was continued for 2 h. The cells were pelleted by centrifugation, washed twice in PBS, and resuspended in 1× sample buffer in preparation for SDS-PAGE. After electrophoresis, proteins were fixed by incubating the gel in 5 to 10 volumes of glacial acetic acid-methanol-water (10:20:70) at room temperature with gentle rocking (23). Gels were impregnated with En³Hance (DuPont NEN), treated with a gel-drying solution, dried, and fluorographed overnight at 80°C.

Antibody production. Peptides were dissolved in deionized water at a concentration of 2 mg/ml. The polyclonal rabbit antiserum was produced at the animal facility at Georgia State University. A cocktail consisting of equimolar amounts of each of five peptides was diluted 1:1 with Freund complete adjuvant (Sigma). The diluted peptide cocktail was injected into New Zealand White rabbits (Myrtle's Rabbitry, Inc., Thompson Station, Tenn.) for antibody production. Animals were boosted after 2 weeks with the peptide cocktail mix at a concentration of 1 mg/ml in Freund incomplete adjuvant (Sigma). Rabbits were bled after 3 weeks, and the sera were purified by centrifugation. For production of polyclonal rabbit antisera to each of the individual peptides, each of the five peptides was administered to rabbits using a protocol similar to that used for the peptide cocktail.

SDS-PAGE. E. coli strains harboring the recombinant plasmids or vectors were induced with 1 mM IPTG prior to SDS-PAGE analysis. Proteins from E. coli and Bartonella strains were solubilized in 1× sample buffer (Novex, San Diego, Calif.) at 100°C for 5 min and subjected to electrophoresis on precast 4 to 20% gradient Tris-glycine gels (Novex). Gels were run in Tris-glycine SDS-PAGE running buffer at 125 V. Separated proteins were either transferred to nitrocellulose, stained with Coomassie brilliant blue, or used for autoradiography as described above.

Western blotting. Proteins for immunoblotting were electrophoretically transferred to 0.45-µm (pore size) nitrocellulose membranes (Novex) according to the protocol of Towbin et al. (27). Transfer was performed in Tris-glycine buffer with 20% methanol for 1 h at 100 V with cooling. Membranes were blocked overnight at 4°C in blocking buffer consisting of 5% nonfat milk powder in Tris-buffered saline-Tween 20 (20 mM Tris, pH 7.5; 150 mM NaCl; 0.05% Tween 20). Membranes were reacted with the primary antibody solutions (in blocking buffer) for 2 h at room temperature. The secondary antibody was either goat anti-rabbit or anti-human immunoglobulin G, conjugated to horseradish peroxidase (Kirkegaard and Perry Laboratories, Inc., Gaithersburg, Md.) and diluted 1:5,000 in blocking buffer. Membranes were developed with a standard chromogenic substrate (TMB Membrane Substrate System; Kirkegaard and Perry).

Nucleotide sequence accession number. The nucleotide sequence of the gene encoding the 43-kDa antigen has been deposited in the GenBnk database and has been given the accession no. AF157831.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Cloning of the 43-kDa antigen gene. A B. bacilliformis genomic library was constructed using the lambda ZapII system as described previously (18). The library was screened with serum from a patient from Ecuador who had the chronic verruga phase of bartonellosis. One of the clones (pBIPIM-17) isolated as a result of this immunoscreening expressed a fusion protein that was encoded by a partial open reading frame of 741 bp and did not contain a putative ATG start codon (data not shown). To obtain the entire open reading frame, the B. bacilliformis genomic library was rescreened by plaque hybridization using the pBIPIM-17 insert as the probe. This screening resulted in the isolation of three hybridizing clones, pBIPH-1, pBIPH-2, and pBIPH-3, all of which were revealed by DNA sequencing to contain the full-length open reading frame. The deduced 1,206-bp open reading frame was capable of encoding a protein of 401 amino acids (ORF-401) with an estimated molecular mass of approximately 43 kDa (Fig. 1). Examination of the sequence (Fig. 1) revealed a second in-frame ATG codon 9 bp downstream of the first ATG codon (ATG-1) (Fig. 1). This ATG codon is preceded by a Shine-Dalgarno (SD) sequence that is identical to the SD sequence of E. coli (Fig. 1). Also, based on E. coli SD sequences, the SD preceding the second ATG is more optimally located. Thus, it is possible that the second ATG within the open reading frame is an alternative start site. We have therefore designated the second ATG codon as ATG-2 to indicate the possibility that it is used as an alternative site for translational initiation, which would result in a protein of 397 amino acids.

View larger version (45K): [in this window] [in a new window]   FIG. 1.   Nucleotide and deduced amino acid sequences of ORF-401, expressing the 43-kDa antigen. Amino acids are shown in single-letter code. An asterisk indicates the stop codon. Translation has been shown from the first in-frame ATG codon (ATG-1), which has been highlighted. A second possible translational start site (ATG-2) downstream of ATG-1 has also been indicated by highlighting. The SD sequences preceding ATG-1 and ATG-2 have been underlined.

The 43-kDa antigen is a homologue of LppB/NlpD. A search through the nucleic acid and protein databases using the BLAST search tool (1) revealed that the predicted amino acid sequence of ORF-401 is homologous to that of a bacterial lipoprotein that was recently identified and designated as novel lipoprotein D (NlpD)/lipoprotein B (LppB). Alignment of the deduced amino acid sequence of ORF-401 with some of the known LppB/NlpD sequences (Fig. 2) revealed that the homology extends throughout the length of the protein and is particularly striking within a region of approximately 100 amino acids near the carboxyl end of the protein (Fig. 2). This strong sequence similarity suggests that the 43-kDa antigen is a homologue of the LppB/NlpD proteins.

View larger version (110K): [in this window] [in a new window]   FIG. 2.   Alignment of the deduced amino acid sequence of the B. bacilliformis NlpD/LppB homologue with other NlpD/LppB proteins. The alignment was performed with the CLUSTAL program, and the shading was done with BOXSHADE. The dark shading indicates identical residues, while the light shading indicates conserved substitutions. The signal peptide sequence is underlined, and the signal peptidase cleavage site is indicated by the arrowhead. The regions corresponding to synthetic peptides are numbered and are indicated by discontinuous lines. The RGD sequence is indicated by the heavy underline. Hi, H. influenzae, Hs, H. somnus, Ec, E. coli, Bb, B. bacilliformis. The overall identity and similarity values for the homology of the B. bacilliformis NlpD/LppB protein with the NlpD/LppB homologues used in the alignment are as follows: H. influenzae, identity, 26.6%, similarity, 42.2%; H. somnus, identity, 19.4%, similarity, 37.6%; E. coli, identity, 24.4%, similarity, 34.7%.

ORF-401 expresses a 43-kDa antigen. To verify that ORF-401 codes for a 43-kDa antigen, the open reading frame was subcloned into pKK223-3 and expressed in Escherichia coli. The resulting recombinant clone, pKIP-7, was found to express a 43-kDa antigen (Fig. 3, lanes 3 and 4). The recombinant antigen migrated with an antigen of the same size in the cell lysates of two strains of B. bacilliformis, KC584 (Fig. 3, lane 5) and KC583 (Fig. 3, lane 6). Lane 2 represents the immunoreactive fusion protein that is expressed from the recombinant vector, pBHIM-17. The 43-kDa band was not present in E. coli cells containing the plasmid vector, pKK223-3 (Fig. 3, lane 1), indicating that the expressed recombinant antigen was specific to pKIP-7. The Western blot shown in Fig. 3 was reacted with a pool of sera from patients with clinical bartonellosis. The other immunoreactive bands seen in lanes 1 to 4 are most likely due to the presence of antibodies in the human sera that cross-react with E. coli antigens since they are also present in E. coli cells containing the plasmid vector (Fig. 3, lane 1). Preadsorption of the sera with E. coli antigens reduced the cross-reactivity but could not eliminate it.

View larger version (53K): [in this window] [in a new window]   FIG. 3.   Western immunoblot showing reactivity of the 43-kDa antigen with a pool of sera from patients with bartonellosis. Lane 1, E. coli JM105 cells containing the plasmid vector, pKK223-3; lane 2, E. coli XL1-Blue cells containing the recombinant plasmid, pBHIM-17, expressing the truncated fusion protein; lanes 3 and 4, E. coli JM105 cells containing the recombinant plasmid, pKIP-7, expressing the 43-kDa antigen under uninduced (lane 3) and induced (lane 4) conditions; lanes 5 and 6, B. bacilliformis strains KC584 and KC583, respectively. Numbers to the right are approximate molecular masses, in kilodaltons (kDa). The arrow indicates the position of the 43-kDa antigen.

The 43-kDa antigen is a lipoprotein. The possibility that, like the LppB and NlpD proteins, the 43-kDa antigen is lipid modified was tested by studying the incorporation of [³H]palmitic acid into the protein. E. coli JM105 cells carrying the recombinant plasmid pKIP-7 were induced with 1 mM IPTG for expression of the 43-kDa antigen and incubated with [³H]palmitic acid. SDS-PAGE analysis of whole-cell lysates of palmitate-labeled E. coli revealed that the 43-kDa antigen is efficiently labeled by [³H]palmitic acid, as is evident from the prominent band migrating at 43 kDa (Fig. 4A, lane 2). This band is absent in cell lysates of E. coli harboring the vector pKK223-3 that were labeled under the same conditions (Fig. 4A, lane 1), confirming that it is specific for E. coli expressing the 43-kDa antigen.

View larger version (45K): [in this window] [in a new window]   FIG. 4.   [3H]palmitate labeling of E. coli cells expressing the 43-kDa antigen. E. coli JM105 cells that contained either pKIP-7 or pKMUT-9 were induced with 1 mM IPTG for expression of the wild-type protein (lane 2) or the mutant protein (with uncleaved signal peptide) (lane 3), respectively. Lane 1 represents a control E. coli JM105 strain containing the plasmid vector, pKK223-3. (A) Cells were labeled with [3H]palmitate as described in the text and analyzed by SDS-PAGE and autoradiography. (B) Cells were analyzed by Western immunoblotting using polyclonal rabbit antiserum raised to synthetic peptides of the 43-kDa antigen. Numbers to the left of panel A are approximate molecular masses, in kilodaltons (kDa). The arrow indicates the position of the wild-type, mature form of the 43-kDa antigen.

Posttranslational modification of the 43-kDa antigen. The amino terminus of the polypeptide encoded by ORF-401 contains a 32-amino-acid sequence bearing strong homology to signal peptides found in secreted bacterial proteins (Fig. 2). A typical bacterial signal sequence consists of three distinct regions: a basic region at the N terminus consisting of two to four basic amino acids, a core region consisting primarily of hydrophobic amino acids, and a cleavage region consisting of the consensus sequence Leu-Ala-Gly-Cys (11, 19). As is evident from Fig. 2, the signal peptide of the 43-kDa antigen has a predominance of positively charged amino acids at the N terminus (Arg2-Arg3 and Lys8) and a core that consists primarily of hydrophobic amino acids. It also contains the sequence Ile-Thr-Gly-Cys, which conforms to the consensus sequence Leu(Ile)-Ala(Ser/Thr)-Gly(Ala)-Cys that is required for cleavage of the signal peptide and lipid modification. In bacterial lipoproteins, the cysteine residue is the site at which posttranslational lipid modification of the protein occurs, which is followed by cleavage of the signal peptide (11, 19). To test the possibility that the 43-kDa antigen was modified at Cys-33 by a similar mechanism, the cysteine codon (TGT) was changed to a serine codon (TCT) by introducing a single point mutation (GC) at the appropriate position in pKIP-7. The expression of the protein from pKMUT-9, the plasmid carrying the mutation, was studied by Western blot analysis (Fig. 4B) using polyclonal antiserum raised against the 43-kDa antigen (described in a following section). The CysSer change was accompanied by a shift in the mobility of the mutant protein, as indicated by the appearance of an immunoreactive protein band (Fig. 4B, lane 3) that migrates more slowly than the band that corresponds to the wild-type protein (lane 2). The change in migration is consistent with the accumulation of a precursor form of the 43-kDa antigen which has an intact signal peptide. The size of the mutant protein estimated by its rate of migration is approximately 46 kDa, which correlates well with the predicted increase (3.5 kDa) in the M_r of the protein due to the intact signal peptide.

Immunoreactivity of the 43-kDa antigen with individual sera from patients with clinical bartonellosis. To study the reactivity of the 43-kDa antigen with individual patient sera, sera from five clinical cases of human bartonellosis from Peru with high IFA titers for B. bacilliformis (512 to 1,024) were reacted with the recombinant 43-kDa antigen expressed in E. coli. The sera were from patients who had either Oroya fever or verruga peruana. As shown in Fig. 5 (lanes 1 to 5), all of the sera tested showed strong reactivity with the 43-kDa antigen. On the other hand, control sera that tested negative for bartonella as determined by IFA (titer of 32) did not recognize the 43-kDa antigen (lanes 6 to 10). We studied the cross-reactivity of the 43-kDa antigen with other Bartonella species by Western blot analysis using sera from patients with CSD as confirmed by a high IFA titer for B. henselae (titer of 2,048). All of the five sera that were tested showed strong reactivity to the 43-kDa antigen (lanes 11 to 15), suggesting that a homologue of the 43-kDa antigen exists in B. henselae.

View larger version (36K): [in this window] [in a new window]   FIG. 5.   Western immunoblot of the 43-kDa antigen reacted with individual serum samples from patients with clinical bartonellosis and CSD. Cell lysates of E. coli JM105 cells containing the recombinant plasmid pKIP-7 were electrophoresed on a 4 to 16%, precast, polyacrylamide gel (Novex). Proteins were transferred onto nitrocellulose and reacted with the different sera by using a multiscreen apparatus (Bio-Rad). Sera were diluted 1:200 in 5% Blotto. The rest of the protocol is as described in Materials and Methods. Lanes 1 to 5, serum samples from patients with clinical bartonellosis; lanes 6 to 10, serum samples from patients with negative IFA titers for Bartonella; lanes 11 to 15, serum samples from patients with CSD. Numbers to the right represent approximate molecular masses, in kilodaltons. The arrow indicates the position of the 43-kDa antigen.

Generation of polyclonal antiserum against the 43-kDa antigen of B. bacilliformis. The antigenic index of the 43-kDa antigen was determined by using the Jameson-Wolf algorithm (DNASTAR). This algorithm uses criteria, such as hydrophilicity, surface probability, and flexibility, to predict regions of the protein that could serve as epitopes involved in generating a humoral immune response. Five regions of the protein were predicted as predominant epitopes of the 43-kDa antigen and peptides corresponding to these regions were synthesized. The positions of these peptides are shown in Fig. 2. A mixture of the five peptides was used to immunize rabbits for the purpose of generating specific antiserum against the 43-kDa antigen of B. bacilliformis. The antiserum was found to react very strongly with a 43-kDa protein in E. coli harboring the recombinant plasmid, pKIP-7, under uninduced (Fig. 6, lane 3) and induced (Fig. 5, lane 4) conditions. This protein band was absent in cell lysates of E. coli harboring pKK223-3 (lane 1), indicating that it is specific to E. coli containing pKIP-7. The rabbit antiserum also reacted with the fusion protein that was expressed from pBIPIM-17 (Fig. 6, lane 2). Furthermore, the rabbit antiserum was also able to recognize 43-kDa antigens in both of the known strains of B. bacilliformis, KC584 and KC583 (lanes 5 and 6).

View larger version (39K): [in this window] [in a new window]   FIG. 6.   Western immunoblot of the 43-kDa antigen reacted with polyclonal rabbit serum raised to a cocktail of synthetic peptides. Lane 1, E. coli JM105 cells containing the plasmid vector, pKK223-3; lane 2, E. coli XL1-Blue cells containing the recombinant plasmid, pBHIM-17, expressing the fusion protein; lanes 3 and 4, E. coli JM105 cells containing the recombinant plasmid, pKIP-7, expressing the 43-kDa antigen under uninduced (lane 3) and induced (lane 4) conditions; lanes 5 and 6, B. bacilliformis strains KC584 and KC583, respectively; lane 7, cell lysate of B. henselae (Houston-1); lanes 8 and 9, cell lysates of B. quintana (Oklahoma) and B. elizabethae (ATCC 49927), respectively. The serum was diluted 1:5,000 in 5% Blotto. Numbers to the left are approximate molecular masses, in kilodaltons (kDa). The arrow indicates the position of the 43-kDa antigen.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

B. bacilliformis, a member of the family Bartonellaceae within the alpha-2 subgroup of Proteobacteria, is the etiologic agent of human bartonellosis. Bartonellosis is a unique disease because of its biphasic nature, in which bacteria exhibit tropism for different cells in each of the two phases. In the primary acute phase (Oroya fever), the bacteria invade nearly 100% of erythrocytes, causing a severe hemolytic anemia. During the secondary chronic phase (verruga peruana), the bartonellae invade endothelial cells, which results in wart-like multiple tumors involving the skin, mucous membranes, and internal organs (3, 7). The remarkable difference in disease manifestation during the two stages of bartonellosis suggests a complex interaction between B. bacilliformis and the human host that may involve a multitude of both bacterial and host proteins. Characterization of antigens expressed during the course of an infection by B. bacilliformis would be helpful in elucidating the pathogenesis of the disease. Identifying immunogenic antigens would also be useful for developing tools for the rapid diagnosis of bartonellosis.

In this study, we have cloned, sequenced, and characterized an immunogenic 43-kDa antigen of B. bacilliformis. The predicted amino acid sequence of the 43-kDa antigen shows homology to the LppB proteins of Haemophilus somnus and H. influenzae and the NlpD protein of E. coli (Fig. 2). LppB/NlpD is a recently identified lipoprotein proposed to be located in the outer membrane of H. somnus, a pathogenic bacterium that causes hemophiliosis in cattle (8). In H. somnus, LppB is an immunodominant protein that has been shown to be able to bind the aromatic dye Congo red, a structural analog of heme (26). Congo red binding (Crb⁺) is a property that has been used as an indicator of virulence for several pathogenic bacteria (5, 13, 20). We have demonstrated that the 43-kDa antigen is able to incorporate [³H]palmitate efficiently, providing experimental evidence that it is indeed a lipoprotein. In addition, mutagenesis of the signal peptidase cleavage site resulted in the accumulation of a precursor form of the protein that could not incorporate [³H]palmitate. These results suggest that the 43-kDa antigen of B. bacilliformis is synthesized by a pathway similar to that used by other major lipoproteins in bacteria (11).

We have used Western blotting to demonstrate that sera from individuals who had classical Peruvian bartonellosis recognize the recombinant 43-kDa antigen that was expressed in E. coli. When reacted with a pool of these sera (Fig. 3), the 43-kDa recombinant antigen migrated with an immunoreactive protein of the same size in the B. bacilliformis cell lysates, suggesting the presence of the LppB protein in B. bacilliformis. The presence of antibodies against the LppB homologue in the patient sera indicates that it is an immunogenic protein. Immunoreactivity of the LppB protein with individual sera from patients with bartonellosis was also demonstrated (Fig. 5). Our Western blot analysis data (Fig. 5) also revealed that positive CSD sera with a high antibody titer to B. henselae recognize the 43-kDa antigen of B. bacilliformis. This suggests that a homologue of the antigen exists in B. henselae, a bacterium that is phylogenetically closely related to B. bacilliformis. The immunogenicity of the 43-kDa antigen in B. bacilliformis and B. henselae may have important implications from the perspective of pathogenesis. Common factors between these two closely related organisms are of special relevance because of the pathological similarities between verruga peruana and BA. Future studies aimed at testing the ability of the 43-kDa antigen of B. bacilliformis to bind to endothelial cells may help to identify a possible role in the endothelial cell proliferation that is a hallmark of verruga peruana and BA.

We generated polyclonal antisera against a cocktail of synthetic peptides corresponding to antigenic regions within the 43-kDa antigen. This serum reacted strongly with the 43-kDa antigen expressed in E. coli and B. bacilliformis but did not cross-react with the E. coli NlpD homologue, as indicated by the absence of an immunoreactive 43-kDa band in the E. coli negative control that harbored the plasmid (Fig. 6). In addition, the serum did not recognize a 43-kDa protein in the B. henselae cell lysate (Fig. 6), although the presence of the antigen in B. henselae was demonstrated by the reactivity of the positive CSD sera with the recombinant 43-kDa antigen expressed in E. coli (Fig. 5). This suggests that the 43-kDa antigens of B. bacilliformis and B. henselae exhibit considerable divergence, at least within the regions where we synthesized peptides, so that they could potentially be useful in the development of diagnostic tools for differentiating between these Bartonella species.

Investigation of the immunogenicity of the five individual peptides revealed that only the antiserum against peptide 2 (Fig. 2) could recognize the 43-kDa antigen in the B. bacilliformis cell lysates (data not shown). This suggests that peptide 2 is the most immunogenic of the five synthetic peptides and is responsible for the major part of the immunogenicity of the peptide cocktail antiserum. It is possible that differences between the B. bacilliformis and B. henselae LppB homologues within the region corresponding to peptide 2 could contribute to the lack of reactivity of the peptide cocktail antiserum with the B. henselae homologue. The inability of the antiserum to recognize 43-kDa antigens in any of the other bartonellae suggests that species-specific synthetic peptides based on the 43-kDa antigen, such as peptide 2 used in this study, would be potentially useful for diagnostic purposes. The usefulness of synthetic peptides as diagnostic reagents has been demonstrated in previous studies (22).

The presence of the 43-kDa antigen in two strains of B. bacilliformis, KC583 and KC584, was demonstrated by immunoblot analysis using the polyclonal antiserum generated against the 43-kDa antigen (Fig. 6). In addition, the 43-kDa antigen is also present in a putative, uncharacterized Ecuadorian strain of B. bacilliformis (2), as demonstrated by its reactivity with sera from a patient with the milder, atypical form of bartonellosis. These results suggest that the 43-kDa antigen is a protein that is highly conserved among different strains of B. bacilliformis.

It is possible that, like other NlpD/LppB homologues the LppB protein of B. bacilliformis is exposed at the cell surface. Evidence from fractionation of B. bacilliformis has shown the presence of antigens of between 42 and 48 kDa, a size range that corresponds to that of the LppB protein, in outer membrane fractions of B. bacilliformis (14, 16). In addition, examination of the amino acid sequence of the LppB homologue of B. bacilliformis revealed the presence of a serine residue following the cysteine at the signal peptide cleavage site, a feature that is characteristic of lipoproteins that are exported to the outer membrane (28). Given the importance of cell-surface-exposed factors to pathogenesis, the possible surface localization of the 43-kDa antigen suggests a potential role in the infection process of B. bacilliformis. A structural feature that supports a role for the 43-kDa antigen in cell adhesion is the presence of the tripeptide motif, arginine-glycine-aspartate (RGD), near the carboxyl-terminal end of the protein (Fig. 2). The RGD motif, a cell adhesion feature present on bacterial and viral virulence factors, has been proposed to facilitate binding of pathogens to host cells, promoting their internalization during the infection process (4, 25). Future studies involving mutagenesis of the RGD sequence of the 43-kDa antigen will help to determine if this protein plays a role in the interaction of B. bacilliformis with human endothelial cells which culminates in verruga peruana.