Use of Multiplex Quantitative PCR To Evaluate the Impact of Pneumococcal Conjugate Vaccine on Nasopharyngeal Pneumococcal Colonization in African Children

This study focused on evaluating the effect of infant vaccination with 7-valent pneumococcal conjugate vaccine (PCV7), using a multiplex qPCR method, on the density of serotype-specific nasopharyngeal colonization in order to delineate the relative role of serotype replacement versus unmasking as the cause for the increase in nonvaccine serotype colonization in PCV7-vaccinated children. This is pertinent in the context of the ongoing deployment of PCV immunization in children, with surveillance of colonization considered an early proxy for disease that might arise from nonvaccine serotypes, as well as the success of childhood vaccination on indirect effect in the community through the interruption of pneumococcal transmission from vaccinated young children.

KEYWORDS PCV7 vaccination, serotype replacement, unmasking T he human nasopharynx is colonized with multiple commensal and some potentially pathogenic organisms, including Streptococcus pneumoniae (1). Vaccination of children with pneumococcal conjugate vaccines (PCV) reduces the risk of S. pneumoniae vaccine serotype (VT) nasopharyngeal colonization but is associated with increased detection of nonvaccine serotypes (NVT) (2). Increased NVT colonization in vaccinees could involve either replacement colonization through reduced VT acquisition or unmasking of previously prevailing NVT that were not identified by traditional culture methods that detect only the dominant colonizing serotype or both mechanisms (3). Molecular detection of pneumococci in the nasopharynx has several advantages over traditional culture-based methods, including the detection of multiple serotypes from a single sample with high sensitivity, as well as quantitative PCR (qPCR) methods being able to measure the density of colonization and relative proportion of colonizing serotypes (4).
The study aimed to use qPCR to evaluate the effect of infant vaccination with 7-valent PCV (PCV7) on the density of serotype-specific nasopharyngeal colonization to delineate the relative roles of serotype replacement and unmasking to explain the increase in NVT colonization in PCV7-vaccinated children.

RESULTS
Quantitative PCR analysis involved 713 (83%) of the initial 857 nasopharyngeal swabs collected from children at either 9 or 16 months of age (Fig. 1). Among the children with samples available for testing, 58% were HIV exposed and 52% were male (Table 1). In the PCV7-vaccinated group, a lower percentage of children were black African compared to the PCV-unvaccinated group (83% versus 100%; P Ͻ 0.001). The percentage of children attending day care and having a smoking household contact was higher in PCV7-vaccinated children at 9 months of age (day care, 52% versus 22.1% [P Ͻ 0.001]; smoking contact, 42.3% versus 36.5% [P ϭ 0.003]) and 16 months of age (day care, 53.6% versus 37.3% [P ϭ 0.002]; smoking contact, 39% versus 33.7% [P ϭ 0.004]), while the mean age (in months) at the time of sample collection was 9 (standard deviation [SD], 0.7) and 15.4 (SD, 0.4) in PCV7-vaccinated children and 9.4 (SD, 0.5; P Ͻ 0.001) and 16 (SD, 1.4; P Ͻ 0.001) in PCV-unvaccinated children. Although the differences in age were small and unlikely to be of epidemiological or clinical significance, we chose to adjust the subsequent analyses for age in addition to the other variables that were different between the two cohorts.
PCV vaccination and density of pneumococcal carriage. When the children were 9 months old, pneumococcal colonization density was higher in PCV7-vaccinated FIG 1 Schematic diagram of study population. Flow diagram indicating the number of children initially enrolled in a PCV-unvaccinated and PCV7-vaccinated cohort of HIV-uninfected children, as well as the number of nasopharyngeal (NP) swabs available for subsequent quantitative PCR (qPCR) analysis. PCV7-vaccinated participants were excluded from analysis if they did not receive all three doses of the pneumococcal conjugate vaccine (PCV) within protocol-defined window periods. The number of NP swabs available for molecular testing was defined by whether there was an adequate volume of sample remaining. Some samples were used for external quality assessment (EQA). Impact of PCV on Pneumococcal Colonization in Children children than in PCV-unvaccinated children (4.68 versus 4.28 CFU/ml; P ϭ 0.007) ( Table 3). This was associated with a higher density overall of PCV7 serogroups/ serotypes (3.8 versus 3.4 CFU/ml; P ϭ 0.048), despite lower prevalence thereof, and higher overall density of NVT (3.6 versus 3.1 CFU/ml; P ϭ 0.018). No difference in density of individual PCV7 serotypes/serogroups was found between PCV-unvaccinated and PCV7-vaccinated groups. The density of nonvaccine serotype 19A was, however, higher in PCV7-vaccinated children than in PCV-unvaccinated children (3.76 versus 2.83 CFU/ ml; P ϭ 0.046), for which there was also a higher colonization prevalence in PCV7vaccinated children. Further, the density of nonvaccine serogroup 23A/B was also   The percentage of children colonized is shown in parentheses. c The odds ratio (OR), the 95% confidence interval (95% CI) for the odds ratio (shown in parentheses), followed by the P value. d The adjusted odds ratio (aOR) for pneumococcal colonization was determined by multivariate logistic regression, controlling for race, smoking household contact, co-trimoxazole use, day care attendance, and mean age at the time of sample collection. higher in PCV7-vaccinated children than in PCV-unvaccinated children (4.48 versus 2.07 CFU/ml; P ϭ 0.002), despite no difference in colonization prevalence.
By the time the children were 16 months old, there was no overall difference in pneumococcal colonization density between the two study groups, although there was a higher density of NVT in PCV7-vaccinated children than in PCV-unvaccinated children, including serotypes/serogroups 5 (2.79 versus 2.01 CFU/ml; P ϭ 0.015), 19A (4.15 versus 3.04 CFU/ml; P ϭ 0.013), and 23A/B (3.59 versus 1.58 CFU/ml; P ϭ 0.013), although there was only a higher prevalence of serotype 19A colonization Cocarriage of multiple serotypes. Overall, qPCR detected one, two, and three or more serotypes in 42.8%, 16.5%, and 5.3% nasopharyngeal swabs, respectively. Additionally, 13.2% of the swabs were lytA positive but negative for all tested serotypes, implying the presence of nontypeable pneumococci, pneumococci belonging to untested serotypes, or lytA-positive nonpneumococcal streptococcal species. Colonized PCV7-vaccinated children were less likely to have only a single serotype identified than PCV-unvaccinated children at 9 months (59.2% versus 64.6%; P ϭ 0.025) and 16 months (44.7% versus 50.3%; P ϭ 0.019) of age (Table 4). The P values were determined by multivariate logistic regression, controlling for race, smoking household contact, co-trimoxazole use, day care attendance, and mean age at the time of sample collection by using generalized estimating equations. P values of Ͻ0.05 were considered significant. PCV7, seven-valent pneumococcal conjugate vaccine.
Among PCV-unvaccinated children, PCV7 serogroups/serotypes 6A/B (prevalence, 52/264, or 19.7%), 19B/F (prevalence, 51/264, or 19.3%), and 23F (prevalence, 40/265, or 15.2%) were the highest-density-colonizing serotypes. Furthermore, nonvaccine serotype 5 was the most prevalent second colonizing serotype, ranked by having a lower density than the primary colonizing serotype but a higher density than the tertiary colonizing serotype. PCV7 serogroup 9A/L/N/V was the most prevalent tertiary colonizing serotype/serogroup ranked by having a lower density than both the primary and secondary colonizing serotypes (Fig. 3A). Among PCV7-vaccinated children, PCV7 serogroups 19B/F (14.7%) and 6A/B (14.7%) and nonvaccine serotype 19A (14.7%) were the highest-density-colonizing serotypes, while nonvaccine serotype 5 was the most common second (16.2%) and third (44.4%) colonizing serotype based on its density of carriage (Fig. 3B). The serotype-specific propensity of whether a given serotype/serogroup is more likely to be found as a primary or nonprimary isolate is shown in Fig. 4, with vaccine serogroups 6A/B and 19B/F and nonvaccine serotypes/groups 15A/B/C/F and 34/37/ 17A being more likely to be identified as primary isolates than nonprimary isolates in both PCV7-vaccinated and PCV-unvaccinated children. Nonvaccine serotype 19A was, however, more likely identified as a primary isolate than as a nonprimary isolate in PCV7-vaccinated children only. Further, nonvaccine serotype 5 was more likely identified as a nonprimary isolate in both PCV7-vaccinated and PCV-unvaccinated groups. These results did not differ between study time points.
PCV7-vaccinated children compared to PCV-unvaccinated children at 9 months of age had a higher mean density of colonization for the first (3.9 versus 4.38 CFU/ml; P ϭ 0.05), second (3.59 versus 3.99 CFU/ml; P ϭ 0.05) and third (1.7 versus 2.35 CFU/ml; P ϭ 0.005) colonizing serotypes; however, no difference in densities were found by the time the children were 16 months old (Fig. 5).

DISCUSSION
In this study, quantitative PCR was used to compare pneumococcal serotype/ serogroup-specific colonization in cohorts of PCV7-vaccinated and PCV-unvaccinated African children. We clearly showed the vaccine effect, typified by a decrease in the prevalence of PCV7 serotype colonization and a corresponding increase in NVT colonization (5-7); this study also showed that both mechanisms of serotype replacement and unmasking led to the increase in NVT colonization in PCV7-vaccinated children. Some serotypes were associated with an absolute increase in colonization (replacement), as both the colonization prevalence and density increased and were commonly found as primary colonizers in PCV-vaccinated children while they were found equally as primary and nonprimary colonizers in PCV-unvaccinated children. Other serotypes were associated with an increase in detection (unmasking), as only the colonization density increased and they were commonly found as second and third colonizers in both PCV-vaccinated and PCV-unvaccinated children.
Although serotype-specific analysis was often limited by a small sample size for a given serotype, the carriage prevalence of most NVT remained unchanged in PCV7-

FIG 4
Serotype-specific propensity of whether a given serotype/serogroup is more likely to be found as a primary or nonprimary isolate in PCV7-vaccinated (red) and PCV-unvaccinated (blue) children. Pneumococcal serotypes identified from nasopharyngeal swabs were classified according to whether they occurred as a primary or nonprimary isolate as determined by carriage density. Single colonizers were included in the analysis as primary isolates. The box at the top of the graph is an expanded view of the block in the lower left-hand corner of the graph. Green dots represent two or more serotypes with the same colonizing profile. Results were considered significant when the P value was Ͻ0.05, and these results are indicated by an asterisk. vaccinated children, with the exception of serotype 19A, for which the prevalence of colonization was higher. Although this difference in serotype 19A has been previously documented in PCV7-vaccinated populations (5,(8)(9)(10), the qPCR method allowed quantification and showed a higher density of this serotype. Also, serotype 19A was more commonly reported as a primary isolate in PCV7-vaccinated children, while being equally a primary (46.2%) or nonprimary (53.8%) isolate in PCV-unvaccinated children. This would indicate that the difference in serotype 19A carriage among PCV7vaccinated children was due to a combination of serotype replacement and unmasking of colonization which would have been missed using conventional culture methods. In addition, the higher carriage prevalence and density of serotype 19A has clinical relevance, as it could explain the emergence of this serotype as the major replacement serotype causing invasive pneumococcal disease (IPD) following PCV7 introduction in several settings, most likely as this serotype commonly has a high prevalence of antimicrobial resistance clones, thus facilitating its survival compared to antibioticsusceptible NVT (11)(12)(13).
The higher colonization density of nonvaccine serotype 5 and serogroup 23A/B was identified in PCV7-vaccinated children compared to PCV-unvaccinated children in the absence of a difference in overall colonization prevalence. As serotype 5 is rarely seen in conventional carriage studies, these results suggest an unmasking of serotype 5 by the qPCR method. Nonetheless, serotype 5 has a high invasive disease potential in our setting (14,15), and a higher carriage density as a result of PCV immunization could enhance its ability to cause mucosal and invasive disease that could potentially offset the effectiveness of PCV7 immunization. Further, a composition shift in cocolonized children from a mixture of NVT and vaccine serotypes to almost pure NVT as a result of PCV7 immunization was observed, supporting an unmasking of serotype carriage. Children with multiple serotype colonization, however, also had a lower density of colonization with NVT than children colonized with only one serotype. These observations therefore further support increases in detection probability (unmasking) and NVT acquisition (true replacement) to explain the higher rate of NVT colonization among PCV7-vaccinated children (2).
PCV7 was introduced into the public immunization program in May 2009 in South Africa and replaced by PCV13, which includes serotypes 19A and 5, in April 2011. These serotypes that have a high invasive disease potential in our settings, were thus considered nonvaccine serotypes at the time of our studies. Further, qPCR increases the detection of serotypes 5 and 19A previously missed by traditional culture methods (4) and highlights the importance of continued surveillance using sensitive molecular methods able to detect serotypes at a low carriage density to more accurately measure the effectiveness of these vaccines in reducing/eliminating carriage or change in the circulating pneumococcal serotypes. For example, after the introduction of PCV13, serotype 12F has emerged as a leading "replacement" serotype, causing IPD in many settings (16)(17)(18).
Of note among PCV7-vaccinated children, despite a lower prevalence of PCV7 serotype colonization at 9 months of age, the density of colonization of both PCV7 serotype and NVT pneumococci was higher than in PCV-unvaccinated children. Although we might expect PCV vaccination to reduce the density of vaccine serotype carriage, the opposite was found. One explanation for this observation is that pneumococcal carriage density may be influenced by antibodies to common pneumococcal surface antigens (CPAs). Ditse et al. (19) found lower titers to CPAs in PCV7-vaccinated infants than in PCV-unvaccinated infants at 10 months of age, most likely from reduced pneumococcal exposure. By the time the infants were 18 months old, however, these differences were not significant (19), which was consistent with colonization density noted in our study. Another explanation could be that PCV is less effective in preventing colonization with PCV7 serotypes among a subset of children who remained colonized, possibly due to poorer immune response or other factors contributing to the colonization by these PCV7 serotypes, which also enables a higher density of colonization by these serotypes.
Although we have shown the value of molecular assays for surveillance of pneumococcal colonization, limitations of our study include the fact that the qPCR assays could not discriminate between all vaccine serotypes within their respective serogroups due to the high genotypic similarities within each group as described previously (4); however, due to the high concordance between serotypes identified by culture and qPCR, we can assume from colonization data using traditional culture methods that 93.3%, 88.9%, and 91.4% of all 9A/L/N/V, 18A/B/C, and 19B/F serogroups identified by qPCR to be vaccine serotypes 9V, 18C, and 19F, respectively. The qPCR assay did not detect all pneumococcal serotypes and nontypeable pneumococci could not be identified. This may have limited our understanding of the roles of nontypeable and other serotypes in limiting vaccine effect for protection against colonization, especially since recent genome sequencing projects have shown an increase in nontypeable isolates following PCV immunization (20). Further, this study did not have adequate power to evaluate serotype-specific differences. Another limitation is that the control group was enrolled after the vaccinated group; however, the limited coverage of PCV in the community at the time made it unlikely that a broader PCV7 indirect effect reduced transmission by young vaccinated children to unvaccinated children in the community. Last, as the study was not a randomized clinical trial and the cohorts were not matched, unmeasured factors could have influenced the results.
In conclusion, molecular qPCR allowed us to gain a better understanding of serotype carriage and indicated that the underlying mechanisms for the increase in NVT colonization in PCV7-vaccinated children was likely due to both serotype replacement and unmasking of underlying preexisting colonizing serotypes which were previously undetected by conventional culture methods.

MATERIALS AND METHODS
Study population. Archived nasopharyngeal swab samples collected from pneumococcal conjugate vaccine (PCV)-unvaccinated and 7-valent PCV (PCV7)-vaccinated cohorts of HIV-uninfected children from Soweto, South Africa, were retrospectively analyzed. Detailed information of the study cohorts has been described previously (21,22). Briefly, the PCV7-vaccinated cohort was enrolled between April 2005 and June 2006 and included 125 HIV-exposed-uninfected (HEU) infants born to HIV-infected mothers and 125 HIV-unexposed infants, with all infants between 6 and 12 weeks old at enrollment. These infants received three doses of PCV7 (Prevnar; Wyeth Vaccines, NJ, USA) at 6, 10, and 14 weeks of age (22,23). From January 2007 through October 2007, 251 PCV7-naïve infants, including 125 HEU infants and 126 HIV-unexposed infants were also enrolled in a separate pneumococcal carriage study (21). During both studies, pneumococcal immunization of children in Soweto, South Africa (birth cohort of approximately 28,000 per annum) was limited mainly to study participants (approximately 600 children in total participated in PCV studies at that time), as PCV7 was introduced into the pubic immunization program in May 2009 (24). Nasopharyngeal swabs were collected from participants in both cohorts at several time points, including at 9 and 15 to 16 months of age. Swabs were stored in skim milk-tryptone-glucose-glycerol (STGG) transport medium at the Respiratory and Meningeal Pathogen Research Unit (RMPRU) in South Africa, as recommended by WHO (25). The samples had been previously cultured for Streptococcus pneumoniae using standard culture methods, and pneumococcal serotyping was undertaken using the Quellung method as described previously (21). Direct comparison of pneumococcal serotype colonization between the PCV7-vaccinated and PCV-unvaccinated cohorts was not performed.
Multiplex quantitative PCR methods. Briefly, stored nasopharyngeal swabs were thawed, and total nucleic acids were extracted using the automated NucliSens easyMAG extraction system (BioMérieux, Marcy l'Etoile, France) according to the manufacturer's instructions; extracted nucleic acids were stored at Ϫ20°C. The quantitative PCR (qPCR) method used in this study has previously been described and validated (4). Briefly, target DNAs were prescreened for the Streptococcus lytA gene (26). All samples with quantification cycle (Cq) values of Ͻ35 were regarded as positive for streptococci and further molecularly serotyped for PCV7 serotypes/serogroups ( Statistical analysis. The Pearson 2 test or Student t test was used to compare baseline characteristics between the vaccinated and unvaccinated cohorts. Comparisons of prevalence of pneumococcal colonization between cohorts were analyzed using multiple logistic regression models adjusted for race, passive smoke exposure, day care attendance, co-trimoxazole usage, and mean age at sample collection; adjusted odd ratios (aOR) were calculated. Colonization density data were presented as CFU/milliliter and geometric mean densities (GMD). Confidence intervals (95% confidence intervals [95% CI]) of pneumococcal concentrations were calculated following log 10 transformation, using analysis of covariance adjusted for possible covariates. Serotype-specific propensity comparing the observed proportion of a given serotype/serogroup found as a primary or nonprimary isolate were analyzed using a two-tail binomial test, in which the primary isolate was defined as the first dominant colonizing serotype that had the highest colonization density among other cocolonizing serotypes identified in the same sample, or was a single colonizer. Nonprimary isolates were defined by having a lower carriage density than the primary isolates. Results were considered significant when the P values were Ͻ0.05. Statistical analysis was performed with Stata version 11.0 (Statacorp, TX, USA).
Ethics. Ethical approval for the original two studies was obtained from the Medical Human Research Ethics Committee (HREC) of the University of the Witwatersrand {vaccinated cohort [HREC (040704)], and Clinical trials registration number NCT00099658; PCV-unvaccinated cohort [HREC (050705)]}. Approval for further testing of samples was obtained from the HREC (M120972). Written, informed consent was obtained from the parents/guardians of the study participants at the time of enrollment.
Data availability. The data that support the findings of this study are available from the corresponding author upon request.