We used 511 specimens from a 1985 study of a random sample of the Nova Scotia population (6); 121 acute- and convalescent-phase serum specimens from a study (Nova Scotia, 1991-1994) of 149 ambulatory patients with community-acquired pneumonia (7); and specimens from a prospective study of community-acquired pneumonia requiring hospitalization conducted at 15 teaching hospitals in eight Canadian provinces (1996-1997).

All serum specimens from both groups of patients with pneumonia were tested for antibodies to Mycoplasma pneumoniae; influenza viruses A and B; parainfluenza viruses 1,2,3; adenovirus; and Respiratory syncytial virus (RSV) by a standard complement fixation technique in microtiter plates. Serum specimens from 60% of the patients (randomly selected from the group of patients with community-acquired pneumonia requiring hospitalization) were tested by the microimmunofluorescence test (8-10) for immunoglobulin (Ig) G and IgM antibodies to Chlamydia pneumoniae (AR 39 strain); C. psittaci (avian strain 6BC, feline pneumonitis strain FP, turkey strain TT 3, and pigeon strain CP 3); C. pecorum (ovine polyarthritis strain); and C. trachomatis (pooled antigens of serovars BED, CJHI, and FGK). Serum specimens from hospitalized patients with pneumonia were tested for antibodies to Streptococcus pneumoniae pneumolysin, pneumolysin immune complexes, C polysaccharide, surface protein A, Haemophilus influenzae, and Branhamella catarrhalis by Dr. M. Leinonen, National Public Health Institute, Oulu, Finland, as reported previously (11-13).

Acute- and convalescent-phase serum specimens from 150 patients also had been previously tested by enzyme-linked immunosorbent assay (ELISA) for antibodies to L. pneumophila serogroups 1-6 by Yu (14). A urine sample collected from each patient within 24 hours of hospitalization was tested for L. pneumophila serogroup 1 antigen by ELISA (15) (Binax, Inc., Portland, ME). Antibodies to Coxiella burnetii phase 1 and 11 antigens and to Chlamydia pneumoniae were determined by a microimmunofluorescence test, as described (16,17).

Antibody titers to LLAPs also were determined by the indirect fluorescent antibody technique. These included Acanthamoeba polyphaga strain Linc AP 1 and LLAP strains 1, 2, 4, 6, 7, 9, 10, 12; L. lytica (strains LLAP 3 and L 2, formerly Sarcobium lyticum); and Parachlamydia acanthamoeba (strain BN 9 and Hall's coccus). A. felis ATCC 53690 was from the American Type Culture Collection. LLAPs were cultured in A. polyphaga in 150 peptone-yeast extract-glucose broth (18) at 30°C. When maximally infected, amoebae were lysed through three cycles of freeze-thawing in liquid nitrogen. This suspension was then resuspended in 30 mL of phosphate-buffered saline and centrifuged at 10,000 rpm for 10 minutes. Supernatant fluid was removed, and pellets containing respective LLAPs were resuspended in the smallest possible volume of sterile distilled water and were adjusted to a concentration of 2 mg/mL, as determined spectrophotometrically. Antigen prepared in this manner was frozen at -20°C until required.

Specimens with an IgM titer of >1:100, seroconversion from 0 to 100, a fourfold rise in antibody titer between acute- and convalescent-phase serum, and a single or stable titer of >400 were considered indicative of recent infection. An antibody titer of >50 was considered seropositive, i.e., evidence of prior but not recent infection.

The seropositivity rate to various antigens is shown in Table 1. The background rate of such infection is low. Only one healthy Nova Scotian had serologic evidence of recent infection with an LLAP, LLAP 4. None of the patients with ambulatory pneumonia had such an infection. LLAP 4 was the most common LLAP-causing pneumonia: four such infections among patients with community-acquired pneumonia required hospitalization. Two of the 58 patients from the Nova Scotia site had infection with LLAP 4, versus one of 511 healthy Nova Scotians (p<0.029, Fisher exact test). BN 9 caused two infections, and LLAP 1 and 12 caused one each among patients with community-acquired pneumonia requiring hospitalization.

In August 1986, Rowbotham (19) isolated LLAP 1 from the sputum of an 82-year-old woman with persistent pneumonia, by cocultivation with A. polyphaga. Seroconversion was demonstrated to LLAP 3. He screened >5,000 serum specimens submitted for Legionella antibody testing and found that 10 patients met the criteria for infection with LLAP 3 (19).

The only other study similar to ours is a study by Benson et al. (20), who examined 500 patients with community-acquired pneumonia and determined antibody titers to LLAP 1,2,3,4,6,7,9, and Hall's coccus; 94 (18.9%) had a fourfold rise in antibody titer of 128 to any LLAP; 36 (7.2%) had a titer rise to >1,024. In contrast, 1.4% of our 255 hospitalized patients with community-acquired pneumonia had evidence of recent infection with a LLAP or Hall's coccus. As in our series, LLAP 4 was the most common cause of infection in the Benson study, which also found that in 10 (10.6%) of 94 patients with LLAP or Hall's coccus infection a copathogen had been implicated as cause of the pneumonia. Likewise, almost all our patients with LLAP infection were infected with another pathogen.

One of the most interesting findings in our study was a fourfold rise in antibody titer to BN 19 in a patient with presumed adult-onset Kawasaki syndrome, an acute vasculitis of unknown cause found predominantly in infants and young children. The diagnostic criteria include fever of >5 days plus four of the following five features: bilateral conjunctivitis without exudate; polymorphous eruption; cervical lymph node >1.5 cm in diameter; changes in the extremities, including edema of the hands or feet, palm or sole erythema, and periungual desquamation during convalescence; and changes in the oropharynx, including fissured red lips, strawberry tongue, and diffuse erythema of the oropharyngeal mucosa (21). Our patient met this definition. An association between an antecedent respiratory infection and Kawasaki syndrome has been described (21,22), as has exposure to freshly cleaned carpets (23,24). It is possible that the gamma globulin administered to our patient contained antibody to BN 19. However, there was no seroconversion or high titer of antibody to any of the other antigens included in our test panel. A possible association between infection with BN 19 and Kawasaki syndrome is easily tested.

Strengths of this study are its size and the comprehensiveness of the diagnostic work-up. Its limitations include the following: the three populations were enrolled in different periods; we tested only a subset of the patients hospitalized with community-acquired pneumonia and these patients were from multiple centers across Canada; our comparison groups (healthy persons and patients with ambulatory pneumonia) were Nova Scotians. However, the inferences that we are making are limited to the rate of infection in these three separate groups and are not intended to indicate differences temporally or geographically.

Our data suggest that LLAPs play a role, albeit an infrequent one, in community-acquired pneumonia. Usually they are a copathogen, but in some cases they are the sole pathogen. The possible association between BN 9 and Kawasaki disease requires further study.