INTRODUCTION

Top Previous Next References

Infective endocarditis (IE) is usually suspected in a patient with fever and a new or changing cardiac murmur and is diagnosed based on the presence of a vegetation on echocardiography and positive blood cultures. Diagnosis of endocarditis is usually easy in febrile patients with a continuous bacteremia and the presence of vegetation on echocardiography or on gross examination or histologic testing of the removed valve. However, although numerous clinical situations lead to a high degree of suspicion of endocarditis, culture or histologic examination does not confirm the diagnosis.

Although fever is the single most common finding in endocarditis, it may be absent in the elderly or in patients given previous antibiotic therapy before presentation or it may be intermittent or low-grade as for Q fever endocarditis 29. Cardiac murmur is the second most frequent finding in endocarditis. However, it is not usually present at the initial stage of right-sided endocarditis, and new or changing murmurs are detected in only 40% of patients with endocarditis 261; this rate is even lower in the elderly. Despite the fact that transesophageal echocardiography is more sensitive than transthoracic echocardiography, a vegetation is rarely detected in Q fever or Whipple's disease 29, 221. Sterile blood cultures have been noted for 2.5 to 31% of patients with endocarditis 286. Blood cultures are frequently sterile when antibiotic therapy was administered before sampling and in patients with subacute right-sided endocarditis, mural endocarditis, and endocarditis caused by slow-growing or fastidious organisms such as anaerobes, the HACEK group, Abiotrophia spp., Brucella spp., Bartonella spp., Legionella spp., and Mycoplasma spp. or when obligate intracellular organisms such as Coxiella burnetii are involved 3, 124. These bacteria require specific media and conditions such as L-cysteine-enriched medium for Abiotrophia spp., buffered charcoal yeast extract (BCYE) agar for Legionella spp., or special culture conditions favorable for anaerobes or intracellular bacteria. Moreover, in some cases, slow-growing bacteria require incubation times as long as 6 weeks 179. In such situations, infective endocarditis remains a diagnostic challenge.

To both assist physicians in establishing the final diagnosis of endocarditis and allow comparisons of published cases, diagnostic criteria have been defined 70, 122, 123, 290. For many years, the Beth Israel criteria 290 were the only recognized diagnostic criteria. In 1994, Durack et al. from the Duke Endocarditis Service 70 added echocardiographic findings and other clinical and laboratory data to the well-established clinical and microbiological criteria. More recently, additional criteria have been proposed to improve the list 147, such as including C. burnetii serology or culture as additional major criteria 86.

According to the Duke Endocarditis Service, the diagnosis of IE is definite (i) when a microorganism is demonstrated by culture or histologic testing in a vegetation, an embolism, or an intracardiac abscess; (ii) when active endocarditis is confirmed by histologic examination of the vegetation or intracardiac abscess; or (iii) in the presence of two major clinical criteria, one major and three minor criteria, or five minor criteria (the major and minor Duke criteria are listed in Table 1). The diagnosis of IE is rejected when a firm alternate diagnosis explains the manifestations of endocarditis, when the fever resolves with antibiotic therapy for 4 days or less, or when no pathologic evidence of infective endocarditis is found at surgery or autopsy after antibiotic therapy for 4 days or less 70.

Consequently, suspected cases of endocarditis should be investigated until all clinical, epidemiologic, echocardiographic, and laboratory data are compiled in order to be able to establish the diagnostic score.

New etiologic agents of endocarditis such as Bartonella spp. and the Whipple's disease bacillus 233 have been identified by PCR amplification and sequence analysis of the amplified DNA. These molecular tools have considerably improved the detection and the identification of noncultivable causative agents of endocarditis 88. This technique is particularly useful for microbial identification in excised infected valves, even after antibiotic therapy. PCR amplification and subsequent sequence analysis of amplified genes has also been used to better identify fastidious organisms such as Gemella spp. 152. New cultivation techniques, especially tissue cell culture, have also been successfully used for isolation of fastidious organism such as Bartonella 65 or strict intracellular pathogens such as C. burnetii and Chlamydia spp. 229, 253. Tissue cell culture has been greatly improved by using the shell vial technique 229. Culture of the removed valve has been shown to be much more efficient than blood culture in recovering the organisms in patients with culture-negative endocarditis 192, 193. The use of enriched media for specific microorganisms such as nutritionally deficient Streptococcus spp., Legionella spp., or Mycobacteria spp. has also been helpful in the etiologic diagnosis of endocarditis. The availability of these improved and new diagnostic tools has considerably enlarged the spectrum of bacteria involved in the etiologic diagnosis of endocarditis. Herein, we review the epidemiology, clinical presentation, and diagnosis of endocarditis caused by rare and fastidious organisms and then focus on the strategies used for the recovery and identification of the etiologic agents of IE. This review has been based on a MEDLINE search from 1966 to 1999 using Internet Grateful Med search V2.6.2 from the National Library of Medicine with the following key words: negative blood culture or culture-negative plus endocarditis or endocarditis plus (name of the bacteria and/or the disease). When a bacterium had different names, such as Abiotrophia spp., all known synonyms were used. The research was carried out without restriction, except that endocarditis or the name of the bacteria should be included in the title. Only papers written in English or French were selected. This research yielded 1,056 references, from which 304 have been selected.

Once the diagnosis of IE is suggested or confirmed, the causative agent can be specifically investigated by evaluation of the medical history, clinical examination, and echographic findings (Fig. 1).

View larger version (49K): [in this window] [in a new window]   FIG. 1.   Probable bacterial etiology of endocarditis depending on the patient history and epidemiologic situation. IVDU/IVDA, intravenous drug user/abusers.

The patient's medical history is an important clue for the etiological diagnosis of IE. Contact with animals is an important feature; for example, contact with a cat (including scratches or bites) suggests Bartonella henselae or Pasteurella spp., while contact with cattle should suggest a role for C. burnetii or Brucella spp. Parrots, pheasants, and pigeons have been reported to play a role in Chlamydia psittaci IE 253. Contact with domestic animals such as swine, fish, and poultry is frequently reported in the history of patients with Erysipelothrix endocarditis. Consequently, occupational exposure is an important finding. For example, farmers and veterinarians are exposed to both C. burnettii and Erysipelothrix. The latter is a causative agent of endocarditis in butchers, fishermen, and even in homemakers by contact with organic matter in which the organism is commonly found 230. Contact with arthropods such as lice and fleas should also be considered. The human body louse is involved in the transmission of Bartonella quintana 27, whereas cat fleas are involved in the epidemiology of B. henselae infections 143. Some pathogens are mostly nosocomial in origin. For example, IE following open heart surgery should suggest Legionella spp., Mycobacterium spp., Nocardia spp., or fastidious fungi, especially if routine blood cultures are negative 160, 281. Most reported cases of nosocomial endocarditis occur as small outbreaks, and all available epidemiologic information should be obtained for a diagnosis of postoperative IE. Since a large number of pathogens involved in IE are of oral origin, a history of a dental procedure is an important feature. HACEK group bacteria are often found to be pathogens in young adults after dental procedures, but Lactobacillus spp. and Gemella spp. have also been reported in such situations 152, 270. All information on underlying conditions should be obtained from the patient or from the patient's medical history. Patients with prosthetic valves are at particular risk for IE. Of note, patients with prosthetic valves for whom blood cultures are negative are likely to be infected by C. burnetii or B. henselae as etiologic agents, since more than 55% of C. burnetii endocarditis occurs in patients with prosthetic valves 266. Underlying immunocompromising conditions, including AIDS, cancer, and lymphoma, have been reported in 9% of C. burnetii endocarditis patients, especially in the absence of underlying valvular disease. A medical history of cancer, especially adenocarcinoma associated with disseminated intravascular coagulation, should also suggest the possibility of marantic or nonthrombotic endocarditis, a diagnosis of exclusion when all other investigations are nonrevealing 161. In a similar manner, IE can be suspected in patients with collagen vascular diseases such as systemic lupus erythematosus, particularly if associated with the presence of antiphopholipid antibodies such as in Libman-Sack endocarditis 238. Even if noninfective endocarditis is suspected, it is important to remember that the valvular lesions induced by inflammatory disease may facilitate the infection of the inflamed valve or graft and cause secondary IE. Patients should be questioned about their life habits, especially their sexual practices. A few decades ago, Neisseria gonorrhoeae IE was reported very frequently in sexually active patients; active surveillance and prompt treatment has now significantly reduced this rate. In a similar way, behaviors such as intravenous drug use and alcohol abuse should be evaluated. Intravenous drug users are at risk for Neisseria sicca, Clostridium spp., and Eikenella corrodens IE related to injection 8, 56, 118. E. corrodens has been reported in patients who clean needles with their saliva. For chronic alcoholics, the involved pathogens are more likely to be related to the underlying immunocompromising condition rather than to specific practices. B. quintana, Erysipelothrix rhusiopathiae, Campylobacter fetus, Aeromonas hydrophila, Pasteurella spp., and Listeria monocytogenes are most commonly found 38, 65, 77, 104. Portal hypertension and splenic shunt, which occurs with cirrhosis in these patients, may allow prolonged bacteremia and subsequent valvular infection. Social conditions may impart specific risks. For example, homeless persons are often chronic alcoholics and may harbor lice, allowing the transmission of B. quintana infection, the major fastidious etiologic agent of IE in the homeless 27. Finally, the relative frequency of the microorganism as a cause of IE should be considered in the laboratory investigation. For example, it is noteworthy that C. burnetii is as frequently recognized as an etiologic agent of IE as are all the HACEK group bacteria together (Table 2).

Some clinical presentations are characteristic and should indicate possible IE. Febrile congestive heart failure is a frequent mode of presentation for patients with endocarditis caused by fastidious organisms, especially C. burnetii and Bartonella spp., but may also be present with Whipple's disease or HACEK group endocarditis. In fact, it is likely that this mode of presentation is related to a diagnostic delay. Emboli are reported in more than 50% of patients infected with microorganisms such as with Corynebacterium diphtheriae, N. sicca, and Haemophilus paraphrophilus 46, 278; R. Lopez-Velez, J. Fortun, C. de Pablo, and J. Martinez Beltran, Letter, Clin. Infect. Dis. 18:660-661, 1994). When these emboli involve the brain, they may mimic an acute cerebrovascular accident or stroke syndrome, but fever should suggest the diagnosis of endocarditis. Pulmonary manifestations such as repetitive bronchopneumonia are common in right-sided endocarditis. Such findings in intravenous drug users or in patients with intravenous catheters should suggest a diagnosis of right-sided endocarditis. Staphyloccocus spp., Candida spp., Eikenella corrodens, and Salmonella enterica serovar enteritidis in patients with AIDS and N. sicca and Clostridium spp. in other patients are potential etiologic agents. It is important to note that the absence of cardiac murmur does not rule out the diagnosis of IE and can suggest ride-side endocarditis. Genital organs should be examined, since most cases of gonoccocal endocarditis are associated with active sexually transmitted diseases. Concomitant gastroenteritis, sinusitis, or pneumonitis should suggest Salmonella spp., H. influenzae, or Chlamydia psittaci as etiologic agents, respectively, while upper respiratory tract infection and pharyngitis is frequently observed in patients with H. aphrophilus endocarditis and in intravenous drug users with C. diphtheriae endocarditis. In patients with Klebsiella pneumoniae endocarditis, urinary tract infection is common 9, and during pregnancy, Listeria monocytogenes is a possible etiologic agent. Finally, culture-negative endocarditis associated with chronic diarrhea should suggest Whipple's disease.

Over the past 15 years, a number of investigations have confirmed the important role of transthoracic echocardiography in the diagnosis and management of IE. Echocardiography must be performed in all patients suspected of having IE. If transthoracic echocardiography is noncontributory but there is still a high suspicion of IE, transesophageal echocardiography should be performed. Compared to transthoracic echocardiography, transesophageal echocardiography has improved the sensitivity in defining both vegetative lesions and perivalvular infections, particularly in mural abscesses in patients with IE (sensitivity, 80 to 90%) 237, 247. Except in a few cases where vegetations are typically large, such as in Brucella endocarditis, echocardiographic findings are more effective in establishing a diagnosis of IE than in determining the microbial etiology. Recently, the diagnostic value of echocardiography was confirmed to improve the sensitivity of the Duke's criteria 111. Transesophageal echocardiography is particularly important in patients with negative blood cultures, but even under these conditions, 24% of patients with proven IE were misclassified as having "possible" IE 111.

STRATEGIES FOR IDENTIFICATION OF THE ETIOLOGIC AGENT OF INFECTIVE ENDOCARDITIS

Top Previous Next References

Culture Methods

Blood cultures. Quantitative culture techniques show that blood from patients with IE contain 1 to 10 bacteria per ml and that this quantity remains constant during the course of the disease 298. In 1947, Salazar Mallen et al. reported that 19 of 24 patients with a positive culture from at least one site had positive venous blood cultures, none with negative blood cultures had a positive arterial culture, and 5 with both negative venous and arterial cultures had positive bone marrow cultures 245. From these studies it appears that arterial cultures add nothing to the diagnosis of IE and that in rare cases with persistently negative venous blood cultures, bone marrow cultures may yield diagnostic information. Moreover, in a series of 82 patients with IE, 52 (63%) were diagnosed from the first culture, 78% from two blood cultures, 82% from three, and 91% from four 36. Most patients with culture-positive endocarditis are persistently culture positive even if only one, two, or three cultures are obtained 296. Culture-negative endocarditis is best defined when negative blood cultures (more that three venous blood culture of at least 5 ml each) are obtained repetitively. Therefore, the data support the concept that cultures (anaerobic and aerobic) of three sets of blood drawn within a 24- to 48-h period are sufficient to establish a diagnosis of culture-positive endocarditis 36. Because of the approximately linear relationship between the yield of bacteria from blood and the volume of blood, some authors recommend that at least 10 ml of blood be obtained for each culture and that as much as 30 ml of blood be obtained for each culture in an adult 184, 296.

View larger version (99K): [in this window] [in a new window]   FIG. 2.   Abiotrophia spp. (formerly known as nutritionally deficient streptococci) showing satellite growth with Staphylococcus aureus. Photo courtesy of B. La Scola (Unité des Rickettsies, Marseilles, France).

Culture of resected valves or biopsy specimens. Pathogens can also be isolated from resected valves or biopsy specimens by inoculation onto agar or into tissue culture. This is particularly helpful and efficient, probably because of the large number of organisms in the tissue. C. burnetii, Mycobacterium spp., Brucella spp., Legionella spp., and numerous other fastidious bacteria have been identified as causative agents of endocarditis in this manner 159, 192, 225, 281. Thus, surgically resected materials should be cultured in appropriate medium when possible. In some instances, positive serologic test results for agents such as C. burnetii, Legionella spp., or Brucella spp. may be helpful in delineating the appropriate media to be employed.

Tissue cell culture. Tissue cell culture is the only method for isolation of obligate intracellular pathogens. This procedure may be hazardous and should be attempted only in laboratories equipped for biosafety level 3 (BSL3) pathogens. The shell vial technique has been reported as the best method for isolation of C. burnetii from heparinized blood or resected heart valves 29, and it has also been successfully used for isolation of Bartonella spp. and Chlamydia psittaci isolation 63, 65, 253. Basically, different cell lines, chosen depending on the potential bacterial etiology, e.g., L929, Vero, and HEL cells for C. burnetii, L929 cells for Chlamydia psittaci, and ECV cells for Bartonella spp., are grown on coverslips in shell vials until a confluent monolayer is obtained. An inoculum of 200 µl of leukocyte-rich plasma from heparinized blood or a portion of the removed, infected valve is inoculated onto the tissue culture cells and centrifuged at 700 × g for 1 h, after which the remaining plasma is removed and replaced with fresh culture medium and the culture is incubated at 37°C in a CO₂ incubator. On days 3, 6, and 15, coverslips are removed and growth is detected by indirect immunofluorescence either with antibodies that react with the bacteria suspected or with a 1:200 dilution of the patient's serum. When growth is observed, bacteria are subcultured and identified.

Histologic examination of the removed valve is critical. Standard stains such as hematoxylin and eosin can reveal whether the histologic findings from the valve are compatible with the diagnosis of IE. If no inflammation is observed in the valve, the diagnosis of IE should be reevaluated or excluded. Histologic parameters are included among the Duke diagnostic criteria for IE. Hematoxylin and eosin staining, the first step in the histologic examination, allows recognition of a consistent pattern of inflammation. Numerous special histologic stains are available, several of which are particularly appropriate for IE. Tissue Gram stains allow the differentiation of gram-positive and gram-negative microorganisms and, in the hands of a skilled observer, allow a rapid preliminary identification of the organisms based on their morphology 301. However, this technique has limitations since the causative bacteria may lack a cell wall (Mycoplasma spp.). The periodic acid-Schiff (PAS) stain is especially valuable for Whipple's disease and demonstrates PAS-positive foamy histiocytes variably surrounded by infiltrates of mixed neutrophils, lymphocytes, and mononuclear cells (Fig. 3) 301. This stain may also be used to detect fungi. The Giemsa stain not only allows the detection of bacteria, including Bartonella spp., but also stains white blood cells and therefore highlights the inflammatory infiltrate in vegetations 301. The Warthin-Starry technique, a silver impregnation method, is among the most sensitive methods for detection of bacteria, even those which stain weakly with a tissue Gram stain, such as Bartonella spp. (Fig. 4). Using this method, we have demonstrated massive vegetations on the valve surface with extensive destruction of the underlying valve tissue and the presence of numerous bacteria in Bartonella endocarditis 225. The Ziehl-Nielsen stain is used for detection of acid-fast bacteria, especially Mycobacterium spp. The Gimenez stain is a good method for the detection of C. burnetii and Legionella spp. In Q fever endocarditis, the vegetations, which are often smooth and nodular, are frequently infiltrated with mononuclear inflammatory cells containing many C. burnetii cells 25. Another stain of interest is the Grocott-Gomori methenamine silver stain, which provides the best contrast of the special fungal stains 301.

View larger version (102K): [in this window] [in a new window]   FIG. 3.   Whipple's disease endocarditis showing foamy PAS-positive macrophages (arrow). Magnification, ×1,000. Photo courtesy of H. Lepidi (Unité des Rickettsies, Marseilles, France).

View larger version (143K): [in this window] [in a new window]   FIG. 4.   Warthin-Starry silver stain of a Bartonella quintana-infected valve. Note the cluster of bacteria in black within the vegetation. Magnification, ×100. Photo courtesy of H. Lepidi.

Other stains are occasionally used. The acridine orange stain, a stain used in the nonspecific-fluorescence diagnostic technique, may also be used to detect any living microorganism in biopsy specimens. The Kinyoun stain or the Machiavello stain may be used to demonstrate large macrophages containing dark red granules in Chlamydia endocarditis 301.

Immunohistologic methods have been used for the detection of C. burnetii, Bartonella spp., and Chlamydia spp. in valvular tissues 25, 75. Several techniques are available, including a capture enzyme-linked immunosorbent/immunofluorescent assay (ELISA/ELIFA) system, direct immunofluorescence using fluorescein-conjugated monoclonal antibodies, and immunoperoxidase staining (Fig. 5) 25.

View larger version (149K): [in this window] [in a new window]   FIG. 5.   Immunohistochemical demonstration of Coxiella burnetii in a heart valve of a patient with Q fever endocarditis. Magnification, ×600. Immumoalkaline phophatase staining was used.

The advent of DNA amplification methods for detecting microorganisms has raised great interest among infectious-disease physicians and clinical microbiologists. Such techniques, particularly PCR, have now been adapted for use with most types of clinical materials, including valvular specimens. For example, detection of a portion of the 16S rRNA gene has been used successful in the identification of the Whipple's disease bacterium as an etiologic agent of endocarditis (297; K. H. Wilson, R. B. Blitchington, R. Frothingham, and J. A. Wilson, Letter, N. Engl. J. Med. 328:62, 1993). This is also true for some cases of Bartonella endocarditis 225. The main advantages of these methods are that they are culture independent and that almost all bacteria can be detected in a single reaction through the incorporation of broad-spectrum primers. The organisms from which amplified DNA is derived can easily be identified by sequence analysis followed by comparison of the nucleotide sequence with a large database that contains sequences of the targeted gene from many other bacteria. Goldenberger et al. amplified bacterial DNA from 15 valvular samples and concluded that this technique was both easy and reliable when applied to surgically removed heart valves from patients with IE 103. However, a limitation of this system is the number and quality of DNA sequences available in databases. For instance, some of the reference sequences in the GenBank and the EMBL databases are too short or contain too many undetermined nucleotides for confident assignment of clinically derived sequences. Moreover, another likely limitation of molecular detection is the quality and nature of the infectious material collected (vegetation or piece of valve). Some clinical materials, especially blood, may contain substances which limit the sensitivity of PCR. Contamination may also occur, and therefore caution must be exercised in the interpretation of PCR-based sequence analyses when the organism has not been observed in stained valve tissues. Currently, specific primers are available for most bacterial genera, including Chlamydia, Brucella, Legionella, Mycobacterium, and Mycoplasma 87.

In contrast to its use directly on clinical samples, PCR with subsequent sequencing of the amplified gene can be used for identification of bacteria isolated by culture methods. This is particularly useful when the bacteria demonstrate unusual biochemical or staining characteristics. For example, molecular identification has successfully identified Gemella spp. in three cases of endocarditis that were reported to be due to small "gram-negative coccobacilli" 152. Such techniques for detection and molecular characterization should be used each time biochemical identification test fail.

In our laboratory, testing for culture-negative endocarditis includes the systematic determination of antibody titers against C. burnetii, Bartonella spp., Mycoplasma pneumoniae, Legionella pneumophila, Chlamydia spp., and Brucella melitensis, which are among the most common pathogens in culture-negative IE. These serologic tests are included as diagnostic criteria for IE in the Duke 70 and the modified Duke 86 criteria. However, in other instances, one must be cautious when relying solely on serologic methods to diagnose endocarditis because of cross-reactions. For example, currently available serologic tests for Bartonella-induced endocarditis may not reliably distinguish between antibody responses to B. henselae and B. quintana or between those to Bartonella spp. and Chlamydia spp. 176, 225. The most common serologic methods used are the tube agglutination test for Brucella melitensis infections, indirect immunofluorescence for Legionella pneumophila, ELISA for Mycoplasma pneumoniae, and complement fixation, ELISA, or indirect immunofluorescence for Chlamydia spp.

ETIOLOGY OF ENDOCARDITIS DUE TO LESS COMMON FASTIDIOUS BACTERIAL AGENTS

Top Previous Next References

Agents of Blood Culture-Negative Endocarditis

The agents of blood culture-negative endocarditis are listed in Table 3 and discussed below.

Bartonella spp. Bartonella spp. are small facultative intracellular bacteria that stain gram negative and belong to the 2 division of the proteobacteria (Fig. 6). These organisms cause various clinical syndromes in immunocompetent and immunocompromised hosts. B. henselae causes cat scratch disease, meningoencephalitis, and prolonged fever in immunocompetent patients and bacillary angiomatosis and hepatic peliosis in human immunodeficiency virus HIV-infected patients 179. B. quintana causes trench fever, lymphadenopathy with fever, and bacillary angiomatosis. Endocarditis is caused by both B. henselae and B. quintana, and it has been found to be caused by B. elizabethae 51 and B. vinsonii 239 in only one reported case each.

View larger version (18K): [in this window] [in a new window]   FIG. 6.   Phylogeny of the representative members of the 2 group of proteobacteria and location of Bartonella spp. based on analysis of the 16S rRNA sequence by the neighbor-joining method.

Mycobacterium spp. Mycobacteria are acid-alcohol-resistant bacteria that are visualized by the Ziehl-Nelsen stain. Tuberculous valvular endocarditis is exceptionally rare; only 16 cases have been reported in the literature 47. It usually presents in the context of miliary tuberculosis, and in all but one case the diagnoses were made at autopsy. The rare cases reported involved congenital defects of the aortic valves, prosthetic valves, and ventriculoatrial shunts. However, because of its rarity, there is still uncertainty whether true tuberculous endocarditis exists as a clinical entity 47.

Mycoplasma spp. Despite the fact that endocarditis due to cell wall-deficient bacteria has been suggested since 1978 215, endocarditis due to Mycoplasma spp. has been reported in the literature only twice 45, 217. The first report described a patient with postrheumatic heart disease; the diagnosis was based on serologic test results alone, and no culture was reported. He was treated with 12 × 10⁶ U of benzylpenicillin and 280 mg of gentamicin daily for 6 weeks and then given a 4-week course of oxytetracycline 217. The other case was reported in 1989 45 and involved a 25-year-old woman with a history of SLE, treated with corticosteroids, in whom Streptoccocus sanguis aortic and mitral valve endocarditis developed. Aortic and mitral prosthetic valve replacements were performed for acute cardiac failure. During penicillin treatment, she developed a sternal wound infection and fever. Swabs of the sternal wound and blood cultures were sterile. The prosthetic valves were replaced owing to significant valve dysfunction and renewed cardiac failure. At surgery there was no evidence of sternal or mediastinal infection, but extensive dehiscence of the aortic valve suture line was noted with multiple small abscesses and fistulous tracts extending from the aortic annulus. The mitral valves appeared grossly infected. Mycoplasma hominis was isolated from a portion of the mitral annulus. The presence of sodium polyanetholsulfonate in the blood culture media might be responsible for the lack of M. hominis growth 55. This patient was cured with clindamycin and rifampin administered for 6 weeks followed by a 4-week course of oral doxycycline 45. The role of Mycoplasma spp. in culture-negative endocarditis is unknown and very probably underestimated. This small organism cannot be detected by Gram stain, and growth is not detected in routine blood culture systems. Likewise, Mycoplasma spp. can be difficult to detect in paraffin-embedded valve tissues when using nonspecific staining techniques. Strategies for isolation of this pathogen in patients with culture-negative endocarditis should include systematic subculture of blood and excised valves. Inoculation should be made onto specific Mycoplasma agar media such as SP 4 glucose (pH 7.5), which can be used for both M. pneumoniae and M. hominis, provided that arginine is added for the latter organism. Mycoplasma serologic testing should be evaluated. PCR techniques will be probably very helpful in detecting Mycoplasma spp. in blood or in infected valves.

Whipple's disease bacterium. Whipple's disease is a rare systemic disorder presenting with arthralgia, diarrhea, abdominal pain, generalized lymphadenopathy, progressive weight loss, and central nervous system involvement. This disease is caused by a gram-positive bacterium that has been identified by sequence analysis of the 16S rRNA gene 234. The name Tropheryma whippelii has been suggested for cases when the bacteria have not successfully been established in culture 234. Recently, the Whipple's disease bacterium was propagated in an interleukin-4-treated macrophage cell culture but showed only limited propagation 251. In 2000, we reported the successful isolation and establishment of a strain of Whipple's disease bacterium obtained from the valve of a patient with blood culture-negative endocarditis 222. Up to one-third of patients with Whipple's disease present with concomitant cardiac symptoms, mostly attributed to endocarditis, myocarditis, pericarditis, or coronary lesions (125, 181, 188; M. Jeserich, C. Ihling, and C. Holubarsch, Letter, Ann. Intern. Med. 126:920, 1997). More recently, the relationship between endocarditis and the Whipple's disease bacillus was established in a patient who presented with diarrhea, fever, grand mal seizure, and a history of a 20-kg weight loss over 4 years 297. The diagnosis was based upon the presence of numerous PAS-positive macrophages in the duodenal biopsy specimen. After successful antibiotic therapy, he developed acute congestive heart failure due to aortic valve endocarditis, requiring urgent prosthetic valve replacement. Histologic examination revealed PAS-positive macrophages scattered throughout the entire valve (Fig. 3). DNA prepared from paraffin-embedded valve tissues identified T. whippelii by PCR and sequencing of the PCR product. Interestingly, PAS-positive granules have been demonstrated in the brains of patients with and without cerebral manifestation of the disease 84, and rheumatoid factor is found in 8% of these patients. It is likely that Whipple's disease endocarditis is much more frequent than currently reported. Recently, four cases were reported in patients without fever or evidence of systemic Whipple's disease 110. Before pathologic examination of the resected valves, which established the diagnosis, one patient fulfilled none of the Duke criteria, two patients fulfilled one major Duke criteria, and one patient fulfilled two Duke criteria (one major and one minor). Without examination of the excised valves, the diagnosis of IE could not have been made, indicating that Duke criteria for endocarditis are not pertinent for Whipple's disease endocarditis 110, 221. This underlined the fact that occurrence of the disease is probably underestimated and must depend on an active search for the cause by physician. In a recent, not yet published review on Whipple's disease endocarditis, we found that data for 34 patients had already been published with enough details to include the valvular lesion as a part of the disease and that those patients presented with aortic damage (14 of 34), mitral damage (5 of 34), tricuspid valve damage (2 of 34), mitral and aortic valve damage (5 of 34), mitral and tricuspid valve damage (3 of 34), and damage to three valves (5 of 34). The diagnosis is currently based on the histologic observation of PAS-positive macrophages in the surgically resected valve tissues and by PCR identification from either the valve or a duodenal biopsy specimen. Cultures on axenic media, using chocolate or Columbia sheep blood agar incubated at 32 or 37°C under 5% CO₂, or in a microaerophilic or anaerobic atmosphere were unsuccessful. Similarly, cultures on cell culture medium and cell culture medium with lysates of HEL cells incubated at 32 or 37°C under 5% CO₂ were not able to recover the bacilli. However, Whipple's bacilli can be cultured using human fibroblast cells lines (HEL and MRC5) which are grown in minimal essential medium with 10% fetal calf serum and 2 mM L-glutamine without antibiotics; a control strain is available for the scientific community (222). By observing the flask monolayer under an inverted microscope, a cytopathic effect could be observed after several weeks; small coarse dark inclusions and large coarse round structures were detected within cells. The cultured bacterium can been identified by PCR and sequencing of the 16S rRNA.

Legionella spp. Legionella spp. are small, facultative, gram-negative intracellular bacteria that cause nosocomial pneumonia and, rarely, disease of extrapulmonary sites 165. Endocarditis due to Legionella spp. has been cited in only three publications. In 1984, McCabe reported the first case of Legionella pneumophila endocarditis in a patient with porcine aortic and mitral valve prostheses 182. In 1988, Tompkins et al. reported seven cases of nosocomial prosthetic-valve endocarditis at Stanford University Hospital Center. L. pneumophila was isolated in two cases, L. dumofii was isolated in three cases, both species were identified in one case, and an undefined Legionella sp. was serologically diagnosed in one case 281. In 1994, Park et al. reported one case of serologically diagnosed Legionella micdadei prosthetic-valve endocarditis 208.

Chlamydia psittaci. Chlamydia spp. are obligate intracellular organisms that can be grown only in tissue culture. The intracellular location explains the difficulty in isolation and identification and the difficulty in investigating its role as a causative agent of endocarditis. Chlamydia spp. have often been suggested to be causative agents of endocarditis (18, 24, 60, 61, 89, 132, 136, 156, 201, 231, 253, 285, 291, 295; R. Norton, S. Schepetiok, and T. W. Kok, Letter, Lancet 345:1376-1377, 1995. D. Dumont, D. Mathieu, M. Alemanni, F. Eb, and G. Manigand, Letter, Press Med. 19:1054, 1990). However, in a recent review of 10 patients reported to have chlamydial endocarditis, 8 were finally diagnosed with Bartonella endocarditis after their sera were tested by antibody cross-absorption and Western immunobloting 176, 225. Interestingly, epidemiologic data for three of these eight patients were comparable to those reported for Bartonella endocarditis patients, including homelessness and alcoholism 172, 176. Therefore, by using serologic testing only, it is difficult to know how many patients really have chlamydial endocarditis.