Blood culture-negative endocarditis remains a diagnostic and therapeutic problem, accounting for 5% to 30% of cases of endocarditis [15-17]. For 10 years, we have conducted surveillance of Q fever [18] as a cause of endocarditis. During the course of this surveillance, we received several blood specimens from patients who had a diagnosis of blood culture-negative endocarditis as defined by the criteria of Durack and colleagues [19]. During the past 2 years, we have systematically searched for Bartonella species as a cause of endocarditis, and we retrospectively studied serum specimens from a series of Canadian and a series of French patients with endocarditis, including patients who had previously received a diagnosis of chlamydial endocarditis [20, 21]. We have previously shown serologic cross-reactions between Chlamydia species and Bartonella species [10].

In this paper, we report 22 new cases of Bartonella endocarditis.

Methods

Top Methods Results Discussion Author & Article Info References

Patients

We prospectively studied 11 patients from France, the United Kingdom, and South Africa, and we retrospectively studied 123 patients from Halifax, Nova Scotia, Canada. Serum specimens for these patients were tested under code at Marseille. We also retrospectively studied 8 patients from France who had blood culture-negative endocarditis and had previously received a diagnosis of Chlamydia species infection on the basis of serologic testing. All patients fulfilled the criteria of Durack and colleagues [19] for the diagnosis of endocarditis.

Data Collection

Demographic and clinical data were abstracted from the medical records of each patient who had Bartonella endocarditis. Three epidemiologic features—homelessness, alcoholism, and contact with cats—were specifically sought.

Serologic Studies

Bartonella quintana, Oklahoma strain; B. henselae, Houston-1 strain (ATCC [American Type Culture Collection] 49882T); B. henselae, serotype Marseille [22]; and B. elizabethae (ATCC 49927T) [13] were used as antigens. The bacteria subcultured between the fourth and the seventh passages in a human endothelial cell line (ECV 304) were harvested, pelleted, and used as crude antigen. Chlamydia antigens included C. psittaci (ATCC-VR601) and C. pneumoniae (ATCC-VR1310), which were grown in HeLa 228 cells, and C. trachomatis (ATCC-VR878), which was grown in McCoy cell line. These antigens were used in a microimmunofluorescence assay that was done as reported elsewhere [23]. To investigate cross-reactivity between Bartonella species and Chlamydia species, serum specimens were adsorbed with Renografin-purified (Shering, Lys-lez-Lannoy, France) C. pneumoniae [24] or with B. quintana. The adsorption was difficult to complete with B. quintana, and several protocols were used. In the first protocol, bacteria were purified on a sucrose cushion and adsorbed three times. Because the adsorption was incomplete in patients 8, 11, and 12, the adsorption process was done a second time using sonicated, infected ECV 304 cells. Adsorbed serum specimens (final dilution, 1:40) were retested using the microimmunofluorescence assay.

Blood and Valve Cultures

We attempted to isolate Bartonella species from whole blood specimens and homogenized valvular biopsy material. Specimens were inoculated onto either 5% sheep blood agar (bioMerieux, Marcy-l'Etoile, France) or into the human endothelial cell line ECV 304 [12]. Inoculated media were incubated at 37 °C in a CO (₂) atmosphere for as long as 60 days. The identity of any isolates obtained was confirmed using species-specific mouse polyvalent antisera [11] or methods based on polymerase chain reaction (PCR) [25, 26].

Molecular Detection and Identification of Isolates

The DNA that was used as a template in PCR amplification was prepared from agar-grown single colonies of the isolate or from inoculated endothelial cells using proteinase K and sodium dodecyl sulfate digestion followed by phenolchloroform extraction and ethanol precipitation of nucleic acids. Crude DNA extracts were prepared from serum and whole blood specimens by using the QI Amp-Blood kit (Qiagen, Inc., Chatsworth, California) according to the recommendations of the supplier and from fresh or fixed paraffin-embedded valvular tissue by boiling with Chelex 100R (BioRad, Richmond, California) as described elsewhere [27]. Citrate synthase gene fragments were amplified from DNA extracts [28], and a semi-nested PCR assay was done by incorporation of internal primer CSBH1: 5'-AGCGTGATAGCAATATCAAG-3' to yield a 234-base pair product. Sterile water, processed concurrently, was used as a negative control in all PCR analyses. The identity of PCR products was confirmed by either Taq I restriction enzyme analysis [25] or base-sequence determination. Sequencing the 16S recombinant DNA gene of the isolates was done as reported elsewhere [10].

Transmission Electron Microscopy of Infected Valves

Small pieces of vegetation from the valves of three patients (patients 9, 11, and 12) were surgically removed and immediately fixed in 4% paraformaldehyde, 2.5% glutaraldehyde solution at 4 °C. Fixed tissue was processed for transmission electron microscopy. Small pieces of vegetation from B. quintana-infected valves were washed in phosphate-buffered saline (pH, 7.4), with 2% osmium tetroxide before dehydration in increasing concentrations (25% to 100%) of ethanol and acetone, and embedded in Epon 812 (Sigma, Saint-Quentin, France). Thin transverse sections were cut on an LKB Ultratome II microtome (Leica, Paris, France) and stained with a saturated solution of methanol-uranyl acetate and lead citrate in water before being examined on a JEOL 1200 electron microscope (Jeol, Ltd., Tokyo, Japan) at 80 kV.

Results

Top Methods Results Discussion Author & Article Info References

Twenty-two patients with Bartonella infection received a diagnosis of definite endocarditis according to the criteria of Durack and colleagues [19] (Table 1). Nineteen of these patients had a pathologic diagnosis of endocarditis after valvular surgery, and 3 (patients 10, 17, and 22) met one major criterion (vegetation) and at least three minor criteria for endocarditis, including body temperature greater than 38 °C, an echocardiogram consistent with infective endocarditis, and embolism. The median age was 47 years, and 17 of the 22 patients (77%) were male. Valvular disease had previously been diagnosed in 13 of the 22 patients (59%). All but 3 patients had body temperatures greater than 38 °C, 9 patients (41%) had embolic phenomena, and all but 2 patients had aortic valve involvement. Echocardiography showed vegetation in 19 patients (86%); death related to endocarditis occurred in 6 of these 19 patients (31%), although 15 of 22 patients (68%) received aminoglycoside therapy and 19 of 22 (86%) had valvular surgery.

Eight patients were homeless, 10 were alcoholic, and 3 owned cats. No patient was immunocompromised. Seventeen patients had preexisting valvular disease. Eight of 11 cases of B. quintana endocarditis occurred in homeless patients (patients 1, 2, 3, 21, 23, 24, 25, 26, 29, 30, and 32), 7 occurred in alcoholic patients, and 9 occurred in patients who had not previously had valvular disease (Table 2). Compared with patients infected with B. henselae, patients infected with B. quintana were more likely to be homeless and less likely to have had valvular disease. The 2 patients who had B. quintana infection and were not homeless were different from the others: Patient 27 lived in Algeria, and patient 28 was HIV positive.

Bartonella infection was diagnosed in 3 of 86 patients with endocarditis in Marseille (3.5%), 3 of 123 patients with endocarditis in Halifax (2.4%), and 4 of 90 patients with endocarditis in Lyon (4.4%). In the same series of patients, Q fever endocarditis was diagnosed in 9 patients in Marseille, 2 patients in Halifax, and 2 patients in Lyon. Bartonella species accounted for about 3% (10 of 299) of all cases of infective endocarditis.

Of 9800 serum specimens that were prospectively tested in our laboratory, 125 (taken from 26 patients) had a Bartonella species titer of 1:1600 or more. Of these 26 patients, 22 had Bartonella endocarditis, 1 had Q fever endocarditis, 1 had meningitis, and 2 had cat-scratch disease without endocarditis. A titer of 1:1600 or more had a positive predictive value for Bartonella endocarditis of 0.884. Serum specimens from patients with B. henselae infection (patients 13, 17, 19, 20, and 21) generally had the strongest reaction with B. henselae, although antibody levels were very low in one case (patient 20). Serum specimens from patients with B. quintana infection (patients 1, 2, 3, and 21) reacted with equal strength to B. quintana and B. henselae. Patients had low titers to B. elizabethae. In all but 1 patient (patient 20), the first available serum specimen yielded a titer of antibodies to Bartonella species of 1:1600 or more. Patient 20 was positive at a titer of 1:100 when B. henselae, serotype Marseille, was used as antigen and was nonreactive when B. henselae, Houston-1 strain, was used. Nine patients who were previously thought to have chlamydial endocarditis (patients 5 to 12 and 15) were reassessed and received a new diagnosis of Bartonella endocarditis—adsorption with Bartonella removed antibodies to both Bartonella and Chlamydia, whereas adsorption with Chlamydia species removed antibodies to Chlamydia only.

Blood culture was positive in five patients, and a B. quintana isolate was obtained from the valvular tissue of one patient (patient 21). An isolate was also obtained from the valvular tissue of patient 19 by using endothelial cell co-culture; however, attempts to subculture this strain failed.

Citrate synthase gene amplicons were obtained from 6 of 16 valves tested. In patient 20, the amplicons were the only evidence of Bartonella infection. Analysis of the results of PCR amplification show that all four valves that were unfixed when tested (those in patients 2, 13, 19, and 21) yielded a PCR product. However, of the slides of the embedded tissues obtained from 12 patients Table 3, only 2 were positive (those for patients 1 and 20). On the basis of culture or PCR analysis, the final diagnosis was determined to be B. quintana infection, B. henselae infection, or, when restriction enzyme analysis and sequence were unavailable, infection with an unidentified Bartonella species.

Electron microscopic studies of cardiac valves from patients with B. quintana endocarditis showed that B. quintana was located intracellularly Figure 1, although many extracellular bacteria grouped in small clusters were seen. The cell wall and the cytoplasmic membrane were preserved. Careful examination showed that these bacteria were surrounded by a thin vacuolar membrane. Dividing organisms were present.

View larger version (147K): [in this window] [in a new window]   Figure 1. Human valve infected with Bartonella quintana. Top. Transmission electron microscopy. Note the intracellular location of the organisms. Bacteria appear as dark dots (arrow) because of a packed DNA (methanol-uranyl acetate and lead citrate; original magnification, x 5000). Bottom. Higher magnification shows the trilamellar membrane (single arrow) and some dividing organisms (double arrows) (methanol-uranyl acetate and lead citrate; original magnification, x 15 000). M = mitochondria. Bar = 1 µm.

Discussion

Top Methods Results Discussion Author & Article Info References

Bartonella species have recently been recognized as etiologic agents of endocarditis. Only 11 cases of endocarditis caused by these organisms have been reported to date, including the 5 cases previously reported by our group [3, 10, 13, 22, 29-33]. We have identified 22 more cases.

The clinical features of these 22 patients were similar to those previously reported for patients with infective endocarditis [19], except that patients with Bartonella endocarditis were more frequently male and were more likely to have had valvular surgery. We could not identify optimal treatment because of the heterogeneity of the series. Because Bartonella species are highly susceptible to aminoglycosides, the usual recommendation of penicillin or ampicillin plus an aminoglycoside to treat blood culture-negative endocarditis will effectively treat Bartonella endocarditis [34, 35]. In one patient (patient 19), however, B. henselae was isolated after the completion of a course of aminoglycoside therapy. Some researchers have reported that ciprofloxacin is more effective [32, 33].

Our data allowed us to estimate that Bartonella species causes 3% of all cases of endocarditis and that many of the previously reported cases of blood culture-negative endocarditis [15, 17] may be due to Bartonella species. We previously reported the cross-reaction between Chlamydia antigens and Bartonella antisera [10], and this finding is confirmed here. In fact, nine patients who had previously been thought to have chlamydial endocarditis received a new diagnosis of Bartonella infection [19, 20]. One of these patients (patient 12) showed bacteria typical of Bartonella on electron microscopy.

It is difficult to isolate Bartonella species from human specimens, particularly blood specimens. In contrast, it is easy to isolate these organisms from the blood of cats [36]. Although many attempts at isolation have been made, only five isolates of B. henselae have been obtained from the lymph nodes of patients with cat-scratch disease [2, 21, 37], and two cases of culture-proven B. henselae endocarditis have been seen (in patients 19 and 22). It has been reported [38] that isolation of these microorganisms is easier in patients with the acquired immunodeficiency syndrome who have bacillary angiomatosis or septicemia induced by Bartonella species [38]. Bartonella species have been cultivated from blood using either Bactec blood culture media (Becton-Dickinson, Paris, France) or blood agar [38, 39]. We report one new isolation from blood. In our study, the best diagnostic tools were serologic tests and the PCR-based methods used on valvular tissue.

When the microimmunofluorescence assay was used, a titer of 1:100 indicated Bartonella infection [23]. A titer of 1:1600 had a positive predictive value for Bartonella endocarditis of 0.884. Some patients with Q fever endocarditis may have cross-reactive antibodies. When the first serologic test was developed for B. quintana-induced trench fever, it was recommended that patients with Q fever be excluded because of cross-reactions [40]. This cross-reaction is unidirectional because patients with Bartonella infection do not have antibodies to C. burnetii (Unpublished data). Patients with Bartonella endocarditis produce antibodies that react with C. pneumoniae, C. psittaci, and C. trachomatis [41]. We suggest adding positive results on a serologic test for Bartonella species as a minor criterion to the criteria proposed by Durack and colleagues for the diagnosis of endocarditis [19]. However, in some cases (such as patient 20 and other patients with reported cases of B. quintana [9] or B. henselae infection [38]), no antibodies were detected. Such findings highlight the fact that serologic tests alone cannot be relied on to provide a diagnosis, especially in immunocompromised patients.

An alternative diagnostic approach is to use PCR-based amplification of Bartonella DNA fragments from infected valves. This method was especially valuable when fresh valvular tissue was used, although paraffin-embedded valves can also be used [27], allowing a retrospective diagnosis to be made. In our series, however, we were unable to amplify DNA from embedded tissues in 8 of 10 attempts.

Clinical and epidemiologic data available for our patients adds weight to the microbiological diagnosis. Clinical presentation and epidemiologic factors depend on the Bartonella species involved.

Bartonella quintana was transmitted by the human body louse during World War I and resulted in trench fever. This could explain why B. quintana infections have been described primarily in patients with poor socioeconomic conditions, especially homelessness and alcoholism. We found more alcoholism among patients with B. quintana and more cat owners among patients with B. henselae. On the basis of the specificity of the B. quintana infections described above, we believe that patients 4, 9, 10, 18, and 22, who were homeless and had not previously had valvular disease, had B. quintana infection.

We conclude that Bartonella species accounted for 3% of all cases of endocarditis seen in our series. Blood cultures, serologic testing, and DNA amplification can be used for diagnosis. Bartonella infection should be considered in all patients with blood culture-negative endocarditis.

Dr. Marrie: Department of Medicine, Division of Infectious Diseases, Victoria General Hospital, Room 5014 ACC, 1278 Tower Road, Halifax, Nova Scotia 83H 2Y9, Canada.

Dr. Etienne: Laboratoire de Bacteriologie, Hopital Louis Pradel, 28, Avenue du Doyen Lepine, 69500 Lyon, France.

Dr. Cosserat: Service de Medecine Interne, Centre Medical-Chirurgical Foch, 40 rue Worth BP 36-92151, Suresnes, France.

Dr. Cacoub: Service de Medecine Interne, Groupe Hospitalier Pitie-Salpetriere, 47-83 Boulevard de l'Hopital, 75651 Paris Cedex 13, France.

Dr. Poinsignon: Service de Medecine Interne et Pathologie Vasculaire, Hopital Saint-Louis, 1 Avenue Claude Vellefaux, 75475 Paris Cedex 10, France.

Dr. Leclerc: Clinique Medicale et des Maladies Infectieuses, Centre Hospitalier Universitaire, BP 217, 38043 Grenoble, Cedex 09, France.

Dr. Sefton: Department of Medical Microbiology, London Hospital Medical College, Turnet Street, London E1 2AD, United Kingdom.