|Home | About | Journals | Submit | Contact Us | Français|
Community-acquired pneumonia (CAP) and sepsis are important causes of morbidity and mortality. We describe the development of two molecular assays for the detection of 11 common viral and bacterial agents of CAP and sepsis: influenza virus A, influenza virus B, respiratory syncytial virus A (RSV A), RSV B, Mycoplasma pneumoniae, Chlamydophila pneumoniae, Legionella pneumophila, Legionella micdadei, Bordetella pertussis, Staphylococcus aureus, and Streptococcus pneumoniae. Further, we report the prevalence of carriage of these pathogens in respiratory, skin, and serum specimens from 243 asymptomatic children and adults. The detection of pathogens was done using both a manual enzyme hybridization assay and an automated electronic microarray following reverse transcription and PCR amplification. The analytical sensitivities ranged between 0.01 and 100 50% tissue culture infective doses, cells, or CFU per ml for both detection methods. Analytical specificity testing demonstrated no significant cross-reactivity among 19 other common respiratory organisms. One hundred spiked “surrogate” clinical specimens were all correctly identified with 100% specificity (95% confidence interval, 100%). Overall, 28 (21.7%) of 129 nasopharyngeal specimens, 11 of 100 skin specimens, and 2 of 100 serum specimens from asymptomatic subjects tested positive for one or more pathogens, with S. pneumoniae and S. aureus giving 89% of the positive results. Our data suggest that asymptomatic carriage makes the use of molecular assays problematic for the detection of S. pneumoniae or S. aureus in upper respiratory tract secretions; however, the specimens tested showed virtually no carriage of the other nine viral and bacterial pathogens, and the detection of these pathogens should not be a significant diagnostic problem. In addition, slightly less sensitive molecular assays may have better correlation with clinical disease in the case of CAP.
Community-acquired pneumonia (CAP) and sepsis are important causes of morbidity and mortality worldwide, including the United States (1, 8, 28, 33). Streptococcus pneumoniae remains the most common bacterial cause of pneumonia in all age groups and is closely followed by Mycoplasma pneumoniae, Chlamydophila pneumoniae, Haemophilus influenzae, Legionella pneumophila, and Staphylococcus aureus (28). The most common viral causes of CAP include influenza viruses A and B and respiratory syncytial virus A (RSV A) and RSV B, followed by parainfluenza virus types 1, 2, and 3, human metapneumovirus, and adenovirus, with many other viruses, such as enteroviruses, rhinoviruses, coronaviruses, and bocavirus, playing a less defined role (18). S. aureus and S. pneumoniae are two of the most common causes of sepsis (1). Despite significant advances in diagnostics, there remain no good diagnostic tests for CAP and sepsis, particularly tests targeting both bacterial and viral agents. Currently approved diagnostic tests for CAP and sepsis (blood culture, urinary antigen tests, Gram staining, culture of sputum, and serology) are associated with significant limitations (28). An increasing number of studies have been published evaluating the use of molecular techniques, most commonly PCR, as diagnostic tools for patients with pneumococcal (27, 31) and staphylococcal (37) invasive infections, as well as atypical pneumonias caused by C. pneumoniae, M. pneumoniae, and L. pneumophila (25, 42) and common community-acquired viruses (10, 11, 19, 24, 36).
At present, no multiplex PCR assay detecting both viruses and bacteria has been approved by the U.S. Food and Drug Administration or is commercially available for clinical use for patients with CAP or sepsis. In addition, molecular testing for this patient population is not standardized, with extensive variations in the methodologies used among studies. Another limitation of such investigations using molecular techniques (and some conventional techniques) has been the detection of some of these pathogens in both respiratory specimens and sera from healthy controls (27, 31), as has been shown previously for S. pneumoniae, the most common etiological agent of bacterial CAP. This finding has been attributed to the carriage of or colonization by bacteria in the upper respiratory tract in healthy subjects, with subsequent presumed intermittent seeding into the blood and resultant detection in sera. The detection of bacterial nucleic acid (NA) in blood specimens may additionally be due to the contamination of these specimens with bacteria either transiently contaminating or more permanently colonizing the skin. This detection of bacterial NA in both healthy subjects and asymptomatic patients has made the interpretation of test results difficult.
Nasopharyngeal (NP) carriage of several bacterial species has been extensively described globally, with widely varying prevalence rates depending on factors including, but not limited to, the diagnostic methodologies employed and the specimen types (9, 22, 38). Carriage rates reported vary from 5 to 90% for S. pneumoniae, depending on the age and geographic location of the population studied and the presence or absence of risk factors (38). S. aureus has been reported to colonize up to 32 to 36% of individuals in healthy populations in the United States (6, 14). Asymptomatic colonization by C. pneumoniae in both adults and children, with the prevalence varying from 2 to 5%, has also been described previously (9, 22). Whether or not asymptomatic carriage of Legionella species occurs remains unknown (39). One study using immunofluorescence methods has previously reported colonization with L. pneumophila in 6% of adults, although none of the results could be confirmed by culture (3). No data exist for asymptomatic infections with Legionella micdadei.
Asymptomatic carriage of or persistent infection with the community-acquired respiratory viruses that cause the majority of viral CAP occurs in immunocompromised individuals and those with chronic respiratory medical conditions (5, 12, 16, 43). Such individuals have been found to be virus positive for many months. Also, individuals have been found to be positive several days prior to the occurrence of symptoms and after the resolution of symptoms too. Healthy children produce much higher viral titers than adults, and NA can persist in children for typically 2 to 3 weeks after the start of infection, while NA persistence in adults rarely goes past 7 days. It is thought that most healthy children and adults do not carry these viruses asymptomatically and that most molecular diagnostic assays do not detect NA from these pathogens in healthy individuals outside of the parameters just described (4). However, mildly symptomatic infections are very common, and many may be reported as asymptomatic or attributed to allergies, causing some diagnostic confusion if careful attention to signs and symptoms is not given.
Although several studies have reported the detection of bacterial DNA in blood samples or sera from healthy subjects and explained such detection as being secondary to carriage, contamination, or simply a reservoir of bacterial DNA in human blood (31, 35), no systematic study that has evaluated the carriage of S. pneumoniae and other common bacteria in the blood or sera of healthy subjects has been published. Knowledge of the background rates of detection of pathogens associated with carriage in various body sites will be necessary and useful for any molecular assays developed for testing patients with invasive infections.
We describe the development of a sensitive and specific multiplex reverse transcription-PCR (RT-PCR) assay targeting 11 common bacterial (S. pneumoniae, C. pneumoniae, M. pneumoniae, S. aureus, L. pneumophila, L. micdadei, and Bordetella pertussis) and viral (influenza virus A and B and RSV A and B) pathogens causing CAP and sepsis by using two different detection methods with different analytical limits of detection (LODs), one manual (the enzyme hybridization assay [EHA]) and one automated (the electronic microarray [eMA]). In addition, we determined the rates of carriage of these pathogens at various sites and in individuals of different ages, as age is an issue that has been previously speculated to impact the clinical use of such molecular diagnostic assays.
Children and adults presenting to Children's Hospital of Wisconsin and Froedtert Hospital, Milwaukee, with a variety of illnesses or injuries were invited to participate in the study. This study was approved by the Institutional Review Board of Children's Hospital of Wisconsin and the Medical College of Wisconsin. Informed consent was obtained from adults and from the parent or legal guardian of each child enrolled in the study, and assent was obtained from all children 7 to 18 years old. Subjects for the study were enrolled from October 2005 to August 2007.
NP swabs, skin swabs, and serum specimens were collected for the study. NP and serum specimens were obtained from patients with no signs or symptoms of an infectious disease, and skin swab specimens were obtained from subjects who had no signs or symptoms of a dermatologic illness but who may or may not have had a respiratory illness. NP specimens were obtained using MicroTest multimicrobe collection and transport systems (Remel, Lenexa, KS) by passing the swab through the nostril until posterior pharyngeal resistance was felt and were placed in 4 ml of M4 transport medium (MicroTest viral transport medium; Remel, Lenexa, KS) in 15-ml Falcon tubes. Skin swabs were obtained by rubbing the dry swab firmly over a 2-in.-long by 1-in.-wide area on the surface of the forearm skin and were immediately placed into 4 ml of M4 transport medium. Sera were obtained in Vacutainer serum tubes (BD, Franklin Lakes, NJ). Specimens were transported to the lab on wet ice and stored at 4°C until further processing. All specimens were processed within 24 h of collection, at which time NP swabs were centrifuged at 580 × g for 3 min and blood samples were centrifuged at 2,030 × g for 10 min. The supernatant or serum was removed and stored at −80°C until further testing.
Aliquots of 400 μl of specimens were processed with a Mini-Beadbeater (Biospec, Bartlesville, OK) to facilitate cell lysis and disruption. Total NA was subsequently extracted using a MagNA Pure Compact system with MagNA Pure Compact NA isolation kit 1 according to the instructions of the manufacturer (Roche Applied Science, Mannheim, Germany) and eluted into a volume of 50 μl. cDNA was synthesized using random hexamers and murine leukemia virus reverse transcriptase with a 20-μl NA input in the 96-well GeneAmp PCR system 9700 (Applied Biosystems, Foster City, CA). Reaction conditions for RT were as follows: 22°C for 5 min, 42°C for 14 min, and 95°C for 1 min. The resultant cDNA was used for both the EHA and eMA detection methods.
Primers and probes were designed for the detection of seven bacterial and four viral species in the assay. This design was carried out by performing alignments of sequences of each pathogen obtained from GenBank and selecting highly conserved regions of genetic sequences for primer and probe design. The in silico coverage rates of most of the primers and probes used in the assay ranged between 98.5 and 100%, except for L. micdadei, for which the coverage rate was 90.9%, and RSV B, for which it was 50%. Although the in silico coverage rates of the RSV B primers and probes was low, these oligonucleotide sets have been tested by us and others on thousands of clinical samples and have been able to correctly identify >95% of RSV B-positive samples (11, 20, 24). The sequences of the primer pairs and probes used in the EHA are listed in Table Table1.1. For the eMA, the same primers shown in Table Table11 were used as in the EHA except that the reverse primers were not biotin labeled.
PCR amplification was carried out with 70-μl reaction volumes containing 30 μl of cDNA input, 5 U of FastStart Taq DNA polymerase (Roche Diagnostics, Indianapolis, IN), 1× PCR buffer II (Applied Biosystems, Foster City, CA), 3.5 mM MgCl2, 2.4 mM deoxynucleoside triphosphate mix, and 0.25 μM (each) primers. Each reaction mixture contained 11.5 pairs of primers corresponding to seven bacterial and four viral species and one internal control. After denaturation at 95°C for 10 min, reaction mixtures were subjected to amplification in 2 cycles at 95°C for 1 min, 55°C for 30 s, and 72°C for 45 s and 38 cycles at 95°C for 15 s, 60°C for 15 s, and 72°C for 30 s and were then held at 72°C for 3 min. Postamplification, PCR products were purified with the QIAquick PCR purification kit (Qiagen, Valencia, CA), and 5-μl aliquots of the PCR products were added to 96-well NeutrAvidin-coated microtiter plates (Pierce, Rockford, IL). Detection was carried out with horseradish peroxidase-labeled probe solutions specific for each pathogen and the internal control (bacteriophage MS2), as described previously (11). The positive cutoff value was calculated to be three times higher than the background signal observed in the negative control, or an optical density (OD) of ≥0.4. Values between 0.3 and 0.4 were considered indeterminate. ODs of <0.3 were considered to indicate negativity for the target NA.
PCR amplification was carried out with 50-μl reaction volumes as described previously (17). PCR mixtures contained 30 μl of cDNA input and concentrations of primers, MgCl2, and deoxynucleoside triphosphates equivalent to those in the multiplex PCR EHA. The reaction cycling conditions were the same as those described above for the EHA. Following amplification, 6 μl of the amplicon NA was mixed with 44 μl of CAPdown sample buffer A (114 mM l-histidine, 142.5 mM 1-thioglycol) in an onboard automated dilution apparatus by the NanoChip 400 eMA instrument. No post-PCR processing of the amplicon was required. The automated detection of study pathogens on the eMA was carried out in three distinct steps: (i) capture oligonucleotide addressing, (ii) amplicon addressing and hybridization, and (iii) reporting. Sequences of the capture and discriminator oligonucleotides used for eMA detection are listed in Table Table2.2. By using the eMA platform with a six-pad capture down assay, two Legionella species and RSV A and RSV B were detected as single analytes. Each 100-μl capture mix containing a combination of two capture oligonucleotides at 100 nM each in a solution of 50 mM histidine and 0.05% ProClin 300 was addressed to the detection pad on the cartridge. The positive cutoffs in the data analysis program were set to use a minimum background of 300 relative fluorescence units and a minimum signal-to-background ratio of >3.0.
Whole bacterial or viral strains used as positive controls were purchased from the American Type Culture Collection (ATCC), Manassas, VA. Medium-level positive controls (viral load, 102 or 103 50% tissue culture infective doses/ml; bacterial load, 101 to 103 cells/ml) and low-level positive controls (viral load, 101 or 102 50% tissue culture infective doses/ml; bacterial load, 1 to 102 cells/ml) were prepared in M4 medium. Medium- and low-level positive controls and a negative control were included in every amplification run. In addition, detection controls, which were previously amplified and detected amplicons at high concentrations, were included at the detection step in the EHA to identify any problems related to the amplicon-probe hybridization process, such as probe degradation or suboptimal hybridization with the probe due to any experimental condition affecting the process.
Analytical sensitivities and specificities were determined using quantitated control organisms purchased from the ATCC.
Analytical sensitivities were determined by making 10-fold serial dilutions of pathogens, extracting NA from the dilutions, and testing with the multiplex PCR and both detection methods. The highest dilution of the pathogen that resulted in a positive test result was considered to be the LOD of the assay for that pathogen.
Analytical specificities were determined using a representative panel of respiratory viruses and bacteria commonly found in the human upper respiratory tract or on the skin. NAs were extracted from Staphylococcus epidermidis ATCC 29641, Streptococcus pyogenes ATCC 19615, Streptococcus mitis ATCC 49456, Streptococcus sanguinis ATCC 10556, Streptococcus anginosus (“Streptococcus milleri”) ATCC 33397, Enterococcus faecalis ATCC 12399, H. influenzae ATCC 10211, Escherichia coli ATCC 11229, Eikenella corrodens ATCC 51724, Bacteroides fragilis ATCC 29762, Peptostreptococcus anaerobius ATCC 27337, Mycobacterium intracellulare ATCC 19077, Neisseria meningitidis ATCC 51287, cytomegalovirus VR-977, herpes simplex virus type 1 VR-1545, enterovirus type 68 VR-1076, and varicella-zoster virus ATCC 49265 and tested with multiplex PCR.
Groups of 20 clinical specimens (12 NP swabs, 5 skin swabs, and 3 serum samples), previously known to be negative for the study pathogens, were each spiked with one of five control organisms, influenza virus A, RSV A, M. pneumoniae, S. aureus, and S. pneumoniae. NA was extracted and tested as before, to rule out the inhibition of PCR by inhibitory substances present in clinical specimens.
Specimens which were positive by either the EHA or the eMA but not by both were reanalyzed using the following strategy.
For those specimens for which original amplicon NA from the initial PCR amplification was left, we repeated the EHA detection step using the original amplicon with freshly made probes. If the repeat EHA testing resulted in an OD reading of ≥0.4, the specimen was counted as a true positive. Previously described work has demonstrated that the EHA detection method is more sensitive than the eMA (17). If the OD was 0.3 to 0.4 (indeterminate), the specimen was further analyzed as described below.
A singleplex RT-PCR assay was performed for the specimens with discrepant results that were shown to be positive upon retesting with the EHA in step 1, with the intent of confirming these results with a second assay. This retesting was done by either using originally extracted NA or reextracting NA from clinical specimens, and detection was performed using both the Agilent assay (see step 3) and the EHA.
For those specimens for which the results of the repeat EHA in step 1 were indeterminate but for which original amplicon NA remained, further testing was done by gel analysis of amplicons on the bioanalyzer 2100 (Agilent Technologies, Palo Alto, CA) using the Agilent DNA 1000 kit according to the manufacturer's instructions. A reading of >1 ng/μl and a band in the appropriate size position were considered to indicate a positive result.
For those specimens for which the original amplicon was not available for repeat testing with the EHA (in step 1) or whose analysis in step 3 revealed a new pathogen not detected in the initial EHA or eMA run, a singleplex RT-PCR assay was performed using originally extracted NA with primers targeting only the pathogen in question to obtain sufficient amplicon NA for further analysis by concurrently performed EHA and Agilent assay runs. Criteria for positives and negatives in these EHA and Agilent assay runs remained the same as before.
The LODs of the RT-PCR-EHA and the RT-PCR-eMA assay are shown in Table Table3.3. Analytical specificities were determined by testing 19 common pathogens (bacterial and viral) that can be found in the respiratory tract for cross-reactivity in the assay. Specificity testing did not reveal any false positives, by either EHA or eMA detection. E. coli, N. meningitidis, and M. intracellulare had minor cross-reactivity with the LM probe but did not test positive by the PCR-EHA even at 103 to 104 CFU/ml and showed no cross-reactivity at lower concentrations. The ODs for these cross-reactions were 0.3, 0.3, and 0.2, respectively, for the three pathogens in the RT-PCR-EHA. Positive test results for the RT-PCR-EHA were ODs of >0.4, and those for the RT-PCR-eMA assay were signal/noise ratios of >3.0. We did not detect any cross-reactivity among the 11 study pathogens by either of the two detection methods in the study.
For each of five pathogens, 20 specimens (12 NP swabs, 5 skin swabs, and 3 serum samples per pathogen, for a total of 100 spiked specimens) were spiked, and no evidence of the inhibition of amplification or a nonspecific signal was observed in any sample, yielding surrogate sensitivities and specificities of 100% as shown in Table Table4.4. The 100 spiked clinical specimens were concurrently tested for false-positive reactions (specificity) with the remaining six targets (with which the clinical specimens were not spiked). None of the 600 reactions showed any evidence of false-positive reactions among the assay targets.
A total of 329 specimens, of which 129 were NP specimens, 100 were skin swabs, and 100 were serum specimens, were collected from 243 subjects in the study. Each of the 329 specimens was tested by both RT-PCR-EHA and eMA for all 11 pathogens. Thus, each detection method tested for 3,619 separate analytes in the clinical samples.
A total of 48 (14.5%) of 329 specimens obtained in the study showed evidence of carriage by one or both technologies. These 48 specimens were positive for 56 pathogens, due to dual (n = 2) or triple (n = 3) positives for some specimens. Of these 56 positive results, 24 were positive by both detection technologies and 32 were detected by only one of the two detection technologies used in the study, 30 by the EHA alone and 2 by microarray alone. Therefore, the RT-PCR-EHA detected a significantly higher level of carriage (54 of 56 carriage positives) than the RT-PCR-eMA (26 of 56 carriage positives).
After a discrepancy analysis performed as described in Materials and Methods, 21 of 30 EHA-positive discrepant results were determined to be true positives, 2 of 30 were indeterminate, and 7 of 30 were possible false positives (indicating that 7 of 3,573 true negatives were falsely positive by the EHA). Of the 2 eMA-positive discrepant results, 1 was confirmed to be a true positive (and thus a false negative for the EHA) and 1 was a possible false positive (indicating that 1 of 3,573 true negatives were falsely positive by eMA). In addition, 10 of the 22 discrepant results that we determined to be true positives were confirmed to be positive by Agilent detection as well, despite the lower sensitivity of this technology than of the EHA. Therefore, 41 of the 48 initially positive specimens were confirmed to be positive for a total of 46 pathogens (24 detected by both methods and 22 detected by the EHA). Based on these results, the true positives for each pathogen by the age groups of the subjects are displayed in Table Table5,5, and a descriptive analysis of the postdiscrepancy data is presented below.
As indicated in Table Table5,5, a total of 28 (21.7%) of 129 NP specimens, 11 of 100 skin specimens, and 2 of 100 serum specimens tested positive for one or more pathogens.
S. pneumoniae was detected in 18 (14%) of 129 NP specimens, 9 (9%) of 100 skin specimens, and 1 (1%) of 100 serum specimens. S. aureus was detected in 11 (8.5%) of 129 NP specimens, 2 (2%) of 100 skin specimens, and none of the serum specimens. In addition, influenza virus A was detected in one NP and one skin specimen, influenza virus B was detected in one NP and one serum specimen, and B. pertussis was detected in one skin specimen. RSV A, RSV B, C. pneumoniae, M. pneumoniae, L. pneumophila, and L. micdadei were not detected in any specimens in the study.
As shown in Table Table5,5, the highest rates of carriage, both on the skin and in the nasopharynx, were detected in the youngest age groups. S. pneumoniae was detected in NP specimens obtained from 12 (48%) of 25, 0 (0%) of 16, and 6 (6.8%) of 88 subjects in the age groups of 0 to 6 years, 7 to 18 years, and 19 to 64 years, respectively, and in 8 (22%) of 36, 1 (5.5%) of 18, and 0 (0%) of 46 skin swabs obtained from the same age groups. S. aureus was detected in 1 (4%) of 25, 2 (12.5%) of 16, and 8 (6.8%) of 88 NP specimens and 0 (0%) of 36, 1 (5.6%) of 18, and 1 (2.2%) of 46 skin swabs obtained from subjects in the age groups of 0 to 6 years, 7 to 18 years, and 19 to 64 years, respectively. Additionally, S. pneumoniae and influenza virus B were each detected in one serum sample among the samples obtained from asymptomatic children.
A total of four specimens showed evidence of more than one pathogen. Three of these were NP swabs, and one was a skin specimen. Of the NP swabs, two were positive for both S. pneumoniae and S. aureus and one was positive for both influenza virus A and influenza virus B. The skin specimen was positive for S. pneumoniae, S. aureus, and influenza virus A.
Eighteen subjects in the study provided more than one specimen for the study. Of these subjects, three provided NP and skin specimens concurrently, eight provided skin and serum specimens, and seven provided NP and serum specimens at the same collection time points. One of the two subjects with positive results in this group had S. pneumoniae detected in the NP sample but not in serum, and the other subject had influenza virus A, S. pneumoniae, and S. aureus detected in a skin specimen but not serum.
We describe the development of a sensitive and specific multiplex RT-PCR targeting 11 common pathogenic agents of CAP and sepsis by using two detection platforms (manual and automated). Additionally, we determined the prevalence of carriage of common respiratory pathogens in the nasopharynges, on the skin, and in the sera of children and adults by using the assays described.
Bioinformatic analysis for primer and probe design is the first critical step of all molecular diagnostic assays. Our primer and probe sets for most targets had in silico coverage rates of 98 to 100%, thus making false negatives highly unlikely. Coverage rates for RSV B were only 50%, but the number of sequences available for primer design was small (n = 9); of these nine available sequences, five did not include any information on the optimal region used for primer design. However, even though only two of the remaining four were predicted to be covered using our stringent coverage calculation algorithm, in actual clinical practice, a much higher percentage (95 to 100%) of RSV B-positive samples have been detected previously (11, 20, 24). Understanding this discrepancy between in silico and real-life coverage rates is important for clinical diagnostics and is addressed in one of our ongoing research programs. Both the RT-PCR-EHA and the RT-PCR-eMA assay were found to have good LODs for all the pathogens, with the LODs of the EHA being equal to (for influenza virus B, RSV B, M. pneumoniae, and C. pneumoniae) or lower than (for influenza virus B, RSV A, S. aureus, S. pneumoniae, L. pneumophila, L. micdadei, and B. pertussis) those of the eMA, consistent with the data previously reported for the respiratory-virus assay and the Hexaplex (17). No inhibition of amplification from clinical specimens spiked with control organisms was observed. Both assays were shown to have no cross-reactivity among study pathogens and analytical and clinical specificities of 100% for the detection of the target pathogens. However, in clinical testing, the EHA detection platform gave 7 false-positive results while the eMA gave 1 false positive out of 3,573 true negatives. The one false-positive eMA result in our testing is the first we have found using this automated detection platform (17). The sample in question was a skin specimen testing as positive for S. aureus by the eMA but as negative in both the initial and subsequent EHA runs, as well as in an Agilent assay run performed with original amplicons from the EHA and eMA and, finally, in the repeat Agilent assay and EHA runs performed with fresh DNA obtained by singleplex S. aureus PCR amplification of the originally extracted NA. In addition, the specimen had a low signal/noise ratio, 5.7, in the initial microarray run. One of the strong advantages of automated detection for molecular amplification assays is the lack of technician handling of samples and opening and closing of tubes. Our results continue to confirm this advantage in clinical testing. Although the overall surrogate clinical specificity (based on results for spiked clinical specimens) was outstanding for both assays, we have not fully validated the clinical sensitivities and specificities of these assays by using culture as the “gold standard” at this time; such testing is planned in the future.
The most common pathogens detected in the present study were S. pneumoniae (in 28 [8.5%] of 329 specimens) and S. aureus (in 13 [3.9%] of 329 specimens), with age- and site-specific prevalences consistent with previously reported data (21, 30). The rates of carriage of S. pneumoniae reported in studies are heavily dependent on the social, demographic, and medical risk factors of study subjects and methodologic variations with respect to the types of specimens and the diagnostic methods used and thus vary widely from study to study. We did not attempt to associate our carriage prevalence rates with specific population characteristics (e.g., immunization history and day care attendance, etc.). Rather, we aimed at determining the overall point prevalences of common respiratory pathogens at various body sites in healthy subjects by using a multiplex PCR assay with two concurrent detection strategies.
Nearly one-fourth of children in the study aged 0 to 6 years and a smaller proportion (5.5%) of children aged 7 to 18 years showed evidence of carriage of S. pneumoniae in skin swabs. We could not find any prospective studies that investigated the carriage of this pathogen on the skin, although a recent retrospective review of pneumococcal skin and soft tissue infections reported that 20 of 56 patients in the study were thought to have positive results as a consequence of colonization of the skin (13). Nearly all S. pneumoniae-positive skin specimens in the study belonged to patients with ongoing bacterial or viral upper and lower respiratory tract infections. Although not surprising given the likely greater presence of respiratory secretions on the skin surfaces of young children than on those of adults and the previously recognized increased rates of detection of S. pneumoniae in patients with upper respiratory tract infections (15), these high rates of detection of S. pneumoniae in skin specimens reiterate the importance of the observance of sterile practices when obtaining blood cultures and/or blood specimens for molecular assays, particularly for patients in these age groups. Interestingly, although the nasopharynx is the only documented niche of S. pneumoniae in humans, many investigators have speculated about skin colonization subsequent to reports of pneumococcal skin and soft tissue infections in healthy adults and children in the absence of preceding systemic disease (13, 23, 34). The rather high rates of detection of S. pneumoniae in skin specimens in the present carriage study call for further thought about the possible existence of a true reservoir of this pathogen on the skin. Additional studies considering serial detection of S. pneumoniae in both skin and NP specimens over a period of time by using molecular typing may help clarify this issue.
In this study, S. pneumoniae was detected in 1 of 100 serum specimens, which were obtained from study subjects with no respiratory symptoms. This result may represent the presence of pneumococcal DNA in blood secondary to carriage in the nasopharynx or on the skin of the child with the positive specimen or may even reflect true asymptomatic pneumococcal bacteremia, which can occur transiently. The use of PCR for the detection of S. pneumoniae in serum has been reported previously, with sensitivities ranging from 38% for patients with lobar pneumonia to 100% for patients with blood and cerebrospinal fluid culture-positive specimens (7, 26). However, the utility of PCR as a diagnostic tool for pneumonia and invasive infections has been limited due to significant rates of serum positivity observed in healthy controls as a consequence of NP carriage, with the highest rates in age groups with the highest NP colonization rates (7). Only seven subjects in the present study had a serum specimen obtained concurrently with an NP specimen, and six of them did not have S. pneumoniae detected in either the NP or serum specimen. The only patient in this group with an S. pneumoniae-positive NP specimen had a serum specimen that tested negative. More importantly, however, only 2 of our 329 specimens revealed evidence of carriage of respiratory pathogens in the serum, 1 (0.3%) of which was positive for S. pneumoniae and 1 (0.3%) of which was positive for influenza virus B. This result is especially noteworthy given the overall rates of pathogen carriage in NP and skin specimens in this study (22 and 11%, respectively, with pathogen-specific rates for S. pneumoniae reaching a high of 48% for NP specimens and 22% for skin specimens obtained from children 0 of 6 years of age).
S. aureus was detected in 11 (8.5%) of 129 NP specimens, which is a lower rate of detection than those in previously published data (6, 14) and may be explained by the use of NP rather than nasal swabs, which have been used in studies investigating specifically the carriage of this pathogen alone. Other bacterial agents detected included B. pertussis, which was detected in one skin specimen in the study, obtained from a 6-year-old patient with pharyngitis, by EHA alone. Upon repeat testing in the discrepancy analysis, the specimen remained positive, although it had a low OD (0.499). We believe the result most likely reflects contamination from the child's respiratory tract since the skin was asymptomatic.
Pathogens typically associated with atypical pneumonia were not detected in any of the 329 specimens in the study. C. pneumoniae has previously been reported to colonize the nasopharynx in 5 to 15% of asymptomatic healthy subjects (9, 22), although recent studies, including the present study, did not detect NP colonization with this pathogen. In addition, C. pneumoniae has been reported to have been detected in peripheral blood mononuclear cells isolated from patients with a variety of illnesses, as well as healthy subjects (17 to 47%), in various studies (reviewed in reference 2) and in sera obtained from healthy subjects in one study (32). The presence of C. pneumoniae in the community may be cyclic, and there may currently be a nadir in the epidemiology of this pathogen, given other reports of low levels of detection by adequate diagnostic tools even in patients with pneumonia (41). However, the absence of detection in our study (with a large number of patients) suggests that detection in samples from asymptomatic subjects by molecular amplification assays similar to our assays would be infrequent at this time. No data on respiratory colonization with M. pneumoniae, L. pneumophila, or L. micdadei exist. Multiplex molecular assays that target these pathogens with or without S. pneumoniae have been described previously (25) and will likely emerge increasingly in the future given the well-known diagnostic difficulties that render the routine diagnosis of infections with these fastidious bacteria difficult. For such emerging assays to be adopted and accepted for routine use, they will need to be tested and validated with both respiratory and nonrespiratory specimens obtained from both patients and healthy controls.
One NP specimen obtained from a 2-year-old child with the diagnosis of a seizure was positive for influenza virus A and influenza virus B (Table (Table6).6). The child may have been recovering from respiratory symptoms and thus been deemed asymptomatic or may have been mildly symptomatic but called asymptomatic by the parents, or the seizure may have preceded the development of overt respiratory symptoms. One skin specimen was positive for influenza virus A, S. aureus, and S. pneumoniae, which was explained by ongoing fever, respiratory symptoms, and vomiting, most likely having been caused by influenza virus A and possibly having resulted in the detection of S. pneumoniae and S. aureus on the skin. The serum specimen positive for influenza virus B had been collected from a child with no respiratory symptoms reported, and thus, possible explanations include all of those given above for the child with the seizure but with the addition of possible contamination of the child's skin (prior to phlebotomy) with influenza virus B-containing secretions from a close contact. Finally, contamination of specimens during collection or handling by various people in the lab remains a possibility.
The amount (burden) of whole pathogens or NA detectable in specimens from persons in healthy or diseased states by using molecular amplification assays will depend on the status of each subject with respect to asymptomatic colonization or infection with a particular organism, the location of the colonization or infection (e.g., the skin versus the nasopharynx), and the severity of any infection. In addition, the analytical sensitivity of the molecular assay will play a critical role in the detection of this carriage. The ideal molecular assay would have sufficient sensitivity to detect all true infections (pathogens or analytes in different clinical samples) yet would not detect colonization or noninfectious NA. This standard is obviously difficult to achieve since we traditionally want very sensitive molecular assays in medicine.
In summary, we believe that this is the first report of using a multiplex RT-PCR assay that targets 11 different bacterial and viral pathogens to test a large number of asymptomatic subjects. We detected the carriage of respiratory pathogens in 22% of NP specimens, 11% of skin swabs, and 2% of serum specimens by using this highly sensitive and specific multiplex molecular assay with both manual (EHA) and automated (eMA) detection strategies. The more sensitive EHA demonstrated a much higher level of carriage detection than the eMA, yet both showed excellent analytical LODs and analytical and clinical specificities. This finding suggests that the eMA may perform better in predicting disease than the EHA in some clinical situations, but more data would be needed to support this proposition. Clearly, the use of molecular assays for the detection of S. pneumoniae or S. aureus in upper respiratory tract secretions from children or adults is problematic, and the results will be difficult to correlate with true infection or disease no matter which detection strategy is used. However, the study specimens showed virtually no carriage of the other nine bacterial and viral pathogens. The findings of this work, combined with those in previous reports showing good sensitivities and specificities for symptomatic patients (11, 20, 24, 25, 29, 40), suggest that molecular assays testing for these nine pathogens in respiratory secretions will have good utility in clinical practice. The low rate of detection of NA in serum is reassuring in the presence of comparably high rates of detection in specimens from the nasopharynx and skin, sites that have been implicated previously as potentially contributing to the detection of pathogens in serum. We have demonstrated that these current multiplex RT-PCR assays with the EHA or eMA method are clinically useful for the detection of all 11 pathogens, depending on the sample site. Further clinical studies using these assays for symptomatic and asymptomatic subjects need to be performed. We recommend that any new molecular diagnostic assays, especially new multiplex assays, be characterized by using specimens from asymptomatic patients to help establish their value as a clinical diagnostic tool.
This study was supported by grant number AI066584 from the National Institute of Allergy and Infectious Diseases, National Institutes of Health.
We thank Christopher Small for technical assistance with the preparation of the manuscript and all of the staff of the emergency departments of Children's Hospital of Wisconsin and Froedtert Hospital.
Published ahead of print on 23 July 2008.