Comparison of In-House and Commercial Slides for Detection by Immunofluorescence of Immunoglobulins G and M against Bartonella henselae and Bartonella quintana

Abstract

We compared the sensitivities and specificities of indirect fluorescent antibody tests developed in our laboratory and commercially available from Focus Technologies (FT; formerly MRL Diagnostic) for detection of serum antibodies to Bartonella spp. Serum samples tested were from patients with culture- or PCR-confirmed Bartonella quintana or B. henselae infections causing cat scratch disease (CSD), chronic bacteremia, or endocarditis. At a cutoff titer of 64, the FT test had higher sensitivity than our in-house test in detecting anti-B. henselae immunoglobulin G (IgG) antibodies in CSD patients (91.2 versus 52.9%; P < 0.001). The specificity in serum samples from 85 control patients was, however, lower with the FT test (87%) than with the in-house test (98.8%) (P = 0.002). A cutoff titer of 128 improves the specificity for the FT test but lowers the sensitivity to 85%. For patients infected with B. henselae, our in-house test, but not the FT test, enabled endocarditis to be detected more reliably. With the in-house test, titers of IgG against B. henselae of ≥1,024 were found only in endocarditis patients and not in CSD patients. With the FT test, 19.1% of CSD patients had titers of IgG against B. henselae of ≥1,024 (P < 0.001). Our in-house technique also improved detection of anti-B. quintana antibodies in homeless patients with endocarditis. IgG titers of ≥1,024 were present in 75% of serum samples, but only in 16.7% of serum samples with the FT test (P = 0.004). Since each test has advantages over the other, the serological diagnosis of Bartonella infections would benefit if both tests were used concurrently.

There are now 18 species within the genus Bartonella. Three are well-known human pathogens: Bartonella bacilliformis (30), B. quintana (43), and B. henselae (3). Others have been isolated from patients on single occasions only: B. elizabethae (one endocarditis case) (20), B. vinsonii subsp. berkhoffii (one endocarditis case) (36, 53), and B. vinsonii subsp. arupensis (one case of bacteremia) (62). B. clarridgeiae (35, 39), and B. washoensis (8, 15) have been tentatively associated with cat scratch disease (CSD) and myocarditis, respectively.

B. quintana and B. henselae are species of worldwide distribution (3, 43). B. quintana was first described as the agent of trench fever in 1918 (58) and is now known to be responsible for louse-borne bacteremia and endocarditis in homeless people and bacillary angiomatosis in AIDS patients (10, 31, 43). Humans are the only known reservoir of B. quintana, and transmission among people occurs via the body louse (43). B. henselae, a species first recognized in 1990 (56), is the main etiological agent of CSD (49) and is also responsible for bacillary angiomatosis and peliosis hepatitis in immunocompromised (mainly AIDS) patients (34), as well as bacteremia and endocarditis (3). Cats are the main reservoir of B. henselae, and people become infected following cat scratches or bites. A role for the cat flea (Ctenocephalides felis) as a vector for human transmission has been proposed (3). B. henselae comprises two distinct genotypes, B. henselae Houston and B. henselae Marseille (7).

Diagnostic techniques for infections with Bartonella spp. include culture of the pathogen (9, 33, 38), molecular biology techniques, especially PCR, to amplify Bartonella sp. genes (2, 28, 32, 51, 52), and serology (54). The usefulness of these various techniques may vary with the disease present. Culture of Bartonella spp. may be successful with blood from bacteremic patients (38), from cardiac valve specimens from endocarditis patients (38), and from skin, lymph node, or other organ biopsy samples from patients with bacillary angiomatosis (38). By contrast, B. henselae is only rarely isolated from patients with CSD (38). Amplification of Bartonella sp. DNA in tissue biopsy samples is an invasive technique and is mainly useful for patients with CSD (lymph node specimen) (2, 28, 52), bacillary angiomatosis (angiomatous lesions of skin or other organs) (51), or endocarditis (cardiac valve when removed) (23, 26, 32). Serology is the only noninvasive diagnostic technique, and it has been evaluated in the diagnosis of CSD (50, 54), bacteremia (11, 17, 18, 31), and endocarditis (23, 42, 47). The indirect fluorescent antibody (IFA) test was first described for CSD by Regnery et al. (50), who used B. henselae grown in Vero cells as antigens. It remains the most frequently used technique, and many laboratories have performed Bartonella serology using tests developed in-house, with reported sensitivities varying considerably, from nearly 100% to less than 30% (54). Commercially prepared antigen slides are now available for B. henselae and B. quintana serology (29, 48, 55, 65), and in this report we compare the sensitivity and specificity of one of these tests with our in-house IFA test, which has been used by our laboratory for 10 years. We compared the abilities of the two tests to detect immunoglobulin G (IgG) and IgM antibodies in serum samples from patients known to have CSD (B. henselae), chronic bacteremia (B. quintana), or endocarditis (B. henselae and B. quintana).

MATERIALS AND METHODS

Patients and serum samples.

Serum samples from 104 patients infected with Bartonella spp., comprising CSD patients (n = 68), patients with endocarditis (n =17; 5 patients with B. henselae endocarditis and 12 with B. quintana endocarditis), and homeless people (n = 19) with chronic B. quintana bacteremia but without endocarditis, were used in the study. Diagnoses of CSD were based on PCR amplification of B. henselae DNA from lymph node samples of patients with regional lymphadenopathy and close contact with cats (67). Bartonella endocarditis was diagnosed by positive blood cultures or PCR amplification of DNA of the organism from valve samples (25). Diagnoses of B. quintana bacteremia for the 19 homeless patients were based on blood culture results (11).

Controls included 40 apparently healthy blood donors (group A), 10 patients initially thought to have CSD because of lymphadenopathy but later, by culture, found to have tuberculosis (group B), and 35 patients who were seropositive for bacterial or viral diseases not related to Bartonella species and for whom stored serum samples were available (group C). Group C included patients seropositive for cytomegalovirus (5), Epstein-Barr virus (5), hepatitis B virus (5), human immunodeficiency virus (5), acute Q fever (5), chronic Q fever (5), or Mediterranean spotted fever caused by Rickettsia conorii (5).

Immunofluorescence assays.

All serum samples were examined by IFA for the presence of antibodies to B. quintana and B. henselae. For antigens we used either our slides prepared in-house or commercial slides from Focus Technologies (FT), Cypress, Calif. (distributed in France by Eurobio, Paris, France).

We prepared in-house slides by using B. henselae strain Houston-1 or B. quintana strain Oklahoma as the antigen, as previously described (23, 37, 47). Briefly, bacteria were grown in the human endothelial cell line ECV 304 for 2 weeks before infected cells were recovered and lysed by sonication. Cell suspensions were centrifuged at 1,500 rpm (700 × g) for 10 min to remove cell debris, and bacteria were purified by using a sucrose gradient (phosphate-buffered saline [PBS] with 25% sucrose; 6,000 × g for 30 min). The resulting pellet was washed in PBS twice (6,000 × g for 10 min), and purified bacteria were pelleted by centrifugation (10,000 × g for 10 min), resuspended in 1 ml of 0.1% paraformaldehyde, and incubated at 4°C for 24 h for fixation. The bacterial suspension was repelleted by centrifugation (10,000 × g for 10 min), resuspended in sterile distilled water, and stored at −80°C before use. This whole-cell antigenic preparation was plated onto Dynatech slides and used for detection of either IgM or IgG antibodies by IFA.

FT uses the same B. henselae Houston-1 and B. quintana Oklahoma strains for preparation of its commercial slides. Bacteria are grown in Vero cells for preparation of slides used to detect IgG and on blood agar before being harvested and diluted in yolk sac fluid for slides used to detect IgM.

For both the FT and in-house tests, IgG titers of ≥64 and IgM titers of ≥20 were used as cutoff titers with B. henselae or B. quintana antigens.

Statistical analysis.

Differences in antibody titers between groups and/or between IFA techniques were analyzed by using the χ² test for categorical variables or Student's t test for continuous variables. A two-sided P value of <0.05 was considered significant.

RESULTS

With serum samples from CSD patients, the sensitivities of the in-house and FT tests using a cutoff titer of 64 for detection of anti-B. henselae IgG antibodies were 52.9 and 91.2%, respectively (P < 0.001). Median IgG titers were 64 with the in-house test (range, <64 to 256) and 256 with the FT test (range, <64 to 1,024) (Fig. 1). IgM antibodies were detected at a low titer (i.e., 20) for only two patients with the in-house test and one patient with the FT test. IgG antibodies cross-reacting with B. quintana were detected in most of the serum samples (35 of 36 [97.2%]; median titer, 64) from CSD patients with the in-house test, but in only one serum sample with the FT test (P < 0.001).

FIG. 1.

Anti-B. henselae IgG titers in CSD patients.

All endocarditis patients had IgG titers of ≥64 against both B. quintana and B. henselae antigens. However, significant differences in antibody titers were found between the in-house and FT tests. With the in-house test, 9 of 12 (75%) serum samples from patients with B. quintana endocarditis had homologous IgG titers of ≥1,024, whereas such high titers were present in only 2 of 12 (16.7%) patients with the FT test (P = 0.004). Also, median homologous IgG titers in serum samples from B. quintana endocarditis patients were 1,024 with the in-house test but only 256 with the FT test (P < 0.01) (Fig. 2). Anti-B. henselae cross-reacting antibodies were found in all B. quintana endocarditis patients (median IgG titers of 1,024 and 256 with the in-house and FT tests, respectively). All five B. henselae endocarditis patients had homologous IgG titers of ≥1,024 with both tests (range, 1,024 to 4,096; median titers of 2,048 and 1,024 with the in-house and FT tests, respectively). Most of them had IgG antibodies which cross-reacted with the B. quintana antigen, albeit at lower titers (median titers of 512 and 256 with the in-house and FT tests, respectively).

FIG. 2.

Homologous IgG titers in patients with B. quintana endocarditis.

No antibodies to B. quintana were found in 2 of 19 (10.5%) and 3 of 19 (15.8%) serum samples from homeless patients with B. quintana bacteremia with the in-house and FT tests, respectively. For the remaining patients, anti-B. quintana IgG antibodies were found at low titers (range, 64 to 512; median, 128 with both tests) (Fig. 3). Cross-reacting anti-B. henselae IgG antibodies were found for most of these patients, at similar titers. The two bacteremic patients with a homologous IgG titer of 512 had no echocardiographic evidence of endocarditis.

FIG. 3.

Anti-B. quintana IgG titers in homeless patients with chronic B. quintana bacteremia.

None of the 40 serum samples from blood donors (group A) were found to contain antibodies to B. quintana or B. henselae by either of the two tests. Among the group B controls, only one had an IgG titer (of 64) detectable by the in-house test, while with the FT test, serum samples from this and two additional controls had IgG titers of 64. No reactive antibodies were found for the 35 group C controls when the in-house test was used. In contrast, serum samples from 8 of 35 (22.8%) group C controls had IgG titers (of 64 to 256) to Bartonella spp. in the FT test. Cross-reacting IgG antibodies were found in serum samples from three cytomegalovirus patients, two chronic Q fever patients, and three patients with Mediterranean spotted fever. Thus, by using a cutoff titer of ≥64, the specificities of the in-house and FT tests for reactive IgG were both 100% for blood donors (group A); 90 and 70%, respectively, for group B; and 100 and 90%, respectively, for group C (P = 0.002). For all 85 controls taken together, the specificities of the FT and in-house tests were 87 and 98.8%, respectively (P = 0.002). To ensure a specificity of ≥95% with the FT test, a titer of ≥128 rather than ≥64 should have been considered the cutoff. With this cutoff titer, the sensitivity would be lower (85.3%) for CSD patients, but this would still be superior to that of the in-house test.

DISCUSSION

Bartonella serology has been performed in our laboratory for several years now, by using B. quintana or B. henselae grown in cell culture as the antigen. The aim of our study was to compare the sensitivity and specificity of our in-house test to that of a recently available commercial test from FT. With CSD patients, our results confirm the previously reported high sensitivity of the FT serological test (29, 48, 55, 65), whereas a lower sensitivity and significantly lower median IgG titers were found with our in-house test. Reactive IgM antibodies were rarely detected for CSD patients by either test, and these antibodies would thus appear to be of little use diagnostically, as previously mentioned (6, 27, 55, 63). The sensitivity of the B. henselae IFA test first described by Regnery et al. (50) for detection of CSD has been reported from different laboratories as 32% (6) to 100% (45, 55). Enzyme immunoassay (4, 6,27, 40, 46, 59) and Western blotting (40, 44) techniques have not been consistently superior to IFA for the diagnosis of CSD.

Several factors may influence the apparent sensitivity of a serological test. First, sensitivity may change according to disease case definition. Historically, a CSD case has been defined as the combination of regional lymphadenopathy with characteristic histopathological features (granuloma), the presence of an inoculation skin lesion (e.g., cat scratch), a history of cat contact, a positive result on a skin test using an antigen prepared from a CSD lymph node exudate, and negative laboratory results for other causes of lymphadenopathy (13). More recently, less restrictive case definitions have been used (e.g., regional lymphadenopathy and a recent contact with a cat [54]), since the skin test antigen is no longer available and lymph node biopsies are performed infrequently. However, a number of patients (e.g., 14% in the study of Flexman et al. [24]) with elevated anti-B. henselae antibody levels report no previous contact with a cat, and many patients present with clinically atypical cases of CSD (14, 41, 61). Use of a restrictive case definition for CSD may erroneously increase the sensitivity of a serological test because only patients with evident epidemiological and clinical findings suggestive of CSD will be considered. Thus, in a series of 354 suspected cases of CSD where serum samples were examined for anti-B. henselae antibodies, Flexman et al. (24) reported a sensitivity of 18% (with a cutoff titer of ≥1:64). However, when only the 98 patients who met the clinical case definition of CSD (lymphadenopathy and recent contact with a cat) were considered, the sensitivity was 54%. Accurate determination of the sensitivity and specificity of a serological test requires well-defined patient groups (infected and uninfected), identified by objective diagnostic procedures rather than clinical and epidemiological data. In a few recent studies, B. henselae DNA amplification from lymph node biopsy samples was used for confirmation of the diagnosis of CSD in a portion of cases (6, 55, 65). Giladi et al. (27) recently evaluated an enzyme immunoassay for detection of anti-B. henselae antibodies in PCR-confirmed cases of CSD. We report the first evaluation of IFA tests for Bartonella-infected patients for whom diagnoses were unambiguously established by either culture or DNA amplification using PCR. Diagnoses of diseases not related to Bartonella spp. in control patients were also carefully documented

The apparent sensitivity of a serological test may also be greatly influenced by the cutoff values used. Usually, cutoff titers are chosen to obtain a specificity of at least 95% in control patients (16). The control groups are usually composed of a mixture of apparently healthy blood donors and patients with a disease distinct from the disease being tested. Using an IFA test for detection of anti-B. henselae IgG antibodies at a cutoff titer of 64 in CSD patients, Regnery et al. (50), and Zangwill et al. (64) in the United States, reported sensitivities of 88 and 84%, respectively, but the specificity was only ∼90%. In each study, specificity was high at ∼99% when a titer of 512 was considered the cutoff, but sensitivities decreased to only 67 and 64%, respectively. Bergmans et al. (6) in The Netherlands reported a sensitivity as low as 31.8% by using a cutoff titer of 512, which was needed to obtain a specificity of ≥95% in blood donors. The cutoff titer may depend on the seroprevalence of antibodies to Bartonella spp. in the population, and this may vary with geographical region (45, 48, 55, 65). We found a specificity of 98.8% with the in-house test but only 87% with the FT test for the 85 controls we studied. With the FT test, the specificity was only 70% for the 10 patients with tuberculous lymphadenopathy. In our laboratory, cultures of about 5% of lymph node biopsy samples submitted yield Mycobacterium tuberculosis, while Bartonella spp. are demonstrated in approximately 30% of samples by culture or PCR. A highly specific test for CSD is critical in order to accurately rule out other, more serious diseases which may present as chronic lymphadenopathy, including tuberculosis or lymphoma. To ensure a specificity of ≥95% with the FT test, a cutoff titer of ≥128 should be used instead of ≥64.

It is becoming clear that some patients with typical CSD have no detectable antibodies to B. henselae, as shown in our study. Demers et al. (21) reported a seronegative patient diagnosed by culture of B. henselae from a lymph node. Serology may be negative early in the course of the disease (5, 19, 21). Also, the antigenic preparation used for serology may greatly influence the results. Low antibody titers are usually found when B. henselae grown on agar is used as the antigen (1) rather than cell culture-grown organisms (19, 50, 64). Poor antibody responses to B. henselae antigen have also been ascribed to the antigenic variability between the two B. henselae serotypes that have been described (22). Both the in-house and FT tests use B. henselae strain Houston-1 grown in cell culture as the antigen. Discrepancies between the two techniques with CSD and endocarditis patients may be explained by the mode of antigen preparation and conservation, as this may greatly influence antigenic-epitope recognition by serum antibodies (16). Finally, other Bartonella species (including B. clarridgeiae) may cause CSD (35, 39), and antibodies to these organisms do not react with B. henselae. This was not the case in our study, since our PCR results showed that all CSD patients were infected with B. henselae, and the lack of detectable antibodies to B. henselae in some patients may then truly represent a lack of sensitivity in the serological tests.

For homeless patients with chronic B. quintana bacteremia, both the FT and in-house tests detected only low levels of homologous IgG antibodies, and both failed to detect any reactive antibodies in many patients with culture-positive B. quintana bacteremia. There are few serological data on B. quintana infections. Although two large epidemics of trench fever occurred during the first and second World Wars (43), B. quintana had not yet been cultured (60) and serology was unavailable (18). More recently, data from homeless people with B. quintana infections indicate that most suffer from chronic, intermittent bacteremia and that many do not have detectable antibodies to B. quintana when infections are diagnosed (11, 31, 57).

The in-house and FT tests both detected homologous antibodies in patients with B. henselae endocarditis. However, only IgG titers obtained with the in-house technique allowed efficient discrimination between endocarditis patients and those with CSD. All B. henselae endocarditis patients presented with homologous IgG titers of ≥1,024, whatever the technique used. In contrast, none of the CSD patients presented with IgG titers of ≥1,024 with the in-house technique, whereas 13 of 68 (19.1%) had IgG titers of ≥1,024 with the FT technique (P < 0.001). Thus, our in-house technique, but not the FT technique, may help define patients with a high likelihood of Bartonella-related endocarditis. In our experience, this may be critical, because Bartonella sp. endocarditis is a chronic, clinically unspecific disease. In this study, an anti-B. henselae IgG titer of ≥1,024 with the in-house technique was 100% predictive of endocarditis (12), which is in accordance with previous reports from our laboratory (25, 47).

The in-house technique was more effective than the FT technique in detecting homologous IgG antibodies in serum samples from homeless patients with B. quintana endocarditis. The strain of B. quintana and the number of passages in culture have previously been reported to influence seroreactivity (19). The FT and in-house techniques use the same B. quintana Oklahoma strain, grown in cell cultures. The use of a cell culture- rather than an agar-grown antigen may favor the detection of specific antibodies, as previously described for B. henselae (1). With the in-house (19, 50, 64) test, a homologous IgG titer of ≥1,024 was 100% predictive of B. quintana endocarditis. Many B. quintana endocarditis patients, however, presented with lower IgG titers, and titers of 256 or 512 were found in 5 of 12 (41.6%) endocarditis patients. Such titers were also found in 7 of 19 (36.8%) homeless people with chronic bacteremia but with no endocarditis. These two groups of patients, then, could not be reliably differentiated by using IgG titers to B. quintana.

Cross-reactions have been described between various Bartonella spp. (19, 55) and with Chlamydia trachomatis (23, 42), Coxiella burnetii (37), and many other species, including Rickettsia rickettsii, Ehrlichia chaffeensis, Treponema pallidum, Francisella tularensis, and Mycoplasma pneumoniae (44). Cross-reactivity of serum samples containing IgM antibodies directed at Epstein-Barr virus capsid antigen has also been reported (66). The cross-reactivity between Bartonella sp. and Chlamydia sp. antigens has been particularly troublesome, and endocarditis supposedly due to a Chlamydia sp. on the basis of serology has later been shown to be due to a Bartonella sp. (42). The possibility of cross-reactions interfering with the serological diagnosis of Bartonella endocarditis is removed if all suspect serum samples are also tested for antibodies to C. burnetii and Chlamydia spp.

In conclusion, our study has shown the high sensitivity of the FT test in detecting IgG against B. henselae in CSD patients. It should, however, be used with a cutoff titer of 128 to raise the specificity of the test to ≥95%. Highly specific tests are essential for CSD, because a misdiagnosis may delay the detection of more serious conditions (e.g., lymphoma). Our in-house test had an advantage over the FT test in that it allowed CSD patients to be reliably differentiated from those with B. henselae endocarditis. An IgG titer of ≥1,024 was 100% predictive of B. henselae endocarditis. Our in-house test was also more effective than the FT test at detecting antibodies to B. quintana in homeless patients, especially those with endocarditis. An anti-B. quintana IgG titer of ≥1,024 was 100% predictive of endocarditis, although many patients with endocarditis presented with lower antibody levels.

Since both tests have advantages, concurrent use of both tests should be considered in order to improve the diagnosis of B. henselae and B. quintana infections.