Diagnosis of Pneumococcal Pneumonia by psaA PCR Analysis of Lung Aspirates from Adult Patients in Kenya

Study setting and populations.Lung puncture specimens were obtained in a prospective etiological study of acute pneumonia in adults in Kenya (25). Patients who presented to either Coast Province General Hospital, Mombasa, Kenya, or Kilifi District Hospital, Kilifi, Kenya, with radiographic evidence of pneumonia and a history of respiratory illness not exceeding 2 weeks were investigated with the following diagnostic tests: blood cultures, lung aspirate cultures, mycobacterial cultures of sputum and lung aspirate fluid, serotype-specific antigen detection for capsular polysaccharide of S. pneumoniae in urine, and serology for viral pneumonia, Chlamydia pneumoniae, Mycoplasma pneumoniae, and Legionella pneumophila serotype 1 (25). Two hundred fifty-nine patients underwent lung aspiration; from the 89th patient onwards a small aliquot of lung aspirate fluid was set aside for PCR and stored at −70°C. The ethics committees of the London School of Hygiene and Tropical Medicine and of Kenya Medical Research Institute (KEMRI), Nairobi, Kenya, approved the study in accordance with KEMRI guidelines for clinical research.

As it is not possible to obtain control aspirates from healthy humans, lung puncture was performed on 10 C57BL6/J mice that formed part of a separate experiment on nasopharyngeal colonization. Five mice were colonized in the nasopharynx with S. pneumoniae serotype 6B, and five were not colonized.

Clinical methods.Lung puncture was performed on patients with peripheral consolidation by using a 35-mm-long 21-gauge needle attached to a 10-ml syringe. All lung punctures were performed within 24 h, and almost all were performed within 4 h of presentation. The area of diseased lung was localized by clinical signs and by two-plane radiography, and the overlying skin was sterilized with alcohol. During suspended respiration, the needle was passed rapidly into the lung parenchyma through an intercostal space, 0.5 ml of isotonic saline was injected, and the needle was withdrawn over 1 to 2 s under suction. The specimen was distributed to slides for Gram and Ziehl-Neelsen stains and broth cultures. Any residuum in the tip of the needle was expelled into a sterile plastic storage vial and frozen at −70°C for up to 4 years. The specimen volume saved was usually <50 μl (range, 20 to 100 μl). All lung punctures were performed by the same operator. To control for the sterility of the immediate operating environment in Kenya, a similar volume of isotonic saline was expelled from the puncture needle immediately prior to lung aspiration and saved in similar vials. In mice, lung aspirates were performed immediately postmortem with a 25-gauge needle and 100 μl of sterile isotonic saline.

Microbiological methods.Lung aspirates and blood samples were cultured in brain heart infusion broth for 7 days and subcultured on 5% horse blood agar on days 2 and 7 or when indicated by broth turbidity. Colonies of S. pneumoniae were identified by optochin sensitivity and by a positive Quellung reaction with antisera from the Danish checkerboard typing system (30). Pneumococcal pneumonia was also diagnosed from urine specimens by a serotype-specific latex agglutination assay for serotypes 1, 4, 5, 6, 7, 9, 12, 14, 19, and 22; the assay had a sensitivity of 0.46 and a specificity of 0.98 (28).

Urine collected before the administration of treatment in the hospital was also assayed for antimicrobial activity (35). Sterile 6-mm-diameter filter paper disks were placed at the center of a Columbia agar plate evenly inoculated with a fully susceptible Staphylococcus aureus (NCTC 6571). A 20-μl aliquot of urine was dropped onto the disk, and the inhibition zone surrounding the disk was read after overnight incubation.

Control strains.The positive control for psaA PCR was S. pneumoniae serotype 6B (ATCC 51937). Reference strains of Haemophilus influenzae were obtained from J. Elliott (Streptococcal Reference Laboratory, Centers for Disease Control and Prevention [CDC], Atlanta, Ga.) and included 2 isolates of serotype b (one each of biotypes I and II), 1 isolate of serotype f (biotype I), and 5 nontypeable strains (4 of biotype II and 1 of biotype III).

PCR.DNA was extracted from lung aspirates with the QIAamp blood kit protocol (QIAGEN, Chatsworth, Calif.) with modifications that included an overnight digestion with proteinase K (Boehringer-Mannheim, Indianapolis, Ind.) at 42°C and ethanol precipitation of the eluate from the QIAamp spin columns (22). DNA was resuspended in 50 μl of 10 mM Tris-HCl (pH 9.0) (22).

Primers 8229.p (5′-CTTTCTGCAATCATTCTTG-3′) and 6496.n (5′-GCCTTCTTTACCTTGTTCTGC-3′), which define an 838-bp fragment of the psaA gene of S. pneumoniae (GenBank accession no. U53509 ), corresponding to positions 1784 and 2624, respectively, were obtained from J. Sampson (Respiratory Diseases Immunology Laboratory, CDC). PCR was conducted by using the method of Morrison et al. (15); briefly, PCR amplification was performed with 5 μl of extracted DNA in a 50-μl reaction mixture containing 50 mM KCl, 10 mM Tris-HCl (pH 8.3), 1.5 mM MgCl₂, 200 μM concentrations (each) of deoxynucleoside triphosphates (dATP, dCTP, dGTP, and dTTP), 1.5 μM concentrations of each primer, and 1.0 U of AmpliTaq polymerase (Perkin-Elmer, Foster City, Calif.). The thermocycler conditions used were as follows: 95°C for 1 min; followed by 35 amplification cycles of 95°C for 30 s, 52°C for 30 s, and 72°C for 2 min; and held finally at 72°C for 8 min. All PCRs were performed with a Cetus 480 thermocycler (Perkin-Elmer Cetus, Norwalk, Conn.). Ten microliters of each amplification reaction was analyzed by electrophoresis on a 2.0% agarose gel in Tris-borate-EDTA buffer. Gels were stained with ethidium bromide and visualized by UV fluorescence.

Sequencing.PCR products were purified from the post-PCR mix with the QIAquick PCR purification kit (Qiagen) and sequenced in both directions by using the dideoxy-sequencing Prism ready reaction rhodamine dye-terminator kit (Applied Biosystems Incorporated, Foster City, Calif.) with a Cetus 9700 thermocycler (Perkin-Elmer Cetus). The sequencing reactions were resolved with a 4.25% polyacrylamide gel at 51°C with constant voltage (1,500 V) by using an ABI Prism model 373 autosequencer.

Synthetic internal control.An internal control for the psaA primers in this PCR was constructed by inserting the target DNA sequence into a DNA plasmid. The psaA primer sequences were retained in the insert, but the body of the target sequence was replaced with a shorter length of foreign DNA, truncating the resulting amplicon and offering an easy means of identification. The psaA gene was first ligated from positions 1784 to 2624. Using restriction enzymes AciI and HinfI (New England Biolabs, Beverley, Mass.), the sequence was cut at positions 29 and 799, respectively. The fragments were resolved in a 1.5% Tris-acetate-EDTA agarose gel and isolated with GeneClean II (Bio101, La Jolla, Calif.). A portion of the gltA gene from Bartonella henselae was amplified by using primers BhCS781.p (5′-GGGGACCAGCTCATGGTGG-3′) and BhCS1137.n (5′-AATGCAAAAAGAACAGTAAAC-3′) (16). The gltA amplicons and purified digest construct were blunted with T4 polymerase and ligated by using standard molecular biology techniques (22). The construct was transformed into competent JM109 Escherichia coli (Promega, Madison, Wis.) and selected on ampicillin-X-Gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside)-IPTG (isopropyl-β-d-thiogalactopyranoside) plates. Plasmids were harvested from 150-ml overnight cultures by using the QIAGEN plasmid midi kit (Qiagen).

Limit of detection of psaA PCR.First, the concentrations of stock solutions of the internal control were estimated by determining the optical density at 260 nm of serial dilutions in triplicate. Next, the dilutions were amplified in duplicate simultaneous runs with the conditions outlined above for psaA PCR. Standard stoichiometric calculations based upon the optical densities at 260 nm of the dilutions were used to determine the minimum number of plasmids required to produce a positive amplicon.

To determine the limit of detection in terms of viable CFU, psaA PCR analysis was performed on DNA extracts from serial 10-fold dilutions of a 4- to 12-h Todd-Hewitt broth culture and compared with colony counts of a 10-μl aliquot of each dilution cultured overnight on 5% sheep blood agar. Previous investigators have reported a detection limit of <1 CFU (37), but this is most likely to be attributable to DNA from nonviable organisms. To minimize this problem, the experiment was repeated with progressively shorter periods of broth incubation until the detection limit estimate stabilized.

Control of DNA extraction.Human β-actin was selected as a control gene to monitor specimen processing and DNA extraction. Primers GH20.p (5′-GAAGAGCCAAGGACAGGTAC-3′) (GenBank accession no. A26624 ) and PC04.n (5′-CAACTTCATCCACGTTCACC-3′) (GenBank accession no. A26623 ), which define a 267-bp fragment of the human β-actin gene (GenBank accession no. L26469 ), were obtained from the Biotechnology Core Facility (CDC). PCR amplification was performed with 5 μl of extracted DNA in a 50-μl reaction mixture containing 50 mM KCl, 10 mM Tris-HCl (pH 8.3), 1.5 mM MgCl₂, 200 μM concentrations (each) of deoxynucleoside triphosphates (dATP, dCTP, dGTP, and dTTP), 1.5 μM concentrations of each primer, and 0.2 U of thermostable AmpliTaq DNA polymerase (Perkin-Elmer). The thermocycler conditions used were as follows: 95°C for 1 min; followed by 35 amplification cycles of 95°C for 30 s, 58°C for 30 s, and 72°C for 2 min; and held finally at 72°C for 5 min.

Sensitivity.To derive a sensitivity estimate for psaA PCR analysis, the appropriate specimen group in Table 1 is lung aspirate specimens from which S. pneumoniae has already been obtained by culture. Lung aspirate fluid from patients diagnosed by other techniques may not necessarily contain any pneumococcal DNA. The sensitivity of psaA-based PCR, 0.83, is similar to that of other diagnostic PCR assays which use pneumococcal gene primers (1, 4, 7, 8, 11-13, 17, 18, 21, 32, 33, 37). Table 4 illustrates the small size of the groups in which true sensitivity estimates have so far been obtained; in fact, the study group reported here provides the most precise estimate yet. Particularly striking is the contrast between the small numbers of observations and the relatively large numbers of test variables (target gene, product size, PCR design, specimen examined, and study population). The weighted mean sensitivity of all studies is 0.84, but there is significant heterogeneity among the different estimates (χ² test, P = 0.001).

Given that, under ideal conditions, PCR detects as little as one copy of target DNA, why is sensitivity of pneumococcal PCR so poor in clinical evaluations? There are several potential explanations. Taq polymerase is highly sensitive to porphyrin inhibitors that are generated from the breakdown of hemoglobin (9). Lung aspirate samples may also contain blood either because of traumatic puncture of a small vessel or simply reflecting the natural invasion of a diseased lung by red blood cells during the pathological phase of pneumonia known as red hepatization. Proteolysis, DNA extraction, and ethanol precipitation should eliminate most inhibitors, but when we evaluated inhibition with a synthetic internal control, one in six of the extracted DNA samples would not initially support PCR.

Southern hybridization can extend the limit of detection of PCR (for the pneumolysin gene) 10-fold and may increase sensitivity (32). Theoretically, nested PCR will also increase sensitivity and specificity, though experience with this is variable and it is prone to contamination problems. Salo and colleagues demonstrated a 1,000-fold gain in sensitivity with nested PCR, but Toikka and colleagues, using the same primer pair, found no additional sensitivity advantage to the second (nested) reaction (21, 32). Selecting the optimal specimens for PCR might also maximize sensitivity; however, studies of humans and bacteremic mice have shown little practical difference in the sensitivity of PCR in whole blood, buffy coat samples, or serum (17, 20, 32).

Although the sensitivity of the psaA PCR was 0.83, the assay detected only 51% of all episodes of pneumococcal pneumonia. This is due to the sensitivity limitations of lung aspiration, sample storage and transport, and DNA extraction and purification. The lung aspirate needle may miss the area of consolidation and retrieve only saline and air (26); the size of the specimen is very small compared, for example, to blood aliquots used for culture and may not contain sufficient DNA. In the present study, storage and transportation of the specimen may have led to degeneration of DNA, and finally, in the process of extraction and purification, the small mass of DNA present may be lost. As lung aspirates should also contain human leucocytes and alveolar cells, we made extracted human DNA the control for these factors by using the β-actin gene as a PCR target; after sampling, storage, transport, processing, and DNA extraction, 30% of the samples taken from patients with negative psaA PCR showed no evidence of human DNA with β-actin-specific primers.

Specificity.The appropriate control material to evaluate the specificity of PCR in lung aspirates is lung aspirate fluid from sick patients without pneumonia. This is normally only available from healthy controls (14, 34) or, more practically, from patients in whom pneumonia has been diagnosed initially and a different diagnosis arrived at subsequently (36); this is an uncommon eventuality and we were not able to find appropriate specimens in Kenya. psaA PCR analysis was not positive in lung aspirate material from mice, even in those who had experimental nasal colonization with S. pneumoniae. In humans, specificity estimates have also been derived by (i) evaluation of culture-negative lung aspirates (7) or (ii) evaluation of lung aspirates that have yielded nonpneumococcal pathogens on culture (18). However, these strategies are undermined by the fact that culture is insensitive for detecting pneumococci and that isolation of two organisms from one aspirate is not uncommon (2, 7, 26, 29); the absence of pneumococci on culture does not prove that they were absent from the original sample, even if an alternative pathogen was cultured.

Despite this, lung aspirate fluid is likely to have a very high specificity as a specimen because it is uncontaminated by upper respiratory tract secretions. Several findings from this report support this contention. First, 73% (19 of 26) of the patients with positive psaA PCR results but negative lung aspirate cultures had alternative evidence of pneumococcal pneumonia from blood cultures or antigen detection. Second, knowing that culture techniques are sensitive to the presence of antibiotics and that psaA PCR analysis is not, we would predict that among patients whose cultures were negative but whose PCR results were positive there would be a concentration of individuals with evidence of antibiotics. In fact 79% of this group had evidence of antimicrobial activity in urine against a background prevalence of 54%.

The specificity of PCR for pneumococcal disease in previous studies is shown in Table 5 (1, 4, 11-13, 21, 32). Among adults, specificity has been uniformly high with a weighted mean of 0.97 and little heterogeneity between studies (Fisher exact test, P = 0.98). In the study with the lowest estimate, 6 of 100 samples from healthy elderly persons were PCR positive; the investigators speculated that these false-positive results were due to laboratory contamination as the serum vials had been opened for use in previous assays (17). Among children in Israel, 20% of sera were positive by pneumolysin PCR and the distribution of false-positive results with age closely followed the prevalence of nasopharyngeal carriage with age, suggesting that intermittent transient bloodstream invasion occurs in colonized children leading to false-positive PCR results (4). For children, PCR of blood clearly lacks specificity; PCR of lung aspirate fluid may yield more-specific results, though it remains to be tested.

This study has demonstrated that psaA PCR analysis can amplify pneumococcal DNA in lung aspirates from adults with pneumococcal pneumonia in Kenya, that the sequence of this PCR product is identical to that of the psaA gene target, and that the limit of detection is appropriate for the diagnosis of S. pneumoniae in clinical specimens. The sensitivity in culture-positive lung aspirates is 0.83, and the combination of psaA-based PCR analysis and culture of lung aspirates diagnosed nearly twice as many patients as did culture of lung aspirates alone.