lytA Quantitative PCR on Sputum and Nasopharyngeal Swab Samples for Detection of Pneumococcal Pneumonia among the Elderly

ABSTRACT Real-time quantitative PCR (qPCR) assay of sputum or nasopharyngeal specimens has shown promising results in the detection of pneumococcal community-acquired pneumonia (PncCAP). We applied qPCR for the autolysin gene (lytA) and compared sputum and nasopharyngeal swab (NPS) pneumococcal loads in elderly patients with community-acquired pneumonia (CAP), and specifically in patients with PncCAP, to those in patient groups with other respiratory diseases. We studied patients aged ≥65 years with radiologically confirmed CAP, clinical CAP not retrospectively radiologically confirmed, other acute respiratory infections, or stable chronic lung disease. Pneumococcal etiology of CAP was ascertained by using a combination of multiple diagnostic methods. We analyzed sputum and NPS specimens by lytA qPCR with 104 pneumococcal genome equivalents (GE)/ml as a cutoff for positivity. Among PncCAP patients, lytA qPCR detected pneumococci in 94% of the sputum samples and in large quantities (mean, 6.82 ± 1.02 log10 GE/ml) but less frequently in NPS (44%) and in smaller quantities (5.55 ± 0.92 log10 GE/ml). In all other patient groups, ≤10% of the sputum samples and <5% of the NPS samples were lytA qPCR positive; but when they were positive, the sputum pneumococcal loads were similar to those in the PncCAP patients, suggesting a pneumococcal etiology in these patients. This was supported by other pneumococcal assay results. Overall, sputum lytA qPCR positivity was more common in PncCAP patients than in the other patient groups, but the quantitative results were mainly similar. NPS lytA qPCR was less sensitive than sputum lytA qPCR in detecting PncCAP.


RESULTS
When 10 4 GE/ml was used as the cutoff for positivity, pneumococci were detected by lytA qPCR in sputum samples from 44/47 (94%) PncCAP cases, 10/103 (10%) CLD cases, and Ͻ5% of the cases in the other groups. NPS specimens were positive for lytA in 23/52 (44%) PncCAP cases and in Ͻ5% of those in all other groups (Table 1). There were 0 PncCAP, 7 non-PncCAP, 2 rejected CAP, 1 ARI, and 1 CLD cases with a lytA qPCR result below the cutoff but greater than zero in sputum and 12, 7, 3, 2, and 11 in NPS, respectively. The log 10 -transformed lytA qPCR results of the sputum and NPS samples with a result greater than zero among the patient groups are shown in Fig. 1A and B, respectively. The samples with a quantitative result below the cutoff were considered negative and were not included in the following analyses.
The diagnostic performance of lytA qPCR for the detection of PncCAP among CAP patients is presented in Fig. 2. At the predefined cutoff of 10 4 GE/ml of sample, the sensitivity and specificity of the lytA qPCR test were 94 and 97% for sputum samples and 44 and 99% for NPS samples, respectively. The criteria for PncCAP (Jokinen et al., submitted) (11) included lytA qPCR from sputum or NPS but only in combination with a Ն2-fold increase in paired serum anti-PsaA and/or anti-CbpA antibodies or a positive pneumococcal urine antigen test. There were four cases in which the PncCAP diagnosis depended on a positive sputum lytA qPCR result, and in three of these, pneumococcal serology or the urine antigen test was missing. When lytA qPCR positivity was left out of the criteria for PncCAP, the sensitivity and specificity of the lytA qPCR test with sputum were similar (93 and 94%, respectively). There were no cases in which the diagnosis depended on a positive NPS lytA qPCR result.
While only Յ10% of the non-PncCAP, rejected CAP, ARI, or CLD cases were positive for pneumococcus by lytA qPCR, when positive their mean pneumococcal load (Ϯ standard deviation) in sputum did not differ statistically significantly from that of the PncCAP cases (Table 1). This was true also when only HQ sputum samples were included in the analysis (data not shown). The five patients with ARI or rejected CAP that had a positive lytA qPCR result and a high pneumococcal density in sputum (Table  1) had encapsulated pneumococci cultured from their sputum samples, which all were of high quality. Two of these patients also had a positive urine antigen test result. This suggests a pneumococcal etiology in these five patients, as they fulfilled the etiological case definition but not the radiological CAP definition.
In the subgroup of patients with positive lytA qPCR results (Ն10 4 GE/ml) with both sputum and NPS, the lytA qPCR results showed no correlation between sputum and NPS pneumococcal genomic loads (n ϭ 23, Pearson's correlation r ϭ 0.19, and P ϭ 0.39 for all cases; n ϭ 19, r ϭ 0.14, and P ϭ 0.56 for CAP cases; n ϭ 18, r ϭ 0.17, and P ϭ 0.51 for PncCAP cases) (Fig. 3). a Only lytA qPCR positive cases included. b Quantitative results compared to PncCAP. c PncCAP was defined as (i) encapsulated pneumococci cultured from blood, (ii) encapsulated pneumococci cultured from HQ sputum (a leukocyte/epithelial cell ratio of Ͼ1), or (iii) at least two of the following: (i) a Ն2-fold increase in serum anti-PsaA and/or anti-CbpA antibodies, (ii) a positive pneumococcal urine antigen test, (iii) detection of pneumococci from sputum of any quality or NPS by culture (encapsulated) or lytA qPCR. d P Ͻ 0.001 compared to PncCAP. e The cutoff for positivity was 10 4 GE/ml.
Pneumococci were cultured from 50 (22%) CAP cases' sputum and 37 (12%) CAP cases' NPS samples, and the cultured pneumococci were encapsulated in 40 (80%) and 32 (86%) of the cases, respectively; the lytA qPCR result was positive in 37 (93%) and 22 (69%) of these, respectively, and negative in all cases with unencapsulated pneumococci cultured from the same sample type. Among the CAP cases with a negative pneumococcal culture, the lytA qPCR was positive in 13 (7%) and 3 (1%) cases with sputum and NPS samples, respectively. In the patients with a positive sputum lytA qPCR result, the mean pneumococcal genomic load in sputum was greater among the cases with pneumococci (encapsulated) cultured from sputum than among those with no pneumococci cultured from sputum ( Table 2). The same was true of NPS samples (5.59 Ϯ 0.93 versus 4.33 Ϯ 0.49 log 10 GE/ml, P ϭ 0.03).
The lytA qPCR positivity and quantitative results of those with a positive lytA qPCR result were also explored for sputum in some selected subgroups of CAP patients ( Table 2). The CAP patients with any respiratory viral coinfection (influenza A or B virus, RSV, PIV1, PIV2, or PIV3) were more often lytA qPCR positive and had a slightly higher pneumococcal genomic load than those with no viruses detected, although the difference in the quantitative result was not statistically significant. However, among the patients with RSV, the mean pneumococcal load was significantly greater ( Table 2). The pneumococcal genomic load of the CAP patients with a CURB-65 score of 3 to 5 did not differ from that of the patients with a CURB-65 score 1 or 2, and the pneumococcal genomic loads in hospitalized and nonhospitalized patients were similar ( Table 2). The CAP patients who had received antibiotics at their acute-phase visit before sputum sampling had a lower mean sputum pneumococcal genomic load than those who were not exposed to antibiotics at the visit or within 2 weeks before the visit ( Table 2). The analyses presented in Table 2 were also conducted by including only the CAP cases with no antimicrobial exposure in the 2 weeks before the visit or at the visit (n ϭ 138; lytA qPCR positive n ϭ 34). The results were similar to those in Table 2, except that there was no difference between the pneumococcal loads of females and males when only patients with no antimicrobial use were included (data not shown).

DISCUSSION
In the present study of elderly patients, pneumococci were detected by lytA qPCR in the majority of the PncCAP patients' sputum samples and less frequently and in smaller quantities in NPS. In all other patient groups, the prevalence of pneumococci by lytA qPCR was low. However, when the test was positive, a high pneumococcal density was observed in the sputum of ARI and rejected CAP patients, suggesting a pneumococcal etiology in these patients.
The diagnostic performance of the lytA qPCR test was studied by comparing it to the criteria used for PncCAP in the FinCAP Epi study (11). The criteria were derived by LCA with a focus on an optimal case definition for vaccine trial purposes (Jokinen et al., submitted), and they included lytA qPCR positivity with sputum or NPS in combination with a positive urine antigen test result or a Ն2-fold increase in paired serum anti-PsaA and/or anti-CbpA antibodies. Therefore, comparison of lytA qPCR test performance against these criteria may not be optimal for the determination of lytA qPCR sensitivity. On the basis of the LCA model, the sensitivity of lytA qPCR with sputum was 90% (Jokinen et al., submitted), which is likely to be closer to the real sensitivity of the test. Also, the criterion for HQ sputum used in our definition of PncCAP (11) and in the present study differs from that described by Murray and Washington (12). We have previously evaluated the significance of sputum quality in sputum culture for the diagnosis of PncCAP by using more stringent criteria for HQ sputum (a leukocyte/ epithelial cell ratio of Ͼ5 and Յ2.5 epithelial cells/field) and found that cultures positive for encapsulated pneumococci from HQ and LQ sputum samples showed a similar concordance with other pneumococcal diagnostic tests if the other test was positive (16). If the other pneumococcal diagnostic test was negative, encapsulated pneumo- cocci were isolated from LQ sputum samples less often than from HQ sputum samples (16). As our previous results did not support the concept that encapsulated pneumococci cultured from an LQ sputum sample would more probably be a false-positive indicator of pneumococcal etiology in CAP than the same finding in an HQ sample (16), we applied a less stringent criterion for HQ sputum in the PncCAP definition (Jokinen et al., submitted) (11) in the present study.
In a recent study, Strålin et al. (8) used a lytA qPCR assay and reported a prevalence of positive results with sputum samples (94% with a cutoff of 10 5 copies/ml and 97% with a cutoff of 10 4 copies/ml) from adult PncCAP patients (median age, 71 years) similar to that in the present study. They also reported a similar mean sputum pneumococcal load (6.71 Ϯ 1.01 log 10 copies/ml) in their qPCR-positive cases. In nasopharyngeal aspirates (NPA), they detected positive results among PncCAP patients by using their lytA qPCR assay more often (62% with a cutoff of 10 4 copies/ml) than we did in NPS specimens, but the mean nasopharyngeal pneumococcal load (5.35 Ϯ 1.69 log 10 copies/ml) among the qPCR-positive cases was similar to that reported here (8). NPA and NPS sampling methods for detecting pneumococcal pharyngeal carriage by culture have been compared in children by Rapola et al. (17), and they found no significant difference in the rate of pneumococcus isolation between NPA and NPS by culture. Among the patients with non-PncCAP and a positive lytA qPCR result, we detected a greater mean pneumococcal load in sputum than Strålin et al. did (8). In agreement with their results, we found no correlation between the pneumococcal genomic copy numbers in sputum and nasopharyngeal specimens. However, Albrich et al. (9) previously noted a good correlation between the genomic pneumococcal loads in sputum and NPS from HIV-infected adults with CAP. The studies of Strålin et al. (8) and Albrich et al. (9) did not include patients with respiratory tract infections other than CAP.
The mean pneumococcal load detected by qPCR was greater in patients with a positive sputum culture (encapsulated pneumococci) than in patients with no pneumococci cultured. This is in accordance with previous studies where greater pneumococcal genomic loads have been observed in culture-positive sputum specimens (5,9,18). Gadsby et al. (5), however, noted that their culture-negative group was more frequently exposed to antibiotics, which was also associated with lower bacterial loads. Werno et al. (18) did not find any significant effect of antibiotic use on genomic pneumococcal loads, even though they detected pneumococci less often in the sputum of patients exposed to antibiotics prior to admission. We have previously studied the effect of antibiotic use on pneumococcal diagnostic tests with the same study population as here (19) and found that antibiotic use within 2 weeks before the acute-phase visit, and specifically when still ongoing at enrollment, but not antibiotic use at the visit, was associated with lower sputum lytA qPCR positivity. In the present study, patients who had received antibiotics at the acute-phase visit, but not those who had gotten antibiotics within 2 weeks before the visit, had a significantly lower mean pneumococcal genomic load in their sputum than those with no antibiotic use.
In all CAP cases in which unencapsulated pneumococci were cultured, the lytA qPCR result with the same sample type was negative. This implies that the unencapsulated cultured isolates were, in fact, not true pneumococci and they were disregarded in the further analyses.
It is well known that respiratory virus infections play a role in respiratory bacterial infections (20). Respiratory virus coinfection has also been associated with greater nasopharyngeal pneumococcal loads in children with CAP (21) and in ARI patients with a high HIV prevalence (22). Alpkvist et al. (23), however, found no association between viral coinfection and greater nasopharyngeal pneumococcal loads in adult patients. In the present study, virus infection was detected by sputum PCR and/or serology and CAP patients with a respiratory viral coinfection had a slightly, although not statistically significantly, greater sputum pneumococcal genomic load than CAP patients with no viruses detected. Among the CAP patients with RSV, the pneumococcal load was, however, significantly greater. Interestingly, in mice, RSV infection has been shown to lytA qPCR for Detection of Pneumococcal Pneumonia Journal of Clinical Microbiology decrease the clearance of pneumococci from the lungs and increase pulmonary inflammation (24). A high nasopharyngeal pneumococcal density has been found to be associated with greater disease severity (23,25). However, when the sputum pneumococcal load and disease severity were compared by Werno et al. (18), no association was found except for those patients who were previous and current smokers; they were more often in a higher pneumonia severity index risk class when the pneumococcal load in sputum was Ͼ10 3 CFU/ml. We analyzed the pneumococcal loads in the sputum samples of CAP patients, their CURB-65 scores, and their hospitalization status for associations and found none.
The present study confirms the results of the study of Strålin et al. (8), which identified the lytA qPCR assay of sputum as a useful method for the diagnosis of PncCAP. In contrast to the study of Albrich et al. (9), the lytA qPCR assay of NPS did not perform as well as the lytA qPCR assay of sputum samples in the present study of elderly persons. The quantitative result obtained with NPS may be less reliable because of variations in the quantity of the actual sample and in the recovery of pneumococci from STGG. The lytA qPCR assay is a more rapid method than culture. We did not try to identify optimal cutoff values for the diagnosis of PncCAP by using lytA qPCR but used a predefined cutoff that was the lowest concentration that could be consistently detected. The strengths of the present study are that it was a prospective follow-up study with systematic data collection and that it included different patient groups. Thus, we had qPCR results available also for patient groups with respiratory tract infections other than CAP or a CLD. However, a limitation of this study was the low number of patients, especially in the PncCAP group. In addition, only a few patients in the present study had pneumococcal bacteremia. A potential limitation of the lytA qPCR assay is that lytA has been found in some nonpneumococcal streptococcal isolates of the Streptococcus mitis group (26,27). Thus, the lytA qPCR assay potentially overestimates the pneumococcus as the etiological agent of CAP. In the present study, 3% of the non-PncCAP cases' sputum samples and 1% of their NPS samples were positive by lytA qPCR assay but the etiological agent in these was not determined.
In conclusion, pneumococci were detected by lytA qPCR in the majority of the PncCAP patients' sputum samples when 10 4 GE/ml was used as the cutoff for positivity. In all other patient groups, the prevalence of pneumococci by lytA qPCR was low. Sputum was superior to NPS as a sample type in the detection of PncCAP by lytA qPCR, and lytA qPCR analysis of sputum continues to be a promising diagnostic tool in the detection of pneumococcal etiology in CAP. A trend toward greater pneumococcal loads in sputum samples from CAP patients with a viral coinfection, particularly RSV, was seen.