Abstract
Burkholderia gladioli colonizes the respiratory tracts of patients with cystic fibrosis and chronic granulomatous disease. However, due to the high degree of phenotypic similarity between this species and closely related species in the Burkholderia cepacia complex, accurate identification is difficult. Incorrect identification of these species may have serious repercussions for the management of patients with cystic fibrosis. To develop an accurate procedure for the identification of B. gladioli, a molecular method to discriminate between this species and other species commonly isolated from the sputa of patients with cystic fibrosis was investigated. The 23S ribosomal DNA was cloned from several clinical isolates of B. gladioli, and the nucleotide sequence was determined. Computer-assisted sequence comparisons indicated four regions of the 23S rRNA specific for this species; these regions were used to design three primer pairs for species-specific PCR. Two of the primer pairs showed 100% sensitivity and specificity for B. gladioli when tested against a panel of 47 isolates comprising 19 B. gladioli isolates and 28 isolates of 16 other bacterial species. One of the primer pairs was further assessed for species specificity by using a panel of 102 isolates obtained from the Burkholderia cepacia Research Laboratory and Repository. The species-specific PCR was positive for 70 of 74 isolates of B. gladioli and was negative for all other bacterial species examined. Overall, this primer pair displayed a sensitivity and specificity of 96% (89 of 93) and 100%, respectively. These data demonstrate the potential of species-specific PCR for the identification of B. gladioli.
Burkholderia gladioli (basonym, Pseudomonas gladioli or Pseudomonas marginata) was originally described in 1921 as a phytopathogen of gladioli and other flowers (11). Initially considered a harmless commensal colonizing the respiratory tracts of cystic fibrosis (CF) patients, B. gladioli has recently been associated with a number of pulmonary infections in both CF (4) and immunocompromised patients (12, 13). However, the extent to which this organism contributes to disease in CF patients is unclear. Recently Khan and colleagues (7) reported empyema and bacteremia caused by B. gladioli; however, the infection was subsequent to lung transplantation, during which the pleural cavity was inadvertently seeded. Evidence of pathogenicity is more forthcoming in patients with chronic granulomatous disease; several reports describe B. gladioli pneumonia (12) and severe sepsis (5).
Currently, there is great interest in the taxonomy of the genus Burkholderia, primarily focused on the characterization of the species of the Burkholderia cepacia complex. Since the designation of five distinct genomovars within this group (14), it is becoming increasingly apparent that the members of genomovar III are responsible for the majority of deaths of patients with CF ascribed to “cepacia syndrome” (14). Therefore, it is important that clinical microbiology laboratories be able to correctly distinguish members of the B. cepacia complex from other closely related species. To facilitate this identification, the Burkholderia cepacia Research Laboratory and Repository (BcRLR) tests all forwarded strains; a high incidence of misidentified isolates (10%) has been detected (9). Of the isolates misidentified as B. cepacia, approximately 50% are confirmed by the BcRLR as B. gladioli (data not shown).
Misidentification of closely related bacteria creates significant clinical dilemmas. Identification may be inaccurate due to the high degree of similarity between the B. cepacia complex and B. gladioli, as well as the existence of strains which appear to be hybrids of the two species (2, 13) and an inability of commercial tests to provide definitive identifications of the related species (8). Some CF centers treat B. cepacia-positive patients as a cohort, and incorrect identification of isolates could result in exposure of non-B. cepacia-colonized individuals to patients colonized with B. cepacia. Misidentification may also have a negative impact on treatment, transplant eligibility, and perceived disease progression. To facilitate the accurate and rapid identification of B. gladioli, we developed a B. gladioli species-specific PCR (SS-PCR) method.
MATERIALS AND METHODS
Bacterial strains.
The specificity of this method was tested by using bacterial species known to colonize CF patients and other Pseudomonas species of medical importance. The following microbial species were tested: Pseudomonas fluorescens ATCC 13525, Pseudomonas stutzeri ATCC 17588, Klebsiella pneumoniae ATCC 13883, Proteus mirabilis ATCC 29906, Stenotrophomonas maltophilia ATCC 13637, Moraxella catarrhalis ATCC 25238, B. gladioli ATCC 10854, ATCC 19302, and ATCC 10248, Ralstonia solanacearum ATCC 10692 and ATCC 11696, Ralstonia pickettii ATCC 27511, Burkholderia caryophylli ATCC 25418 and ATCC 11441, a clinical isolate of Pseudomonas aeruginosa, and a clinical isolate of Acinetobacter anitratus. In addition, a total of 16 clinical B. gladioli isolates, 13 B. cepacia complex isolates, and 2 R. pickettii isolates were used to establish sensitivity and specificity (for simplicity, the five genomovars of the B. cepacia complex are treated as separate species within this article). The identities of the clinical isolates of B. cepacia complex were confirmed at the BcRLR (9). The identities of the clinical B. gladioli isolates were confirmed by using Burkholderia/Ralstonia-specific PCR (9) in conjunction with biochemical reactions: oxidase negative, o-nitrophenyl-β-d-galactopyranoside (ONPG) positive, lysine decarboxylase negative, and no oxidation of lactose or sucrose. When the identity was still questionable, it was confirmed by whole-cell protein sodium dodecyl sulfate-polyacrylamide gel electrophoresis as described by LiPuma et al. (9).
Isolation of genomic DNA.
Genomic DNA was purified from cultures of each individual strain with the QIAamp Tissue Kit (QIAGEN, Valencia, Calif.). DNAs were resuspended in sterile high-pressure liquid chromatography-grade water at a final concentration of 1 μg/ml.
Cloning of the 23S ribosomal DNA and determination of the nucleotide sequence.
PCR primers were designed based on published B. cepacia 23S rRNA gene sequence data (6) (Table 1). These primers, 23RNA3 and 23RNA5, were used in separate PCRs to amplify an approximately 2,900-bp product from three B. gladioli clinical isolates designated BG2, BG3, and BG4. The products were cloned into the pCR-2.1 TOPO vector (Topo TA Cloning Kit; Invitrogen, Carlsbad, Calif.) to produce pBG2A, pBG3A, and pBG4A. The insert of each clone was sequenced with an ABI 373 automated sequencer (Perkin-Elmer Applied Biosystems, Foster City, Calif.). DNA sequences were assembled and analyzed with the GCG Wisconsin Package (Genetics Computer Group Inc., Madison, Wis.).
TABLE 1.
Oligonucleotide primers selected for PCR
| Primer | Sequencea | Target siteb |
|---|---|---|
| LP1 | GGGGGGTCCATTGCG | 872→886 |
| LP2 | TAGCAAGCTTAACCGA | 595→610 |
| LP3 | CTGCGGATGCACATTTATG | 1151→1170 |
| LP4 | AGAAGCTCGCGCCACG | 1523→1508 |
| 23RNA11 | GGGACTTCAATCAAGAGCTTGCAC | 1869→1846 |
| 23RNA5 | GGTCAAGCGAACAAGTGC | 1→18 |
| 23RNA3 | AGGATCAAGCCTTACGGGC | 2865→2847 |
Underlined nucleotides are putatively specific for B. gladioli.
Numbering corresponds to the B. gladioli 23S rRNA nucleotide sequence (EMBL accession no. Y17182). The arrow indicates the direction of the oligonucleotide (5′ to 3′).
Development of B. gladioli-specific PCR.
The B. gladioli 23S rRNA consensus nucleotide sequence was determined from the inserts of pBG2A, pBG3A, and pBG4A. Differences between this consensus and the published B. cepacia 23S rRNA gene sequence (6) were determined. These regions were further analyzed with the BLASTN nucleotide sequence alignment algorithm to determine if the putatively specific B. gladioli regions had homology with other 23S rRNA genes. Four regions putatively specific for B. gladioli were determined. Oligonucleotide PCR primers were designed at these regions, incorporating the B. gladioli-specific nucleotides at the 3′ end; these were designated LP1, LP2, LP3, and LP4 (Table 1). An additional primer, 23RNA11, was used. This primer is complementary to a conserved region of the 23S rRNA gene. LP1, LP2, and LP3 initiate extension in the same direction, while LP4 and 23RNA11 initiate extension in the opposite direction. Individual PCRs were optimized for the primer pairs LP1-23RNA11, LP2-23RNA11, and LP3-23RNA11 by using genomic DNA derived from BG2 and BG4. PCRs were performed with the Rapid Cycler thermocycler (Idaho Technologies, Idaho Falls, Idaho). All PCRs had an initial denaturation of 95°C for 5 min with a subsequent 30-cycle amplification, and the PCR mixture contained 1 μM each primer, 10 ng of genomic DNA, 200 μM each deoxynucleoside triphosphate, and 1.25 U of Taq DNA polymerase (Boehringer Mannheim, Indianapolis, Ind.) in a 3 mM MgCl2 PCR buffer (Idaho Technologies), in a total volume of 50 μl. For LP1–23RNA11, the cycle parameters consisted of annealing at 69°C for 10 s, extension at 72°C for 60 s, and denaturation at 95°C for 10 s. For the primers LP2–23RNA11 and LP3–23RNA11, the PCR parameters consisted of annealing at 60°C for 10 s, extension at 72°C for 60 s, and denaturation at 95°C for 10 s. Similarly, reactions were optimized for the LP1–LP4, LP2–LP4, and LP3–LP4 primer pairs. For these primer pairs the PCR parameters were identical: annealing at 60°C for 10 s, denaturation at 95°C for 10 s, and extension at 72°C for 60 s. Following amplification, 20 μl of each reaction mixture was subjected to electrophoresis in a 0.8% agarose gel in 0.5× Tris-borate-EDTA buffer (pH 8.0) alongside a 100-bp molecular ladder. The PCR products were visualized and photographed after ethidium bromide staining. Positive results were assessed by the amplification of a band of the correct size for the primer pair used. For LP1–LP4, LP2–LP4, and LP3–LP4, the amplified bands were 700, 900, and 400 bp, respectively. Samples that failed to amplify a product with these primer pairs were further analyzed with the PSL and PSR primers (3) as a control.
RESULTS
Cloning and sequencing of the 23S rRNA gene.
The 2.9-kb 23S rRNA genes from three independent clinical isolates of B. gladioli, were amplified, and the products were cloned into the TOPO TA vector (Invitrogen) and sequenced in both directions. The resulting nucleotide sequences were aligned by using the GCG PILEUP algorithm, yielding a consensus sequence for the B. gladioli 23S rRNA gene.
Development of a B. gladioli-specific PCR.
The consensus sequence for the B. gladioli 23S rRNA gene was compared with the published sequence for the B. cepacia 23S rRNA gene described by Hopfl et al. (6). Four regions of the 23S rRNA were putatively specific for B. gladioli. Oligonucleotides were designed at these regions to encompass the variable nucleotide(s) at the 3′ end for use in SS-PCR. These primers were named LP1, LP2, LP3, and LP4 (Table 1).
The specificities of LP1, LP2, and LP3 were tested by performing individual PCRs with a second oligonucleotide primer, 23RNA11, designed to hybridize to a conserved region of B. gladioli and B. cepacia. Initially, template DNA for the PCR was purified genomic DNA from two clinical isolates of B. gladioli and two clinical isolates of B. cepacia. Following optimization of the PCRs, each primer pair was subsequently tested, blindly, against a panel of 47 different isolates representing 17 species; at least one isolate of each of the five genomovars of the B. cepacia complex was included in this study. PCRs with the primer pair LP1–23RNA11 identified all the B. gladioli isolates; however, this primer pair also gave positive results for seven B. cepacia isolates. LP2–23RNA11 also gave positive results for all the B. gladioli isolates; however, this primer pair gave a positive result with one of the B. caryophylli isolates, two B. cepacia isolates, and three R. pickettii isolates. LP3–23RNA11 was the only primer pair that did not identify all the B. gladioli isolates, missing 2 of 19. This primer pair also gave a positive result with one isolate of B. cepacia. Interestingly, this was the B. cepacia isolate also identified by primer pairs LP1–23RNA11 and LP2–23RNA11.
To improve the sensitivity, the PCRs were repeated, again blindly, with the putatively B. gladioli-specific primer LP4 substituted for the universal primer 23RNA11. Both primer pairs LP1–LP4 and LP2–LP4 yielded identical results, identifying all of the 19 B. gladioli isolates and 1 of the B. cepacia isolates. LP3–LP4 gave negative results for the same two B. gladioli isolates that were negative with the LP3–23RNA11 primer pair and gave false-positive results for isolates (one each) of P. aeruginosa, R. solanacearum, R. pickettii, and B. cepacia. The latter isolate was PCR positive with all primer pairs tested. The results of the PCRs are shown in Table 2.
TABLE 2.
Identification of B. gladioli by PCR
| Speciesa | No. of isolates with the indicated PCR result | Identification as B. gladioli by PCR with the indicated primer pair
|
|||||
|---|---|---|---|---|---|---|---|
| 23RNA11 and:
|
LP4 and:
|
||||||
| LP1 | LP2 | LP3 | LP1 | LP2 | LP3 | ||
| B. gladioli | 15 | + | + | + | + | + | + |
| 2 | + | + | − | + | + | + | |
| 2 | + | + | − | + | + | − | |
| B. cepacia complex | 6 | − | − | − | − | − | − |
| 5 | + | − | − | − | − | − | |
| 1 | + | + | − | − | − | − | |
| 1 | + | + | + | + | + | + | |
| B. caryophylli | 1 | − | + | − | − | − | − |
| R. solanacearum | 1 | − | − | − | − | − | + |
| R. pickettii | 2 | − | + | − | − | − | − |
| 1 | − | + | − | − | − | + | |
| P. aeruginosa | 1 | − | − | − | − | − | + |
The identity of each isolate was confirmed by the BcRLR. Other isolates tested include the following: R. solanacearum, B. caryophylli ATCC 25418, P. stutzeri ATCC 17588, P. mirabilis ATCC 29906, K. pneumoniae ATCC 13883, S. maltophilia ATCC 13637, P. fluorescens ATCC 13525, A. anitratus, and M. catarrhalis ATCC 25238. All were negative in all PCRs.
To further investigate this isolate, 10 single colonies were isolated and each was subjected to PCR using both B. gladioli- and B. cepacia-specific primers (10, 15). Of these, eight were positive with the B. cepacia primers while two were positive with both the LP1–LP4 and the LP2–LP4 primer pair. No isolate was positive with both B. gladioli- and B. cepacia-specific PCR. Thus, we conclude that the original isolate was a mixed culture of these two species.
The performance of the B. gladioli SS-PCR was further assessed by the BcRLR, which blindly analyzed isolates from its collection. A total of 102 clinical isolates, consisting of 74 B. gladioli isolates and 28 other isolates from CF patients, were examined. The SS-PCR with LP1–LP4 correctly identified 70 of the B. gladioli isolates and was uniformly negative for the other isolates.
DISCUSSION
Previous attempts to distinguish B. gladioli from the closely related species B. cepacia used biochemical tests. However, these tests are inaccurate in differentiating the species but have shown how closely they resemble each other. (2, 13). The failure of such an approach, coupled with the success of molecular approaches to distinguish the five genomovars of the B. cepacia complex (10, 15), led us to examine the possibility of a molecular analysis for the identification of B. gladioli.
During the development of the protocols described in this report, Bauernfeind et al. published a molecular method for distinguishing B. gladioli from other Burkholderia spp. by SS-PCR of the 16S and 23S rRNA genes (1). This report identifies a single site showing interspecies variation in the 23S rRNA, at the 45a/b variable region. Our study identified three additional variable regions. The site identified by Bauernfeind et al. corresponds to the region of variability that primer LP3 targets. While this region has the largest number of nucleotide differences between B. gladioli and B. cepacia, in our study it turned out to be the least sensitive when analyzed in PCRs using the LP3 primer and the universal primer 23RNA11. This may be due, in part, to the high degree of variability at this site. Primers LP1 and LP2 were designed to be complementary to regions of lower variability; LP1 has two nucleotide positions specific for B. gladioli, while LP2 has only the terminal nucleotide specific for B. gladioli. Use of these primers in combination with the universal primer 23RNA11 identified species other than B. gladioli. However, the combination of the specific primers with another specific primer, LP4, improved both the sensitivity and specificity. Thus, the use of two species-specific primers may be advantageous in SS-PCR.
Strains of B. gladioli reported in previous publications were also analyzed. Nine isolates reported by Wilsher et al. in their study on the nosocomial acquisition of B. gladioli (17) were tested in PCRs with the primer pairs described above. Of the nine isolates, two were confirmed by the BcRLR to be B. gladioli. The other isolates were B. cepacia and, in one instance, P. aeruginosa. In concordance with the BcRLR findings, all the PCRs detected the two confirmed B. gladioli isolates; with the exception of one other isolate, all were uniformly negative for the other species examined. One isolate was determined by the BcRLR to be B. cepacia but appeared to be B. gladioli by the PCRs. This sample was further analyzed by B. cepacia-specific PCR as described by LiPuma et al. and Whitby et al. (10, 16). This isolate was positive with B. cepacia complex-specific PCR and was putatively ascribed to genomovar I or IV (data not shown). This was the only sample for which the PCR protocols gave a result that conflicted with the isolate identity. This sample was further analyzed by isolating 10 colonies from the parent culture (10). The results indicated that the stock culture was a mixture of both B. cepacia and B. gladioli. One isolate that was PCR positive with the B. gladioli primers and two that were positive with the B. cepacia primers were forwarded to the BcRLR. The reported identities of these isolates correlated with the PCR results, confirming an original mixed culture.
The previous misidentification of some of the organisms used in this study emphasizes the requirement for a highly discriminatory protocol for the accurate identification of bacterial isolates. Without accurate determination of an isolate's identity, it is impossible to determine the epidemiology of infection. In this instance Wilsher et al. (17) may have inadvertently reported person-to-person transmission of B. cepacia as B. gladioli. This study also demonstrates that the PCR protocols described here are an effective tool for the identification of B. gladioli.
A review of the strains forwarded to the BcRLR has indicated a high incidence of misidentified isolates originally designated B. cepacia by the clinical laboratory. Of these, approximately 50% have been shown by phenotypic analysis to be B. gladioli. By using the SS-PCR with primer pair LP1–LP4, 102 isolates were examined. The SS-PCR accurately identified 70 of 74 B. gladioli isolates and was negative for 13 R. pickettii isolates, 1 S. maltophilia isolate, 1 Alcaligenes xylosoxidans isolate, 2 Pseudomonas sp. isolates, 8 B. cepacia complex isolates, and 3 isolates of unknown identity (data not shown). Overall, the sensitivity and specificity of the LP1–LP4 SS-PCR for B. gladioli are 96 and 100%, respectively. In addition, this experiment indicates that the protocol developed for the SS-PCR can be readily performed in different laboratories.
In summary, an SS-PCR for B. gladioli has been developed. Use of this technique in combination with phenotypic analysis will facilitate the accurate identification of B. gladioli.
ACKNOWLEDGMENTS
This work was supported by Cystic Fibrosis Foundation grant STULL99GO, awarded to P.W.W., T.L.S., and J.J.L. P.W.W. and T.L.S. acknowledge the support of the Children's Medical Research Institute.
We thank Jennifer McMenamin for technical support and Kenneth Hatter for critical review of the manuscript.
REFERENCES
- 1.Bauernfeind A, Schneider I, Jungwirth R, Roller C. Discrimination of Burkholderia gladioli from other Burkholderia species detectable in cystic fibrosis patients by PCR. J Clin Microbiol. 1998;36:2748–2751. doi: 10.1128/jcm.36.9.2748-2751.1998. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Baxter I A, Lambert P A, Simpson I N. Isolation from clinical sources of Burkholderia cepacia possessing characteristics of Burkholderia gladioli. J Antimicrob Chemother. 1997;39:169–175. doi: 10.1093/jac/39.2.169. [DOI] [PubMed] [Google Scholar]
- 3.Campbell P W, III, Phillips III J A, Heidecker G J, Krishnamani M R S, Zahorchak R, Stull T L. Detection of Pseudomonas (Burkholderia) cepacia using PCR. Pediatr Pulmonol. 1995;20:44–49. doi: 10.1002/ppul.1950200109. [DOI] [PubMed] [Google Scholar]
- 4.Christenson J C, Welch D F, Mukwaya G, Muszynski M J, Weaver R E, Brenner D J. Recovery of Pseudomonas gladioli from respiratory tract specimens of patients with cystic fibrosis. J Clin Microbiol. 1989;27:270–273. doi: 10.1128/jcm.27.2.270-273.1989. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Hoare S, Cant A J. Chronic granulomatous disease presenting as severe sepsis due to Burkholderia gladioli. Clin Infect Dis. 1996;23:411. doi: 10.1093/clinids/23.2.411. [DOI] [PubMed] [Google Scholar]
- 6.Hopfl P, Ludwig W, Schleifer K H, Larsen N. The 23S ribosomal RNA higher-order structure of Pseudomonas cepacia and other prokaryotes. Eur J Biochem. 1989;185:355–364. doi: 10.1111/j.1432-1033.1989.tb15123.x. [DOI] [PubMed] [Google Scholar]
- 7.Khan S U, Gordon S M, Stillwell P C, Kirby T J, Arroliga A C. Empyema and bloodstream infection caused by Burkholderia gladioli in a patient with cystic fibrosis after lung transplantation. Pediatr Infect Dis J. 1996;15:637–639. doi: 10.1097/00006454-199607000-00020. [DOI] [PubMed] [Google Scholar]
- 8.Kiska D L, Kerr A, Jones M C, Caracciolo J A, Eskridge B, Jordan M, Miller S, Hughes D, King N, Gilligan P H. Accuracy of four commercial systems for identification of Burkholderia cepacia and other gram-negative nonfermenting bacilli recovered from patients with cystic fibrosis. J Clin Microbiol. 1996;34:886–891. doi: 10.1128/jcm.34.4.886-891.1996. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.LiPuma J J, Dulaney B J, McMenamin J D, Whitby P W, Stull T L, Coenye T, Vandamme P. Development of rRNA-based PCR assays for identification of Burkholderia cepacia complex isolates recovered from cystic fibrosis patients. J Clin Microbiol. 1999;37:3167–3170. doi: 10.1128/jcm.37.10.3167-3170.1999. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.LiPuma J J, Dulaney B J, McMenamin J D, Whitby P W, Stull T L, Coenye T, Vandamme P. Polymerase chain reaction identification of species within the Burkholderia cepacia complex. Pediatr Pulmonol. 1998;15(Suppl.):S306. . (Abstract 367.) [Google Scholar]
- 11.McCulloch L. A bacterial disease of gladiolus. Science. 1921;54:115–116. doi: 10.1126/science.54.1388.115. [DOI] [PubMed] [Google Scholar]
- 12.Ross J P, Holland S M, Gill V J, DeCarlo E S, Gallin J I. Severe Burkholderia (Pseudomonas) gladioli infection in chronic granulomatous disease: report of two successfully treated cases. Clin Infect Dis. 1995;21:1291–1293. doi: 10.1093/clinids/21.5.1291. [DOI] [PubMed] [Google Scholar]
- 13.Simpson I N, Finlay J, Winstanley D J, Dewhurst N, Nelson J W, Butler S L, Govan J R W. Multi-resistance isolates possessing characteristics of both Burkholderia (Pseudomonas) cepacia and Burkholderia gladioli from patients with cystic fibrosis. J Antimicrob Chemother. 1994;34:353–361. doi: 10.1093/jac/34.3.353. [DOI] [PubMed] [Google Scholar]
- 14.Vandamme P, Holmes B, Vancanneyt M, Coenye T, Hoste B, Coopman R, Revets H, Lauwers S, Gillis M, Kersters K, Govan J R W. Occurrence of multiple genomovars of Burkholderia cepacia in cystic fibrosis patients and proposal of Burkholderia multivorans sp. nov. Int J Syst Bacteriol. 1997;47:1188–1200. doi: 10.1099/00207713-47-4-1188. [DOI] [PubMed] [Google Scholar]
- 15.Whitby P W, Carter K B, Coenye T, LiPuma J J, Vandamme P, Stull T L. Identification of Burkholderia cepacia genomovars I, III and IV by nested polymerase chain reaction. Pediatr Pulmonol. 1998;15(Suppl.):S324. . (Abstract 429.) [Google Scholar]
- 16.Whitby P W, Dick H L, Campbell P W, Tullis D E, Matlow A, Stull T L. Comparison of culture and PCR for detection of Burkholderia cepacia in sputum samples of patients with cystic fibrosis. J Clin Microbiol. 1998;36:1642–1645. doi: 10.1128/jcm.36.6.1642-1645.1998. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Wilsher M L, Kolbe J, Morris A J, Welch D F. Nosocomial acquisition of Burkholderia gladioli in patients with cystic fibrosis. Am J Respir Crit Care Med. 1997;155:1436–1440. doi: 10.1164/ajrccm.155.4.9105090. [DOI] [PubMed] [Google Scholar]