|Home | About | Journals | Submit | Contact Us | Français|
Bartonella quintana is a fastidious, gram-negative, rod-shaped bacterium that causes prolonged bacteremia in immunocompetent humans and severe infections in immunocompromised individuals. We sought to define the outer membrane subproteome of B. quintana in order to obtain insight into the biology and pathogenesis of this emerging pathogen and to identify the predominant B. quintana antigens targeted by the human immune system during infection. We isolated the total membrane proteins of B. quintana and identified 60 proteins by two-dimensional sodium dodecyl sulfate-polyacrylamide gel electrophoresis and peptide mass fingerprinting. Using the newly constructed proteome map, we then utilized two-dimensional immunoblotting with sera from 21 B. quintana-infected patients to identify 24 consistently recognized, immunoreactive B. quintana antigens that have potential relevance for pathogenesis and diagnosis. Among the outer membrane proteins, the variably expressed outer membrane protein adhesins (VompA and VompB), peptidyl-prolyl cis-trans-isomerase (PpI), and hemin-binding protein E (HbpE) were recognized most frequently by sera from patients, which is consistent with surface expression of these virulence factors during human infection.
Bartonella quintana, the agent of trench fever, is a fastidious, gram-negative, rod-shaped organism that can cause prolonged bacteremia in immunocompetent humans and severe infections in immunocompromised individuals. Humans are the only known reservoir for B. quintana (12), and the vector for transmission is the human body louse, Pediculus humanus corporis (38). B. quintana infections have occurred worldwide, and severe, potentially lethal complications, such as endocarditis and bacillary angiomatosis, can develop in immunocompromised patients with AIDS, cancer, and organ transplants. However, little is known about the pathogenesis of B. quintana, and diagnosis of human infection remains extremely challenging. To address this paucity of knowledge, we sought to identify potential membrane-associated virulence factors, as well as protective and diagnostically relevant B. quintana antigens, by characterizing the total membrane fraction and immunome of B. quintana.
Bacterial outer membrane proteins (OMP) can be important virulence factors, playing a critical role in adherence, invasion, and immune evasion during infection of the host, as well as during transmission via arthropod vectors. Many outer membrane-associated proteins that are important for pathogenesis also are consistent targets for the host immune system after infection. Workers in our lab previously identified a family of variably expressed outer membrane proteins (Vomp) that play a role in adhesion and autoaggregation (45). To initially identify the Vomp family, we used two-dimensional (2D) sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis (PAGE) to visualize changes in expression of membrane proteins in sequential isolates from animals experimentally infected with B. quintana.
To identify additional membrane proteins of B. quintana, we utilized 2D SDS-PAGE to systematically characterize the total membrane protein (TMP) subproteome and to determine whether the Vomp and other identified B. quintana membrane proteins are recognized by sera from patients naturally infected with B. quintana. We constructed a protein map of the TMP of B. quintana by 2D gel electrophoresis and then identified individual proteins by peptide mass fingerprinting (PMF). We next performed a 2D immunoblot analysis using sera from 21 B. quintana-infected patients to identify the membrane-associated antigens consistently recognized by the human immune system during B. quintana infection. Analysis of these membrane-associated proteins provided insight into the identities of virulence factors, as well as protective and diagnostic antigens, of B. quintana.
B. quintana strains were isolated from Bartonella-infected patients with concomitant human immunodeficiency virus infections (Table (Table1)1) and were passaged fewer than three times from frozen stocks before use. Strains were streaked onto chocolate agar plates, incubated at 37°C in candle extinction jars, and harvested for protein preparation after 7 days. Bartonella antibody titers were determined for each patient serum sample by indirect immunofluorescent antibody (IFA) testing. The IFA test for Bartonella antibodies was developed at the CDC (10, 40). Patient serum was diluted twofold to 1:1,024, and a reciprocal titer to Bartonella henselae or B. quintana of ≥64 was considered a positive result based on previous studies (10, 40). Although the antigenic profile of B. quintana grown with Vero cells for IFA analysis and the antigenic profile of bacteria grown on agar for immunoblotting may differ somewhat, culture on agar was necessary to generate a sufficient mass of bacteria and to maintain bacterial cell fractions that did not contain eukaryotic cells.
Subcellular fractions were obtained by using methods described previously (32, 42). Bacteria were harvested from chocolate agar plates, washed twice with phosphate-buffered saline (PBS) (pH 7.4), and pelleted by centrifugation with a tabletop microcentrifuge for 2 to 3 min at 4°C at the maximum speed. The final pellet was resuspended in 10 mM HEPES buffer. We added a protease inhibitor cocktail [20 mM 4-(2-aminoethyl)-benzenesulfonyl fluoride (Calbiochem, San Diego, CA); 1 mg/ml leupeptin, 0.36 mg/ml E-64, and 5.6 mg/ml benzamidine, all obtained from Sigma, St. Louis, MO; 50 mM EDTA] and incubated the preparation on ice for 10 min. The bacterial cells were disrupted with five 1-min bursts using a sonicator (Labsonic U sonicator; B. Braun Biotech, Inc., Allentown, PA) with cooling on ice between bursts. A small aliquot was removed and saved for whole-cell lysate preparation. Cellular debris was pelleted by centrifugation at 4,300 × g for 30 min at 4°C with a Sorvall centrifuge (SS-34 rotor; Thermo Electron, Asheville, NC). The supernatant was transferred to Ultra-Clear ultracentrifuge tubes (13 by 51 mm; Beckman, Palo Alto, CA) and centrifuged at 100,000 × g with an L8-M ultracentrifuge for 1.5 h at 4°C using an SW55Ti rotor (Beckman). The supernatant was removed and saved for cytosolic protein preparation. The pellet was resuspended either in 10 mM HEPES for TMP preparation or in 1% (wt/vol) N-lauryl sarcosine (Sigma) in 10 mM HEPES to separate the OMP. The lauryl sarcosine suspension was incubated at room temperature for 30 min and then pelleted by ultracentrifugation at 100,000 × g for 1.5 h at 4°C. The supernatant containing the inner membrane proteins (IMP) was saved, and the OMP pellet was resuspended in 10 mM HEPES and treated with nuclease (50 mM MgCl2, 100 mM Tris [pH 7.0], 500 μg/ml RNase, 1 mg/ml DNase [Sigma]). OMP were then washed twice in 10 mM HEPES and pelleted by centrifugation at 40,000 × g for 30 min at 4°C with a Sorvall centrifuge (SS-34 rotor). The final TMP preparation was treated with nuclease and pelleted in the same way. The cytosolic preparation was precipitated with 45% ammonium sulfate (Sigma) in 0.01 M Tris (pH 7.0) and incubated on ice for 45 min. The precipitated proteins were pelleted by centrifugation at 19,000 × g for 30 min at 4°C. The resulting pellet was resuspended in cold PBS and dialyzed overnight against PBS using a D-tube dialyzer maxi (molecular weight cutoff, 3,500 Da; Novagen, Darmstadt, Germany). The dialyzed proteins were concentrated using an Amicon Ultra-4 centrifuge filter (Millipore, Bedford, MA). The IMP preparation was concentrated in the same way. Protein concentrations were determined using a MicroBCA protein assay (Pierce, Rockford, IL), and proteins were separated by one-dimensional (1D) SDS-PAGE to confirm that subcellular fractions were separated. All fractions were frozen at −80°C until they were used.
2D gel electrophoresis was performed using the method of O'Farrell (35) by Kendrick Labs, Inc. (Madison, WI), as follows. Protein pellets were dissolved in 200 ml of SDS boiling buffer (5% SDS, 10% glycerol, 60 mM Tris [pH 6.8]) without β-mercaptoethanol, and protein concentrations were determined using a bicinchoninic acid assay (Pierce). Protein samples were then diluted to obtain a concentration of 2.0 or 4.0 mg/ml in SDS boiling buffer containing 5% β-mercaptoethanol and boiled for 5 min. Isoelectric focusing was carried out in glass tubes with an inside diameter of 2.0 mm using 2% pH 4 to 8 ampholines (BDH; obtained from Hoefer Scientific Instruments, San Francisco, CA) for 9,600 V·h. For the TMP preparations 100 μg of protein was loaded, and for the OMP preparations 200 μg was loaded. After equilibration for 10 min in buffer O (10% glycerol, 50 mM dithiothreitol, 2.3% SDS, 0.0625 M Tris [pH 6.8]), the tube gels were laid on top of 10% acrylamide slab gels (thickness, 0.75 mm), and SDS slab gel electrophoresis was carried out for about 4 h at 12.5 mA/gel. The gels were stained with either Coomassie brilliant blue R-250 or silver stain (34). The Coomassie brilliant blue-stained gels were maintained wet in 10% acetic acid between sheets of filter paper until spot excision and subsequent mass spectrometry analysis; the silver-stained gels were dried between sheets of cellophane. After slab gel electrophoresis, a duplicate gel was transblotted onto a polyvinylidene difluoride membrane (Immobilon-P; Millipore).
To identify immunoreactive proteins in the TMP fraction of B. quintana, proteins separated by 2D SDS-PAGE were transferred to polyvinylidene difluoride membranes, which were then blocked overnight at 4°C with 5% milk in TBST (150 mM NaCl, 10 mM Tris-HCl [pH 8.0], 0.5% Tween 20, 0.2% sodium azide). The membranes then were washed three times (5 min each time) in TBST containing 0.5% bovine serum albumin (Sigma) and once in TBST. Patient sera (primary antibody) were inactivated with 0.5% Nonidet P-40 (Roche, Mannheim, Germany) and diluted 1:50 in TBST. Each membrane was placed in a heat-sealable bag containing the primary antibody and vigorously shaken for 2 h at room temperature. The membranes were washed, blocked for 30 min in 1% milk in TBST, and then washed again. Secondary antibody (alkaline phosphatase-conjugated goat anti-human immunoglobulin G; Zymed Laboratories, Inc., South San Francisco, CA) was diluted 1:5,000 and incubated with the membranes for 30 min at room temperature. The membranes were washed and developed using alkaline phosphatase buffer (100 mM Tris-HCl [pH 9.5], 100 mM NaCl, 5 mM MgCl2), nitroblue tetrazolium, and 5-bromo-4-chloro-3-indolylphosphate (BCIP) (Promega, Madison, WI).
To corroborate the immunoreactivity of B. quintana proteins identified by the immunoblotting described above, a λ genomic expression library of B. quintana strain JK31 was screened using serum from a Bartonella-infected patient. The library was generated with a Sau3AI partial digest of B. quintana chromosomal DNA and the lambda-ZAP Express vector used according to the manufacturer's instructions (Stratagene, La Jolla, CA) and was screened by lifting plaques onto isopropyl-β-d-thiogalactopyranoside (IPTG)-impregnated nitrocellulose (27), followed by immunoblotting, as previously described (30). The initial screening was performed for 3 h at 25°C using polyclonal rabbit anti-B. quintana antiserum (1:1,000 dilution of serum generated by intravenous immunization with B. quintana, as described previously , except that the formalin treatment was omitted). Human antibody recognition of positive plaques was verified by using serum (1:50 dilution) from a patient with a Bartonella infection; plaque lifts were probed for 16 h at 25°C. Plaques identified as positive for both human and rabbit antisera were isolated, replaqued, and rescreened to ensure clonality. Phagemid contents were excised and rescued with Escherichia coli XLOLR (Stratagene), and then plasmids were purified and sequenced (29). Data were analyzed using the Chromas (Technelysium), MacVector (Accelrys, San Diego, CA), and BLAST (http://www.ncbi.nlm.nih.gov/BLAST/) software. The protein profiles for each strain were analyzed to identify seroreactive protein species by SDS-PAGE and immunoblotting, as previously described (27). Molecular mass values for reactive proteins were cross-referenced to predicted mass values for plasmid-encoded proteins determined by sequence analysis.
Individual protein spots were excised from the Coomassie brilliant blue-stained 2D gels and cut into 1-mm2 pieces. The stain was removed from the gel pieces by incubating the pieces with 25 mM ammonium bicarbonate (NH4HCO3) (Sigma) in 50% acetonitrile (Sigma) overnight with gentle vortexing. After removal of the supernatant, gel pieces were dried with a SpeedVac and then reduced with 10 mM dithiothreitol (Sigma) for 1 h at 56°C, followed by alkylation with 55 mM iodoacetamide (Sigma) for 45 min at room temperature in the dark. Next, the gel pieces were washed with 25 mM NH4HCO3, dehydrated with 100% acetonitrile and dried with a SpeedVac. Porcine trypsin (12.5 ng/μl in 25 mM NH4HCO3; Promega) was added, and the gel pieces were allowed to rehydrate for 1 h on ice. The excess trypsin solution was removed, and 25 mM NH4HCO3 was added to cover the gel pieces. The preparations were digested at 37°C overnight. Next, the supernatant (postdigestion) was transferred to a clean Eppendorf tube, and tryptic peptides were extracted from the gel pieces by vortexing them for 15 min with 50% acetonitrile-50% H2O-0.1% trifluoroacetic acid (Pierce). The peptide extract was combined with the postdigestion supernatant, and the total volume was reduced to approximately 10 μl with a SpeedVac. The concentrated peptide extracts were desalted using C18 ZipTips (Millipore) and were eluted with 3 to 5 μl of 50% acetonitrile-0.1% trifluoroacetic acid. The peptides were stored at −80°C until they were used.
PMF was used for protein identification. Peptide extracts were mixed with a matrix solution containing α-cyano-4-hydroxycinnamic acid (2 mg/ml in 50% acetonitrile-0.1% trifluoroacetic acid) at a 1:1 (vol/vol) ratio directly on a stainless steel target. A matrix-assisted laser desorption ionization—time of flight mass spectrometry analysis was performed in the reflector, positive-ion mode in the mass-to-charge ratio (m/z) range from m/z 800 to m/z 4000 utilizing a Voyager DE STR matrix-assisted laser desorption ionization—time of flight mass spectrometer (Applied Biosystems, Foster City, CA). Each mass spectrum was calibrated internally using trypsin autolysis product masses. Mass spectra were processed (baseline adjustment, noise filtering, and de-isotoping) to produce a list of monoisotopic, monoprotonated molecular ion masses. Monoisotopic peak lists were submitted to the Mascot Peptide Mass Fingerprint (http://www.matrixscience.com) search engine for analysis. Searches that interrogated the Eubacteria protein database within the MSDB (ftp://ftp.ncbi.nih.gov/repository/MSDB/msdb.nam) sequence database were performed. Validation of the results was based on the top hit score, a requirement for high precision of mass measurement (defined as a low standard deviation, <25 ppm, of mass assignment errors for all matching peptide masses detected within a sample spot) and a minimum of 30% sequence coverage.
Selected protein identities were confirmed by high-performance liquid chromatography-tandem mass spectrometry. Liquid chromatographic separation was performed with an Ultimate capillary high-performance liquid chromatography system (Dionex/LC Packings, Sunnyvale, CA) equipped with a PepMap trap column (Dionex/LC Packings) and a reversed-phase C18 nanocolumn (packed in house; inside diameter,75 μm; length, 15 cm; pore size, 100 Å; particle size, 3 μm) and a Famos Micro autosampler. A 3- to 4-μl aliquot of peptide extract was loaded onto the trap column with loading solvent (0.1% formic acid) at a flow rate of 20 μl/min. The trap column was washed with the loading solvent for 3 min before it was switched in line with the reversed-phase nanocolumn. The nanocolumn mobile phase flow rate was 325 nl/min, and the nanocolumn was maintained at the ambient temperature. The nanocolumn was equilibrated with 2% solvent B (80% acetonitrile, 20% H2O, 0.08% formic acid) and 98% solvent A (2% acetonitrile, 98% H2O, 0.1% formic acid) for 20 min prior to sample injection. Peptides were separated using a binary gradient that consisted of a 5-min isocratic wash with 2% solvent B, followed by a linear gradient from 2% solvent B to 50% solvent B over 45 min and then by a column cleanup step consisting of 95% solvent B for 7 min. The column effluent flowed directly into a nanoelectrospray ion source (Protana, Odense, Denmark) in a QSTAR XL quadrupole/quadrupole time of flight mass spectrometer (Applied Biosystems). Proteins were identified by isolating sequentially eluting peptide populations with a single m/z value in the mass spectrometer, fragmenting each population, and determining the masses of the peptide fragment ions. The experimentally determined peptide fragment ion masses were matched, within a window of ±0.2 Da, to theoretical fragment ion masses generated by in silico fragmentation of all theoretical tryptic peptides derived from Eubacteria protein sequences in the MSDB database.
In addition to spot identification by PMF, we used PSORTb v.2.0 to predict protein localization based on signal peptides, transmembrane helices, homology to proteins whose localization is known, and amino acid composition and motifs (14; http://www.psort.org/psortb/). Identities of protein families were determined using Pfam (4, 11; http://www.sanger.ac.uk/Software/Pfam/). Grand average of hydropathy was used to evaluate the hydrophilicity and hydrophobicity of each protein along its amino acid sequence (23; http://us.expasy.org/tools/protparam.html).
Spot detection for gels and blots was performed using the 2D Evolution software (Nonlinear Dynamics, Durham, NC). Spot detection for the master gel was performed manually due to nonspecific spot detection. A silver-stained gel of B. quintana JK31 proteins was used as the master protein profile, and each spot was assigned a number. Immunoreactive spots identified on the blots were automatically matched with the master gel, and additional spots detected visually were added manually. To allow comparisons across immunoblots, the background was subtracted using a fully automated method with the Evolution software, called Lowest on Boundary, and was determined by tracing a line just outside the boundary of each spot and then using the lowest pixel intensity that was encountered during this process as the background intensity for that spot. Additionally, each blot was normalized to a single common spot in the blot to eliminate differences in spot intensity due to immunoblot development. The volume of each spot was then calculated by dividing the pixel intensity by the area of the spot. Means, medians, and ranges of volumes were determined.
OMP were isolated from B. quintana JK31 using lauryl sarcosine fractionation. This method has been used to enrich OMP from a number of bacterial species (3, 31, 39) and was also utilized to separate the OMP from the IMP of B. henselae (42). We initially compared the protein profiles of the subcellular fractions of B. quintana after separation by 1D SDS-PAGE, followed by Coomassie blue staining (Fig. (Fig.1).1). Enrichment of the OMP and IMP from the TMP preparation was evident when the proteins were compared with proteins from the cytosolic preparation. There were prominent bands in the OMP preparation at approximately 116, 93, 45, 40, and 34 kDa (Fig. (Fig.1).1). These prominent OMP bands either were observed exclusively with the OMP fraction or were highly enriched in the OMP fraction compared with the IMP fraction. Coomassie blue staining of proteins separated on a gel containing a lower percentage of acrylamide revealed that the 116-kDa bands in the OMP and IMP fractions were actually at slightly different molecular masses (data not shown).
The TMP and OMP preparations were resolved by 2D SDS-PAGE, followed by silver staining. The resolution of individual spots from gel to gel and for different protein preparations was highly reproducible. We visualized more than 300 distinct protein spots in the TMP preparation in a pI range from 4.5 to 9.5 and in a molecular mass range from 14 to 220 kDa (Fig. (Fig.2).2). Protein spots were numbered, excised from Coomassie blue-stained gels, subjected to in-gel tryptic digestion, and submitted for protein identification by PMF. We excised 137 spots and identified 110 separate protein spots (Table (Table2).2). Some spots were not positively identified because of the low concentration of protein and/or contamination with human keratin. The 110 spots identified by PMF correspond to 60 B. quintana genes. With the exception of a few spots that stained intensely with silver stain but not with Coomassie blue, most protein spots that were visualized were identified.
A number of the B. quintana membrane proteins identified appeared as protein isoforms or families. The protein product of a single gene can appear as several protein spots on a 2D gel due to posttranslational modifications; these isoforms are usually visualized as a horizontal pattern of spots at the same molecular weight. The modifications of bacterial proteins can include phosphorylation, glycosylation, methylation, deamidation, and biotinylation, each of which can affect the charge and the isoelectric point. Of the 60 unique proteins that we identified, 18 had at least two isoforms, and these proteins included proteins that play a role in energy metabolism (AcnA, AtpA, AtpD, PpdK, SucB, and SucD), protein fate (ClpB and MopA), protein synthesis (FusA, RpsA, and Tuf1), transcription (Pnp and Rho), purine ribonucleotide synthesis (GuaB and GlyA), and virulence (HbpE, HbpD, VompA, VompB, and VompC).
Twenty-four percent (26/110) of the proteins identified were predicted by PSORTb to localize to the outer membrane, 65% (72/110) of the proteins were predicted to localize to the cytoplasm, and the localizations of 11% (12/110) of the proteins are not known. The 26 proteins localized to the outer membrane correspond to 10 distinct gene products, including three hemin-binding proteins (HbpA, HbpD, and HbpE), Omp43, Omp89, peptidyl-prolyl cis-trans-isomerase (Ppi), BQ08370 (a putative OMP), and three adhesins (VompA, VompB, and VompC). The spots with the greatest apparent protein concentration correspond to GroEL (MopA) (at 57.6 kDa; spots JB51, JB52, JB74, JB78, JB109, and JB121), EF-Tu (Tuf1) (at 42.9 kDa; spots JB1/108, JB37, JB38, JB79, and JB115), HbpA (at 29.3 kDa; spot JB87), HbpD (at 32.7 kDa; spots JB2, JB3/31, JB30, JB88, JB119, and JB137), HbpE (at 33 kDa; spots JB4/107, JB5/113, and JB114), VompA (97.0 kDa), VompB (100.5 kDa), and VompC (99.8 kDa). Pfam predictions and grand average of hydropathy values for all of the spots identified are shown in Tables Tables22 and and33.
We identified 24 B. quintana proteins that are recognized consistently by sera from humans with documented B. quintana infections. TMP from the same preparation that was used for PMF were separated simultaneously by 2D SDS-PAGE to produce two identical 2D gels, and then TMP from one gel were transferred and immunoblotted with sera from each of 21 patients from whom B. quintana was isolated and whose sera were positive for Bartonella antibodies as determined by IFA analysis (10, 40) (Table (Table1).1). Each patient's serum was analyzed on a separate immunoblot, and immunoreactive antigens were identified by alignment with a simultaneously prepared silver-stained gel using the 2D Evolution software (Nonlinear Dynamics). For negative controls, two blots with 2D-separated TMP were immunoblotted with sera that were from Bartonella IFA-negative, culture-negative patients. These control sera detected a few B. quintana proteins, usually the protein spots that had the highest protein concentrations and were most dense. The proteins that were immunoreactive on these two negative control blots were considered false positives and were not included in the analysis of positive sera.
To identify the B. quintana antigens most commonly recognized by sera from patients infected with B. quintana, we established a positive cutoff value of 24, representing the B. quintana TMP antigens recognized by sera from 24% or more of the patients infected with B. quintana (at least 5 of the 21 patients analyzed). Using this cutoff value, we identified 24 immunodominant B. quintana proteins recognized by sera from these patients (Table (Table4).4). Figure Figure33 shows a representative 2D immunoblot of B. quintana TMP probed with serum from patient 4 (Table (Table1).1). The pI values of these immunoreactive antigens ranged from 4 to 7, and the predicted molecular masses ranged from 20 to 100 kDa. Four of the immunodominant antigens were OMP (VompA, VompB, HbpE, and Ppi). The remainder were predicted to be cytoplasmic proteins, and many of these cytoplasmic proteins have been identified previously in the outer membrane fractions of other gram-negative organisms, including other Bartonella species (6, 16, 42). Each of the 24 immunoreactive antigens commonly recognized by patients’ sera was labeled in the immunoblot shown in Fig. Fig.3.3. Note that the serum from B. quintana-infected patient 4 recognized all of the 24 immunoreactive proteins whose values were above the 24% cutoff value for all patients. Antibodies in the serum of patient 4 also recognized several B. quintana protein spots whose values were below the cutoff value of 24%; therefore, although these proteins (e.g., NusA, VompC, CarB, SecA, SdhA, and PdhB) were strongly immunoreactive on the blot, they were not labeled or included in Table Table4.4. Note that in some cases (e.g., GroEL [MopA], AtpA, EF-Tu [Tuf1], RpsA, HbpE, and Pnp), the most dominant member of a protein family did not meet the inclusion criteria because the spot was so highly concentrated that it was also immunoreactive with the negative sera. Therefore, these proteins were not included in the analysis.
Spots JB15 and JB17 had low sequence coverage as determined by PMF, and spot JB16 had no significant hits in the database search but was found to have monoisotopic peaks similar to those of spots JB15 and JB17. Because these proteins were found to be highly immunogenic, we confirmed their identities by submitting the peptides for mass spectrometry/mass spectrometry analysis. Validation of the liquid chromatography-mass spectrometry/mass spectrometry results confirmed that these three protein spots (spots JB15, JB16, and JB17) were dihydrolipoamide succinyltransferase (SucB) from B. quintana.
In addition to the immunoblot analysis, screening of an expression library was performed to identify antigenic proteins using a λ phage genomic library of B. quintana JK31. The primary screen included ~3,000 plaques per plate and four plates (a total of 12,000 plaques). Twelve plaques were identified as plaques that were reactive with human and rabbit antisera, and they were replaqued and rescreened to ensure clonality. Characterization of two positive clones resulted in identification of SucB and GroEL (MopA), confirming the immunoreactivities of these two proteins observed by immunoblot analysis.
The outer membrane of a bacterium forms the interface between the microorganism and the host and plays an essential role in adhesion and host immune evasion, two important virulence mechanisms utilized by B. quintana. The bacterial proteins mediating these interactions between B. quintana and its host are critical targets of the host immune response and often have diagnostic relevance, and they are useful candidate antigens for vaccine development. Our goals were to characterize the TMP subproteome of B. quintana and to further identify the immunome subset of proteins recognized by sera from humans infected with B. quintana. The membrane proteins that we identified are involved in pathogenesis and also are candidate antigens for diagnostic evaluation, treatment, and prevention of B. quintana infection, especially in patients with concomitant human immunodeficiency virus infections.
One-quarter of the proteins that we identified are predicted to be membrane proteins, nearly one-third of the proteins are predicted to be cytoplasmic proteins, and the remainder have unknown localizations. As found in other gram-negative bacteria (6, 16), including B. henselae (42), many of the proteins that fractionated with the TMP fraction are not membrane proteins. We performed a search of the B. quintana genome in the TIGR database for “membrane proteins” and identified 55 membrane-associated proteins. Of these 55 proteins, 61.8% were found to have a predicted pI of ~9.0 or higher. With our 2D gel system, we were able to resolve proteins with pI values ranging from 4.5 to 9.5, and therefore we identified only a few OMP with a pI near 9.5, including Omp43 (pI 9.5), Omp89 (pI 9.3), and HbpA (pI 9.5). Identification of the more basic membrane proteins in the pI range from 9.0 to 12.0 requires a different method to improve resolution, and indeed, these very basic membrane proteins of gram-negative bacteria are often refractory to fractionation regardless of the method used (42).
Of the cytosolic proteins that fractionated in the 2D TMP fraction, GroEL (MopA) and EF-Tu (Tuf1) are commonly found in membrane preparations of other gram-negative bacteria (2, 41). Two of the proteins that we identified, GroEL and DnaK, are common heat shock proteins that also function as chaperones and thus are often membrane associated (7, 37). Indeed, Bartonella bacilliformis has been shown to actively secrete GroEL (30). Other cytosolic proteins, including FusA, TypA, EF-Tu, and Tig, are ribosome-associated proteins that can be membrane associated during the biosynthesis of proteins destined for the periplasm or outer membrane (17). Additional cytoplasmic proteins are associated with the membrane either transiently or while they are functioning as chaperones (7), and thus our detection of these proteins in the membrane fraction is not unexpected or unprecedented.
By using 2D SDS-PAGE, we identified 19 membrane proteins that were present in both B. quintana and B. henselae (42), another species of Bartonella that infects AIDS patients. Of the 60 unique membrane-associated B. quintana proteins, 13 have characteristics of a prototypical OMP as determined by PSORTb analysis: HbpA, HbpD, HbpE, Omp43, Omp89, Ppi, BQ08370, and six Vomp paralogs. Five of the OMP identified in B. quintana have orthologs that also were identified in the sarcosine-insoluble fraction of B. henselae (42): HbpA, HbpD, Omp43, Omp89, and Ppi. Two additional OMP were identified in B. henselae but not in B. quintana: HutA and BH00450. Although the latter two OMP are present in the B. quintana genome, they have predicted pI values of 9.5 and 9.9, respectively, and did not resolve well in our system.
Finally, our subproteome analysis identified the following OMP virulence factors that were unique to B. quintana and were not found in B. henselae (42): VompA, VompB, VompC, and six additional isoforms of Vomp (spots JB75, JB77, JB116, JB117, JB120, and JB123), in addition to HbpE (spots JB4/107, JB5/113, and JB114) (Fig. (Fig.2).2). The vomp genes encode a family of four OMP adhesins that contribute to binding of B. quintana to collagen and to autoaggregation (45). The Vomp proteins are members of the newly described trimeric autotransporter adhesin family that includes YadA of Yersinia enterocolitica (9). Each Vomp has a major variable region near the adhesin tip. The major variable region of each Vomp confers a specific and different virulence phenotype on B. quintana (e.g., VompA is necessary and sufficient to mediate autoaggregation). In addition, in a recent study Schulte et al. (44) suggested a specific role for B. quintana Vomp in the angiogenic reprogramming of host cells. Infection of human macrophages (THP-1) and epithelial cells (HeLa 229) with B. quintana JK31 (a Vomp-expressing strain) induced secretion of vascular endothelial growth factor from both cell types. Strains lacking Vomp expression (BQ2-D70, B. quintana Toulouse, and B. quintana Munich) did not induce secretion of vascular endothelial growth factor (44). This suggests that the Vomp proteins have a specific pathogenic role in the angiogenesis response that occurs in bacillary angiomatosis lesions.
Five hemin-binding proteins (Hbp) have been described in B. quintana (HbpA to HbpE) (29) and are encoded by a five-member gene family comprised of hbpA to hbpE (29). Five orthologs also occur in B. henselae (HbpA to HbpD and Bh10780) (1). We identified three B. quintana Hbp by PMF: HbpA (spot JB87), HbpD (spots JB2, JB3/31, JB30, JB88, and JB119), and HbpE (spots JB4/107, JB5/113, JB84, and JB114) (Fig. (Fig.22 and Table Table2).2). Because heme is essential for B. quintana (33), it is not surprising to find that Hbp are very abundant OMP. Recent studies have shown that the hbp gene family exhibits differential expression in response to environmental cues such as temperature, oxygen, and heme concentration (5), and the hbpADE subfamily is markedly induced under conditions that simulate the conditions in the human host. HbpA, HbpD, and HbpE are the most prominent Hbp, and although this is in agreement with previous reports regarding hbp gene induction, the pI values of HbpB and HbpC (10.2 and 10.1, respectively) are higher than the highest pI resolved by our system (pI 9.5).
We identified 24 B. quintana proteins that are consistently recognized by sera from patients infected with B. quintana, using the 2D Evolution software to identify immunoreactive spots, and evaluated spot size relative to intensity. Of these 24 proteins, 4 were proteins that that we identified in this study as OMP: VompA, VompB, HbpE, and Ppi (Table (Table44 and Fig. Fig.3).3). Twenty are cytosolic proteins or IMP that also are recognized by sera from patients infected with B. quintana. OMP and non-OMP are listed in Table Table44 in order of frequency of recognition by human serum.
Of the four immunoreactive OMP that we identified by PMF that are consistently immunoreactive with patient sera, three are known virulence factors: VompA, VompB, and HbpE. The Vomp adhesins are of particular interest because they are unique to B. quintana and they play a significant role in virulence during infection (45). Bartonella species can survive in the bloodstream for weeks and even months and can adhere to host cell erythrocytes. We demonstrated that the Vomp are surface exposed, using binding of fluorescent antibodies. We found that some vomp genes undergo phase variation in vivo and are not expressed during prolonged bloodstream infection (45). It is advantageous for the bacterium to be able to alter the expression of the Vomp adhesins and other virulence-associated factors in order to evade the host immune response. It is therefore noteworthy that the surface-expressed Vomp adhesins are targeted by the human immune system in many patients infected with B. quintana, which could be important in generating phase and/or antigenic variation of the vomp gene expression in this bacterium. VompC was not recognized by a sufficient number of patients, however, and thus did not meet the criteria for inclusion. It is therefore possible that not all B. quintana isolates express all four Vomp proteins, preventing targeting of the Vomp proteins by the host immune system. The lack of antibodies recognizing VompC could be the result of attenuated expression of VompC compared to the expression of VompA and VompB; VompC also may not be comparably immunogenic. Finally, VompC expression could be turned off due to phase variation, before antibodies are elicited in the host. It will be interesting to examine the isolates from the patients whose sera recognized only one or two of the four Vomp proteins to see if a corresponding isolate from a patient has the full complement of four vomp genes and, if so, whether the genes are expressed. We used a single strain, B. quintana JK31, for antigen preparation, against which we blotted each patient's serum. This enabled us to directly compare the antibody responses of the patients and to determine which antigens are most consistently recognized. However, it is possible that JK31 does not have the same protein profile as the B. quintana strain from an individual patient, and this should be investigated further. For instance, from the standpoint of both virulence and diagnosis, it is important to determine if one specific Vomp is always expressed during human infection with B. quintana.
HbpE was the only other OMP that was identified as a protein that was immunoreactive with at least 24% of the patient sera. Although differences in gene expression have been noted for hbpE (5), more notable is the level of immunogenicity of the HbpE protein compared to the levels of immunogenicity of other members of the Hbp gene family. It should be interesting to further evaluate differences in this protein and to determine the specific role of HbpE in Bartonella pathogenesis.
One of our goals was to identify relevant B. quintana antigens recognized by the human immune response during the natural course of infection which could lead to both improved diagnosis and an understanding of Bartonella infections. Antibody detection is the most widely used diagnostic test for B. quintana infection; an IFA test is the current reference method. However, IFA tests are performed in only a few laboratories, and the Bartonella IFA test is subjective and extremely laborious. Antigen must be prepared by cocultivation of Bartonella with Vero cells on slides, the serum must be serially diluted, and the assay results must be manually screened and graded by highly trained personnel. In addition to the difficulty in performing the IFA test, cross-reactions with other Bartonella species can occur. As shown in Table Table1,1, it is apparent that the IFA titers for both B. henselae and B. quintana are positive in nearly all patients with a documented B. quintana infection, and in some cases the titers for the two species are nearly the same, preventing identification of the infecting Bartonella strain to the species level. In addition, cross-reactivity can occur with Coxiella burnetii and Chlamydia species (25, 28). Culture-based diagnosis of Bartonella infection is even more difficult and time-consuming (21, 24), and molecular biology techniques have little practical application outside the research lab. In most of the immunoscreens for diagnostic antigens in Bartonella workers have used pooled sera from a small number of patients without culture-proven infections or sera collected from experimentally infected small animals (8, 13).
In this study, we utilized a large collection of sera from B. quintana culture-positive humans in conjunction with 2D SDS-PAGE and PMF subproteome data for a virulent strain to systematically characterize the total membrane immunome of B. quintana, and we identified proteins that are recognized during human infection. All 21 patients had naturally acquired B. quintana infections, as documented by isolation of the bacterium from blood or tissue or both (21, 22) and by positive reciprocal IFA titers for B. quintana of ≥64 (10, 40) (Table (Table1).1). Using these sera, 44 immunoreactive B. quintana TMP were identified by 2D immunoblot analysis. In addition to these immunodominant TMP, we also found 20 non-TMP that were reactive with one-quarter of the patient sera. We found that SucB (spots JB15, JB16, and JB17) had the highest frequency of recognition, and we also identified SucB by immunoscreening of an expression library. SucB (dihydrolipoamide succinyltransferase), a 43.8-kDa cytosolic protein, was recognized by 76% of our patient sera (Table (Table4)4) and has been detected in immunoscreens of genomic expression libraries for both B. henselae and Bartonella vinsonii subsp. berkhoffii (15, 18, 26). Another cytosolic protein, the 63.8-kDa protein FtsZ (spots JB48 and JB56), reacted with sera from 24% of our patients and has been identified previously as a potential diagnostic antigen for Bartonella infection (19, 20, 36). However, because both these cytosolic proteins are highly conserved among bacteria, they are unlikely to be useful for Bartonella-specific diagnosis. Considering the frequent recognition and the unique presence of the Vomp and Hbp in Bartonella, these antigens are likely to be the most useful antigens for diagnosis of B. quintana infections in humans.
In summary, we established a 2D map of the total membrane subproteome of B. quintana. We identified 60 unique B. quintana proteins by 2D gel electrophoresis and PMF, including OMP virulence factors. Using this newly constructed subproteome map, we identified 24 immunodominant antigens after performing 2D immunoblotting with sera from 21 naturally infected patients. Our goal was to perform a general screen for B. quintana antigens that reacted with serum from 24% or more of the patients, as a prelude to more definitive future testing of proteins that appear to be candidate diagnostic and/or vaccine antigens. Additionally, characterization of the B. quintana immunome demonstrated that the Vomp virulence factors are frequently recognized by the host immune system, supporting the hypothesis that anti-Vomp antibodies can stimulate the phase variation that occurs in vivo. The identification and evaluation of these B. quintana proteins should not only aid in the development of better diagnostic tests and better disease prevention but also provide insight into the pathogenesis of Bartonella.
We gratefully acknowledge Jon Sargent at the Biomolecular Resource Center Mass Spectrometry Facility at the University of California, San Francisco, for assistance with the mass spectrometry analysis and protein identification.
This work was supported by NIH grants R01 AI43703 and R01 AI52813 and by a Burroughs Wellcome Fund Clinical Scientist Award in Translational Research (to J.E.K.). This investigation was also supported by the Sandler New Technology Fund (to the UCSF Biomolecular Resource Center Mass Spectrometry Facility). M.F.M. was supported by NIH grants R01 AI053111 and U54 AI065357.
Editor: W. A. Petri, Jr.
Published ahead of print on 16 February 2007.