|Home | About | Journals | Submit | Contact Us | Français|
Better understanding of the epidemiology and transmission patterns of human Pneumocystis should lead to improved strategies for preventing Pneumocystis pneumonia (PCP). We have developed a typing method for Pneumocystis jirovecii based on restriction fragment length polymorphism (RFLP) analysis following PCR amplification of an ~1300 bp region of the msg gene family, which comprises an estimated 50 to 100 genes/genome. The RFLP pattern was reproducible in samples containing >1,000 msg copies/reaction, and stable over time based on analysis of serial samples from the same patient. In our initial analysis of 48 samples, we found that samples obtained from different individuals showed distinct banding patterns; only samples obtained from the same patient showed an identical RFLP pattern. Despite this substantial diversity, samples tended to cluster based on country of origin. In evaluating samples obtained from an outbreak of PCP in kidney transplant patients in Germany, RFLP analysis demonstrated identical patterns in samples that were from 12 patients previously linked to this outbreak, as well as 2 additional patients. Our results highlight the presence of a remarkable diversity in human Pneumocystis strains. RFLP may be very useful for studying clusters of PCP in immunosuppressed patients, to determine if they have a common source.
Pneumocystis jirovecii is an opportunistic fungus that causes pneumonia in immunocompromised patients, especially those with HIV infection, in whom it remains a major cause of morbidity and mortality [1–3]. Understanding the epidemiology of human Pneumocystis infection can be important in minimizing the clinical impact of this pathogen.
Recently, a number of molecular biologic methods have been developed to study the epidemiology, transmission patterns, and potential emergence of antimicrobial resistance of this organism [4–10]. At least 14 unique genetic loci have been evaluated using a variety of typing methods, including DNA sequence, single strand conformation polymorphism (SSCP), and restriction fragment length polymorphism (RFLP). In general, genes that evolve very slowly, such as the mitochondrial small subunit rRNA and 5.8S rRNA genes, have been utilized in evolutionary studies or to examine genetic diversity. Genes that show greater diversity, such as the internal transcribed spacer region of the nuclear rRNA genes, have been used to evaluate case clustering and to look for evidence of direct person-to-person transmission. Genes that are targets of therapeutic drugs, such as the dihydropteroate synthetase (DHPS) and dihydrofolate reductase (DHFR) genes, have been used to study the development of drug resistance.
The major surface glycoprotein is a Pneumocystis surface protein encoded by a multi-copy msg gene family, with an estimated 50 to 100 copies of related but unique msg genes per genome; only a single msg variant, which is present at a unique (single copy) expression site termed the upstream conserved sequence (UCS), is expressed in a given organism [11–17]. We have previously shown that a region in the intron of the UCS of P. jirovecii contains variable numbers of 10 base-pair tandem repeats that can be easily quantitated and used in epidemiologic studies . More recently, we have demonstrated by RFLP analysis that the repertoire of msg genes (the 50–100 genes per genome) shows substantial variation among P. jirovecii isolates, but not P. carinii (infecting rats) or P. murina (infecting mice) isolates . This suggested that RFLP analysis of msg genes could be used as an epidemiologic tool to investigate human Pneumocystis infection. The current study was undertaken to determine if msg repertoire variability could be used as a typing method to distinguish among different human isolates of Pneumocystis.
A total of 48 isolates of P. jirovecii, which included autopsy lung, bronchoalveolar lavage (BAL), sputum, and oral wash samples, were obtained from 40 patients from San Francisco (n=8), other parts of the United States (n=8), the Netherlands (n=3)  and Milan (n=21) . An additional 26 amplifiable samples from 22 patients were obtained from Germany . Guidelines of the US Department of Health and Human Services were followed in the conduct of these studies.
Genomic DNA was extracted using QIAamp DNA mini Kit (Qiagen, Valencia, California), according to the manufacturer's instructions. Nested (initial studies) or semi-nested PCR was utilized to amplify ~1.3 kb of the msg variable region using the following primers: for the first round, GK 126, 5' – GTGGCGCGGGCGGT-3' (corresponding to 2138–2151 bp of P. jirovecii msg UCS, GB#AF367050, initial studies) or GK 472, 5'- TGGATCAAAAGMGAGAYTTYCCRACAG-3' (corresponding to 1576–1602 bp of P. jirovecii msg HuMSG14, GB#AF033209, later studies) and GK 452, 5'-AATGCACTTTCMATTGATGCT-3' (complementary to 479–499 bp of P. jirovecii msg, partial cds, GB#AF372980); for the second round (all studies), GK 472 and GK 195, 5'GTGTGTGTCGATGTCTGTG-3' (complementary to 2875–2893 bp of P. jirovecii msg HuMSG14). The PCR was performed using HotStart Taq (Qiagen) and the conditions were 15 min at 95°C followed by 35 cycles of 30 s at 94°C, 30 s at 60°C, and 4 min (for the first round) or 2 min (for the second round) at 72°C, with a finally extension of 10 min at 72°C. msg copy number was quantitated by a previously described real-time Q-PCR assay .
PCR products were purified using QuickStep 2 PCR Purification Kit (Edge BioSystems, Gaithersburg, Maryland) according to the manufacturer's instructions. The purified products were digested with Dra1, HindIII, or Xba1 restriction enzymes for 5–6 hours at 37°C. The digested products were analyzed on a 1.2 % TBE agarose gel and visualized by SYBR green staining (Molecular Probes, Eugene, Oregon). Following transfer to a Nytran membrane (Whatman, Sanford, Maine), the blot was hybridized with a digoxigenin-labeled DNA probe (PCR DIG Probe Synthesis Kit, Roche, Indianapolis, Indiana). The PCR probe (spanning ~ 1.3 Kb) was an equal mixture of 4 products obtained by PCR amplification (as above) of lung samples from 4 P. jirovecii-infected individuals. The hybridization signal was detected by chemiluminescence using alkaline phosphatase conjugated anti-digoxigenin antibody and CDPstar (Roche, Indianapolis, Indiana). The results were recorded with a Luminescent Imager (Kodak Image Station 440CF, PerkinElmer, Waltham, Massachusetts).
The gels and blots were analyzed using BioNumerics software version 4.01 (Applied Maths, Inc., Austin, Texas). The pattern of banding among different gels/blots was normalized using internal standards that were included in each run: Lambda/HindIII molecular weight markers for gels, and a clinical sample (sample number 385) for blots. Molecular weights were assigned to the bands of the standards, and sample bands were identified manually. The Dice coefficient was used to analyze the similarity of the patterns of bands with a position tolerance of 1.9% . The un-weighted pair group method with average linkages (UPGMA) was utilized by the BioNumerics software for cluster analysis. DNA samples with banding patterns with 100% similarity (Dice coefficient=1) were considered to be identical.
We have previously demonstrated substantial variation in RFLP patterns among a small number of P. jirovecii isolates when analyzing the entire msg sequence (~3200 nucleotides) from autopsy lung samples . Due to difficulties in amplifying the entire sequence when using samples with lower organism loads (e.g. induced sputum), we developed a semi-nested PCR that amplified an ~1300 nucleotide region of msg, using primers from conserved regions of msg (based on alignment of available P. jirovecii msg clones).
In preliminary studies RFLP analysis of this shorter msg fragment also showed substantial variation among isolates. To investigate the reproducibility of this assay, we ran replicate PCR reactions using Pneumocystis DNA extracted from lung tissue and oral wash samples. The PCR products were digested with HindIII or Dra1 restriction enzymes, and analyzed by agarose gel electrophoresis or Southern blotting. We found that the pattern was reproducible using lung tissue but not consistently with oral wash samples (data not shown) suggesting that the reproducibility of the RFLP pattern is dependent on organism load. By serial dilution of a single sample, we found that this assay lost reproducibility at an msg copy number (as determined by Q-PCR ) of less than ~1,000 copies per PCR reaction (data not shown). Based on these data, only samples with msg copies above ~1,000/copies per PCR reaction were considered to be reliably reproducible, while samples with msg copies number below that cut off were utilized with caution.
To investigate the stability of the RFLP pattern in samples obtained from one individual over time, we examined paired samples collected over varying periods of time. Those collected at close time-points (e.g. < 3 months) likely represent the same episode of PCP and likely would exhibit the greatest stability. We analyzed 16 samples from 8 individuals (10 BAL samples, 2 sputum samples and 4 oral wash samples). The collection time between samples ranged from 1 day to 111 days. All BAL pairs showed an identical RFLP pattern (Figure 1, patients 1 and 2) and the pair of sputum samples were identical (Figure 1, patient 3), although there were differences in banding intensity among the paired isolates that may represent variation in the proportions of co-infecting P. jirovecii types. The oral wash pairs (Figure 1, patients 4 and 5) were highly similar, but, for patient 4, not identical, possibly due to low msg copy numbers. These data demonstrate that the RFLP pattern is stable over at least a period of days to weeks, and that recurrent episodes of PCP (patient 2, figure 1) can result from relapse rather than reinfection with a new strain.
We then undertook RFLP studies to compare a larger number (n=48) of Pneumocystis isolates that we had collected over time. Because RFLP analysis with Dra1 alone appeared sufficient to distinguish among isolates, these studies utilized only Dra1 in the initial analysis. Because samples needed to be run on different gels, a known sample (sample number 385) that would hybridize to the probe during Southern blotting, as well as commercial molecular weight standards, were included in each gel to allow comparison among different runs.
Based on visual examination, gels and blots showed an identical RFLP banding pattern only for samples collected from the same individual, and a distinct RFLP pattern in samples obtained from different individuals. To allow comparison of samples run on different gels, we utilized BioNumerics software, with standardization using molecular weight markers (for gels) and sample 385 (for blots). Figure 2 shows a dendrogram created by BioNumerics software from different gels, after normalization utilizing the internal standards. For occasional samples, there were differences in the precise clustering between gels and blots, which may be related to differences following hybridization with the probe specific for P. jirovecii (data not shown).
The dendrogram demonstrates that in general only samples obtained from the same patient showed 100% similarity. However, this analysis identified 4 Italian samples that form two pairs, with each pair showing 100% similarity (samples: 2999–3960 and 2428–7780). All the Italian samples were collected in the same city, Milan, during the period of time 1994–1999, but due to unlinking we cannot go back to the patient data to know if these samples were collected from the same patient or at close time points. To explore this further, we digested one pair of samples (2999–3960) with another enzyme, Xba1. Xba1was used rather than HindIII as the second enzyme in our later studies because it generated greater variability in the RFLP pattern than HindIII. These samples clearly showed different RFLP patterns (data not shown), demonstrating that they were not identical. Unfortunately, there was inadequate material to run the other 2 samples.
The Pneumocystis isolates we analyzed were collected in the United States, Italy and the Netherlands. Interestingly, the samples collected in the same country clustered more closely to each other than to samples from other countries (Figure 2). In particular, the samples from Italy and the United States, which accounted for the majority of isolates, were not randomly intermingled.
Given the substantial variability among Pneumocystis isolates from different patients and the stability of RFLP patterns within individuals over time, RFLP analysis appeared to provide a method for easily demonstrating whether isolates from a potential outbreak of PCP were identical. A recent study of an outbreak of PCP in renal transplant patients in Germany provided molecular evidence, primarily by single nucleotide polymorphism (SNP) analysis, that all patients were infected with a single Pneumocystis strain . Pneumocystis DNA from 22 German patients was provided to us for RFLP analysis: 13 were from patients linked to the outbreak and 9 were unrelated to the outbreak. In 12 isolates previously identified as a single strain of Pneumocystis by SNP analysis, RFLP analysis showed an identical banding pattern (Figure 3A), although patient 8 (<1,000 msg copies/assay) had an additional band not seen in the other 11 isolates. RFLP analysis with a second enzyme, Xba1, confirmed that the banding pattern was identical in all 12 isolates (data not shown). One additional isolate from the outbreak had a different banding pattern but the sample contained <200 msg copies/assay. Of the 9 patients that were not previously linked to the outbreak, 6 isolates showed a unique pattern, while 2 isolates showed a pattern identical to the outbreak pattern (Figure 3B) and one was similar but had <1,000 msg copies/assay; RFLP analysis with Xba1 again confirmed these results (data not shown). Clinical history obtained after these results confirmed that all 3 of the latter patients were renal transplant patients who had been seen in the same clinic as the other outbreak patients and who underwent bronchoscopy within the time-frame of the outbreak.
When the samples from Germany were included in the dendrogram, the non-outbreak samples did not cluster with the outbreak strain, but tended to cluster with samples from Italy (data not shown).
We have developed a reproducible and easy to perform method to type human Pneumocystis strains using RFLP analysis. By this method we have demonstrated a remarkable diversity among human Pneumocystis strains: no two isolates from different patients showed an identical RFLP pattern, other than those from a cohort of German patients previously linked as part of a nosocomial outbreak of PCP. In contrast, samples from the same patient (that were obtained within 111 days of each other) showed an identical pattern. Thus this method may be very useful for studying transmission patterns as well as potential outbreaks of PCP in immunosuppressed patients. Moreover, our data strongly support the previously published conclusions that the renal transplant patients from Germany were infected with the same Pneumocystis strain, and identified at least 2 additional renal transplant patients that were likely part of the same outbreak.
A strength of RFLP analysis is that rather than examining a single or very limited number of nucleotide polymorphisms, as is the case with many available typing methods, it interrogates the entire msg repertoire of the Pneumocystis genome, which is estimated to include 50 to 100 genes, with ~ 40% of each msg of ~3200 nucleotides being evaluated in the RFLP analysis. Given the multiple msg copies per genome, and the high level of sequence conservation in short stretches across msg genes, it is likely that recombination in Pneumocystis can lead to rearrangements and establishment of unique msg repertoires, as we and others have previously shown . The conservation in RFLP pattern among isolates from the same patient, as well as the conserved pattern among a cohort of patients linked epidemiologically, suggest that recombination does not occur within a period of days to weeks. Previously we have shown that the RFLP pattern in mouse and rat Pneumocystis isolates obtained from inbred animals, at two locations, over a period of years, were identical or highly similar. Assuming human Pneumocystis are biologically similar, it appears likely that repertoire evolution is not rapid, and the observed diversity is related to recombination that has evolved over many, perhaps thousands or millions of years.
One potential disadvantage of RFLP analysis is that it is unable to distinguish, in patients infected with more than one isolate, the contributions of individual isolates to the banding pattern. If only one of multiple strains is transmitted to a new host, however, the RFLP bands should be easily distinguishable as a subset of those in the first host. In addition, the reproducibility of RFLP analysis was lost in samples with msg copies below 1000. This likely represents a sampling bias in a specimen with a low organism load that result in uneven distribution of msg variants in different aliquots.
Based on the dendrogram analysis isolates obtained from patients from the same geographic area at approximately the same time did not show 100% similarity (other than 1 pair of Italian samples and samples from the renal transplant patients), suggesting that inter-human transmission among these patients did not occur to any significant extent. These data are in agreement with the results of a previous study  in which the authors showed, with PCR-SSCP typing, that transmission of Pneumocystis from patients with active PCP to susceptible persons is rare. However, outbreaks with the same strain can occur, as demonstrated by the renal transplant patients. It is noteworthy that samples collected in the same country seemed to cluster more closely with each other than with samples from other countries. This clustering may represent local strain variation. It is intriguing to speculate that host immune pressures at the population level (e.g. HLA-mediated) are driving the diversity of the msg repertoire, as has been reported for HIV .
In summary, this study has demonstrated a broad diversity in Pneumocystis strains, has provided a method for rapidly typing strains, and has provided confirmatory evidence that an outbreak of PCP was caused by a single strain of Pneumocystis. Larger studies utilizing this approach are needed to better define the epidemiology of Pneumocystis pneumonia, and to determine whether any predominant strains, as defined by RFLP analysis, can be identified.
We would like to thank Dr. Pieter J.A. Beckers, University Medical Centre St Radboud, The Netherlands for providing samples of P. jirovecii, and allowing us to include them in this study.
This research was supported in part by the Intramural Research Program of the NIH Clinical Center. Dr. Huang was supported by NIH K24 HL087713 and R01 HL090335.
The authors have no conflicting financial interests.