A Newly Recognized Fastidious Gram-Negative Pathogen as a Cause of Fever and Bacteremia

Table 1. Table 1. Epidemiologic and Clinical Characteristics of the Patients at Presentation.*

Two illustrative case reports are provided. Pertinent data regarding all patients are given in Table 1.

Patient 1

A 31-year-old man infected with human immunodeficiency virus (HIV) was seen with a fever (temperature to 40.8°C), chills, sweats, and weight loss in November 1986. Fever persisted despite empirical therapy with erythromycin and then cephalexin. A chest radiograph revealed no active disease. Sputum examinations revealed no malignant cells or potential pathogens. Blood cultures were negative after a 48-hour incubation. Fever remitted during two weeks of empirical treatment with oral trimethoprim–sulfamethoxazole but recurred after its discontinuation.

Examination in January 1987 revealed fever (temperature, 38.3°C), marked cachexia, and weakness. He had pancytopenia, lymphopenia, and hypoxemia. Liver enzyme levels in serum were elevated. Bronchoscopy revealed no pathogens or histopathologic features. Empirical treatment with parenteral trimethoprim–sulfamethoxazole induced severe nausea and was soon stopped. Hepatomegaly was the sole abnormality detected by computed tomography of the head and abdomen. Liver biopsy disclosed normal architecture with interspersed granulomata. Aspiration and biopsy of bone marrow revealed dyserythropoiesis without megaloblastic or microcytic changes. No organisms were seen in or cultured from liver or marrow. The results of lumbar puncture and cardiac sonography were normal. Studies for antibodies to brucella, tularemia, and fungi were negative; cultures of urine and the bully coat for cytomegalovirus were also negative.

After 15 days of incubation, a culture of blood subjected to blood-cell lysis and then centrifugation (lysis—centrifugation) grew a curved gram-negative bacillus. Two cultures obtained six days after the first also grew the same organism after a long incubation. Parenteral erythromycin therapy was begun, and the fever declined slowly over a two-week period. Five weeks later, the fever recurred despite renewed treatment with oral erythromycin, but further blood cultures were negative. Norfloxacin was substituted, and fever remitted permanently during the four weeks of therapy. Further blood cultures were negative.

The patient died less than a year later of progressive HIV-related neurologic disease and disseminated Mycobacterium kansasii infection.

Patient 5

A 40-year-old woman presented in August 1989 with a four-day history of fever (temperature to 39.4°C), chills, sweating, nausea, vomiting, frontal headache, back pain, and weight loss. Examination disclosed fever (temperature, 38.8°C), liver tenderness to percussion, and lower-extremity petechiae. She had thrombocytopenia, granulocytosis, and mild lymphopenia, but hemoglobin levels and coagulation and serum-chemistry profiles were normal. Aspiration and examination of bone marrow demonstrated increased numbers of megakaryocytes. Cerebrospinal fluid was normal. Autoantibody screening was negative, and complement levels were normal. Serologic testing revealed high antibody titers to Epstein–Barr virus and cytomegalovirus, but negative reactions for HIV, brucella species, Francisella tularensis, Rickettsia rickettsii, and Ehrlichia canis (serum samples obtained during the acute phase of the disease and during convalescence were used to test for the last two microorganisms). Urine cultures were negative for bacteria and cytomegalovirus. Blood for culturing was drawn once on hospital days 1 and 2 and twice on day 4, when the patient was still febrile. A 10-day course of oral tetracycline was started on day 4. The fever subsided in 24 hours, and platelet counts soon returned to normal.

After incubation periods of 10 to 14 days, all lysis—centrifugation blood cultures yielded a curved gram-negative bacillus. All biphasic blood cultures had been negative when discarded after seven days. The patient has remained well on follow-up.

Isolation and Characterization

Blood was inoculated into 10-ml lysis—centrifugation tubes (Isolator, Dupont, Wilmington, Del.) for Patients 1, 2, and 3. Culture sets consisting of a biphasic medium (Septi-chek, Roche Diagnostics, Nutley, N.J.) and a lysis—centrifugation tube were used for Patients 4 and 5. Biphasic cultures were kept for seven days. Cultures derived by centrifugation after blood-cell lysis were plated on fresh medium, including chocolate agar (proteose peptone number 3 agar base) and sheep's-blood agar (Columbia blood agar base), and incubated at 35°C in 5 percent carbon dioxide for at least 14 days. Suspect colonies were subcultured on heart-infusion rabbit-blood agar.

Motility was determined by microscopical examination of saline wet-mount preparations. Catalase testing was done with 3 percent hydrogen peroxide. The oxidase test was performed with O-paraphenylene-diamine (Marion Scientific, Kansas City, Mo.). Further biochemical testing was attempted with both conventional¹ and commercially available (Micro-Scan, Baxter-Travenol, Sacramento, Calif.; API Rapid-Strep, Analytab Products, Plainview, N.Y.) approaches. The Centers for Disease Control in Atlanta further evaluated isolates 1, 2, and 3.

Direct fluorescent antibody testing for Legionella pneumophila groups 1 through 6 and L. micdadei conjugate A (with use of a commercial kit, Wampole Laboratories, Cranbury, N.J.) and for F. tularensis (with use of a reagent provided by the Centers for Disease Control) was performed. Susceptibility testing was done with single concentrations of antimicrobial agents incorporated in chocolate agar (prepared from GC medium base).² The following concentrations (in milligrams per liter) of antimicrobial agents were used: penicillin, 0.12; ampicillin, 2.0; aztreonam, 8.0; gentamicin, 4.0; tobramycin, 4.0; trimethoprim–sulfamethoxazole, 2.0 and 40, respectively; norfloxacin, 4.0; ciprofloxacin, 2.0; chloramphenicol, 8.0; erythromycin, 0.5; rifampin, 2.0; tetracycline, 4.0; and vancomycin, 4.0. Sensitivity to nalidixic acid and cephalothin was determined by disk diffusion with 30-μg disks.

Analyses of Whole-Cell Fatty Acids

Cultures of each isolate and of strains of Helicobacter pylori, formerly Campylobacter pylori ³ (strain 7878, American Type Culture Collection [ATCC], Rockville, Md.), and Rochalimaea quintana (ATCC strain RV358) were harvested after a seven-day incubation at 35°C in 5 percent carbon dioxide on plates containing heart-infusion 5 percent rabbit-blood agar. Fatty acid methyl esters were made and chromatographed according to the techniques developed by Miller and Berger.⁴ Fatty acid methyl ester profiles were identified by a computer-assisted comparison of the retention times of the specimens with that of a standard quantitative mixture of fatty acid methyl esters (Microbial-ID, Newark, Del.).

Preparation of Cell-Membrane Proteins

A modification of a technique for the isolation of outer-membrane proteins from Haemophilus influenzae was used.⁵ Bacteria grown after a one-week incubation on chocolate agar at 37°C in 5 percent carbon dioxide were suspended in HEPES buffer and sonicated at 70 W in 15-second bursts until the solution was clear. After centrifugation at 1700×g, the remaining supernatant was ultracentrifuged at 100,000×g. The resultant pellet was resuspended in HEPES buffer with 1 percent sodium lauroylsarcosine and, after a 30-minute incubation at room temperature, ultracentrifuged. The final gelatinous pellet of cell-membrane protein was resuspended in distilled water and stored at -70°C after colorimetric measurement of protein content (Bio-Rad, Richmond, Calif.).

Electrophoresis

Sodium dodecyl sulfate–polyacrylamide-gel electrophoresis was performed with a resolving gel with a gradient of 7.5 to 15 percent and a 4 percent stacking gel. Samples containing 2 to 3 μg of cell-membrane protein were run in each lane, with molecular-weight standards in adjacent lanes. Electrophoresis was performed at 15 mA per gel in a water-cooled chamber and was halted when the tracking dye reached the lower edge of the gel. The resolving gel was then fixed and stained with silver.⁶

Patterns of DNA Digestion by Restriction Endonuclease

The technique of van Ketel et al.⁷ was used for restriction-endonuclease digestion. Bacteria grown as for the preparation of cell-membrane protein were suspended, incubated with lysozyme for 30 minutes, and then lysed with 1 percent sodium dodecyl sulfate for 15 minutes before overnight digestion with pronase (Sigma, St. Louis). Nucleic acids were twice extracted with phenol—chloroform—isoamyl alcohol (25:24:1), precipitated overnight in ethanol at -20°C, and centrifuged. The washed pellet was resuspended and incubated with RNase (Sigma) for one hour. DNA purification was completed with an additional extraction with phenol—chloroform—isoamyl alcohol, followed by chloroform extraction and overnight precipitation in ethanol. A total of 4 μg of each DNA specimen was digested for two hours with 40 units of EcoRV (Bethesda Research Laboratories, Gaithersburg, Md.). Digests loaded onto a 0.7 percent agarose gel were electrophoresed overnight at 40 V, stained with ethidium bromide, and transilluminated with ultraviolet A for observation.

Immunodiffusion Studies

Serum samples obtained from Patients 2, 3, and 5 at least a month after the resolution of their febrile illnesses were stored at -20°C. No earlier serum samples were used. Aliquots were thawed and placed in the central wells of agarose immunodiffusion plates. The peripheral wells contained whole-cell sonicates of suspensions of isolates 1 through 5 and, as controls, supernatant from a broth culture of Staphylococcus aureus (ATCC strain 29213) or a whole-cell sonicate of a suspension of R. quintana. A control plate contained serum from a patient with S. aureus bacteremia in the center well and the same peripheral-well contents as in the first plate. The plates were incubated in a humidified container at 37°C for a week, then viewed by transmitted indirect lighting.

Isolates were recovered by extended incubation of lysis—centrifugation blood cultures (Table 1). Where there were paired lysis—centrifugation and biphasic cultures, two or more lysis—centrifugation cultures were positive in each case, whereas no biphasic cultures yielded organisms during the seven days they were incubated.

Subcultures from primary culture plates grew in four days at 35 or 37°C in 5 to 10 percent carbon dioxide (but not at 30 or 42°C or in the absence of oxygen or carbon dioxide) on chocolate, charcoal—yeast extract, or blood agar. Rabbit blood supported growth better than sheep's blood. Growth was best on fresh chocolate agar and poor on medium more than three weeks old. Hemin disks (0.01 or 0.1 percent) enhanced growth on older mediums. Satellite growth occurred around micrococcus species and S. aureus but not Escherichia coli or Pseudomonas aeruginosa. Growth was absent on Thayer Martin or campylobacter blood agar, or on agar lacking heme compounds, including mediums for mycobacteria and MacConkey agar.

Figure 1. Figure 1. Gram Stain of a Clinical Isolate, Revealing Fine, Curved, Lightly Staining Gram-Negative Rods (×610).

Adherent white iridescent colonies were morphologically heterogeneous. The colonies consisted of small, slightly curved gram-negative rods (Fig. 1) measuring 0.6 by 1.0 μm on electron microscopy. The morphologic features suggested a species of campylobacter. When mounted in saline, all isolates displayed twitching motility that was especially notable with short forms. Attempts to stain flagella were unsatisfactory because of the autoadherence of the colonies. Flagella were not seen on electron microscopy. Organisms stained poorly with safranin and were non—acid-fast. Fluorescent antibody tests for legionella species and F. tularensis were negative.

Isolate 1 was weakly catalase-positive, but isolates 2 through 5 were catalase-negative; none of them were oxidase-positive. Inadequate growth made all attempts at further biochemical characterization equivocal; all other tests — response to triple sugar iron agar and other measures of carbohydrate fermentation or oxidation, nitrate reduction, salt tolerance, and hippurate hydrolysis — were negative. The CDC found that the organism was oxidase-positive according to Kovac's method,⁸ but otherwise confirmed our findings pertaining to motility and the lack of other discernible biochemical activities. No classification of the organisms could be made.

All isolates were susceptible in vitro to cephalothin, aztreonam, trimethoprim–sulfamethoxazole, chloramphenicol, erythromycin, aminoglycosides, fluoroquinolones, and rifampin. All were resistant to nalidixic acid. Isolate 1 was resistant to penicillin and ampicillin (because of the production of beta-lactamase), as well as to vancomycin. Isolate 2 was resistant to tetracycline.

Table 2. Table 2. Cellular Fatty Acid Composition of Isolates 1 through 5, Brucella, Rochalimaea, Helicobacter, Campylobacter, and Legionella.

The whole-cell fatty acid composition of each isolate was determined as an alternative approach to biochemical characterization. The proportions of fatty acids were similar for all isolates (octadecenoic acid, 48 to 58 percent; octadecanoic acid, 21 to 26 percent; and hexadecanoic acid, 16 to 23 percent), closely resembling those of R. quintana and somewhat like those of brucella species,⁹ but unlike those of H. pylori, campylobacter species,^{10 , 11} or legionella species¹² (Table 2).

Figure 2. Figure 2. Sodium Dodecyl Sulfate–Polyacrylamide-Gel Electrophoresis of Preparations of Cell-Membrane Protein (Silver Stain).

There is great similarity in the protein patterns of the isolates from Patients 1 through 5, whereas the pattern of R. quintana is very different. The first and eighth lanes contain molecular-weight (MW) standards.

Figure 3. Figure 3. DNA-Fingerprint Study.

Lane 1 contains DNA from a lambda phage (as an endonuclease quality control); lane 2, R. quintana; and lanes 3, 4, 5, 6, and 7, clinical isolates 1, 2, 3, 4, and 5, respectively, after EcoRV cleavage, agarose electrophoresis, and staining with ethidium bromide under ultraviolet A transillumination. The patterns of the clinical isolates are virtually identical and distinct from that of R. quintana.

For all isolates, preparations of cell-membrane protein resolved by sodium dodecyl sulfate–polyacrylamide-gel electrophoresis contained similar major bands between 31 and 42 kd and minor bands at 19 kd and between 45 and 150 kd — a pattern that differed from that of R. quintana (Fig. 2). Patterns of EcoRV-produced DNA cleavage of all isolates were virtually identical and differed from that of R. quintana (Fig. 3).

Figure 4. Figure 4. Immunodiffusion Study.

A serum sample obtained during convalescence from Patient 2 was placed in the center well. Whole-cell sonicates of isolates from Patients 1 through 4 and S. aureus (Sa) were placed in the peripheral wells. A line of identity connects all the clinical isolates. No immunoprecipitation occurred with S. aureus.

By immunodiffusion, serum samples obtained during convalescence from Patient 2, who had a prolonged febrile illness, produced a line of identity with all five isolates (shown in Fig. 4 with isolates 1 through 4) but did not react with R. quintana or S. aureus. Serum samples obtained during convalescence from Patients 3 and 5, who had febrile illnesses lasting a week or less, and serum from a patient with S. aureus bacteremia did not react with any of the isolates or control organisms. In addition, a human serum specimen reactive with E. canis (at a dilution of 1:1024 by indirect immunofluorescence) did not immunoprecipitate with isolates 1 and 2.

We have isolated a previously unrecognized pathogen that caused bacteremic disease in five patients, three with immunologic dysfunction and two without. The organism is a highly fastidious, slender, curved, gram-negative motile bacillus that forms heterogeneous colonies under microaerophilic conditions. It requires an extended incubation for primary isolation from lysis—centrifugation blood cultures. It was susceptible in vitro to most antimicrobial agents tested. Highly similar concentrations of total cellular fatty acid, profiles of cell-membrane protein, and DNA-cleavage patterns were demonstrated among the isolates, and the antigenic identity of the isolates was shown by immunoprecipitation with serum samples obtained during convalescence from one patient.

The isolation of this bacterium resulted from laboratory practices designed to enhance the detection of disseminated pathogens, such as mycobacteria^{13 14 15} and Histoplasma capsulatum, ¹⁶ by lengthening the routine maintenance of plates of lysis—centrifugation cultures to at least two weeks. Before 1989 at the Oklahoma Medical Center, a routine blood culture from an adult consisted of inoculation of the specimen into a pair of conventional broth bottles. Lysis—centrifugation cultures were available by request when indicated. This was the case for Patients 1 through 3. In 1989 the routine blood-culture system for adults was changed to include one lysis—centrifugation tube and one biphasic culture to maximize microbial recovery.¹⁷

This pathogen may have been previously unrecognized because of its fastidious nature. Its low frequency of isolation is underscored by the fact that 6096 lysis—centrifugation blood cultures were processed at the Oklahoma Medical Center in 1989. Recognition also may have been retarded by the widespread use of antimicrobial agents, since it is broadly susceptible to most common agents.

Although sharing characteristics with pathogens such as R. quintana, Brucella canis, and H. pylori, this organism is clearly different. R. quintana, the cause of trench fever and the only known species of rickettsia cultivable on artificial medium,¹⁸ had a fatty acid profile most similar to that of our isolates. The chief difference was a ratio of C16:0 to C18:0 of greater than 1.0 for R. quintana and less than 1.0 for the isolates from the patients. However, R. quintana grew faster than our isolates and had substantially different colonial and microscopic morphologic features, profile of cell-membrane protein, and DNA-cleavage pattern.

No epidemiologic or clinical clues have suggested the environmental origin of this pathogen or the mechanism by which infection occurred. None of the patients knew each other. Their illnesses were sporadic, and they came from diverse areas of Oklahoma. All but one used water from municipal water supplies, and the well water used by the fifth patient did not harbor the organism. Some had healthy domestic pets. None recalled exposure to ticks or wild animals — a key issue in the light of the high rates of rickettsial diseases (Rocky Mountain spotted fever and human ehrlichiosis) in Oklahoma¹⁹ and the similarity of this organism to R. quintana.

No portal of infection was evident. Patients had normal chest examinations and radiographs. None had meningeal signs; cerebrospinal fluid was normal in the two who underwent lumbar puncture. Only one had skin lesions — petechiae resulting from platelet consumption. One patient with hepatomegaly and abnormal liver function had granulomata within normal liver architecture, but no organisms were found by special staining or culture. Results of urinalyses were normal; urine cultures were negative.

Responses to therapy differed notably. Before their presentation at our hospital, both the HIV-infected patients had received courses of antimicrobial agents to which their isolates ultimately demonstrated susceptibility in vitro, yet they remained febrile and had persistent bacteremia. Treatment lasting at least a month appeared to be necessary to eradicate the fever, although the organisms could no longer be cultured after the institution of antimicrobial therapy at our hospital. Septicemia developed in the bone marrow—transplant recipient who had been treated with steroids despite long-term therapy with an antimicrobial agent to which her isolate also was susceptible in vitro. However, she had a very prompt response to therapy with other agents. The immunocompetent hosts, who had febrile illnesses lasting several days, also had prompt disappearance of their symptoms within days of starting antimicrobial therapy.

Although the immunodiffusion technique was useful in demonstrating the antigenic similarity of the isolates, it was able to detect an antibody response only in the patient with prolonged fever (and likely prolonged antigenic exposure). Countercurrent immunoelectrophoresis was attempted with whole-cell sonicates of the five isolates, as well as of a clinical isolate of R. quintana and the ATCC type strain (VR-358). All yielded lines of precipitation with serum samples obtained during convalescence from Patients 2, 3, and 5; however, indistinguishable lines of precipitation also developed with several randomly selected serum samples from patients with bacteremia due to other organisms. This suggests that one or more ubiquitous antigens may resemble or be shared by the clinical isolates. Preliminary immunoblot studies using cell-membrane proteins as antigens have suggested to us that some normal human serum specimens also contain antibodies reactive with antigens of this organism.

Since this paper's submission, we have seen a previously healthy 30-year-old man in whom fever developed after a tick bite. He was initially treated empirically with tetracycline, and his symptoms disappeared. A lysis—centrifugation culture of blood obtained after his fever recurred in association with headache, photophobia, confusion, myalgia, and arthralgia yielded the organism described above (9 colony-forming units per milliliter) after eight days of incubation. The results of computed tomography of the brain and an examination of the cerebrospinal fluid were normal. The patient became afebrile after one dose of ceftriaxone and a day of parenteral chloramphenicol, and he completed treatment with oral doxycycline. A case of aseptic meningitis and persistent bacteremic illness due to an organism morphologically and biochemically comparable to the one described in this paper has since been described in a 30-year-old man after exposure to ticks in Arkansas.²⁰

In summary, we have identified a pathogen capable of causing febrile illness associated with persistent bacteremia in immunocompromised and immunocompetent hosts. Its recognition in the laboratory is based on optimal handling of lysis—centrifugation blood cultures. It differs from previously described pathogens, and its prevalence, habitat, and mechanism of pathogenesis remain undetermined. It should be sought in the setting of cryptogenic fever, especially in persons with defective cell-mediated immunity.