This study is the largest investigation of the prevalence and possible function of the PI-1-encoded pilus in carriage isolates of S. pneumoniae. It also provides pneumococcal strain data from Southeast Asia, a densely populated region that is under-represented in the pneumococcal carriage and disease epidemiology literature. As serotypes are commonly used as the basis for defining transmission, we were interested in the frequency with which a molecular analysis may confound the analysis of transmission. By genotyping, discrepancies were discovered in only 3.5% of transmission events classified by serotype. We also demonstrated that pneumococci of the same serotype are predominantly clonal within a discrete carriage episode. We therefore concluded that serotype can be used to define transmission events and carriage episodes with relatively high confidence, but that the inclusion of genotyping is vital to ensure complete accuracy.
Because of the selection criteria for the isolates included in the study, we cannot describe the overall pilus prevalence for the population. However, in the 887 carriage isolates analysed, the pilus prevalence was 35.2%. Among the 34 serotypes analysed, we found that PI-1 was present only in NT isolates and nine other serotypes: 65% of PCV7 isolates were PI-1-positive, as compared with 9% of non-vaccine serotypes. Interestingly, Basset et al.
, examining nasopharyngeal and invasive pneumococcal isolates from the American Indian collection, also found that PCV7 strains were significantly more likely to be PI-1-positive than non-vaccine serotypes, but that the overall proportion of PI-1-positive isolates was slightly lower [13
]. They, along with other authors, demonstrated that PI-1-positive strains were contained within a small number of CCs [14
]. Indeed, in the current study, we demonstrated that PI-1-positive isolates were clustered predominantly within four dominant CCs (CC15, CC271, CC315, and CC802). The strong association between serotype/CC and pilus presence, along with strain selection criteria, and the different regional distribution of the clones, may explain the variability in PI-1 prevalence between different studies.
We did not demonstrate a pilus-attributable effect on pneumococcal transmissibility. However, there are limitations to our study that may be important confounders for this analysis. The carriage study was carried out in a densely populated refugee camp where 13% of the population are <5 years old, and there is likely to be frequent transmission of nasopharyngeal organisms both within families and between members of the community. Despite the use of a combination of serotype and ST to increase the accuracy of our transmissibility categories, we cannot exclude the possibility that ‘non-transmitted’ strains were effectively transmitted between mother or infant and others but not detected by us. Several studies have documented the clustering of pneumococcal serotypes and genotypes within families, which highlights the difficulty of assigning a definitive ‘non-transmitted’ label to isolates collected from an incomplete household group [26
]. However, we focused on mother–infant transmission, because we felt that the absence of a strain in one member of this pair would be the best marker for relative non-transmissibility in the early months of life. In addition, we included both concordant and discordant time-point pneumococcal serotype/ST identifications in the mother–infant pair as ‘transmitted’, as long as they occurred no more than 2 months apart. However, the study sampling frequency may have been too low to demonstrate transmission of serotypes carried for very short durations. Also, in the presence of multiple serotype carriage, a common occurrence in infancy and one that is underestimated by standard culture protocols [21
], a particular serotype may become undetectable for a period of time before re-emerging as the dominant serotype, and this may result in incorrect categorization regarding transmissibility.
Although, in the crude analysis, PI-1 presence was associated with longer first pneumococcal carriage episodes in infants, we could find no significant association between carriage duration and PI-1 at the individual serotype level. This is likely to be the result of the low numbers of carriage episodes of individual serotypes (resulting in wide CIs around the carriage duration estimates) and the restricted number of clones within each serotype. We could have increased the number of carriage episodes included in our analysis by looking at all carriage episodes of each serotype rather than restricting our investigation to each infant’s first-ever carriage episodes, but this would have introduced other confounding factors, such as the impact of previous carriage and immune factors on subsequent carriage episode duration. Therefore, it is possible that, as a result the study sample size, a small effect of pilus on either transmission or carriage duration may have been missed.
In conclusion, we found that Southeast Asian pneumococcal carriage isolates had a similar pilus prevalence to to that in previously described strain collections, which is helpful in the ongoing assessment of likely global coverage of a pilus subunit-containing vaccine. Despite its known role in pneumococcal attachment, we could not determine a clear impact of PI-1 presence on transmissibility or carriage duration.