Received 5 November 1996/Returned for modification 18 February 1997/Accepted 27 April 1997

Transcription

1 JOURNAL OF CLINICAL MICROBIOLOGY, Aug. 1997, p Vol. 35, No /97/$ Copyright 1997, American Society for Microbiology Pitfalls and Fallacies of Cat Scratch Disease Serology: Evaluation of Bartonella henselae-based Indirect Fluorescence Assay and Enzyme-Linked Immunoassay A. M. C. BERGMANS, 1 M. F. PEETERS, 2 J. F. P. SCHELLEKENS, 3 M. C. VOS, 4 L. J. M. SABBE, 5 J. M. OSSEWAARDE, 1 H. VERBAKEL, 2 H. J. HOOFT, 3 AND L. M. SCHOULS 1 * Research Laboratory for Infectious Diseases 1 and Laboratory for Infectious Diseases Research and Screening, 3 National Institute of Public Health and the Environment, Bilthoven, Laboratory of Medical Microbiology, St. Elisabeth Hospital, Tilburg, 2 Department of Bacteriology, University Hospital Rotterdam, Rotterdam, 4 and Department of Bacteriology, Regional Public Health Laboratory Zeeland, Goes, 5 The Netherlands Received 5 November 1996/Returned for modification 18 February 1997/Accepted 27 April 1997 The diagnostic value of the detection of immunoglobulin G (IgG) and IgM by Bartonella henselae-based indirect fluorescence assay (IFA) and enzyme-linked immunoassay (EIA) for the diagnosis of cat scratch disease (CSD) was evaluated. The IFA was performed either with B. henselae that was cocultivated for a few hours with Vero cells or with noncocultivated B. henselae as the antigen. Additionally, the performance of a Bartonella PCR hybridization assay based on the 16S rrna gene was determined and compared with those of the serologic assays. The study group consisted of 45 patients suspected of suffering from CSD by fulfilling one or more of the classical criteria. The specificities of the immunoassays were set at >95% by analysis of sera from 60 healthy blood donors. It is shown that the sensitivities of the IgG assays are very low (40.9% for the IFA with noncocultivated B. henselae as antigen) and that those of the IgM assays are higher (71.4% for the EIA) for patients who fulfilled two or more criteria for CSD. The IgM EIA showed the highest sensitivity: 71.4% in patients with two or more criteria for CSD and 80.6% for patients with a positive Bartonella PCR result. The results indicate that the specificities of both IFA and EIA IgG serologies and the sensitivity of the IFA IgM serology need to be improved. The PCR hybridization assay showed a sensitivity of 86.4% for patients who fulfilled two or more criteria for CSD and 100% for seven patients who fulfilled three or more criteria. The kinetics of IgG and IgM antibody production were studied in 18 patients with CSD on the basis of a positive B. henselae IFA IgM serology. The results indicate that there is no standard course of anti-b. henselae IgG and IgM production in patients with CSD, because some patients produced high levels of both IgG and IgM, others produced only high levels of IgM, and a few patients produced only low levels of antibodies. During the last few years it has become clear that the majority of cat scratch disease (CSD) cases are caused by Bartonella henselae (2, 6, 10, 13, 19). Culture of B. henselae from the lymph nodes of CSD patients has proven to be extremely difficult and therefore cannot be used for the routine diagnosis of CSD. Several reports describing the detection of B. henselae DNA in pus or lymph node biopsy specimens by PCR have been published (2, 6). Thus, tests based on the detection of antigens or DNA sequences from B. henselae are useful for the diagnosis of CSD. However, such tests require invasive procedures. Therefore, many investigators used serologic methods such as indirect fluorescence assay (IFA) and enzyme-linked immunoassay (EIA) to diagnose CSD (1, 4, 6, 8, 9, 14, 17 22). In all IFAs described, immunoglobulin G (IgG) or total antibody titers and no IgM titers were determined. By the IFA, anti-b. henselae IgG antibodies were found in sera from patients with CSD, but they were also found in 2 to 6% of the sera from healthy controls and in 29% of sera from healthy family members of CSD patients (1, 9, 19, 21). As a consequence, diagnosis of active CSD infection on the basis of the anti-b. henselae IgG antibody titer is questionable. Additionally, the sensitivity of the IgG EIA has proven to be low (20). The * Corresponding author. Mailing address: Research Laboratory for Infectious Diseases, National Institute of Public Health and the Environment, P.O. Box 1, 3720 BA Bilthoven, The Netherlands. Phone: (31) Fax: (31) address: detection of anti-b. henselae IgM may be more indicative of an active B. henselae infection. The performances of the IgM EIAs described so far have proven to be better than those of most IgG assays, but the sensitivities need to be increased (4, 18, 20). No studies have compared the performances of B. henselae-based PCR and serologic assays for the diagnosis of CSD. In this study we evaluated the diagnostic performance of B. henselae-based IgG and IgM IFAs and IgG and IgM EIAs with sera from patients with clinical CSD. Additionally, the performance of a Bartonella PCR hybridization assay with a lymph node aspirate or a biopsy specimen was determined and was compared with those of the serologic assays. It is shown that the sensitivities of all serologic assays that we evaluated except for the B. henselae IgM EIA are rather low. The Bartonella PCR hybridization assay showed the highest sensitivity. MATERIALS AND METHODS Patients and samples. From December 1993 to May 1994, clinical material was collected from 45 Dutch patients suspected of having CSD on the basis of clinical symptoms. All patients had regional lymphadenopathy and were suspected of having CSD on the basis of one or more of the following four classical criteria for having CSD: (i) cat contact with the presence of a visible scratch or bite or a papule or vesicle at the primary inoculation site, (ii) negative results of routine cultures of a lymph node specimen, (iii) granulomatous abnormalities, characteristic of CSD, in a histopathologic analysis of lymph node material, and (iv) positive CSD skin test. Two patients with lymphadenopathy and cat contact were included, but no additional laboratory information with regard to the four criteria was available for these two patients. Pus aspirates or lymph node biopsy and serum samples were taken from all patients. The time of sampling ranged from 1931

2 1932 BERGMANS ET AL. J. CLIN. MICROBIOL. approximately 10 days to 5 months after the onset of the lymphadenopathy. Relevant information for each patient was collected by means of a questionnaire, filled out by the clinician. The form comprised questions about sex, age, cat contact, date of onset of lymphadenopathy, presence of a visible scratch or bite, presence of a papule or vesicle, site of lymphadenopathy, hepato- or splenomegaly, fever and other symptoms, and previous antibiotic treatment. Data were supplemented with laboratory results (if present) for human immunodeficiency virus (HIV) serology, CSD skin test, bacterial culture, histopathology, cytomegalovirus serology, Epstein-Barr virus serology, and toxoplasma serology. In addition, three or more successive serum samples were obtained from each of 18 other patients suspected of suffering from CSD on the basis of clinical manifestations and confirmed by positive IFA IgM serology with an acute-phase serum sample. These sera were collected during the period from August 1994 to October Sera from 60 healthy blood donors of 20 to 60 years of age were used to test the specificities of both IFAs and the EIA. PCR detection of Bartonella DNA in clinical material. Clinical samples that were used for the detection of Bartonella DNA consisted of pus aspirates and biopsy specimens from lymph nodes from the 45 patients suspected of having CSD mentioned above. Extraction of DNA from clinical material and subsequent PCR detection of Bartonella DNA were performed as described previously (6). Serologic assays. Analysis of sera was performed at the Laboratory of Virology of the National Institute of Public Health and the Environment in Bilthoven, The Netherlands, and at the Regional Public Health Laboratory in Tilburg, The Netherlands. Two different IFA protocols and one EIA protocol were used. In the first IFA protocol, B. henselae ATCC was cocultivated with Vero cells to detect anti-b. henselae IgG (6). No penicillin or streptomycin was added to the bacterial cells, and a 1:2,000 dilution of a goat anti-human IgG fluorescein conjugate (Cappel, Boxtel, The Netherlands) was used. For the detection of anti-b. henselae IgM, sera were treated with Quik-Sep IgM rprotein G affinity resin (Isolab, Mechelen, Belgium) to remove serum IgG. A 1:64-diluted rabbit anti-human IgM fluorescein conjugate (Dako, Glostrup, Denmark) was used to detect IgM antibodies. The lowest serum dilution tested was 1:16. For the second IFA protocol, the following modifications were made to the first protocol. No cocultivation of B. henselae ATCC cells with Vero cells was performed, but instead, a bacterial suspension of approximately 10 8 cells/ml was made from bacteria grown on Columbia agar supplemented with 5% sheep blood. The suspension was mixed with egg yolk emulsion (Unipath Ltd., Basingstoke, Hampshire, England), and the mixture was spotted onto Teflon-coated slides. The slides were air dried for 30 min and fixed in acetone for 30 min. Sera and conjugates were diluted in phosphate-buffered saline (PBS; ph 7.2) containing 0.05% bovine serum albumin (BSA) and 0.1% NaN 3. For detection of IgM antibodies, sera were pretreated with IgG blocking solution (Incstar Corporation, Stillwater, Minn.) to remove serum IgG. Antibodies to B. henselae were also determined by an EIA as described by Barka et al. (4), with minor modifications. Briefly, B. henselae ATCC was grown on Columbia agar plates (Unipath) containing 5% sheep blood for 7 days at 35 C in an atmosphere containing 5% CO 2. Colonies were scraped from two plates, suspended in 5 ml of 0.04 M Na 2 CO 3 (ph 9.6), diluted to an optical density at 620 nm of 1.0 to 1.4, and sonicated for 3 min. After sonication a 1:200 dilution in 0.01 M PBS (ph 7.2) was used to coat the microtiter plates (medium capacity binding; Greiner, Alphen a/d Rijn, The Netherlands). After overnight incubation at 4 C, the wells were blocked for 1hat37 C with 1% BSA in PBS. Prior to IgM analysis, the IgG was removed from the serum specimens by using Quik-Sep IgM rprotein G affinity columns (Isolab). Serum specimens were diluted 1:200 in PBS with 0.05% Tween 20, 1% BSA, and 40% Chanock-PPLO medium. Rabbit anti-human IgG or rabbit anti-human IgM labeled with horseradish peroxidase was used to detect bound IgG or IgM, respectively. Incubations were carried out in duplicate, with and without antigen, for 1 h at 37 C. The titers were calculated relative to the titers for negative and highly positive reference sera from CSD patients by linear regression. RESULTS Patients data and PCR. Forty-five samples and questionnaires from patients suspected of suffering from CSD were examined. Forty-two patients had a history of cat contact, and 6 of the 45 patients reported a visible cat scratch with the presence of a papule or vesicle at the primary inoculation site. Thirty-six of the 38 cultured lymph node or pus specimens were negative, one culture yielded Staphylococcus aureus, and one (from an HIV-seropositive patient) yielded Staphylococcus epidermidis. Histopathologic analysis of lymph node specimens was performed for 25 patients, and 14 of the specimens showed granulomatous inflammation indicative of CSD. CSD skin tests were performed for 17 patients, and 7 were positive. Eleven patients were tested for the presence of antibodies to toxoplasma, 7 patients were tested for antibodies to cytomegalovirus, and 5 patients were tested for antibodies to Epstein-Barr virus infections. None of these patients had IgM antibody titers indicative of an active infection. All patients had regional lymphadenopathy. On the basis of clinical and laboratory data, patients were classified into three groups. The first group (CSD patients; 7 patients) consisted of patients who fulfilled three of the four classical criteria for having CSD. The second group (probable CSD patients; 15 patients) consisted of patients who fulfilled two of those criteria, and the third group (possible CSD patients; 21 patients) consisted of patients who fulfilled only one of the criteria. Patients for whom few data were available thus belonged to group 2 or 3, although they might have fulfilled three or more of the four criteria indicative of CSD. Data for two patients that fulfilled only part of one criterion for CSD were not included for determination of assay sensitivities. Comparison of PCR and clinical data. In 33 of 45 pus aspirates or lymph node biopsy specimens, Bartonella DNA could be detected by PCR (73%). The distribution of PCRpositive samples was almost equal among male and female patients, it was highest in patients of 20 years of age and younger (16 of 18 [89%]), and it was lowest in patients older than 40 years of age (4 of 7 [57%]). Almost all patients with a positive PCR result reported cat contact (32 of 33 [97%]). The majority of the patients with a visible cat scratch had a positive PCR result (14 of 16 [88%]). PCR positivity was highest for patients with axillary lymphadenopathy (15 of 16 [94%]) and lowest for patients with cervical lymphadenopathy (7 of 15 [47%]). All seven patients with a positive skin test result also had a positive PCR result. Only 7 of the 15 patients (47%) with a positive PCR result had a positive skin test result. Both HIV-seropositive patients had a negative PCR result. One 11-year-old patient with a positive PCR result had splenomegaly, in addition to axillary and inguinal lymphadenopathies. The rate of PCR positivity was relatively high for patients with malaise and fatigue (10 of 12 [83%]). Eighteen patients were treated with antibiotics, and 67% of these 18 patients had a positive PCR result after treatment, while 79% of the patients without antecedent antibiotic treatment had a positive PCR result. Performances of serologic assays and PCR. The cutoff values for positive serology in the different assays were determined by analyzing 60 serum samples from healthy blood donors in the assays and choosing the titer at which a minimum of 95% of the donor sera were negative. The cutoff values for positive serology were 512 for the IgG IFA with cocultivated B. henselae as the antigen, 128 for the IgG IFA with noncocultivated B. henselae as the antigen, 16 for both IgM IFAs, 900 for the IgG EIA, and 250 for the IgM EIA. The serologic results were classified into three categories: (i) negative serology (IgG IFA and IgM-IFA titers, 16; IgG EIA and IgM EIA titers, 200, (ii) ambiguous serology, and (iii) positive serology. In the IgM IFAs no ambiguous antibody titers were distinguished. The sensitivities of the different serologic assays were determined for the groups of patients described above. Only a positive (and not ambiguous) serology was assumed to be diagnostic and was used to calculate sensitivities. As group 1 (CSD patients, fulfilling a minimum of three of four criteria for CSD) consisted of only seven patients, patients in groups 1 and 2 were joined together and are referred to as probable CSD patients. The performances of the serologic assays and the PCR hybridization assay are presented in Table 1. The sensitivity of

3 VOL. 35, 1997 PITFALLS AND FALLACIES OF CAT SCRATCH DISEASE SEROLOGY 1933 TABLE 1. Performances of IFAs with either cocultivated or noncocultivated B. henselae as antigen, EIA with noncocultivated B. henselae as antigen, and Bartonella PCR Assay and antibody Test group a No. of patients No. of patients with the following result: Positive Ambiguous Negative Sensitivity (%) b Specificity (%) IFA (cocultivation) IgG Probable CSD patients Possible CSD patients Donors IgM Probable CSD patients Possible CSD patients Donors IFA (noncocultivation) IgG Probable CSD patients Possible CSD patients Donors IgM Probable CSD patients Possible CSD patients Donors EIA (noncocultivation) IgG Probable CSD patients Possible CSD patients Donors IgM Probable CSD patients Possible CSD patients Donors PCR Probable CSD patients Possible CSD patients a Test groups were (i) patients who fulfilled a minimum of two of the four classical criteria for CSD (probable CSD patients), (ii) patients who fulfilled only one of the classical criteria for CSD (possible CSD), and (iii) healthy blood donors. b For calculation of the sensitivities, only positive test results were used. the PCR proved to be 100% for the seven patients who fulfilled three or four of the four criteria for CSD. The sensitivities of the serologic assays for the latter group of seven patients were 28.6, 42.9, 16.7, 42.9, 42.9, and 83.3%, respectively, for IgG IFA (cocultivation), IgG IFA (noncocultivation), IgG EIA, IgM IFA (cocultivation), IgM IFA (noncocultivation), and IgM EIA (data not shown). Comparison of serologic assays and PCR. Comparisons between results of the serologic assays and the PCR hybridization assay are presented in Table 2. In the comparisons of serologic test results, 45 serum samples from the patients suspected of having CSD and 60 serum samples from healthy blood donors were used. When the serologic test results were compared with the results of the PCR, only specimens from the 45 patients suspected of having CSD were used. It is shown that concordance between the results of the three (IgG and IgM) assays is rather high. However, when the results of the separate serologic assays are compared with those of the PCR, concordance is low except for that for the IgM EIA and PCR. The discordance between serology and PCR was mainly due to the high number of patients (approximately 50% of all patients) with a positive PCR result and negative serology. Few patients had a negative PCR result and positive IgG serology. Concordance was highest between the IgM EIA and the PCR (76.7%), which is mainly due to the high sensitivities observed for both assays. Correlation of PCR and serologic assay results with duration of disease. The correlation between serologic assay and PCR results with the interval between the onset of lymphadenopathy and sampling was determined in probable CSD patients as described above. The time intervals were known for 17 patients. The duration of disease at the time of sampling varied between 10 and 103 days. No significant differences in duration of disease were found between patients with positive serology and patients with negative or ambiguous serology, nor was this the case in IgG assays and IgM assays. Two of the 17 patients with known durations of disease had a negative PCR result. Samples from the latter two patients were taken for PCR at approximately 78 and 103 days after the onset of lymphadenopathy. The 15 PCR-positive samples were taken at approximately 10 to 94 days after the onset of the disease. Antibody kinetics in patients with CSD. A longitudinal IFA study was performed with sera from 18 patients suspected of having CSD; their acute-phase sera had positive IgM serology, as determined by IFA with noncocultivated B. henselae, and three or more serum samples were available from these patients for IFA analysis. Pus samples from 5 of the 18 patients were available for the Bartonella PCR, and all 5 samples were

4 1934 BERGMANS ET AL. J. CLIN. MICROBIOL. TABLE 2. Concordant and discordant test results of IFA with either cocultivated or noncocultivated B. henselae as antigen, EIA with noncocultivated B. henselae as antigen, and Bartonella PCR for pairs of serologic and PCR assays a % Concordant % Discordant c Antibody and assay pairs b Total no. of pairs Positive- Negative- Total Positive- Negative- Total (100%) Positive Negative (no. of specimens) Negative Positive (no. of specimens) IgG IFA cocult.-ifa non-cocult (91) (14) 105 IFA cocult.-eia (93) (10) 103 IFA non-cocult.-eia (85) (18) 103 IFA cocult.-pcr (20) (25) 45 IFA non-cocult.-pcr (24) (21) 45 EIA-PCR (17) (26) 43 IgM IFA cocult.-ifa non-cocult (96) (9) 105 IFA cocult.-eia (84) (19) 103 IFA non-cocult.-eia (85) (18) 103 IFA cocult.-pcr (24) (21) 45 IFA non-cocult.-pcr (25) (20) 45 EIA-PCR (33) (10) 43 a Ambiguous serologic assay results were considered negative. b cocult., cocultivation; non-cocult., noncocultivation. c For discordant results, positive-negative and negative-positive refer to the order of assays as listed. positive. The IFA results for all 18 patients are depicted in Fig. 1. For only 5 of the 18 patients, the date of onset of disease was known; these varied from approximately 7 to 44 days before the first serum sample was taken. Therefore, the date of the first serum sample from each patient is designated day 0 in Fig. 1. Different courses of IgG and IgM antibody titers were found in sera from different patients. Figure 1 indicates that if a patient had a positive B. henselae IgM serology, he or she did not always have a positive B. henselae IgG serology (titers, 128) (patients B, F, J, L, N, Q, and R). It is possible that some patients with CSD produced detectable amounts of IgM but insignificant titers of IgG against B. henselae. Only 11 of the 18 patients (61%) had at least one serum sample with an IgG titer of 128, and in 6 of these 11 patients (55%), the maximum titer was only 128. Analysis of the antibody kinetics revealed that in almost all patients tested, anti-b. henselae IgM antibodies are produced earlier than IgG antibodies and that the IgM antibodies disappear within approximately 100 days after collection of the first serum sample. IgG titers remained elevated for longer periods. DISCUSSION Several serologic and molecular studies have recently established that B. henselae is the major cause of CSD (2, 6, 19). PCR has proven to be a reliable tool for the diagnosis of CSD (2, 6). However, because the PCR assay requires invasive methods to obtain clinical material, many physicians will prefer serologic analysis. In this study, the validity of serologic tests for the diagnosis of CSD was determined with samples from well-documented patients suspected of having CSD. Cutoff values for the serologic assays were set by determining antibody titers in healthy blood donors, with the requirement that the specificities had to be 95%. Many of the blood donors had moderately high IgG antibody titers. Probably, those IgG-seropositive donors were cat owners with past subclinical B. henselae infection. To obtain assays that are highly diagnostic for active CSD infection, the cutoffs for the IgG assays were necessarily set very high, resulting in low sensitivities, especially for the IgG EIA (9.5%). These findings result in the implication that the IgG-based serologic assays described in this report are unsuitable for the reliable diagnosis of active B. henselae infection. The results of other investigators also indicate that diagnosis of CSD by determining only anti-b. henselae IgG antibody titers may not be reliable (9, 21). Demers et al. (9) found by IFA positive IgG serology in 29% of asymptomatic family members of CSD patients, and Zangwil et al. (21) found positive IgG serology in 18% of asymptomatic family members of CSD patients, as determined by IFA. In most studies only the presence of IgG antibodies is determined (9, 14, 19, 21, 22). The sensitivity of the IgM EIA (71.4%) was higher than that of the IgM IFA (50.0%). The PCR showed the highest sensitivity for CSD patients (86.4%). For 10 patients discordant results were found when the results of PCR were compared with those of the IgM EIA. For only three of those patients, the discrepancies may be due to a difference in the time of collection of serum and lymph node samples. It is possible that at the time of serum sampling, the IgM antibodies had already disappeared. However, time intervals between the onset of lymphadenopathy and sampling for patients positive for IgM by EIA were similar to those for patients negative for IgM by EIA. The discrepancies between the PCR and the IgM EIA can also be explained by the lower sensitivity of the EIA. Alternatively, some patients produce low-level anti-b. henselae IgM antibodies or none at all. A negative PCR result in combination with positive EIA IgM serology may be explained by the elimination of the organism before IgM antibodies disappear. Barka et al. (4) found detectable IgM in only 55% of patients with confirmed CSD. They also reported that sera from patients with suspected CSD of very recent onset may contain only anti-b. henselae IgM and not anti-b. henselae IgG. Szelc- Kelly et al. (20) reported that IgM and IgG antibodies to B. henselae and B. quintana as determined by EIA were significantly higher in CSD patients than in control subjects (20). Their findings were comparable to ours: their Bartonella IgM EIA had sensitivities of 66% for patients with clinical CSD and 73% for skin test-positive CSD patients and a specificity of 95%. The specificity of their IgG EIA was similar to that of their IgM EIA. However, the sensitivity of their IgG EIA was as low as that of our IgG EIA: 16% for patients with clinical CSD and 18% for skin test-positive patients. Cocultivation of

5 VOL. 35, 1997 PITFALLS AND FALLACIES OF CAT SCRATCH DISEASE SEROLOGY 1935 FIG. 1. Antibody kinetics in 18 patients (patients A to R) with CSD on the basis of positive B. henselae IgM serology as determined by IFA.

6 1936 BERGMANS ET AL. J. CLIN. MICROBIOL. B. henselae with Vero cells increased the sensitivity of the IgG EIA to 32% for patients with clinical CSD and to 35% for skin test-positive CSD patients. Patnaik and Peter (18) found by EIA that 94% of sera from CSD patients had IgG and/or IgM antibodies against B. henselae, while 10% of the control sera contained IgG. Only 2% of 200 healthy blood donors had IgG antibodies to B. henselae, and none of the donors had IgM antibodies. In the IgG IFA with cocultivated B. henselae as the antigen, 25% of the sera from PCR-negative patients had high antibody titers. It is possible that for some PCR-negative patients the bacterium had already been eliminated at the time of sampling or that the PCR was false negative due to a sampling error. The discrepancy could also be explained by assuming that those patients were infected with a bacterium that was closely related to B. henselae and that raised IgGs cross-reacting with the B. henselae antigens used in that IFA. The cross-reactivity of B. quintana with several rickettsiae was reported in 1978 (15). Other reports also mention cross-reactivity between B. henselae and Chlamydia species (11). It is striking that with this IgG IFA, seropositivity for the seven PCR-positive patients that fulfilled three or more criteria for CSD (group 1; 28.6%) was not higher than seropositivity for patients fulfilling only one or two criteria (groups 3 and 2; 33.3%). These data result in the implication that this IgG IFA with cocultivated B. henselae as the antigen is not informative for the diagnosis of CSD. The study on antibody kinetics indicate that it is not sufficient to test one serum sample from patients suspected of having CSD for the presence of anti-b. henselae IgG. However, even if two consecutive serum samples from one patient are tested to observe an IgG titer increase to or above the cutoff, several cases of CSD will be missed. In this study only sera from patients with positive IgM serology were evaluated. We also found patients with PCR-proven CSD who had elevated IgG titers and no detectable IgM antibodies and patients with no detectable anti-b. henselae antibodies at all. From the latter results we conclude that the B. henselae IgG and IgM antibody kinetics vary widely between patients with CSD. In contrast, Barka et al. (4) found that none of the 40 serum samples from CSD patients tested in an EIA contained IgM in the absence of IgG. Only 58% of all patients produced measurable IgM antibodies. From our results it can be concluded that in the majority of CSD patients with positive IgM serology, the anti-b. henselae IgM response disappears within approximately 100 days after the onset of the symptoms and that the IgG response, if present, generally appears later and remains elevated during a longer period, as is normal in bacterial infections. Dalton et al. (8) described a longitudinal study in which they used sera from 420 IgG-seropositive CSD patients to construct a geometric mean IgG titer curve. However, they did not include the results for patients with a negative IgG serology, and no antibody kinetics for individual patients were shown. Therefore, one should be careful in drawing conclusions on anti-b. henselae IgG kinetics from those results. Szelc- Kelly et al. (20) tested antibody responses in paired serum specimens from 63 patients with clinical CSD by EIA and found that the IgM responses had significant declines between the first and second samples. No differences among the median IgG levels were noted between the first and second serum samples from these patients. Again, no serologic results for individual patients were shown, and the antibody titer ranges that they found were very wide. In our view, the most appropriate strategy for the diagnosis of CSD would be as follows. First perform a B. henselae IgM EIA. If the IgM serology by EIA is positive, the patient suffers from CSD. If the IgM serology by EIA is negative, a Bartonella PCR with pus or a lymph node biopsy specimen should be performed. A positive PCR result will confirm that the patient suffers from CSD. However, a negative PCR result does not exclude the possibility that the patient has CSD. The results of the study presented here indicate that the serologic tests for B. henselae need to be improved. Two or more consecutive serum samples from patients need to be tested when IgG antibody titers are measured to observe a rise in the antibody titer. However, even with consecutive serum samples, a significant change in IgG titer is not always observed. Therefore, one must be very careful in using IgG titers for the diagnosis of CSD. IgG-based serology may be improved by using modifications that enhance its specificity. This may be achieved by adsorption of serum samples with cross-reacting antigens before B. henselae serology is performed. IgM serology proved to be highly specific and would be a valuable diagnostic tool if its sensitivity could be improved. In a recent report the existence of different serotypes of B. henselae is described (12). Thus, the use of local B. henselae strains or a mixture of different strains as antigen in an EIA or IFA may improve serologic diagnosis of CSD. Dalton et al. (8) described a lowered seroreactivity of the B. quintana Fuller strain compared to that of the B. quintana OK strain. The Fuller strain has an extensive passage history and may have lost immunodominant antigens. Additionally, Batterman et al. (5) showed by transmission electron microscopy that low-passagenumber, piliated B. henselae adhered to and invaded HEp-2 cells to a greater extent than did multiply passaged B. henselae with reduced pilus expression. Thus, the sensitivity of B. henselae serology may be improved by using bacterial strains that have had few in vitro passages or by using B. henselae strains that have been cocultivated with Vero cells for a few days. Major improvements may also be expected from the use of purified antigens in serologic assays. Recently, a 17-kDa antigen of B. henselae that was highly reactive with serum samples from patients with CSD has been characterized and expressed in Escherichia coli (3). The use of recombinantexpressed antigens such as the 17-kDa protein might be of value for the serologic diagnosis of CSD and other Bartonella infections. ACKNOWLEDGMENTS We acknowledge B. A. M. van der Zeijst, J. D. A. van Embden, and J. W. Dorigo-Zetsma for critically reviewing the manuscript. We also thank H. Krul for technical assistance. REFERENCES 1. Amerein, M. P., D. de Briel, B. Jaulhac, P. Meyer, H. Monteil, and Y. Piemont Diagnostic value of the indirect immunofluorescence assay in cat scratch disease with Bartonella henselae and Afipia felis antigens. Clin. Diagn. Lab. Immunol. 3: Anderson, B., K. Sims, R. L. Regnery, L. Robinson, M. J. Schmidt, S. Goral, C. Hager, and K. Edwards Detection of Rochalimaea henselae DNA in specimens from cat scratch disease patients by PCR. J. Clin. Microbiol. 32: Anderson, B., E. Lu, D. Jones, and R. Regnery Characterization of a 17-kilodalton antigen of Bartonella henselae reactive with sera from patients with cat scratch disease. J. Clin. Microbiol. 33: Barka, N. E., T. Hadfield, M. Patnaik, W. A. Schwartzman, and J. B. Peter EIA for detection of Rochalimaea henselae-reactive IgG, IgM, and IgA antibodies in patients with suspected cat-scratch disease. J. Infect. Dis. 167: Batterman, H. J., J. A. Peek, J. S. Loutit, S. Falkow, and L. S. Tompkins Bartonella henselae and Bartonella quintana adherence to and entry into cultured human epithelial cells. Infect. Immun. 63: Bergmans, A. M. C., J. W. Groothedde, J. F. P. Schellekens, J. D. A. van Embden, J. M. Ossewaarde, and L. M. Schouls Etiology of cat scratch disease: comparison of polymerase chain reaction detection of Bartonella (formerly Rochalimaea) and Afipia felis DNA with serology and skin tests. J. Infect. Dis. 171:

7 VOL. 35, 1997 PITFALLS AND FALLACIES OF CAT SCRATCH DISEASE SEROLOGY Campbell, J. A. H Cat scratch disease. Pathol. Annu. 12(Part 1): Dalton, M. J., L. E. Robinson, J. Cooper, R. L. Regnery, J. G. Olson, and J. E. Childs Use of Bartonella antigens for serologic diagnosis of cat-scratch disease at a national referral center. Arch. Intern. Med. 155: Demers, D. M., J. W. Bass, J. M. Vincent, D. A. Person, D. K. Noyes, C. M. Staege, C. P. Samlaska, N. H. Lockwood, R. L. Regnery, and B. E. Anderson Cat-scratch disease in Hawaii: etiology and seroepidemiology. J. Pediatr. 127: Dolan, M. J., M. T. Wong, R. L. Regnery, J. H. Jorgensen, M. Garcia, J. Peters, and D. Drehner Syndrome of Rochalimaea henselae adenitis suggesting cat scratch disease. Ann. Intern. Med. 118: Drancourt, M., J. L. Mainardi, P. Brouqui, et al Bartonella (Rochalimaea) quintana endocarditis in three homeless men. N. Engl. J. Med. 332: Drancourt, M., R. Birtles, G. Chaumentin, F. Vandenesch, J. Etienne, and D. Raoult New serotype of Bartonella henselae in endocarditis and catscratch disease. Lancet 347: English, C. K., D. J. Wear, A. M. Margileth, C. R. Lissner, and G. P. Walsh Cat-scratch disease. Isolation and culture of the bacterial agent. JAMA 259: Hamilton, D. H., K. M. Zangwil, J. L. Hadler, and M. L. Cartter. Cat-scratch disease Connecticut, J. Infect. Dis. 172: Hollingdale, M. R., J. E. Herrmann, and J. W. Vinson Enzyme immunoassay of antibody to Rochalimaea quintana: diagnosis of trench fever and serologic cross-reactions among other rickettsiae. J. Infect. Dis. 137: Kalter, S. S., A. R. Rodriguez, and R. L. Heberling Cat-scratch disease skin-test antigen preparation. Lancet ii: Nadal, D., and R. Zbinden Serology to Bartonella (Rochalimaea) henselae may replace traditional diagnostic criteria for cat-scratch disease. Eur. J. Pediatr. 154: Patnaik, M., and J. B. Peter Cat-scratch disease, Bartonella henselae, and the usefulness of routine serological testing for Afipia felis. Clin. Infect. Dis. 21: Regnery, R. L., J. G. Olson, B. A. Perkins, and W. Bibb Serological response to Rochalimaea henselae antigen in suspected cat-scratch disease. Lancet 339: Szelc-Kelly, C. M., S. Goral, G. I. Perez-Perez, B. A. Perkins, R. L. Regnery, and K. M. Edwards Serologic responses to Bartonella and Afipia antigens in patients with cat scratch disease. Pediatrics 96: Zangwil, K. M., D. H. Hamilton, B. A. Perkins, R. L. Regnery, B. D. Plikaytis, J. L. Hadler, M. L. Cartter, and J. D. Wenger Cat scratch disease in Connecticut. Epidemiology, risk factors, and evaluation of a new diagnostic test. N. Engl. J. Med. 329: Zbinden, R., M. Höchli, and D. Nadal Intracellular location of Bartonella henselae cocultivated with Vero cells and used for an indirect fluorescent-antibody test. Clin. Diagn. Lab. Immunol. 2: Downloaded from on December 30, 2018 by guest