Bacillary angiomatosis in immunocompromised patients

Abstract

Objectives: 

To report seven cases of bacillary angiomatosis; to evaluate the most useful diagnostic tools; to analyse the clinical and epidemiological features associated with Bartonella quintana or Bartonella henselae infections.

Design: 

Clinical, diagnostic and epidemiological evaluation of 37 speciated bacillary angiomatosis cases in the literature, including the seven patients in our study.

Methods: 

Pathological examination of tissue samples, including Warthin–Starry staining and immunohistology; titre of antibodies to Bartonella sp.; detection of Bartonella sp. in blood and biopsy materials by culture or PCR; and statistical analysis of clinical and epidemiological features associated with B. quintana or B. henselae bacillary angiomatosis cases.

Results: 

Seven immunocompromised patients (six with AIDS and one patient with acute leukaemia) had bacillary angiomatosis confirmed by histology. B. quintana was cultured in three patients, whereas B. henselae DNA was amplified by PCR in the remaining four patients. Serum from only one patient reacted with Bartonella antigens. Amongst the 14 B. quintana and 23 B. henselae bacillary angiomatosis cases now reported in the literature, lymphadenopathies were significantly more frequent in B. henselae-infected patients, and neurological disorders of the central nervous system in B. quintana-infected patients. Risk factors were contact with cats, and homelessness or poor socioeconomic status in B. henselae and B. quintana bacillary angiomatosis cases, respectively.

Conclusions: 

Although diagnosis of bacillary angiomatosis often remains solely based upon histology, culture or PCR-based methods are useful for the detection of Bartonella sp., and allow identification of the species involved, which is necessary to further characterize clinical and epidemiological features associated with B. quintana or B. henselae infections.

Introduction

Bacillary angiomatosis (BA) is a vascular proliferative disease most often described in patients with AIDS or other causes of immunodeficiency. The disease usually manifests as cutaneous tumours, but may also involve other organs. Bacteria were initially observed on skin sections using the Warthin–Starry stain [¹]. In 1990, Relman et al. [²] used PCR incorporating broad-specificity primers to amplify bacterial 16S rRNA fragments directly from cutaneous, spleen, and bone-marrow specimens obtained from BA patients. The nucleotide base sequence of these fragments was determined and alignment demonstrated that they were closely related to the bacterium Rochalimaea quintana (now Bartonella quintana), the agent of trench fever. BA-associated bacteria were subsequently isolated [^3,4] and further characterized as a new species, Rochalimaea henselae (now Bartonella henselae) [^5,6]. B. quintana was also isolated from the blood, cutaneous tissue, and bone tissue from BA patients [³], demonstrating that both B. henselae and B. quintana are aetiological agents of BA. B. quintana is also recognized as an aetiological agent of trench fever, bacteraemia, endocarditis and chronic lymphadenopathy, whereas B. henselae has also been isolated from patients with peliosis hepatitis, relapsing fever with bacteraemia, endocarditis, encephalitis and chronic fatigue-like syndrome [^7–9]. Importantly, B. henselae is now recognized as the main aetiological agent of cat scratch disease (CSD) [^10–13].

This study reports seven BA cases in immunocompromised patients, in which diagnosis was established both by histology and demonstration of the presence of Bartonella sp. by culture or PCR-based methods. We have also reviewed 37 speciated BA cases in the literature, including our seven cases, in order to establish the most useful diagnostic tools for the disease and to identify clinical features or risk factors associated with either B. quintana or B. henselae infection.

Materials and methods

Patients

Patient 1

This case has been previously reported in part [¹⁴]. A 33-year-old, homosexual, HIV-infected man presented in March 1994 with diffuse cutaneous nodules on the back and head. His CD4 lymphocyte count was below 20 × 10⁶/l. Histology of skin biopsy specimens was compatible with BA, including a positive Warthin–Starry stain. Culture was not performed. Skin nodules resolved on administration of erythromycin (3 g daily) for 2 months. However, the patient experienced two relapses of cutaneous BA with general symptoms in July and October 1994, with resolution of the symptoms on successive administration of erythromycin (3 g daily for 2 months) and spiramycin (3 × 10⁶ U daily for 2 months). In January 1995, the cutaneous lesions reappeared, with costal pain and pain in the legs. Thoracic standard radiographs and computed tomography (CT) scan were normal, but a bone scintigraphy using technetium Tc 99m revealed the presence of diffuse uptake, with hyperfixation in two right ribs, the rachis, and the right femur. Blood cultures remained sterile. Histological examination of bone-marrow biopsy specimens was normal, including a negative Warthin–Starry stain, whereas a bone-marrow culture yielded B. quintana. Since antibiotic susceptibility testing demonstrated the strain to be highly resistant to macrolides, spiramycin was changed to oral ciprofloxacin (750 mg daily) with rifampin (900 mg daily) administered for 4 months, which resulted in rapid improvement of clinical symptoms.

The patient was readmitted in July 1995 following relapse with subcutaneous nodules, fever, chills, and pains in the legs and neck. A blood culture yielded B. quintana. A new bone scintigraphy using technetium Tc 99m revealed the presence of hyperfixation in the cervical rachis, and magnetic resonance imagery showed a lytic C4 lesion which was biopsied. Histological examination of cervical bone specimens was compatible with BA, although Warthin–Starry staining was negative. Culture of bone specimens remained sterile. The combination of clarithromycin (2 g daily) with ofloxacin (600 mg daily) was administered. The patient's condition worsened in October 1995, despite continuation of the antibiotic therapy, with the appearance of psychiatric disorders. A brain CT scan revealed the presence of a frontal mass. The patient died from neurological disorders. An autopsy was not performed.

Patient 2

This case has been previously reported in part [¹⁵]. A 42-year-old, homosexual, drug-using, homeless man presented in March 1992 with skin lesions on the right arm. He had been HIV antibody-positive since 1985. Histological examination of skin biopsy specimens was compatible with BA, including a positive Warthin–Starry stain, and the presence of bacilli on electron microscopic examination of skin sections. Culture of skin specimens remained sterile. A blood sample yielded B. quintana. Pristinamycin (2 g daily) was administered for 1 week, with resolution of cutaneous lesions. The patient was readmitted in June 1992 following relapse, with multiple cutaneous lesions on the arms and trunk, fever, weight loss, and neurological involvement with right hemiparesia and aphasia. A cerebral CT scan showed multiple central hypodense lesions compatible with neurotoxoplasmosis. Sulphadiazine-pyrimethamine therapy, and then high doses of sulphamethoxazole-trimethoprim were administered with resolution of skin lesions, but worsening of neurological symptoms. The patient died with major neurological disorders and without a final diagnosis. An autopsy was not performed.

Patient 3

This case has been previously reported in part [¹⁶]. A 53-year-old, homosexual man with AIDS presented in January 1993 with cutaneous nodules on the face and trunk. He had been diagnosed as HIV antibody-positive in 1987, when a concomitant cytomegalovirus infection was diagnosed. The patient was readmitted in April 1993 with a digestive haemorrhage resulting from peptic ulcer. His CD4 lymphocyte count was below 10 × 10⁶/l. Three new cutaneous nodules were found on his face. Cytomegalovirus chorioretinitis was diagnosed. In June 1993 multiple cutaneous nodules, up to 4 cm in diameter, were found on his left leg, right arm and shoulder. Histological examination of cutaneous specimens biopsied on the leg and arm was compatible with BA, including a positive Warthin–Starry stain. Culture of a skin biopsy yielded B. quintana. Two months of erythromycin therapy led to a reduction of skin lesions. The patient was then lost to follow-up.

Patient 4

A 26-year-old, drug-using man with AIDS was hospitalized in March 1995 with fever (39°C), multiple cervical lymphadenopathies, hepatosplenomegaly and general malaise, but without skin lesions. Histological examination of liver biopsy sections was unspecific, and a Warthin–Starry stain was negative. Culture and PCR amplification applied to a liver biopsy specimen were negative. Histological examination of biopsy sections of a cervical lymph node was compatible with BA, but a Warthin–Starry stain was negative. Specimens from the same cervical lymph node were used for culture and PCR techniques. Although culture remained sterile, B. henselae citrate synthase gene was amplified by PCR. The patient received erythromycin 3 g daily, but was lost to follow-up.

Patient 5

A 35-year-old, drug-using patient with AIDS presented in December 1993 with a cutaneous tumour on a finger of his left hand, which had persisted despite 1 month's therapy with ciprofloxacin (500 mg twice daily). Osteolytic lesions were found on radiography. A skin biopsy was not performed. The patient was read-mitted 5 months later when new skin lesions were found on his hand. These were biopsied for histological examination, culture and PCR. Histology was compatible with BA, including a positive Warthin–Starry stain. Cultures of skin biopsy specimens were contaminated with Staphylococcus aureus. B. henselae citrate synthase gene was amplified by PCR. The patient died during the second hospitalization from an undiagnosed, opportunist AIDS-related complication. An autopsy was not performed.

Patient 6

A 61-year-old patient, with acute myeloblastic leukaemia-transformed chronic myeloid leukaemia, was hospitalized in November 1994 with erythematous cutaneous lesions on the arms, with fever, cervical lymphadenopathies, and splenomegaly. His medical history included a Rhodococcus sp.-related septicaemia a few months earlier for which he had received imipenem and vancomycin therapy. Blood cultures remained sterile. Skin lesions of the arms were biopsied. Histology was compatible with BA, but Warthin–Starry staining and cultures were negative. A bone-marrow biopsy was also performed. Both histology, including Warthin–Starry staining, and culture were negative. However, B. henselae citrate synthase gene was amplified by PCR from both skin biopsy and bone-marrow specimens. The patient died a few weeks later. An autopsy was not performed.

Patient 7

A 40-year-old, drug-using man with AIDS presented in January 1996 with cutaneous nodules. He had been known to be HIV-infected since 1987 when pulmonary tuberculosis was diagnosed. At that time, a CD4 cell count was 350 × 10⁶/l. He received anti-tuberculous therapy combined with zidovudine. Zidovudine was stopped in 1990 because his CD4 lymphocyte count was stable and above 500 × 10⁶/l. The patient remained well until the end of 1995 when multiple superinfected cutaneous nodules of 0.5 cm in diameter were found on his elbows, knees, legs and chest, but without general symptoms. His CD4 lymphocyte count was 400 × 10⁶/l and HIV antigenaemia was negative. Antibiotic therapy with pristinamycin 2 g daily was administered for 4 weeks. Skin biopsies were performed in February 1996 for histological examination, culture and PCR. Histology was compatible with BA, although a Warthin–Starry stain was negative. Cultures remained sterile, but B. henselae citrate synthase gene was amplified by PCR.

Bartonella cultures

Blood samples obtained from patients 1 and 6 were collected on Hemoline (Biomerieux, Marcy l'Etoile, France) and Bactec NR860 (Becton Dickinson, Cocheysville, Maryland, USA) blood culture bottles, respectively. After 15 days incubation of blood culture bottles at 37°C, supernatants were systematically subcultured onto Columbia blood agar. The lysis-centrifugation method, using Isolator tubes, was also applied to a blood sample from patient 1. A blood sample from patient 2 and the bone-marrow sample from patient 1 were inoculated directly onto Columbia blood agar. All other biopsy specimens obtained from the seven patients (five skin biopsies, one bone-marrow sample, one liver biopsy, one cervical lymph-node biopsy, and one C4 cervical bone lesion) were crushed into brain heart infusion broth, and then inoculated onto Columbia blood agar or Polyvitex-supplemented chocolate agar (Biomerieux). All agar plates were incubated at 37°C in CO₂-enriched atmosphere until colonies became apparent, or for 45 days before they were discarded. Biopsy specimens from patients 4, 5 and 7 were also inoculated onto human endothelial cell cultures (cell line ECV 304) maintained in Eagle's minimum essential medium (10% fetal calf serum, 2 mmol/l glutamine) as previously described [³].

PCR amplification

Crude DNA extracts were prepared from blood agar-grown colonies of each isolate, using the Chelex 100R (BioRad, Richmond, California, USA) procedure as previously described [¹⁷]. DNA extracts were also prepared directly from biopsy specimens using QIA-amp tissue kit (Qiagen, Inc., Chatsworth, California, USA) according to the manufacturer's instructions. These extracts were used as templates in a previously described PCR assay, amplifying a fragment of the citrate synthase gene (CS877p and CS1258n) [¹⁸].

Identification of Bartonella sp.

The identification of Bartonella isolates was performed by molecular techniques, including determination of the base sequence [¹⁹] and restriction fragment-length polymorphism (RFLP) analysis using TaqI and AciI restriction enzymes [¹⁹] of PCR amplicons. B. quintana Fuller, B. henselae Houston-1, B. elizabethae F9251, B. vinsonii Baker, and B. bacilliformis served as controls.

Serology

A serum sample was collected from the seven patients at the time of admission to hospital. They were tested in an immunofluorescence assay (IFA) as previously described [²⁰]. Briefly, antigens were prepared from B. quintana Oklahoma strain [American Type Culture Collection (ATCC) VR-358] and B. henselae Houston-1 strain (ATCC 49882 T) propagated onto human endothelial cell monolayers (ECV 304). After four cell passages, supernatants were harvested and bacteria were purified by low speed centrifugation (700 g for 10 min); they were then resuspended in formaldehyde. These antigens were used to perform an indirect IFA.

Pathology and immunohistology

Biopsy specimens were fixed in 10% formaldehyde, paraffin-embedded and sectioned at 5 µm. They were stained either with haematoxylin and eosin, or with the Warthin–Starry silver stain. Immunohistology was performed using a previously described procedure [²¹], with the following modifications. Briefly, after deparaffinization of 5 µm thick tissue section, silane-coated slides were rehydrated in graded alcohols and incubated for 15 min at 37°C with 0.5 mg/ml pronase E in phosphate-buffered saline (Sigma, St Louis, Missouri, USA). Slides were prepared using the immunoperoxidase kit (LSAB K 680 kit, Dako, Trappes, France) following the manufacturer's recommendations. Bartonella-specific polyclonal antibodies were locally prepared by injecting B. henselae (Houston-1 strain) or B. quintana (Oklahoma strain) antigens into rabbits. These polyclonal antibodies were used (1 : 400 final dilution) for immunohistologic staining of tissue preparations. Mayer's haematoxylin was used as a counterstain. Negative controls were performed for each tissue section using normal rabbit serum.

Clinical and epidemiological analysis

We analysed the seven BA cases in our study and 30 speciated BA cases in the literature in order to investigate differences in clinical presentations or epidemiological risk factors associated with B. quintana and B. henselae-infected BA cases. The presence of fever, weight loss, skin lesions, lymphadenopathy, hepatomegaly, splenomegaly, and central nervous system involvement were available as clinical criteria. Age, sex, and the immune status were also analysed. Investigated risk factors were cat contact and homelessness or poor living status, as evidenced by the report of scabies or very low economic conditions. Means were analysed by Student's t-test and proportions were analysed by Fisher's exact test. A difference was considered significant when P < 0.05.

Results

Microbial isolation

B. quintana was isolated in three patients, including from a bone-marrow biopsy and a blood sample from patient 1, from a blood sample from patient 2, and from cutaneous specimens from patient 3 (Table 1). In patient 1, B. quintana was first isolated from the bone-marrow specimen inoculated into a blood culture bottle. Gram-negative bacilli were seen on smear examination prepared from the blood culture supernatant after 15 days of incubation. Colonies were obtained on blood agar 20 days after subculture of supernatant. B. quintana was later isolated from a blood specimen, but only when the lysis-centrifugation technique was performed. Culture of a blood sample from patient 2 yielded characteristic Bartonella-like colonies after 45 days of incubation. Culture of subcutaneous nodule biopsies from patient 3 was positive after 15 days of incubation. Cultures in axenic medium or on cell monolayers from patients 4, 5, 6 and 7 did not allow the isolation of Bartonella spp.

Table 1: . Histological and microbiological findings in the seven cases of bacillary angiomatosis (BA).

Molecular detection

B. henselae citrate synthase gene fragments were detected directly from clinical specimens on five occasions: skin biopsy specimens from patients 5, 6 and 7; a cervical lymph-node biopsy from patient 4; and a bone-marrow biopsy from patient 6.

Identification of Bartonella sp.

Using TaqI and AciI restriction enzymes, RFLP patterns specific to B. quintana were observed from positive cultures from patients 1, 2 and 3. Citrate synthase gene PCR amplification products from the three isolates were sequenced and were indistinguishable from that of the B. quintana Fuller strain. This sequence differed from that of the B. henselae Houston-1 strain in 31 base positions. Bartonella DNA obtained by direct PCR amplification of the citrate synthase gene from specimens from patients 4, 5, 6 and 7 were also characterized using the sequencing method. In all cases, a 100% sequence similarity with the published B. henselae Houston-1 strain citrate synthase gene sequence was obtained.

Serological study

Serum samples from all patients were tested for the presence of IgG and IgM antibodies directed at B. quintana or B. henselae antigens. Specific antibodies were detected only in the sample from patient 3, with IgG titres of 1 : 200 for B. quintana antigens and 1 : 25 for B. henselae antigens.

Pathological findings

The pathological data of BA patients are summarized in Table 1. Histological examination of skin biopsy sections from patients 1, 2, 3, 5, 6 and 7 showed typical aspects of BA (Fig. 1). Numerous clusters of bacteria were revealed on a Warthin–Starry stain, in patients 1, 2, 3 and 5, but not in patient 6. Histological examination of cervical bone specimens from patient 1 and lymph-node biopsy sections from patient 4 were compatible with BA, although Warthin–Starry staining was negative in both cases.

Fig. 1: . Patient 5. Cutaneous bacillary angiomatosis involving superficial and deep dermis. A thinning epidermis is seen as epithelial collarette. Lobular capillary proliferation is visible in the dermis (*). The small vascular channels are lined with epithelioid endothelial cells that protrude into vascular lumina (arrow), but without nuclear atypia. An inflammatory infiltrate with numerous neutrophils is scattered throughout the lesion (**). Haematoxylin–phloxin–saffron stain; original magnification × 250.

Immunohistology

Bacteria were also detected on skin biopsy specimens from patients 2, 3 and 5 by immunohistochemistry (Table 1; Fig. 2). In these cases, the immunodetection was revealed with both anti-B. quintana and anti-B. henselae antibodies. Bartonella spp. were apparently located within cells and were visualized as fine grains.

Fig. 2: . Patient 5. Immunohistochemical demonstration of Bartonella henselae in cutaneous bacillary angiomatosis. Bacteria are revealed by fine granular immunopositive material (arrow). Original magnification × 400.

Immunopositive cells were proliferative endothelial cells, which were localized in the upper reticular dermis. Immunohistochemistry examinations were negative in the other samples, including skin biopsies from patients 6 and 7, bone-marrow samples from patients 1 and 6, and the cervical lymph-node and liver biopsy from patient 4.

Clinical and epidemiological analysis

Retrospective analysis of clinical symptoms in the 37 speciated BA cases revealed that fever was present in most patients (23 out of 37, 62.1%), and weight loss was recorded in 13 (35.1%) out of 37 patients. Skin lesions were usually present (31 out of 37, 83.7%). Involvement of bone, liver, spleen, lymph nodes, and central nervous system was recorded in 16.2, 21.6, 32.4, 45.9 and 8.1% of the patients, respectively. The presence of lymphadenopathies was significantly (P < 0.02, Fisher's exact test) more often associated with B. henselae infection (14 out of 23, 60.8%) than with B. quintana infection (three out of 14, 21.4%). Neurological disorders were significantly (P < 0.05, Fisher's exact test) more frequently associated with B. quintana-related BA cases (three out of 14, 21.4%) than with B. henselae (none out of 23). Other clinical data did not differ significantly between B. henselae-infected BA patients and B. quintana-infected patients (Table 2).

Table 2: . Clinical features of the 37 speciated bacillary angiomatosis (BA) cases.

Epidemiological risk factor analysis (Table 3) revealed that cat exposure was specified in 25 of the 37 cases. Cat contact was recorded in 11 (84.6%) out of 13 B. henselae-infected patients, compared with three (25%) out of 12 infected with B. quintana (P < 0.01, Fisher's exact test). Homelessness and poor living conditions were recorded in only 11 BA patients. However, such epidemiological features were significantly (P < 0.05, Fisher's test) more frequently reported in B. quintana-infected BA patients (four out of six, 66.6%) than in B. henselae BA patients (none out of six).

Table 3: . Epidemiological features of the 37 speciated bacillary angiomatosis (BA) cases.

Discussion

BA has been most often reported in AIDS patients [^1,22,23], less frequently in patients with other causes of immunosuppression such as chronic lymphocytic leukaemia [²⁴], cytotoxic chemotherapy [²⁵] or in transplant recipients [^4,26,27], and exceptionally in immunocompetent patients [^28,29]. BA manifests as cutaneous tumours or as a systemic disease with various organs' involvement. Mohle-Boetani et al. [³⁰] have proposed that BA be considered as an AIDS opportunistic infection, with three main clinical presentations that may be combined, including cutaneous and subcutaneous lesions, chronic lymphadenopathy, and fever with abdominal symptoms. Although BA and peliosis hepatitis may represent different clinical manifestations of a same disease, peliosis hepatitis lesions correspond to blood-filled spaces within the liver or spleen parenchyma without endothelial cell proliferation, suggesting a different pathogenesis for both diseases [³¹]. Peliosis hepatitis was not considered in the present study.

All the seven BA cases presented in our study were immunocompromised, including six HIV-infected patients and one patient with leukaemia. In six patients, skin angiomatous lesions were the conditions that came to medical attention, whereas patient 4 was hospitalized because of cervical lymphadenopathies with hepatosplenomegaly but without skin involvement. Extracutaneous BA was present in patients 1 and 6, as shown by histology of the C4 cervical bone lesion and culture of B. quintana from a bone-marrow biopsy in patient 1, and by amplification of B. henselae DNA from the bone marrow of patient 6. Amongst the 37 speciated BA cases in the literature, fever with skin lesions were the most frequent clinical presentation of the disease. However, statistical analysis of the clinical presentation of either B. henselae or B. quintana-infected BA patients revealed that lymphadenopathies were more frequent in the former, confirming a recent investigation by Koehler et al. [³²]. Interestingly, chronic lymphadenopathies are the primary clinical manifestation of CSD, which corresponds to inoculation of B. henselae via cat scratches. Although chronic lymphadenopathies have occasionally been reported in B. quintana-infected people [^33,34], this species has never been isolated from cats and is not considered aetiological agent of CSD. The role of cats in the transmission of B. henselae-related diseases, including BA, may explain in part why lymphadenopathies are more frequently observed in BA patients infected with this species. Neurological disorders were more often statistically associated with B. quintana BA cases than with B. henselae cases. Neurological manifestations associated with Bartonella infection include aseptic meningitis [³⁵], meningoencephalitis [³⁶], neuroretinitis [³⁵], brain tumours [³⁷], and potentially encephalopathy with dementia in HIV-infected people [^38,39]. Bartonella-related neurological manifestations are not currently considered species-specific. Although the data presented in this study are preliminary, attention should be paid to the apparent specificity of B. quintana in future studies.

Before elucidation of the aetiology of BA, diagnosis of the disease was established by histological examination of skin or other organs biopsy sections, showing the presence of angioproliferative lesions with eosinophilic material revealed as masses of bacteria by Warthin–Starry staining [^{4,24,27,29,39,40–49}]. Bacteria can now be visualized in biopsy specimens by direct IFA or immunohistochemistry using anti-Bartonella sp. immune serums. However, characteristic histological findings may be occasionally absent, especially when antibiotics have been previously administered. In our study, all patients presented histological findings compatible with BA in at least one biopsy specimen. Warthin–Starry staining was positive in four of the seven positive specimens, whereas the immunohistochemistry technique was positive in three skin specimens. Although immunohistochemistry is more specific than Warthin–Starry staining, the sensitivity of both techniques was found to be similar. The fact that the bacteria were visualized using either anti-B. quintana or anti-B. henselae antibodies confirms the existence of serological cross-reactions between both species.

Culture of Bartonella spp. in axenic medium or on cell monolayers remains fastidious and insensitive, and very few isolates are available worldwide. In the present study, specific cultures allowed the isolation of B. quintana in three patients. However, a 15–45-day incubation time was required for primary isolation, and an even longer time was required for definite species identification. Although, B. henselae DNA could be amplified by PCR from clinical specimens from the remaining four patients, specific cultures remained negative. It is interesting to note that amongst the 37 speciated BA cases reported in the literature, a specific diagnosis was established by culture in 15 patients, including in 12 (85.7%) of the 14 B. quintana-infected BA cases, but in only three (13%) of the 23 B. henselae-infected BA cases, suggesting that culture of the later species is more fastidious. Serological techniques have recently been evaluated in CSD patients, with sensitivities ranging from 50% [⁵⁰] to 95% [⁵¹]. On the other hand, cross-reactions have been reported between Bartonella sp. and Coxiella burnetii or Chlamydia sp. [^52,53]. In our experience, however, the use of serology to diagnose BA is mostly limited because the majority of immunocompromised patients do not mount an antibody response. In our series, only a serum sample from patient 3 was positive by IFA, confirming the low contribution of serology to diagnose BA. Our results confirm the difficulty of obtaining a specific diagnosis in Bartonella-infected BA patients. Culture and PCR-based methods to detect Bartonella sp. in biopsy materials, although more invasive than serology, are more suitable in immunocompromised BA patients, and allow accurate identification of the Bartonella sp. involved.

Risk factors associated with B. quintana or B. henselae infections still need to be firmly established, but are considered to be dramatically different. Cats may represent the primary reservoir of B. henselae, which may be transmitted to humans by cat scratches or bites, or potentially by the bite of cat fleas (Xenopsilla cheopis). A strong association has been found between cat scratches and cat bites and both CSD and BA [^{4,28,32,54–59}]. In contrast, no animal reservoir has been shown for B. quintana. The association between B. quintana infection and homelessness, poor socioeconomic status, and louse infestation was first demonstrated in patients suffering from trench fever, and has been confirmed in modern cases of B. quintana infection, including urban trench fever, bacteraemia, endocarditis [^32,60–62], and more recently in patients with BA or peliosis hepatitis [³²]. Our results confirm that B. henselae BA cases are significantly more frequently associated with cat exposure, whereas those resulting from B. quintana are significantly more frequently associated with homelessness and poor living conditions. The precise definition of epidemiological data associated with both species is of primary interest, and will only be possible if accurate identification of Bartonella at the species level is obtained.

Because most patients can be cured with appropriate antibiotic therapy, recognition of BA is of critical importance. Erythromycin (500 mg four times daily) remains the first-line antibiotic therapy for BA [¹]. Tetracyclines are currently proposed as the first alternative. Other antibiotics display in vitro bacteriostatic activity [^63,64], including β-lactam compounds, co-trimoxazole, rifampin, ciprofloxacin and aminoglycosides, but in vitro susceptibility data do not accurately predict the in vivo situation. Failures have recently been reported with β-lactam compounds [⁶⁵], and Koehler et al. [³²] reported the isolation of B. henselae in two patients taking prophylactic oral co-trimoxazole. In the present study, patients 3 and 7 were cured with erythromycin therapy administered for 8 and 4 weeks, respectively. In contrast, relapses were observed in patient 1 despite prolonged antibiotic therapy successively using erythromycin, ciprofloxacin with rifampin, and clarithromycin with ofloxacin, and in patient 5 despite 4 weeks of ciprofloxacin therapy. A 2-month antibiotic therapy is a minimum in the immunocompromised host [³], but relapses may still occur on antibiotic withdrawal. The recent demonstration that only aminoglycosides display in vitro bactericidal activity against Bartonella sp., both in axenic medium and in an eukaryotic cell system [⁶⁶], should prompt clinical investigations in BA-infected immunocompromised patients.

The above epidemiological considerations suggest that immunocompromised patients, especially those infected with HIV, should avoid cat contact and control flea infestation in cats. Delousing procedures recommended for trench fever may also be useful in the prevention of B. quintana-related BA cases. Permethrin dusting powder (1%) is the drug of choice, applied in a dose of 30–50 g per adult. Clothing should be dusted inside and out (125–205 mg/m² of clothing) and bedding should also be treated.

Acknowledgements

The authors thank Drs P. Nordmann, L. Lebrun, J.P. Brion, O. Lortholary, G. Marx, and D. Fasquelle for their collaboration in the clinical studies of the patients. The authors thank Dr R. Birtles for reviewing the manuscript and Dr H. Tissot Dupont for his help with statistical analysis.