Diagnosis of Invasive Pneumococcal Infection by Serotype-Specific Urinary Antigen Detection

Abstract

Widespread use of conjugate pneumococcal polysaccharide-protein vaccines may alter the spectrum of pneumococci producing invasive disease. Novel sensitive diagnostic methods would be valuable for monitoring the epidemiology of pneumococcal disease within populations and vaccine recipients. Ideally, these methods should allow determination of the serotype of the infecting clone. Serotype-specific enzyme-linked immunosorbent assays (ELISA) for 13 capsular polysaccharides (types 1, 3, 4, 5, 6A, 6B, 7A, 9V, 14, 18C, 19A, 19F, and 23F) were developed. Experiments with pure capsular polysaccharide demonstrated that the assays were sensitive (0.01 to 1.0 ng/ml) and specific. These assays were used to detect capsular polysaccharide in urine from 263 adult patients with proven (blood culture-positive) invasive pneumococcal disease and pneumonia of unknown etiology and from patients with positive blood cultures yielding bacteria other than pneumococci (control group). Among 76 patients with invasive pneumococcal disease from whom blood culture isolates had been serotyped, 62 (82%) had infections with pneumococci of serotypes represented in the ELISA panel. Capsular antigen matching the serotype of the blood culture isolate was detected in the urine of 52 of these patients, giving a sensitivity of 83.9% for the target serotypes. The tests were significantly more sensitive for urine from patients with pneumococcal pneumonia (89.8%) than for urine from patients with nonpneumonic invasive infection (61.5%; P < 0.05). Data from the control group indicated a specificity of 98.8%. These assays should prove valuable in epidemiological investigation of invasive pneumococcal infection in adults, particularly if combined with a sensitive C-polysaccharide detection assay to screen for positive samples.

The new generation of pneumococcal vaccines based on conjugate polysaccharide-protein preparations promises improved protection against a variety of pneumococcal infections; notably, a heptavalent preparation has been demonstrated to reduce the incidences of both otitis media and invasive pneumococcal disease in childhood (1, 24). However, concern has been expressed that the use of these vaccines may alter the spectrum of disease-producing pneumococci, and evidence that they cause a shift in the serotypes found at mucosal surfaces of the upper respiratory tract has been presented (10, 11, 16).

The diagnosis of Streptococcus pneumoniae infection is frequently problematic. The clinical signs and symptoms of pneumococcal infections cannot be differentiated reliably from a disease of alternative etiology. The “gold standard” diagnostic method is still culture, but good-quality samples are not always available. Furthermore, cultures are not infrequently negative in infections considered likely on clinical grounds to be of pneumococcal origin, particularly after antibiotic administration (6, 14). New sensitive diagnostic methods would be valuable not only for determining the etiology of individual infections but also for monitoring the epidemiology of pneumococcal disease within the general population and vaccine recipients in particular. Application of PCR assays for the diagnosis of invasive pneumococcal disease has proven to be of limited success because they are insufficiently sensitive when applied to blood or urine and are not infection specific when applied to respiratory samples (12, 19). A number of publications have described antigen detection assays (7, 9). Several have targeted C polysaccharide in urine, and recent evaluations have reported favorable sensitivity and specificity data for commercial kits using this strategy in adults (13, 21), although they lack specificity in children (4, 5). These kits do not, however, give information on the capsular serotype of causative organisms, data which would be valuable for epidemiological purposes and for assessing the extent of postvaccination serotype replacement among pneumococci causing invasive infections.

We report here the development and clinical application of serotype-specific enzyme-linked immunosorbent assays (ELISA) for the detection of capsular polysaccharide in urine.

MATERIALS AND METHODS

ELISA.

A common assay strategy (indirect sandwich ELISA) was used for all serotypes. Assays were carried out in polystyrene microtiter tray wells (Microstrip 8EB; Labsystems Oy, Helsinki, Finland). The sequence of reagents employed was as follows: 75 μl of group-specific (types 1, 3, 4, 5, 6, 7, 9, 14, 18, 19, and 23) antiserum (Statens Serum Institut, Copenhagen, Denmark) diluted in 0.2 M carbonate buffer (pH 9.6) (1 to 6 days of incubation at 4°C) to coat the plate, 100 μl of phosphate-buffered saline (PBS) containing 5% skim milk powder (Oxoid, Basingstoke, United Kingdom) to block remaining binding sites, 60 μl of urine (or other antigen source) diluted 50:50 in PBS containing 1% skim milk powder (incubated overnight) to capture urinary antigen, 70 μl of type-specific (types 1, 3, 4, 5, 6A, 6B, 7F, 9V, 14, 18C, 19A, 19F, and 23F) monoclonal antibodies (donated by Wyeth Vaccines Research) diluted in PBS containing 0.2% Tween 20 and 1% skim milk, 80 μl of polyclonal rabbit anti-mouse immunoglobulin antibody conjugated with alkaline phosphatase (DakoCytomation D0314) diluted 1/10,000 in PBS containing 0.2% Tween 20 and 1% skim milk, and finally, a commercial signal amplification substrate added to develop the reaction (DakoCytomation Ampli-Q, 30 min). All incubations were 1 h in duration at room temperature unless stated otherwise, and wells were washed four times with 300 μl PBS containing 0.2% Tween 20 between each reagent incubation. Reagent dilutions were optimized separately for each assay and for each batch of reagents. The sensitivity of each assay was determined by testing serial 10-fold dilutions of purified capsular polysaccharide (ATCC, Manassas, VA) (type 6A was donated by Wyeth Vaccines Research).

Patient samples.

Samples were obtained between January 1999 and May 2002 from adult (>16 years) patients admitted to hospitals in South West England and recruited as part of a prospective, controlled clinical study investigating novel diagnostic tools to diagnose serious community-acquired bacterial infection (21). Ethical approval was obtained from the South and West Multicenter Research Ethics Committee, and patients or relatives granted informed consent. Urine samples were collected from patients with blood culture-confirmed invasive pneumococcal infection (with or without pneumonia), from patients with pneumonia of unknown etiology (most with negative blood cultures), and from a control group who had positive blood cultures yielding clinically significant bacteria other than S. pneumoniae. Patients were classified as having pneumonia if they presented with an acute illness with the presence of a new or progressive infiltrate on chest radiograph plus at least two of the following symptoms: fever, cough, shortness of breath, and pleuritic chest pain. Urine was routinely collected either before or within 24 h of the start of antibacterial therapy (treatment day 1). These specimens were stored at or below −20°C before submission of batches for ELISA analysis by an operator who was blind to culture results, serotype data, and clinical status of the donor patient. Additionally, sequential urine samples collected on subsequent days (up to treatment day 7) were available from 10 patients; in these cases, the expected serotype was identified before ELISAs were done. Before analysis, samples were thawed at room temperature, a sample was spun at 15,000 × g for 1 min, and the supernatant was taken for assay.

Results of Binax NOW S. pneumoniae urinary antigen tests for C polysaccharide were available for 232 of the 263 samples examined in this study (21).

Statistical analysis.

Confidence intervals were calculated by Wilson's method using CIA software (2). Frequencies were compared using the χ² test with Yates correction.

RESULTS

Patients recruited.

A total of 263 urine samples were analyzed. Eighty-six samples were from patients with blood culture-confirmed invasive pneumococcal disease (64 with pneumonia as the apparent primary focus), 96 were from patients with pneumonia of unknown etiology, and 81 were from control patients with blood culture-confirmed community-acquired infection (bacteremia) of alternative etiology and no evidence of pneumonia. The majority of infections were community acquired, but five culture-proven pneumococcal pneumonia cases were judged to be hospital acquired according to the interval between hospitalization and onset of symptoms (>48 h). Demographic information for recruited patients is given in Table 1.

TABLE 1.

Characteristics of patients recruited to each study group

Patient group	Total no. of patients	Male-female ratio	Age range (median)	No. of patients with positive blood culture
				Pnuemococcus	Other bacterium
Invasive pneumococcal disease, no pneumonia	22	13:9	29-95 (67)	22	0
Invasive pneumococcal disease, pneumonia	64	30:34	31-90 (62)	64	0
Pneumonia of unknown etiology	96	44:52	21-102 (67)	0a	9b
Control	81	36:45	18-92 (74)	0	81

^aNo blood culture was taken from 17 of these patients.

^bIsolates from these blood cultures (six coagulase-negative staphylococci, one Proteus mirabilis isolate, one coagulase-negative staphylococcus plus Streptococcus sanguis, and one Aerococcus sp. plus Clostridium sp.) were not considered clinically significant.

Assay performance.

All assays had a sensitivity greater than or equal to 1 ng/ml purified capsular polysaccharide. Assays for capsular types 3, 6B, and 7F had a sensitivity of approximately 1 ng/ml, assays for capsular types 14, 18C, and 23F had sensitivities of 0.01 ng/ml, and the remaining assays had sensitivities of 0.1 ng/ml. Negligible cross-reaction was noted when each assay was challenged with 100 ng/ml purified polysaccharide of each of the 13 serotypes in the assay panel.

Table 2 displays the sensitivity and specificity of the ELISA tests for the various groups of bacteremic patients investigated. Of the 88 isolates from patients with blood culture-confirmed invasive pneumococcal disease, serotype was not available for 12 isolates, and data from these patients were excluded from further analysis. Isolates from a further 14 patients had serotypes not included in the panel of 13 ELISAs. Capsular antigen matching the serotype of the blood culture isolate was detected in the urine of approximately 84% (52) of the remaining 62 subjects with invasive pneumococcal infection. The tests were significantly more sensitive for urine from patients with pneumococcal pneumonia (89.8%) than for urine from patients with nonpneumonic invasive pneumococcal infection (61.5%; P < 0.05). Data from the control group indicated a specificity of 98.8% (confidence interval [CI], 93.3 to 99.8%); the false-positive result was a weak but reproducible serotype 4 signal in an 84-year-old male with Escherichia coli bacteremia.

TABLE 2.

ELISA data for urine samples from patients with culture-proven bacteremia^a

Group and characteristics	ELISA resultb		Descriptive statistics (95% CI)
	No. +ve	No. −ve
All invasive pneumococcal disease
Serotype in ELISA panel (n = 62)	52c	10	Sensitivity, 83.9% (72.8-91.0)
Serotype not in ELISA panel (n = 14)	1d	13	Specificity, 92.9% (68.5-98.7)
Total (n = 76)	53	23	PPV, 98.1% (90.1-99.7)
Invasive pneumococcal disease, pneumonia
Serotype in ELISA panel (n = 49)	44c	5	Sensitivity, 89.8% (78.2-95.6)
Serotype not in ELISA panel (n = 8)	1d	7	Specificity, 87.5% (52.9-97.8)
Total (n = 57)	45	12	PPV, 97.8% (88.4-99.6)
Invasive pneumococcal disease, no pneumonia
Serotype in ELISA panel (n = 13)	8c	5	Sensitivity, 61.5% (35.5-82.3)
Serotype not in ELISA panel (n = 6)	0	6	Specificity, 100% (61.0-100)
Total (n = 19)	8	11	PPV, 100% (67.6-100)
Control group (nonpneumococcal bacteremia)
Pneumococcal antigen presumed absent from sample (n = 81)	1e	80	Specificity, 98.8% (93.3-99.8)

^aSamples from patients with invasive pneumococcal disease where the serotype of the pneumococcus isolated was unknown were excluded. PPV, positive predictive value.

^b+ve, positive; −ve, negative.

^cThe capsular antigen detected in urine matched the serotype of the blood culture isolate in all these cases.

^dCapsular antigen 6A was detected in a pneumonia patient with S. pneumoniae serotype 8 isolated from blood culture.

^eCapsular antigen 4 was detected in an 84-year-old male with Escherichia coli bacteremia.

The performances of the 13 component ELISAs are shown in Table 3. Numbers of samples containing each serotype were insufficient to permit sound statistical comparison of each assay, but it was notable that none of the three samples containing 19F were ELISA positive. No samples were recruited from adults with blood culture-proven serotype 18C or 5 infection. However, capsular antigen 18C was appropriately identified in a number of samples recruited from children using a similar protocol (data not shown), and capsular antigen 5 was detected in one of the blood culture-negative pneumonia samples studied (see below).

TABLE 3.

Performances of the 13 component ELISAs for detecting pneumococcal capsular antigen in urine of patients with blood culture-proven invasive pneumococcal disease caused by a serotype present in the ELISA panel

Performance	Serotype
	1	3	4	5	6A	6B	7F	9V	14	18C	19A	19F	23F
No. detected	7	1	4	0	3	2	2	3	22	0	2	0	6
No. missed	0	1	0	0	0	2	1	1	0	0	0	3	2

Persistence of antigen after treatment.

The persistence of detectable capsular antigen in urine after the commencement of antibiotic treatment was determined in 10 patients. Positivity declined with time, but antigen was still detected in urine from 7/9 patients at day 3 and 6/9 patients at days 5 and 7.

Pneumonia of uncertain etiology.

Capsular antigen was detected in 11 of 96 (11.5% [CI, 6.5 to 19.4%]) urine samples from patients with pneumonia of uncertain etiology. Capsular antigens detected in these patients were serotypes 1 (one patient), 5 (one patient), 9V (four patients), 14 (three patients), and 23F (two patients).

Concordance with commercial C-polysaccharide assay.

The capsular polysaccharide ELISA results corresponded well with the Binax assay (C polysaccharide) data for serotypes contained in the ELISA panel, but the Binax test detected a higher proportion of infections overall (81.7% [CI, 71.2 to 89.0%] versus 73.2% [CI, 61.9 to 82.1%]) because approximately 20% of infections in this study group were caused by pneumococci with serotypes not represented in the ELISA panel (Table 4). For this sample, if a strategy of using the Binax assay as a primary screen with the serotype-specific ELISAs used only on Binax-positive urine was utilized, approximately 82% of pneumococcal infection would have been detected, of which 74% would have been assigned a serotype. An extra seven (10%) infections would be detected and serotyped if, as in this study, all Binax-negative samples were also subjected to serotype-specific assay.

TABLE 4.

Comparison of Binax NOW Streptococcus pneumoniae urinary antigen test results with serotype-specific ELISA analysis of the same urine samples^a

Group	Serotype-specific ELISA analysis
	No. −ve	No. +ve
Patients with invasive pneumococcal infection, serotype in ELISA panel
Binax −ve	2	7
Binax +ve	5	43
Patients with invasive pneumococcal infection, serotype not in ELISA panel
Binax −ve	4	0
Binax +ve	9	1
Patients with bacteremia of nonpneumococcal etiology
Binax −ve	76	1
Binax +ve	2	0

^aSamples from patients for which the serotype of the pneumococcus isolated was unknown were excluded. −ve, negative; +ve, positive.

DISCUSSION

The method described above has a high sensitivity for the serotypes in the test panel and very high specificity among adults with invasive pneumococcal disease and patients with community-acquired nonpneumococcal bacteremia. The method as described above would fail to detect approximately 20% of patients with invasive disease caused by pneumococcal serotypes other than those in the test panel, but if used in conjunction with a sensitive test for C polysaccharides, it should prove to be a successful method of determining the etiology of the majority of serious pneumococcal diseases in adults. Notably, the panel of serotypes covered includes all those incorporated in both seven-valent and nine-valent conjugate vaccine preparations, thus offering a minimally invasive method for monitoring serotype replacement following introduction of these vaccines. The panel also covers the major causes of invasive pneumococcal infections both in the United Kingdom and internationally (20, 22). Less common causes of invasive pneumococcal infection, including serotypes 8, 9N, 12F, and 22F, could readily be added to the assay panel if suitable monoclonal antibodies become available.

As would be expected for a series of assays of this type, their sensitivities varied when titered using purified capsular polysaccharide, and there was some evidence that this variation in sensitivity was also seen in the positivity rate of individual types in urine samples, although sample sizes for each type were too small to allow meaningful statistical comparison. Improvements in the sensitivity of the less sensitive assays are likely to improve the overall sensitivity of the method further and reduce bias which might otherwise result in an underestimation of the prevalence of some serotypes. The assay for type 19F polysaccharide was capable of detecting 0.1 ng/ml purified antigen but (reproducibly) failed to detect antigen in the urine of three patients with proven invasive infection with this serotype. The assay has been used successfully to detect 19F polysaccharide in other body fluids (cerebral spinal fluid and pleural fluid) (data not given). It is possible that type 19F polysaccharide was degraded during storage, as it has been reported to be unstable in mild acid or alkali (15). In this event, greater sensitivity would be anticipated in samples processed after brief storage (some samples were stored frozen for up to 3 years before testing).

The use of serotype-specific assays for pneumococcal antigen detection in urine has been reported in three previous publications, but all those publications reported markedly lower sensitivity than observed in this study. Capeding and colleagues used seven tube latex agglutination assays to detect 10 capsular polysaccharide antigens in urine from 26 children with lower respiratory tract infection and reported a positivity rate of 42% (3), although independent confirmation of the microbiological status of the patients was not available for many of subjects. Scott and coworkers also used tube latex agglutination to assay 10 capsular serotypes in the urine of patients with pneumonia and, in 72 samples, obtained a sensitivity of 46% overall and 57% among patients from whom blood cultures or lung aspirates yielded pneumococci with a serotype in their assay panel (18). Strålin et al. (23) used 22 latex-bound antibody preparations, covering the 23 serotypes in the current plain polysaccharide vaccine, in slide agglutination assays with patients' urine. In 24 patients with pneumococcus-positive blood cultures, they achieved a sensitivity of 54%. In contrast to our observations, only 39% of patients with strongly positive Binax NOW Streptococcus pneumoniae urinary antigen tests were positive in that study. Interestingly, the study of Scott et al. (18) reported no significant difference in the positivity rate among blood culture-positive and blood culture-negative patients, whereas in this investigation, the positivity rate was considerably lower in pneumonia patients with no confirmatory blood culture. The significance of this disparity is unclear but may indicate that the proportion of pneumonias of pneumococcal etiology may genuinely be low in this sample.

There are some limitations to the application of this approach for diagnosing pneumococcal disease. First, performing multiple ELISAs on each sample is relatively labor intensive, although some automation of the process would be relatively straightforward, and the assays could potentially be “multiplexed” using coated-bead suspension array technology as described previously for the assay of antibodies against pneumococcal capsular polysaccharides (17). Second, previous studies have demonstrated that detectible pneumococcal C polysaccharide is present in the urine of healthy infants with upper respiratory tract colonization (4, 5, 8). While there is currently no evidence that capsular polysaccharide is similarly present in the urine of colonized infants, it would not be surprising if this were the case, and this requires further investigation before these ELISAs could be recommended in a pediatric population. Finally, these assays are dependent on the availability of high-quality monoclonal antibodies with specificity for pneumococcal capsular polysaccharides. These are not currently available commercially for the majority of the most common serotypes. This is likely to preclude the use of these methods for routine diagnostic protocols. The method would, however, be accessible to reference and research laboratories for the improved assignation of invasive pneumococcal disease. Our data indicate that the Binax NOW Streptococcus pneumoniae urinary antigen test would be a suitable complementary method for the rapid screening of urine samples. Using this strategy, these assays should prove valuable in epidemiological investigation of pneumococcal infection in adults and enhanced surveillance of pneumococcal infection following introduction of conjugate vaccines.