P. carinii DNA was extracted from bronchoalveolar lavage specimens by using the Qiamp Blood Kit (QIAGEN GmbH, Hilden, Germany). Bronchoalveolar lavage specimens from 13 patients with recurrent PCP episodes were collected from four European hospitals (Lyon, France; Copenhagen, Denmark; Lausanne, Switzerland; and La Chaux-de-Fonds, Switzerland). To determine the prevalence of the different P. carinii molecular types, we analyzed bronchoalveolar lavage specimens from 310 PCP patients from two Swiss hospitals (Lausanne, 111 patients; Zurich, 64 patients) and Lyon’s hospital (135 patients). Specific information on demographic and clinical characteristics, chemoprophylaxis, and treatment regimens were obtained from the medical charts. P. carinii infecting humans (now named P. jiroveci [7]) was typed by using the multilocus method developed in our laboratory as previously described (8–10). In this method, four variable regions of the P. carinii genome are amplified by polymerase chain reaction (PCR), followed by the detection of polymorphisms using single-strand conformation polymorphism (SSCP). A P. carinii type is defined by a combination of four alleles corresponding to the four genomic regions. If a specimen harbored two alleles of one or more of the four genomic regions, the patient was considered to be co-infected with two or more P. carinii types (9). This typing system has been validated and the stability of its markers assessed; its index of discriminatory power has been estimated to be 0.93 (10). The full length of the DHPS gene was amplified by PCR as described previously (1). PCR products (765 bp) were cloned, and both strands were sequenced (5 clones per sample). The five clones had identical sequences for all samples, except for those from patients 3 and 4, which contained a mixture of DHPS sequences.

Thirteen patients with two separate PCP episodes were analyzed (Table). All patients had recovered between episodes. The intervals between the episodes ranged from 4 to 25 months. All patients had AIDS and all, except patients 8 and 9, were men, with a median age of 35 (range 23–51) and median CD4 cell count of 9.5 cells/μL (range 0–98). Some patients were co-infected with two different P. carinii types, as shown by PCR-SSCP multilocus typing method (patients 4, 5, 8, 11, and 13) or DHPS genotyping (patients 3 and 4). In seven (54%) patients (patients 1–7), the same PCR-SSCP type was observed in both episodes of PCP; six (46%) patients (patients 8–13) had different types in the first and second episodes. This rate of genotype switch is similar to that reported in previous studies, in which such a change was observed in approximately half of recurrent episodes (11–14). The importance of a genotype switch remains uncertain. Indeed, the switch might be due to a de novo infection or to the reactivation of a genotype not detected in the first episode because of the compartmentalization of different co-infecting P. carinii types in the lungs (15).

Figure

Figure. Frequency distribution of Pneumocystis carinii types observed in different locations. Each type, co-infecting or not, was considered as one isolate.

A second episode of PCP could result either from reactivation of organisms that caused the first episode or from de novo infection with a new P. carinii type acquired from an exogenous source. In seven patients (patients 1–7), reactivation was strongly suggested by the detection of identical SSCP types in both episodes of PCP. An alternative explanation could be de novo infection in the second episode by the same P. carinii PCR-SSCP type as that which caused the first episode. However, the prevalence of the types observed in the seven “reactivation” cases was low in Lyon and Switzerland during the study period (types no. 2, 5, and 7 represented 7%, 6%, and 10%, respectively, of Lyon’s isolates; type 6 represented only 3.5% of the Swiss isolates [Figure]). Thus, reinfection with these specific types was unlikely. All Danish patients (1, 3, and 6) were infected with type 6. Although no prevalence data for SSCP genotypes in Denmark are available, no indication of possible contact between these patients, overlap in hospitalization dates, or similar zip codes for home address suggested transmission of type 6 between these patients.

A change of P. carinii DHPS genotype between the two episodes was observed in three reactivation cases, either from wild type in the first episode to DHPS mutations in the second one (patients 1 and 5) or from DHPS with a single mutation (at position 57) in the first episode to a double mutation in the second one (patient 2). In two patients (3 and 4), the DHPS mutant strain was selected out of a mixture of wild-type and DHPS mutant strains. Because both episodes of each patient were caused by the same P. carinii types and because all patients received co-trimoxazole or dapsone as treatment, maintenance therapy, or both, these results strongly suggest that selection of P. carinii DHPS mutations occurred within the patients. The results of tests on patients 3 and 4 isolates highlight the fact that some patients may harbor genetically different strains of P. carinii and that the mutant strain may be readily selected when drug pressure is exerted. In the two remaining patients (6 and 7), the P. carinii DHPS mutant found in the bronchoalveolar lavage specimen from the second episode was already present in the first episode.

The wild-type DHPS allele was more frequently observed in the six reinfection cases than in the reactivation cases (8 wild-type alleles among 12 genotypes versus 4 among 16, Table). This finding is probably related to the fact that, with the exception of the second episode of patient 13, patients who were reinfected had no prophylaxis or did not receive sulfa drugs for prophylaxis.

In all the second episodes caused by reactivation, mutant DHPS strains were observed (7/7), compared to only two of six second episodes caused by reinfection (Table). This observation suggests an association between mutant DHPS and second episodes attributable to reactivation (p<0.02, Fisher exact test).

Our study suggests that P. carinii DHPS mutants may be selected in vivo (within a given patient) under the pressure of co-trimoxazole or dapsone and that DHPS mutations may be associated with reactivation of P. carinii. Whether DHPS mutations are induced by the pressure of the drug or preexisting and selected out by the pressure of the drug remains to be determined. Physicians should be alert to the increased risk for drug resistance during recurrence of PCP infection, although the impact of DHPS mutations on retreatment with sulfa or sulfone drugs remains to be determined. De novo selection of P. carinii DHPS strongly favors the hypothesis that P. carinii is developing sulfa and sulfone resistance.