Longitudinal characterization of nasopharyngeal colonization with Streptococcus pneumoniae in a South African birth cohort post 13-valent pneumococcal conjugate vaccine implementation

Monitoring changes in pneumococcal carriage is key to understanding vaccination-induced shifts in the ecology of carriage and impact on health. We longitudinally investigated pneumococcal carriage dynamics in infants. Pneumococcal isolates were obtained from nasopharyngeal (NP) swabs collected 2-weekly from 137 infants enrolled from birth through their first year of life. Pneumococci were serotyped by sequetyping, confirmed by Quellung. Pneumococci were isolated from 54% (1809/3331) of infants. Median time to first acquisition was 63 days. Serotype-specific acquisition rates ranged from 0.01 to 0.88 events/child-year and did not differ between PCV13 and non-PCV13 serotypes (0.11 events/child-year [95% CI 0.07–0.18] vs. 0.11 events/child-year [95% CI 0.06–0.18]). There was no difference in carriage duration between individual PCV13 and non-PCV13 serotypes (40.6 days [95% CI 31.9–49.4] vs. 38.6 days [95% CI 35.1–42.1]), however cumulatively the duration of carriage of non-PCV13 serotypes was greater than PCV13 serotypes (141.2 days (95% CI 126.6–155.8) vs. 30.7 days (95% CI 22.3–39.0). Frequently carried PCV13 serotypes included 19F, 9V, 19A and 6A, while non-PCV13 serotypes included 15B/15C, 21, 10A, 16F, 35B, 9N and 15A. Despite high immunization coverage in our setting, PCV13 serotypes remain in circulation in this cohort, comprising 22% of isolates. Individual PCV13 serotypes were acquired, on average, at equivalent rate to non-PCV13 serotypes, and carried for a similar duration, although the most common non-PCV13 serotypes were more frequently acquired than PCV13 serotypes.

900,000 of the estimated 6.3 million global deaths in children under 5 years of age in 2013 3 . The incidence of pneumonia-related mortality is particularly high in Africa with more than 600,000 pneumonia-related deaths reported in children each year 4 . Nasopharyngeal (NP) colonization by pneumococci is a necessary first step in progression to pneumococcal pneumonia and yet the dynamic nature of colonization remains incompletely understood. Children under the age of two years are more frequently colonized by pneumococci than older children and adults 5 . The nasopharynx also serves as a reservoir and source for transmission of pneumococci 6 .
The duration of pneumococcal carriage varies across epidemiological settings, with the median duration ranging from 60 days (serotype 11A) to 212 days (serotype 19F) when sampled at monthly intervals 12 . Recent acquisition of pneumococci in the NP has been associated with progression to disease 13 .
Pneumococcal carriage dynamics are best understood from longitudinal rather than cross-sectional studies 12,14,15 . Longitudinal study designs allow for the estimation of acquisition rates, carriage duration and risk factors with optimized sampling frequency, length of follow-up and clinical data collection 12 . We aim to describe the dynamics of pneumococcal NP carriage during the first year of life in an intensively sampled PCV13-vaccinated paediatric birth cohort in South Africa.

Materials and Methods
Study population and sampling. One hundred and thirty seven (137) infants from the Drakenstein community in South Africa were enrolled between May 29 th 2012 and May 31 st 2014, as part of longitudinal, prospective birth-cohort study 16 . The community is a stable, semi-urban, poor community in South Africa, receiving routine immunization with Haemophilus influenzae type b [Hib] and PCV13 conjugate vaccines as part of the national immunisation programme 17 . A 7-valent pneumococcal conjugate vaccine (PCV7) was introduced into the South African Expanded Program on Immunisation (EPI) in April 2009 with no catch-up immunisation but replaced by PCV13 (Prevnar®, Wyeth Pharmaceuticals Inc.) in June 2011 with provision for catch-up vaccination of unimmunised children at 18 months. PCV13 is administered in a 2+1-dosing schedule at 6 weeks, 14 weeks and 9 months of age 18 .
This study was approved by the Human Research Ethics Committee of the Faculty of Health Sciences, University of Cape Town (HREC ref: 401/2009 and 740/2013) and the Western Cape Provincial Child Health Research Committee. Written, informed consent was obtained from the participants' parents or guardian at recruitment and annually thereafter. Enrolment of participants and all procedures were conducted in accordance with the relevant regulations and guidelines.
The details of the birth cohort study population and study design are described elsewhere 16 . Briefly, pregnant women (>18 years), between 20 and 28 weeks' gestation, attending antenatal care at two primary care clinics (Mbekweni and TC Newman) were enrolled and prospectively followed-up through pregnancy and childbirth (n = 1143 live births). NP swabs were collected from infants at birth and every alternate week during the first year of life. The subset of 137 children included in this analysis were selected consecutively from the first child completing 1 year of follow up in the birth cohort, and with adequate specimen collections (at least 23 of the 26 possible NP swabs collected). The collected NP swabs were immediately placed into 1 ml skim milk-tryptone-glucose-glycerol (STGG), transported at 4 °C to the laboratory within 2 hours of collection and frozen at −80 °C for later batch culture. Presumptive pneumococcal isolates were identified by colony morphology, α-hemolysis, optochin disk susceptibility (Oxoid, Basingstoke, UK) and confirmed using lytA PCR 19 . Serotyping was performed on a single morphologically distinct pneumococcal colony per sample by sequetyping 20 and confirmation using the Quellung method.
Statistical analysis. Exploratory statistics were performed using STATA software (Stata Corporation, College Station, TX) and the rest of the analyses were performed using R, version 3.1.1 21 . A pneumococcal acquisition event was defined by the detection of a pneumococcal serotype for the first time in an infant, and when a pneumococcal serotype was recovered following two consecutive negative NP swab cultures for that specific serotype ( Supplementary Fig. S1) 12 . A pneumococcal carriage episode, representing ongoing NP colonization, was defined as the period between acquisition and loss of the same pneumococcal serotype. Acquisition was presumed to start at the midpoint between the last of two negative NP swabs and the first positive NP swab for a specific serotype, whilst clearance was considered as the midpoint between the last positive swab and the first of two consecutive negative NP swabs for that specific serotype ( Supplementary Fig. S1). Carriage duration was right censored at the last NP sampling point to account for uncertainty in how long pneumococci may have been carried after the last visit. Time to pneumococcal acquisition and carriage duration were determined by Kaplan-Meier survival estimates and recurrent colonization episodes were determined by the conditional gap-time model 22 . The model investigates time to pneumococcal acquisition on condition that the infant had previously acquired a pneumococcus. Here the survival function measures the probability of not experiencing the second event before time t 2 − t 1 where the individual experienced the first acquisition event at time t 1 , where t 1 and t 2 are the first and second acquisition events respectively. We report both cumulative (averaged by the number of children) and adjusted rates of acquisition and carriage duration (averaged by the number of serotypes) in order to reflect child-specific carriage dynamics and serotype-specific carriage dynamics, respectively. Data availability. The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.
A total of 3331 NP swabs were collected (median number of swabs per infant, 25 [IQR, [23][24][25][26]). Pneumococci were isolated from 54% (1809/3331) of samples. A pneumococcal serotype was successfully assigned to 91% (1637/1809) of isolates; 9% (172/1809) were non-typeable. Immunization coverage was 100% at the scheduled 6, 14 and 40 weeks visits respectively. However, vaccination was delayed by more than 2 weeks in 6% (8/137) and 18% (24/137) of children at 6 and 40 weeks respectively. Acquisition and prevalence of pneumococcal carriage. Figure 1 shows the conditional-gap time model for the time-to-acquisition of first and recurrent pneumococcal carriage. All but seven (5%, 7/137) children were colonized at least once by 260 days of life. Pneumococci were not detected in three (2%, 3/137) children at any time point during the first year of life. Time to first colonization (median age, 63 days [IQR 55-90 days]) was longer than the time between first and 2 nd acquisition (36 days [IQR 28-47 days]) ( Fig. 1 and Supplementary  Table S1). Thereafter, subsequent acquisition events occurred at similar intervals. There was no difference between PCV13 and non-PCV13 serotypes in time to first pneumococcal acquisition (p = 0.69).

Discussion
This study describes longitudinal patterns of NP pneumococcal colonization over the first year of life in an intensively sampled, PCV13 vaccinated South African birth cohort. Despite the high immunization coverage, pneumococcal carriage prevalence and duration of carriage (on average one third of the first year of life) were high. Children were less likely to be colonized by PCV13 serotypes compared to non-PCV13 serotypes at all time points. The pneumococcal carriage point prevalence increased with age, reaching a maximum at 64% after 24 weeks. We showed no differences in the overall serotype-specific acquisition rates or duration of carriage between PCV13 and non-PCV13 serotypes.
The median age at first pneumococcal acquisition (63 days) is similar to that observed in Bangladeshi 14 and Australian Aboriginal 23 birth cohorts (50% by at 56 days and 39% by 60 days respectively), but higher than those reported for infants in other African countries, including The Gambia (24 days) 15 and Kenya (39 days) 24 . However, both African studies were conducted prior to the introduction of PCV-7 into national immunisation schedules. These differences in age to first acquisition may reflect the impact of herd protection from PCV vaccination on force of exposure, differences in living conditions, or host susceptibility. NP sampling frequency may also affect measures of carriage; Vives et al. showed that pneumococci were isolated at least once from 36% vs. 79% of Costa Rican children who were sampled quarterly vs. weekly 25 . In our cohort, after the first acquisition of NP carriage, time to second acquisition was much reduced, perhaps reflecting that children are repeatedly exposed to pneumococci after reaching a certain age 26 .
Although the cumulative acquisition rates showed that children were 4 times more likely to acquire non-PCV13 serotypes than PCV13 serotypes, on average, individual PCV13 and non-PCV13 serotypes were acquired at an equivalent rate. Pre-PCV data from Kilifi, Kenya, showed higher acquisition rates for some PCV serotypes compared to our data (e.g., serotype 19F, 0.80 episodes/child-year vs. 0.22 episodes/child-year; serotype 6A, 0.66 episodes/child-year vs. 0.13 episodes/child-year). The converse was true for non-PCV serotypes (e.g., serotype 15B, 0.40 episodes/child-year vs. 0.88 episodes/child-year; serotype 21, 0.07 episodes/child-year vs. 0.33 episodes/child-year) 26 . Overall, in our PCV13 vaccinated study population, the risk of acquiring individual PCV serotypes was similar to that of acquiring non-PCV serotypes, although non-PCV13 serotypes were acquired more frequently than PCV13 serotypes.
Our current understanding of the serotype-specific immunity afforded by PCV is incomplete. In our study, 'residual' carriage of PCV13 pneumococci as children aged did not appear to be affected by sequential doses of PCV13. As reported elsewhere, 9V, 19F, 19A and 6A are the most predominant PCV13 serotypes in many settings with high PCV coverage 27 . Data from native American populations has shown that although PCV13 induces higher immunoglobulin G (IgG) concentrations and functional activity against 19F, the vaccine had no additional impact on 19F carriage compared to PCV7 28 . Residual 19F carriage and disease has similarly been documented in other populations that have introduced a 19F-containing vaccine [29][30][31] . It has been suggested that the more prevalent colonisers such as serogroups 6, 18, 19, and 23 incur lower metabolic costs associated with capsule expression and may have greater ability to form biofilms thereby resisting elimination by host mediated immune responses 32 . In addition, the continued circulation of PCV13 serotypes might be attributed to incompletely immunised older siblings or adults with waning immunity (particularly HIV-infected adults) who serve as reservoirs for transmission. Our finding that very few children were colonized by PCV13 serotypes 1 and 3, serotypes particularly associated with invasive disease, is consistent with data from elsewhere showing they are rarely detected among carriage isolates 33 .
Serotypes 15B/15C, 10A, 21, 16F, 9N, 13, 11A, 15A, 17F, 31 and 22A are currently among the most prevalent replacement, non-PCV13, serotypes globally 15,[34][35][36] . Serotype 15B/15C has emerged as one of the predominant serotypes recovered after PCV roll out in both low and high-income countries [37][38][39] . Much of this increase has been linked to the clonal expansion of dominant strains 37,39,40 . Data from serially sampled communities in the United Kingdom and the United States of America have shown that serotype 15B/15C became more dominant post PCV13 implementation where initially PCV13 serotype 19A was circulating at equal proportions at baseline 37,40 . Animal models suggest that 15B/15C is equally capable of causing middle ear infections as 19A 41 . An increase in 15B/15C IPD cases have been noted from the UK and USA 42,43 . A higher valency PCV vaccine, PCV15, is currently under development, and includes all PCV13 serotypes plus 22F and 33F 44 . In our cohort, serotypes 22F and 33F were uncommon, detected in two and three children respectively. However, these two serotypes are more commonly associated with invasive infection rather than colonization.
Children are often colonized by several different pneumococcal serotypes over the first years of life, and the less immunogenic serotypes (e.g., 6, 14, 19, and 23) tend to be carried within the nasopharynx for prolonged periods of time compared to the more immunogenic strains (e.g., 3, 12, and 33) 45 . The longer duration of carriage observed in our settings for PCV13 serotypes 9V, 19F, 19A, 6A, 6B and 14 are consistent with these findings (carriage durations ranging from 30 to 56 days). Serotypes 1 and 4 were carried for shorter periods of times (carriage duration 14 days each). Non-PCV13 serotypes were generally carried for similar periods of time as PCV13 serotypes with exception of serotype 35A which was carried for 65 days with wide confidence intervals. A recent genome-wide study from Malawi has shown that recombination rates in pneumococcal lineages increase with carriage duration and size of the capsular polysaccharide 32,46 . Continued surveillance of pneumococcal carriage and invasive disease is needed to monitor the impact of targeted vaccine strategies.
Limitations of this study include the inability to detect multiple pneumococcal carriage. Serotyping was performed on a single morphologically distinct pneumococcal colony per sample. It is hard to detect serotype mixtures in this way due to the subjectivity in picking colonies and the potentially low abundance of a second serotype. We are in the process of investigating multiple pneumococcal serotypes using multiplex PCR 47 and whole genome shotgun sequencing of total nucleic acid directly extracted from NP swabs. The clinical and public health implication of co-colonization by multiple pneumococcal serotypes is not well described and therefore warrants further investigation. Studies have suggested that multiple carriage is key in understanding microbial interactions, transfer of genetic material, impact of selective pressure on the broader microbiome as well as improving accuracy of vaccination-induced shifts in the NP ecology 48 .
We did not perform any genomic characterisation of our isolates and assumed that if an identical serotype was isolated at consecutive time points this represented the same strain. We may therefore have underestimated acquisition rates. We are currently undertaking whole genome sequence analysis of all pneumococci in the full birth cohort in order to address this issue. The lack of pneumococcal carriage or vaccination data from mothers as well as siblings living in the same household prevents detailed analysis of pneumococcal transmission patterns. Detailed analysis of risk factors for pneumococcal carriage was beyond the scope of the present study but is addressed in detail in another manuscript under review. In addition, the relatively small sample size (albeit very intensively sampled), and the small numbers of less-frequently detected serotypes detected affected the precision of some of our estimates of carriage.
In conclusion, our data show that the rate of pneumococcal acquisition and duration of carriage is serotype-specific, with residual PCV13 pneumococci still circulating despite high immunisation coverage. We detected no overall differences in time to first acquisition, acquisition rate or duration of carriage between PCV13 and non-PCV13 serotypes, however the most prevalent non-PCV13 serotypes were acquired more commonly than PCV13 serotypes, and non-PCV13 serotypes were carried cumulatively for longer than PCV13 serotypes.