Streptococcus pneumoniae is a common commensal of the human upper respiratory tract and nasopharynx, but can also cause pneumonia and invasive diseases such as sepsis or meningitis (Bogaert et al., 2004). Invasive pneumococcal disease (IPD) has a high mortality, and the overall mortality rate from IPD is higher in extreme age ranges, such as infants and the elderly (Wahl et al., 2018; O Brien et al., 2009). In the Netherlands, pneumococcal carriage rates are higher in children than in adults, with a prevalence of up to 80% at 2 years of age (Wyllie et al., 2016).

Pneumococcal carriage manifests as multiple carriage episodes of different serotypes. From birth, mucosal immunity builds up against different serotypes due to exposure, while immunity from maternal antibodies wanes. Host age is known to affect carriage prevalence and carriage duration of different serotypes (Stearns et al., 2015; Turner et al., 2012b), which is suggested to be driven by differences in immunity.(Wyllie et al., 2020) Studies in mice and humans showed evidence for age-dependent host–pathogen interactions involving interleukin (IL)-1 response in reaction to the pore-forming pneumolysin (ply) toxin (Kuipers et al., 2018). IgA secretion is important in clearing S. pneumoniae from host upper respiratory tract mucosa, and this secretion is more effective in previously exposed individuals, the adults (Binsker et al., 2020). Bacterial genetics has been shown to explain over 60% of the variability in carriage duration, and specifically that the presence of a bacteriophage inserted in a mediator of genomic competence was associated with a decreased carriage duration (Lees et al., 2017b).

Pneumococci are highly genetically variable, displaying over 100 diverse capsular serotypes (Ganaie et al., 2020), which are a major antigen and the strongest predictor of carriage prevalence (Croucher et al., 2018). Pneumococcal conjugate vaccines (PCVs), targeting up to 13 capsule serotypes with high burden of invasive disease, decrease the rate of nasopharyngeal carriage and invasive disease (Whitney et al., 2003; Poehling et al., 2006). Besides a direct effect of vaccination with a PCV on the disease burden in the target population, that is, young children, it also reduces the disease burden caused by pneumococci with vaccine serotypes in the population not eligible for vaccination through indirect protection from colonization – reducing carriage rates in children reduces overall transmission of the most invasive serotypes (Croucher et al., 2018; Desai et al., 2015; von Gottberg et al., 2014). However, the introduction of PCV has resulted in the replacement of serotypes not covered by the vaccine (Croucher et al., 2013; Corander et al., 2017), which in some countries reaches levels of invasive disease return towards pre-vaccine levels (Ladhani et al., 2018; Koelman et al., 2020).

As not all serotypes can be included in a conjugate vaccine, three perspectives leading to improved pneumococcal vaccination have been proposed: whole-cell vaccines (Malley et al., 2001; Campo et al., 2018), protein vaccines (Moffitt and Malley, 2016), or changing components in the conjugate vaccine in response to the circulating population.(Colijn et al., 2020) Whole-cell vaccination trials are ongoing, but efficacy remains unproven in human populations (Morais et al., 2019). Protein vaccines contain antigens that elicit a strong mucosal immune response, with their targets chosen to be common or conserved in the target population, and ideally reducing onward transmission (Pichichero, 2017). In their current form, protein vaccines are not thought to be effective on their own, but if administered with serotype conjugates (possibly by replacing the carrier protein) they may help to reduce serotype replacement. Detailed modeling of the dynamics of pneumococcal population genetics has shown that targeting these vaccines towards serotypes prevalent in specific populations would likely be a superior strategy. This work further shows that providing age-specific vaccine design using complementary adult-administered vaccines (CAVs) is predicted to have the greatest effect on total IPD burden (Colijn et al., 2020). These authors also modeled including pilus in the vaccine, which is more frequently present in a few key invasive serotypes and in infant carriage, but found this to be inferior to conjugate vaccines.

If proposing a future pneumococcal vaccination strategy based on host age, we should aim to better understand the differences between infant and adult carriage. Differences between other host niches have been found, some with a potential effect on onward transmission (Lees et al., 2017c; Lees et al., 2017a; Zafar et al., 2019). In particular, a previous study has suggested that the presence of pilus, which is found in a minority of pneumococcal isolates, has a selective advantage in infant carriage (Binsker et al., 2020).

We therefore wished to test three hypotheses. Firstly, carriage rates of individual strains or serotypes vary substantially between infants and adults in the same contact networks. Secondly, this variation is attributable at least in part due to pathogen genetic adaptation to either the infant or adult nasopharynx, which are immunologically different niches. Finally, this adaptation is due both to serotype and genetic background, and that some of the genetic effects are resolvable to individual genes, alleles, or regulatory variants that arise on multiple different genetic backgrounds due to a selective advantage. If we find clear associations, this would support proposals for age-specific vaccine design and may also suggest specific protein components that more broadly suppress carriage in the target age group than multivalent PCV alone.

For the first hypothesis, to test varying rates of carriage we can swab children and adults from the same population (and likely exposed to the same infection pressures) and quantify serotype and strain prevalence. For the second hypothesis, if we whole-genome sequence bacteria from these swabs, we can model the effects of all genomic variants on host age in a heritability analysis. This accounts for genetic variation that arose long ago and is fixed in particular lineage, but cannot map it to a particular region of the genome (Earle et al., 2016). To find individual genetic effects, which have necessarily occurred more recently and frequently, one study design would be to identify adult–infant transmission pairs and find variation that consistently occurs in localized regions of the genome, which would be particularly informative if also associated with a particular direction of transmission (Lees et al., 2017c). This removes genetic background as a confounder, giving a clean signal (Young et al., 2012). However, the identification of such pairs is very challenging for S. pneumoniae, and even when possible the small numbers limit power. We propose taking the more ‘opportunistic’ approach taken in genome-wide association studies (GWAS), where as many cases and controls (in this case, infant and adult samples) as possible are accumulated to boost statistical power, and genetic background is then controlled for in the association analysis. Where variation associated with age has arisen independently on multiple genetic backgrounds, GWAS has the ability to find these signals among the many lineage associations tested in the second step. In all cases, analysis can be improved by studying more than one population to determine whether findings are consistent among different host and pathogen populations.

To carry out these analyses, we used pneumococci isolated from nasopharyngeal swabs of 4320 infants and adults from the Netherlands (2009–2013) and Myanmar (October 2007–November 2008). Each cohort contains infant and adult samples from carriage, and there are significant differences between the host populations. This allows us to follow the above approach in each population and compare our findings between the populations. We present our findings with respect to each above hypothesis in turn and interpret them through the lens of using population genomics to determine optimal vaccination strategies.

We first analyzed the observed distribution of serotypes and strains in each of the two cohorts to assess overall trends of differences in carriage between adults and children exposed to similar forces of infection and look at the pathogen population’s genetic heterogeneity between the two cohorts. Although our cohorts were broadly matched in the primary phenotype, age, large differences between the pathogen population are expected due to different geographies, social backgrounds, and only children in one cohort being vaccinated. Nucleotide variation across the entire genome can be used to cluster genetically related isolates into consistently named strains, called global pneumococcal sequence clusters (GPSC) (Gladstone et al., 2019). We used this over older gene-by-gene approaches such as MLST as it has been shown to represent biologically discrete clusters in the population, uses the full resolution available from whole-genome sequencing, and has strong community support (Gladstone et al., 2020). For each sample, we enumerate the serotype (which is targeted by the vaccine) and GPSC membership, and count the number of each serotype observed in adult and child carriage.

Serotypes and strains are variably carried between age groups and between cohorts

The Dutch cohort was made up of 1329 S. pneumoniae isolates comprising 41 unique serotypes (Supplementary file 1). Of these isolates, 689 (52%) comprised seven serotypes: 19A (225; 17%), 11A (111; 8%), 6C (97; 7%), 23B (84; 6%), 10A (67; 5%), 16F (54; 4%), and 21 (51; 4%). In this cohort of which the children were vaccinated, a minority of isolates (101; 8%) belonged to one of the vaccine serotypes (Supplementary file 2). A total of 3085 pneumococcal isolates of the Maela (unvaccinated) cohort comprised 64 unique serotype groups (Supplementary file 3). Of these isolates, 1631 (53%) comprised five serotypes: non-typable (511; 17%), 19F (402, 13%), 23F (307, 10%), 6B (236; 8%), and 14 (175; 6%). In the Dutch cohort, there were 59 unique sequence clusters of which the four largest sequence clusters were GPSC 4 (171; 13%), GPSC 3 (156; 12%), GPSC 7 (131; 10%), and GPSC 11 (119; 9%) (Supplementary file 4). There were 127 unique sequence clusters found in the Maela cohort (Supplementary file 5). The four largest sequence clusters were GPSC 1 (352; 13%), GPSC 28 (190; 7%), GPSC 20 (168; 6%), and GPSC 42 (123; 5%). We also looked at a subset of the Maela cohort, which included only the earliest obtained samples from unique individuals (mothers and children). This subset consisted of 762 isolates, including 380 from mother–child pairs. Isolates in this subset had the same serotypes among the most common serotypes as in the full dataset (Supplementary file 6 and Supplementary file 7). Restricting this subset to mother–child paired samples only included the same serotypes and sequence clusters among the most prevalent (Supplementary file 6 and Supplementary file 7).

Some serotypes exhibited a large difference in colonization frequency between the two age groups (Figure 1). In the Dutch cohort, serotype 6C (chi-squared test, p=0.02, not corrected for multiple testing) and serotype 15B (p=0.02) were overrepresented in children relative to adults, serotype 3 was overrepresented in adults relative to children (p=2.5 × 10^–5), while in the Maela cohort, serotype groups overrepresented in children were serotype 23F (chi-squared test, p=0.04) and serotype 6B (chi-squared test, p=0.04); while non-typeable serogroup was overrepresented in adults (chi-squared test, p=3.0 × 10^–4) (Table 1). None of the 20 largest groups of sequence clusters overlapped between the cohorts. In the Dutch cohort, only GPSC 11 was significantly associated with carriage in children (chi-squared test, p=0.03, not corrected for multiple testing), while GPSC 12 (chi-squared test, p=1.2 × 10^–4) and GPSC 38 (chi-squared test, p=2.1 × 10^–4) were overrepresented in adults. In the Maela cohort, only sequence cluster GPSC 128 was overrepresented in children compared to adults (chi-squared test, p=0.04), while GPSC 20 (chi-squared test, p=7.0 × 10^–3) and GPSC 74 (chi-squared test, p=0.04) were overrepresented in adults (Table 2).

Figure 1 with 1 supplement see all

Download asset Open asset

Table 1

Serotype	Dutch cohort		Maela cohort
	χ2 p-value	Age group	χ2 p-value	Age group
Non-typeable	0.188	Adults	3.0 × 10–4	Adults
19A	0.089	Children	0.690	Children
11A	0.591	Children	0285	Adults
19F	1	Adults	0.131	Children
6C	0.022	Children	1	Adults
6B	0.099	Children	0.040	Children
35F	0.279	Children	0.100	Children
3	2.5 × 10–5	Adults	0.129	Adults
6A	0.709	Children	1	Children
23A	1	Adults	-	-
15B	0.023	Children	-	-
17F	0.943	Children	-	-
23B	0.727	Children	-	-
10A	0.155	Adults	-	-
15C	1.000	Adults	-	-
35B	0.775	Adults	-	-
22F	1	Adults	-	-
33F	0.132	Adults	-	-
23F	-	-	0.040	Children
14	-	-	0.949	Children
35C	-	-	0.961	Children
34	-	-	0.690	Children
13	-	-	0.756	Adults
10B	-	-	0.756	Adults
4	-	-	0.966	Children
5	-	-	0.710	Adults
33B	-	-	1	Children
28F	-	-	0.652	Children
19B	-	-	0.710	Adults
7F	-	-	0.971	Children
20	-	-	0.971	Children
18C	-	-	1	Adults

1. χ², chi-square; -, not applicable.

Table 2

GPSC	Dutch cohort		Maela cohort
	χ2 p-value	Age group	χ2 p-value	Age group
60	0.568	Adults	0.727	Adults
4	0.298	Children	-	-
3	0.392	Adults	-	-
7	0.858	Children	-	-
11	0.03	Children	-	-
35 and 36	0.617	Adults	-	-
29	0.049	Children	-	-
46	0.563	Children	-	-
75	0.666	Adults	-	-
19	0.978	Children	-	-
12	1.2 × 10–4	Adults	-	-
44	1	Adults	-	-
24	0.094	Children	-	-
49	1	Children	-	-
109	0.817	Adults	-	-
16	0.249	Adults	-	-
38	2.1 × 10–4	Adults	-	-
146	0.489	Children	-	-
99	1	Children	-	-
15	0.22	Adults	-	-
42	-	-	0.134	Children
1	-	-	0.276	Adults
28	-	-	0.110	Children
73	-	-	0.253	Children
10	-	-	0.777	Adults
9	-	-	1	Children
30	-	-	0.993	Children
20	-	-	7.0 × 10–3	Adults
128	-	-	0.042	Children
66	-	-	1	Children
87	-	-	0.450	Adults
63	-	-	1	Adults
45	-	-	0.129	Adults
130	-	-	1	Adults
74	-	-	0.040	Adults
149	-	-	0.686	Adults
8	-	-	0.364	Children
25	-	-	1	Adults
187	-	-	0.371	Adults
154	-	-	1	Adults
118	-	-	0995	Children
110	-	-	0.995	Children
106	-	-	0.073	Adults

1. χ², chi-square; -, not applicable; GPSC, global pneumococcal sequence clusters.

A phylogenetic tree of pooled sequences from both cohorts, with serotype, sequence cluster, age group, and cohort for each sequence, revealed clonal discrimination between cohorts (Figure 2). Combined with the effects shown in Figure 1, this highlighted a key feature of our analysis of these datasets, which was the genetic heterogeneity between the two cohorts. Individually, each dataset clearly has strains and serotypes with strong signals of host age differences, but the overall makeup of each dataset is very different (nine common serotypes are shared, but only a single common GPSC), and where there are shared serotypes these can have different effect directions between the two cohorts.

Figure 2 with 1 supplement see all

Download asset Open asset

Host age is heritable and mostly explained by strain and serotype

To quantify the amount of variability in carriage age explained by variability in the genome, we calculated a heritability estimate (h²) for each cohort. For isolates in the Dutch cohort, we did not find strong evidence that genetic variability in bacteria was related to variance in host age (h² = 0.10, 95% CI 0.00–0.69). In the Maela cohort, we found significant evidence that affinity with host age was heritable (h² = 0.56, 95% CI 0.23–0.87) and thus genetic variation in this cohort explained variation in carriage age to a greater degree. In both cohorts, pan-genomic variation could be used to predict host age to some degree of accuracy (area under the receiver-operating characteristic [ROC] curve 0.82 [Dutch cohort]; 0.91 [Maela cohort]), suggestive of some level of heritability and association of host age with strain (Figure 3). Prediction between cohorts using a simple linear model failed as the genetic variants chosen as predictors were not found in the other cohort – again highlighting the high level of genetic heterogeneity between cohorts.

Figure 3

Download asset Open asset

To further investigate the association of serotype and sequence cluster to carriage age, we determined the proportion of variation in carriage age explained by serotype and sequence cluster alone. Here, we estimated h²_serotype = 0.07 (95% CI 0.04–0.14) and h²_GPSC = 0.06 (95% CI 0.03–0.13) for the Dutch cohort and h²_serotype = 0.11 (95% CI 0.05–0.21) and h²_GPSC = 0.20 (95% CI 0.12–0.31) for the Maela cohort, confirming a larger contribution of serotype and sequence cluster to carriage age heritability in sequences from the Dutch cohort. We also performed a genome-wide association analysis, but without controlling for population structure. This reveals genetic variants specific to serotype as determinants for carriage age (p-values<5.0 × 10^–8) in both cohorts (Supplementary file 8, Dutch cohort, and Supplementary file 9, Maela cohort). Among the genetic variants with the lowest p-values were variants in capsule locus genes (Cps) in both cohorts. This further supports a role of strain and serotype in association with host age, but does not distinguish between the two.

Genome-wide association analysis does not find genetic variants independent of strain

Following these observations that serotype and strain do not explain the full heritability, specifically in the Maela cohort, we performed a pathogen genome-wide association analysis to investigate whether we can detect genetic variants irrespective of the genetic background that are associated with carriage in children or adults. Though the cohorts have little genetic overlap in terms of genetic background, we would be well-powered to detect genetic variation independent of background (‘locus’ associations) (Earle et al., 2016; Lees et al., 2016). In the Dutch cohort, none of the unitigs, SNPs, COGs, or rare variants surpassed the threshold for multiple testing correction (Figure 4—figure supplement 1). The burden (sum) of rare variants in a gene for tryptophan synthase, trpB, approaches the multiple testing threshold, but was not significant. In the Maela cohort, unitigs in the ugpA gene surpassed the threshold for statistical significance (Figure 4—figure supplement 2), but these did not hold after meta-analysis. After meta-analysis, there were two hits that surpassed the threshold for statistical significance (Figure 4).

Figure 4 with 2 supplements see all

Download asset Open asset

The first is a nucleotide sequence marked by multiple unitigs of which the lowest has a p-value of 1.2 × 10^–9 (Supplementary file 10). This sequence does not map to the S. pneumoniae D39V reference sequence (Altschul et al., 1990) For this reason, it is not visualized on the Manhattan plot, for which unitigs were mapped to the S. pneumoniae D39V reference sequence (Figure 4). Upon inspection of the individual sequences these unitigs are called from, we find them to map in the intergenic region between open-reading frames encoding the accessory Sec-dependent serine-rich glycoprotein adhesin and a MarR-like regulator, respectively. This region contains sequences resembling transposable elements and an open-reading frame encoding a transposase. The unitigs map upstream of the start codon of the accessory Sec-dependent serine-rich glycoprotein adhesin. The sequence is present in 169 out of 1282 (13%) sequences in the Dutch cohort and in 241 out of 3085 (8%) in the Maela cohort. The sequence is present in isolates dispersed over the phylogenetic tree and associated with carriage in children (Figure 2—figure supplement 1). This protein is involved in adhesion to epithelial cells and biofilm formation (Chan et al., 2020; Middleton et al., 2021; Weiser et al., 2018). Given that this sequence lies just upstream of the start codon, it is plausible that variation of this sequence alters the expression of the Sec-dependent adhesin protein, and therefore affects carriage.

The second hit is a burden of rare variants in a gene for tryptophan synthase, trpB, that surpass the threshold for statistical significance at a p-value of 5.0 × 10^–5. The variants are two frameshift variants of very low frequency. These result in a predicted dysfunctional trpB gene in 9 out of 1282 (1%) sequences in the Dutch cohort and in 12 out of 3073 (0.4%) sequences in the Maela cohort. This association of the trpB gene is likely to be an artifact of low allele frequency as we estimate we are only powered to detect variation in at least 5% of isolates.

The age of the host is known to have an important effect on pneumococcal colonization (Croucher et al., 2018). Observational studies have demonstrated variation in serotype prevalence and carriage duration between infants and adults. Mechanistic studies in mice and humans have shown examples of differing immune responses depending both on host factors and pathogen factors. Findings from these studies include the observation that capsular polysaccharides (determinants of serotype) inhibit phagocytic clearance in animal models of upper respiratory tract colonization (Nelson et al., 2007). A pneumolysin-induced IL-1 response determined colonization persistence in an age-dependent manner Kuipers et al., 2018; and pilus-expressing strains were found to preferentially colonize children because of immune exclusion via secretory IgA in non-naïve hosts (Binsker et al., 2020).

Building upon these observations, we sought to investigate and quantify the contribution of pathogen genetic variation to carriage in infant versus adult hosts using a top-down approach to systematically test the effect of pathogen genome variation on niche specificity (with age as the niche). We used whole-genome sequencing and applied statistical genetic methods to two large S. pneumoniae carriage cohorts. We aimed to quantify the differing patterns of serotype and strain prevalence between the two age classes during carriage and search for other genetic factors associated with host age. We show evidence that bacterial genetic variability indeed influences predilection for host age, though the effect size appears to be highly variable between populations.

Strain, or genetic background, explains 60% of the total heritability in the Dutch cohort, but only a minority in the Maela cohort. We found sequences in one region that map closely to the start codon of the accessory Sec-dependent serine-rich glycoprotein adhesin to be associated with carriage age independent of genetic background, in a meta-analysis of the two cohorts.

In previous bacterial GWAS of antimicrobial resistance (such as a single gene that causes antibiotic resistance), large monogenic effects have typically been found to have high heritabilities close to one, and the GWAS identify the causal variant precisely.(Earle et al., 2016; Chewapreecha et al., 2014b; Lees et al., 2020). When applied to virulence and carriage duration phenotypes, heritable effects have also been found, but these only explained some of the variation in the phenotype. These appeared to be caused by many weaker effects, known as a polygenic trait, not all of which could be detected using the relatively small cohorts available (Lees et al., 2017b; Lees et al., 2019a). Polygenic effects were also seen in another study on the contribution of genetic variation to disease severity of IPD (Cremers et al., 2019). We found similar results for host age heritability in our two cohorts.

We could not distinguish between genetic background or serotype being the primary effect due to their correlation. We did note a difference in effect size of serotype between the two cohorts, which may make it unlikely to be the single largest effect on host age. This difference in cohorts could be explained by strain/GPSC being the main and consistent effect on host age. As strains are different between cohorts and each serotype appears in multiple strains, combining them in different amounts would create different directions of effect for serotype.

Our results are therefore suggestive of the following genetic architecture for association with host age. Primarily, whether a particular infant is colonized with a pneumococcus, when compared to an adult from the same population, is not due to the pathogen’s genetics. This may be due to technical factors such as detectability related to pathogen load (which varies between adults and children, as well as between serotypes), different local forces of exposure, or other environmental or host factors such as diet (which may affect survival to trpB-defective pneumococci). However, some of the variation of these patterns can be explained by the pathogen’s genetics. This appears largely to be driven by the fact that some strains and serotypes are more likely to be found in an infant or adult nasopharynx. In addition, there are likely to be many variants that each contribute a small amount to host age preference, but no single universally important gene or variant (a polygenic trait).

From this study, we cannot say specifically which regions of the genome contain these small effects, but it is useful to rule out recently adaptive variants with individually strong effects. We did not replicate the association of piliated genomes in infant hosts in our newly sequenced cohort, further demonstrating important differences between populations. An important corollary of our work is on future pneumococcal vaccine optimization efforts. A promising approach for future vaccination strategies is to target the different age groups (Colijn et al., 2020). Whether these should consist of the dominant disease-causing serotypes overrepresented in carriage by each age group or whether there are age-specific pathogen proteins that should be included is an open question. Our study therefore suggests that targeting these age groups using serotype makeup alone would be sufficient and supports previous observational and modeling studies that advise targeting the serotype makeup in the vaccine at specific populations to maximize their effect.

Three reasons can contribute to not finding individual effects: a high proportion of the heritability being caused by lineage effects; rare locus effects that could not be detected with the current sample size; and by sampling from a cohort with vaccinated children and unvaccinated adults and comparing with a cohort of unvaccinated children and adults, we had lower power due to the reduced overlap within and between cohorts in pan-genome content. Although differences in vaccination status between cohorts is one plausible explanation for interpreting our findings, we were unable to rule out other factors, for example, a population-specific host effect, geographical differences (Li et al., 2019), or the broad effects of different socioeconomic status between these cohorts. Pneumococcal factors such as differences in detectability due to carriage density could also influence results.

One important difference between our study cohorts was that children from the Dutch cohort were vaccinated, while children from the Maela cohort were not. While our findings demonstrate that vaccinated versus unvaccinated children were colonized with different bacterial serotypes and different sequence clusters, we observed differences in prevalence beyond just the serotypes included in the vaccine. Another difference between the cohorts was that adults from the Dutch cohort were males and females, while adults from the Maela cohort were females only.

In summary, we found an effect of pneumococcal genetics on carriage in children versus adult hosts, which varies between cohorts, and is likely primarily driven by serotype or strain (lineage) effects rather than large population-wide effects of individual genes.

Fastq sequences of bacterial isolates from the Dutch cohort were deposited in the European Nucleotide Archive (ENA, study and accession numbers in Supplementary file 12). Sequences of bacterial isolates in the Maela cohort are available at ENA under study numbers ERP000435, ERP000483, ERP000485, ERP000487, ERP000598 and ERP000599 (Supplementary file 13). Summary statistics for the results from the genome wide association studies can be found at https://figshare.com/articles/dataset/S_pneumoniae_carriage_GWAS/14431313.

1. 1. Kremer PHC

   (2020) ENA

   ID PRJEB2357. Streptococcus pneumoniae evolution and population structure during longitudinal sampling in a defined human population.

   https://www.ebi.ac.uk/ena/browser/view/PRJEB2357

1. 1. Altschul SF
   2. Gish W
   3. Miller W
   4. Myers EW
   5. Lipman DJ

   (1990) Basic local alignment search tool

   Journal of Molecular Biology 215:403–410.

   https://doi.org/10.1016/S0022-2836(05)80360-2

   • PubMed
   • Google Scholar
2. 1. Binsker U
   2. Lees JA
   3. Hammond AJ
   4. Weiser JN

   (2020) Immune exclusion by naturally acquired secretory IgA against pneumococcal pilus-1

   The Journal of Clinical Investigation 130:927–941.

   https://doi.org/10.1172/JCI132005

   • PubMed
   • Google Scholar
3. 1. Bogaert D
   2. van Belkum A
   3. Sluijter M
   4. Luijendijk A
   5. de Groot R
   6. Rümke HC
   7. Verbrugh HA
   8. Hermans PWM

   (2004) Colonisation by Streptococcus pneumoniae and Staphylococcus aureus in healthy children

   Lancet 363:1871–1872.

   https://doi.org/10.1016/S0140-6736(04)16357-5

   • PubMed
   • Google Scholar
4. 1. Bosch AATM
   2. van Houten MA
   3. Bruin JP
   4. Wijmenga-Monsuur AJ
   5. Trzciński K
   6. Bogaert D
   7. Rots NY
   8. Sanders EAM

   (2016) Nasopharyngeal carriage of Streptococcus pneumoniae and other bacteria in the 7th year after implementation of the pneumococcal conjugate vaccine in the Netherlands

   Vaccine 34:531–539.

   https://doi.org/10.1016/j.vaccine.2015.11.060

   • PubMed
   • Google Scholar
5. 1. Campo JJ
   2. Le TQ
   3. Pablo JV
   4. Hung C
   5. Teng AA
   6. Tettelin H
   7. Tate A
   8. Hanage WP
   9. Alderson MR
   10. Liang X
   11. Malley R
   12. Lipsitch M
   13. Croucher NJ

   (2018) Panproteome-wide analysis of antibody responses to whole cell pneumococcal vaccination

   eLife 7:e37015.

   https://doi.org/10.7554/eLife.37015

   • PubMed
   • Google Scholar
6. 1. Chan JM
   2. Gori A
   3. Nobbs AH
   4. Heyderman RS

   (2020) Streptococcal serine-rich repeat proteins in colonization and disease

   Frontiers in Microbiology 11:593356.

   https://doi.org/10.3389/fmicb.2020.593356

   • PubMed
   • Google Scholar
7. 1. Chewapreecha C
   2. Harris SR
   3. Croucher NJ
   4. Turner C
   5. Marttinen P
   6. Cheng L
   7. Pessia A
   8. Aanensen DM
   9. Mather AE
   10. Page AJ
   11. Salter SJ
   12. Harris D
   13. Nosten F
   14. Goldblatt D
   15. Corander J
   16. Parkhill J
   17. Turner P
   18. Bentley SD

   (2014a) Dense genomic sampling identifies highways of pneumococcal recombination

   Nature Genetics 46:305–309.

   https://doi.org/10.1038/ng.2895

   • PubMed
   • Google Scholar
8. 1. Chewapreecha C
   2. Marttinen P
   3. Croucher NJ
   4. Salter SJ
   5. Harris SR
   6. Mather AE
   7. Hanage WP
   8. Goldblatt D
   9. Nosten FH
   10. Turner C
   11. Turner P
   12. Bentley SD
   13. Parkhill J

   (2014b) Comprehensive identification of single nucleotide polymorphisms associated with beta-lactam resistance within pneumococcal mosaic genes

   PLOS Genetics 10:e1004547.

   https://doi.org/10.1371/journal.pgen.1004547

   • PubMed
   • Google Scholar
9. 1. Colijn C
   2. Corander J
   3. Croucher NJ

   (2020) Designing ecologically optimized pneumococcal vaccines using population genomics

   Nature Microbiology 5:473–485.

   https://doi.org/10.1038/s41564-019-0651-y

   • PubMed
   • Google Scholar
10. 1. Corander J
    2. Fraser C
    3. Gutmann MU
    4. Arnold B
    5. Hanage WP
    6. Bentley SD
    7. Lipsitch M
    8. Croucher NJ

    (2017) Frequency-dependent selection in vaccine-associated pneumococcal population dynamics

    Nature Ecology & Evolution 1:1950–1960.

    https://doi.org/10.1038/s41559-017-0337-x

    • PubMed
    • Google Scholar
11. 1. Cremers AJH
    2. Mobegi FM
    3. van der Gaast-de Jongh C
    4. van Weert M
    5. van Opzeeland FJ
    6. Vehkala M
    7. Knol MJ
    8. Bootsma HJ
    9. Välimäki N
    10. Croucher NJ
    11. Meis JF
    12. Bentley S
    13. van Hijum S
    14. Corander J
    15. Zomer AL
    16. Ferwerda G
    17. de Jonge MI

    (2019) The contribution of genetic variation of Streptococcus pneumoniae to the clinical manifestation of invasive pneumococcal disease

    Clinical Infectious Diseases 68:61–69.

    https://doi.org/10.1093/cid/ciy417

    • PubMed
    • Google Scholar
12. 1. Croucher NJ
    2. Harris SR
    3. Fraser C
    4. Quail MA
    5. Burton J
    6. van der Linden M
    7. McGee L
    8. von Gottberg A
    9. Song JH
    10. Ko KS
    11. Pichon B
    12. Baker S
    13. Parry CM
    14. Lambertsen LM
    15. Shahinas D
    16. Pillai DR
    17. Mitchell TJ
    18. Dougan G
    19. Tomasz A
    20. Klugman KP
    21. Parkhill J
    22. Hanage WP
    23. Bentley SD

    (2011) Rapid pneumococcal evolution in response to clinical interventions

    Science 331:430–434.

    https://doi.org/10.1126/science.1198545

    • PubMed
    • Google Scholar
13. 1. Croucher NJ
    2. Finkelstein JA
    3. Pelton SI
    4. Mitchell PK
    5. Lee GM
    6. Parkhill J
    7. Bentley SD
    8. Hanage WP
    9. Lipsitch M

    (2013) Population genomics of post-vaccine changes in pneumococcal epidemiology

    Nature Genetics 45:656–663.

    https://doi.org/10.1038/ng.2625

    • PubMed
    • Google Scholar
14. 1. Croucher NJ
    2. Løchen A
    3. Bentley SD

    (2018) Pneumococcal vaccines: Host interactions, population dynamics, and design principles

    Annual Review of Microbiology 72:521–549.

    https://doi.org/10.1146/annurev-micro-090817-062338

    • PubMed
    • Google Scholar
15. 1. Desai AP
    2. Sharma D
    3. Crispell EK
    4. Baughman W
    5. Thomas S
    6. Tunali A
    7. Sherwood L
    8. Zmitrovich A
    9. Jerris R
    10. Satola SW
    11. Beall B
    12. Moore MR
    13. Jain S
    14. Farley MM

    (2015) Decline in pneumococcal nasopharyngeal carriage of vaccine serotypes after the introduction of the 13-valent pneumococcal conjugate vaccine in children in atlanta, georgia

    The Pediatric Infectious Disease Journal 34:1168–1174.

    https://doi.org/10.1097/INF.0000000000000849

    • PubMed
    • Google Scholar
16. 1. Earle SG
    2. Wu CH
    3. Charlesworth J
    4. Stoesser N
    5. Gordon NC
    6. Walker TM
    7. Spencer CCA
    8. Iqbal Z
    9. Clifton DA
    10. Hopkins KL
    11. Woodford N
    12. Smith EG
    13. Ismail N
    14. Llewelyn MJ
    15. Peto TE
    16. Crook DW
    17. McVean G
    18. Walker AS
    19. Wilson DJ

    (2016) Identifying lineage effects when controlling for population structure improves power in bacterial association studies

    Nature Microbiology 1:16041.

    https://doi.org/10.1038/nmicrobiol.2016.41

    • PubMed
    • Google Scholar
17. 1. Ganaie F
    2. Saad JS
    3. McGee L
    4. van Tonder AJ
    5. Bentley SD
    6. Lo SW
    7. Gladstone RA
    8. Turner P
    9. Keenan JD
    10. Breiman RF
    11. Nahm MH

    (2020) A new pneumococcal capsule type, 10D, is the 100th serotype and has A large cps fragment from an oral streptococcus

    MBio 11:e00937-20.

    https://doi.org/10.1128/mBio.00937-20

    • PubMed
    • Google Scholar
18. 1. Gladstone RA
    2. Lo SW
    3. Lees JA
    4. Croucher NJ
    5. van Tonder AJ
    6. Corander J
    7. Page AJ
    8. Marttinen P
    9. Bentley LJ
    10. Ochoa TJ
    11. Ho PL
    12. du Plessis M
    13. Cornick JE
    14. Kwambana-Adams B
    15. Benisty R
    16. Nzenze SA
    17. Madhi SA
    18. Hawkins PA
    19. Everett DB
    20. Antonio M
    21. Dagan R
    22. Klugman KP
    23. von Gottberg A
    24. McGee L
    25. Breiman RF
    26. Bentley SD
    27. Global Pneumococcal Sequencing Consortium

    (2019) International genomic definition of pneumococcal lineages, to contextualise disease, antibiotic resistance and vaccine impact

    EBioMedicine 43:338–346.

    https://doi.org/10.1016/j.ebiom.2019.04.021

    • PubMed
    • Google Scholar
19. 1. Gladstone RA
    2. Lo SW
    3. Goater R
    4. Yeats C
    5. Taylor B
    6. Hadfield J
    7. Lees JA
    8. Croucher NJ
    9. van Tonder AJ
    10. Bentley LJ
    11. Quah FX
    12. Blaschke AJ
    13. Pershing NL
    14. Byington CL
    15. Balaji V
    16. Hryniewicz W
    17. Sigauque B
    18. Ravikumar KL
    19. Almeida SCG
    20. Ochoa TJ
    21. Ho PL
    22. du Plessis M
    23. Ndlangisa KM
    24. Cornick JE
    25. Kwambana-Adams B
    26. Benisty R
    27. Nzenze SA
    28. Madhi SA
    29. Hawkins PA
    30. Pollard AJ
    31. Everett DB
    32. Antonio M
    33. Dagan R
    34. Klugman KP
    35. von Gottberg A
    36. Metcalf BJ
    37. Li Y
    38. Beall BW
    39. McGee L
    40. Breiman RF
    41. Aanensen DM
    42. Bentley SD
    43. The Global Pneumococcal Sequencing Consortium

    (2020) Visualizing variation within Global Pneumococcal Sequence Clusters (GPSCS) and country population snapshots to contextualize pneumococcal isolates

    Microbial Genomics 6:1–13.

    https://doi.org/10.1099/mgen.0.000357

    • PubMed
    • Google Scholar
20. 1. Jaillard M
    2. Lima L
    3. Tournoud M
    4. Mahé P
    5. van Belkum A
    6. Lacroix V
    7. Jacob L

    (2018) A fast and agnostic method for bacterial genome-wide association studies: Bridging the gap between k-mers and genetic events

    PLOS Genetics 14:e1007758.

    https://doi.org/10.1371/journal.pgen.1007758

    • PubMed
    • Google Scholar
21. 1. Koelman DLH
    2. Brouwer MC
    3. van de Beek D

    (2020) Resurgence of pneumococcal meningitis in Europe and Northern America

    Clinical Microbiology and Infection 26:199–204.

    https://doi.org/10.1016/j.cmi.2019.04.032

    • PubMed
    • Google Scholar
22. Software

    1. Kremer PHC

    (2022a) Carriage_pneumo_heritability, version swh:1:rev:73c4fa5c8d24d76945308b2616fbb5572d0c39b4

    Software Heritage.

    https://archive.softwareheritage.org/swh:1:dir:9aff8f6bfe3d0e01ac82473092c0f7563264e8ae;origin=https://github.com/philipkremer123/carriage_pneumo_heritability;visit=swh:1:snp:ad5cacce05539197fc2af78f9fee57c3acee8c91;anchor=swh:1:rev:73c4fa5c8d24d76945308b2616fbb5572d0c39b4
23. Software

    1. Kremer PHC

    (2022b) Carriage-age-plots, version swh:1:rev:f9477d6b8382fee6926be7fd29b99afc15873fe8

    Software Heritage.

    https://archive.softwareheritage.org/swh:1:dir:bbeeac923adc3f5311e80ecd6ef3b1d229d76913;origin=https://github.com/johnlees/carriage-age-plots;visit=swh:1:snp:6e0685b66477f0fffcb2e8ff0e5aadedfe4a3e0c;anchor=swh:1:rev:f9477d6b8382fee6926be7fd29b99afc15873fe8
24. 1. Kuipers K
    2. Lokken KL
    3. Zangari T
    4. Boyer MA
    5. Shin S
    6. Weiser JN

    (2018) Age-related differences in IL-1 signaling and capsule serotype affect persistence of Streptococcus pneumoniae colonization

    PLOS Pathogens 14:10.

    https://doi.org/10.1371/journal.ppat.1007396

    • PubMed
    • Google Scholar
25. 1. Ladhani SN
    2. Collins S
    3. Djennad A
    4. Sheppard CL
    5. Borrow R
    6. Fry NK
    7. Andrews NJ
    8. Miller E
    9. Ramsay ME

    (2018) Rapid increase in non-vaccine serotypes causing invasive pneumococcal disease in England and Wales, 2000-17: a prospective national observational cohort study

    The Lancet. Infectious Diseases 18:441–451.

    https://doi.org/10.1016/S1473-3099(18)30052-5

    • PubMed
    • Google Scholar
26. 1. Lees JA
    2. Vehkala M
    3. Välimäki N
    4. Harris SR
    5. Chewapreecha C
    6. Croucher NJ
    7. Marttinen P
    8. Davies MR
    9. Steer AC
    10. Tong SYC
    11. Honkela A
    12. Parkhill J
    13. Bentley SD
    14. Corander J

    (2016) Sequence element enrichment analysis to determine the genetic basis of bacterial phenotypes

    Nature Communications 7:12797.

    https://doi.org/10.1038/ncomms12797

    • PubMed
    • Google Scholar
27. 1. Lees JA
    2. Brouwer M
    3. van der Ende A
    4. Parkhill J
    5. van de Beek D
    6. Bentley SD

    (2017a) Within-host sampling of a natural population shows signs of selection on pde1 during bacterial meningitis

    Infection and Immunity 85:e01061-16.

    https://doi.org/10.1128/IAI.01061-16

    • PubMed
    • Google Scholar
28. 1. Lees JA
    2. Croucher NJ
    3. Goldblatt D
    4. Nosten F
    5. Parkhill J
    6. Turner C
    7. Turner P
    8. Bentley SD

    (2017b) Genome-wide identification of lineage and locus specific variation associated with pneumococcal carriage duration

    eLife 6:e26255.

    https://doi.org/10.7554/eLife.26255

    • PubMed
    • Google Scholar
29. 1. Lees JA
    2. Kremer PHC
    3. Manso AS
    4. Croucher NJ
    5. Ferwerda B
    6. Serón MV
    7. Oggioni MR
    8. Parkhill J
    9. Brouwer MC
    10. van der Ende A
    11. van de Beek D
    12. Bentley SD

    (2017c) Large scale genomic analysis shows no evidence for pathogen adaptation between the blood and cerebrospinal fluid niches during bacterial meningitis

    Microbial Genomics 3:e000103.

    https://doi.org/10.1099/mgen.0.000103

    • PubMed
    • Google Scholar
30. 1. Lees JA
    2. Galardini M
    3. Bentley SD
    4. Weiser JN
    5. Corander J

    (2018) pyseer: a comprehensive tool for microbial pangenome-wide association studies

    Bioinformatics 34:4310–4312.

    https://doi.org/10.1093/bioinformatics/bty539

    • PubMed
    • Google Scholar
31. 1. Lees JA
    2. Ferwerda B
    3. Kremer PHC
    4. Wheeler NE
    5. Serón MV
    6. Croucher NJ
    7. Gladstone RA
    8. Bootsma HJ
    9. Rots NY
    10. Wijmega-Monsuur AJ
    11. Sanders EAM
    12. Trzciński K
    13. Wyllie AL
    14. Zwinderman AH
    15. van den Berg LH
    16. van Rheenen W
    17. Veldink JH
    18. Harboe ZB
    19. Lundbo LF
    20. de Groot L
    21. van Schoor NM
    22. van der Velde N
    23. Ängquist LH
    24. Sørensen TIA
    25. Nohr EA
    26. Mentzer AJ
    27. Mills TC
    28. Knight JC
    29. du Plessis M
    30. Nzenze S
    31. Weiser JN
    32. Parkhill J
    33. Madhi S
    34. Benfield T
    35. von Gottberg A
    36. van der Ende A
    37. Brouwer MC
    38. Barrett JC
    39. Bentley SD
    40. van de Beek D

    (2019a) Joint sequencing of human and pathogen genomes reveals the genetics of pneumococcal meningitis

    Nature Communications 10:2176.

    https://doi.org/10.1038/s41467-019-09976-3

    • PubMed
    • Google Scholar
32. 1. Lees JA
    2. Harris SR
    3. Tonkin-Hill G
    4. Gladstone RA
    5. Lo SW
    6. Weiser JN
    7. Corander J
    8. Bentley SD
    9. Croucher NJ

    (2019b) Fast and flexible bacterial genomic epidemiology with PopPUNK

    Genome Research 29:304–316.

    https://doi.org/10.1101/gr.241455.118

    • PubMed
    • Google Scholar
33. 1. Lees JA
    2. Mai TT
    3. Galardini M
    4. Wheeler NE
    5. Horsfield ST
    6. Parkhill J
    7. Corander J

    (2020) Improved prediction of bacterial genotype-phenotype associations using interpretable pangenome-spanning regressions

    MBio 11:e01344-20.

    https://doi.org/10.1128/mBio.01344-20

    • PubMed
    • Google Scholar
34. 1. Li Y
    2. Metcalf BJ
    3. Chochua S
    4. Li Z
    5. Walker H
    6. Tran T
    7. Hawkins PA
    8. Gierke R
    9. Pilishvili T
    10. McGee L
    11. Beall BW

    (2019) Genome-wide association analyses of invasive pneumococcal isolates identify a missense bacterial mutation associated with meningitis

    Nature Communications 10:178.

    https://doi.org/10.1038/s41467-018-07997-y

    • PubMed
    • Google Scholar
35. Preprint

    1. Lippert C
    2. Casale FP
    3. Rakitsch B
    4. Stegle O

    (2014) LIMIX: Genetic Analysis of Multiple Traits

    bioRxiv.

    https://doi.org/10.1101/003905

    • Google Scholar
36. 1. Malley R
    2. Lipsitch M
    3. Stack A
    4. Saladino R
    5. Fleisher G
    6. Pelton S
    7. Thompson C
    8. Briles D
    9. Anderson P

    (2001) Intranasal immunization with killed unencapsulated whole cells prevents colonization and invasive disease by capsulated pneumococci

    Infection and Immunity 69:4870–4873.

    https://doi.org/10.1128/IAI.69.8.4870-4873.2001

    • PubMed
    • Google Scholar
37. 1. Middleton DR
    2. Aceil J
    3. Mustafa S
    4. Paschall AV
    5. Avci FY

    (2021) Glycosyltransferases within the psrP locus facilitate pneumococcal virulence

    Journal of Bacteriology 203:e00389-20.

    https://doi.org/10.1128/JB.00389-20

    • PubMed
    • Google Scholar
38. 1. Moffitt K
    2. Malley R

    (2016) Rationale and prospects for novel pneumococcal vaccines

    Human Vaccines & Immunotherapeutics 12:383–392.

    https://doi.org/10.1080/21645515.2015.1087625

    • PubMed
    • Google Scholar
39. 1. Morais V
    2. Texeira E
    3. Suarez N

    (2019) Next-Generation Whole-Cell Pneumococcal Vaccine

    Vaccines 7:E151.

    https://doi.org/10.3390/vaccines7040151

    • PubMed
    • Google Scholar
40. 1. Nelson AL
    2. Roche AM
    3. Gould JM
    4. Chim K
    5. Ratner AJ
    6. Weiser JN

    (2007) Capsule enhances pneumococcal colonization by limiting mucus-mediated clearance

    Infection and Immunity 75:83–90.

    https://doi.org/10.1128/IAI.01475-06

    • PubMed
    • Google Scholar
41. 1. Nguyen L-T
    2. Schmidt HA
    3. von Haeseler A
    4. Minh BQ

    (2015) IQ-TREE: A fast and effective stochastic algorithm for estimating maximum-likelihood phylogenies

    Molecular Biology and Evolution 32:268–274.

    https://doi.org/10.1093/molbev/msu300

    • PubMed
    • Google Scholar
42. 1. O Brien KL
    2. Nohynek H
    3. Watt JP
    4. World Health Organization Pneumococcal Vaccine Trials Carriage Working Group

    (2003) Report from a WHO Working Group: standard method for detecting upper respiratory carriage of Streptococcus pneumoniae

    The Pediatric Infectious Disease Journal 22:e1–e11.

    https://doi.org/10.1097/01.inf.0000049347.42983.77

    • PubMed
    • Google Scholar
43. 1. O Brien KL
    2. Wolfson LJ
    3. Watt JP
    4. Henkle E
    5. Deloria-Knoll M
    6. McCall N
    7. Lee E
    8. Mulholland K
    9. Levine OS
    10. Cherian T
    11. Hib and Pneumococcal Global Burden of Disease Study Team

    (2009) Burden of disease caused by streptococcus pneumoniae in children younger than 5 years: global estimates

    Lancet 374:893–902.

    https://doi.org/10.1016/S0140-6736(09)61204-6

    • PubMed
    • Google Scholar
44. 1. Page AJ
    2. Cummins CA
    3. Hunt M
    4. Wong VK
    5. Reuter S
    6. Holden MTG
    7. Fookes M
    8. Falush D
    9. Keane JA
    10. Parkhill J

    (2015) Roary: rapid large-scale prokaryote pan genome analysis

    Bioinformatics 31:3691–3693.

    https://doi.org/10.1093/bioinformatics/btv421

    • PubMed
    • Google Scholar
45. 1. Page AJ
    2. De Silva N
    3. Hunt M
    4. Quail MA
    5. Parkhill J
    6. Harris SR
    7. Otto TD
    8. Keane JA

    (2016) Robust high-throughput prokaryote de novo assembly and improvement pipeline for Illumina data

    Microbial Genomics 2:e000083.

    https://doi.org/10.1099/mgen.0.000083

    • PubMed
    • Google Scholar
46. 1. Pichichero ME

    (2017) Pneumococcal whole-cell and protein-based vaccines: changing the paradigm

    Expert Review of Vaccines 16:1181–1190.

    https://doi.org/10.1080/14760584.2017.1393335

    • PubMed
    • Google Scholar
47. 1. Poehling KA
    2. Talbot TR
    3. Griffin MR
    4. Craig AS
    5. Whitney CG
    6. Zell E
    7. Lexau CA
    8. Thomas AR
    9. Harrison LH
    10. Reingold AL
    11. Hadler JL
    12. Farley MM
    13. Anderson BJ
    14. Schaffner W

    (2006) Invasive pneumococcal disease among infants before and after introduction of pneumococcal conjugate vaccine

    JAMA 295:1668–1674.

    https://doi.org/10.1001/jama.295.14.1668

    • PubMed
    • Google Scholar
48. 1. Schweiger R
    2. Fisher E
    3. Rahmani E
    4. Shenhav L
    5. Rosset S
    6. Halperin E

    (2018) Using stochastic approximation techniques to efficiently construct confidence intervals for heritability

    Journal of Computational Biology 25:794–808.

    https://doi.org/10.1089/cmb.2018.0047

    • PubMed
    • Google Scholar
49. 1. Seemann T

    (2014) Prokka: rapid prokaryotic genome annotation

    Bioinformatics 30:2068–2069.

    https://doi.org/10.1093/bioinformatics/btu153

    • PubMed
    • Google Scholar
50. 1. Spijkerman J
    2. Prevaes S
    3. van Gils EJM
    4. Veenhoven RH
    5. Bruin JP
    6. Bogaert D

    (2012) Long-term effects of pneumococcal conjugate vaccine on nasopharyngeal carriage of S pneumoniae, S aureus, H influenzae and M catarrhalis

    PLOS ONE 7:e39730.

    https://doi.org/10.1371/journal.pone.0039730

    • Google Scholar
51. 1. Stearns JC
    2. Davidson CJ
    3. McKeon S
    4. Whelan FJ
    5. Fontes ME
    6. Schryvers AB
    7. Bowdish DME
    8. Kellner JD
    9. Surette MG

    (2015) Culture and molecular-based profiles show shifts in bacterial communities of the upper respiratory tract that occur with age

    The ISME Journal 9:1246–1259.

    https://doi.org/10.1038/ismej.2014.250

    • PubMed
    • Google Scholar
52. 1. Trzciński K
    2. Bogaert D
    3. Wyllie A
    4. Chu MLJN
    5. van der Ende A
    6. Bruin JP
    7. van den Dobbelsteen G
    8. Veenhoven RH
    9. Sanders EAM

    (2013) Superiority of trans-oral over trans-nasal sampling in detecting Streptococcus pneumoniae colonization in adults

    PLOS ONE 8:e60520.

    https://doi.org/10.1371/journal.pone.0060520

    • PubMed
    • Google Scholar
53. 1. Turner P
    2. Melchiorre S
    3. Moschioni M
    4. Barocchi MA
    5. Turner C
    6. Watthanaworawit W
    7. Kaewcharernnet N
    8. Nosten F
    9. Goldblatt D

    (2012a) Assessment of Streptococcus pneumoniae pilus islet-1 prevalence in carried and transmitted isolates from mother-infant pairs on the Thailand-Burma border

    Clinical Microbiology and Infection 18:970–975.

    https://doi.org/10.1111/j.1469-0691.2011.03711.x

    • PubMed
    • Google Scholar
54. 1. Turner P
    2. Turner C
    3. Jankhot A
    4. Helen N
    5. Lee SJ
    6. Day NP
    7. White NJ
    8. Nosten F
    9. Goldblatt D

    (2012b) A longitudinal study of Streptococcus pneumoniae carriage in A cohort of infants and their mothers on the Thailand-Myanmar border

    PLOS ONE 7:e38271.

    https://doi.org/10.1371/journal.pone.0038271

    • PubMed
    • Google Scholar
55. 1. Turner P
    2. Turner C
    3. Jankhot A
    4. Phakaudom K
    5. Nosten F
    6. Goldblatt D

    (2013) Field evaluation of culture plus latex sweep serotyping for detection of multiple pneumococcal serotype colonisation in infants and young children

    PLOS ONE 8:e67933.

    https://doi.org/10.1371/journal.pone.0067933

    • PubMed
    • Google Scholar
56. 1. van Beek J
    2. Veenhoven RH
    3. Bruin JP
    4. van Boxtel RAJ
    5. de Lange MMA
    6. Meijer A
    7. Sanders EAM
    8. Rots NY
    9. Luytjes W

    (2017) Influenza-like illness incidence is not reduced by influenza vaccination in a cohort of older adults, despite effectively reducing laboratory-confirmed influenza virus infections

    The Journal of Infectious Diseases 216:415–424.

    https://doi.org/10.1093/infdis/jix268

    • PubMed
    • Google Scholar
57. 1. von Gottberg A
    2. de Gouveia L
    3. Tempia S
    4. Quan V
    5. Meiring S
    6. von Mollendorf C
    7. Madhi SA
    8. Zell ER
    9. Verani JR
    10. O’Brien KL
    11. Whitney CG
    12. Klugman KP
    13. Cohen C
    14. GERMS-SA Investigators

    (2014) Effects of vaccination on invasive pneumococcal disease in South Africa

    The New England Journal of Medicine 371:1889–1899.

    https://doi.org/10.1056/NEJMoa1401914

    • PubMed
    • Google Scholar
58. 1. Wahl B
    2. O’Brien KL
    3. Greenbaum A
    4. Majumder A
    5. Liu L
    6. Chu Y
    7. Lukšić I
    8. Nair H
    9. McAllister DA
    10. Campbell H
    11. Rudan I
    12. Black R
    13. Knoll MD

    (2018) Burden of Streptococcus pneumoniae and Haemophilus influenzae type b disease in children in the era of conjugate vaccines: global, regional, and national estimates for 2000-15

    The Lancet. Global Health 6:e744–e757.

    https://doi.org/10.1016/S2214-109X(18)30247-X

    • PubMed
    • Google Scholar
59. 1. Weiser JN
    2. Ferreira DM
    3. Paton JC

    (2018) Streptococcus pneumoniae: transmission, colonization and invasion

    Nature Reviews. Microbiology 16:355–367.

    https://doi.org/10.1038/s41579-018-0001-8

    • PubMed
    • Google Scholar
60. 1. Whitney CG
    2. Farley MM
    3. Hadler J
    4. Harrison LH
    5. Bennett NM
    6. Lynfield R
    7. Reingold A
    8. Cieslak PR
    9. Pilishvili T
    10. Jackson D
    11. Facklam RR
    12. Jorgensen JH
    13. Schuchat A
    14. Active Bacterial Core Surveillance of the Emerging Infections Program Network

    (2003) Decline in invasive pneumococcal disease after the introduction of protein-polysaccharide conjugate vaccine

    The New England Journal of Medicine 348:1737–1746.

    https://doi.org/10.1056/NEJMoa022823

    • PubMed
    • Google Scholar
61. 1. Willer CJ
    2. Li Y
    3. Abecasis GR

    (2010) METAL: fast and efficient meta-analysis of genomewide association scans

    Bioinformatics 26:2190–2191.

    https://doi.org/10.1093/bioinformatics/btq340

    • PubMed
    • Google Scholar
62. 1. Wyllie AL
    2. Wijmenga-Monsuur AJ
    3. van Houten MA
    4. Bosch AATM
    5. Groot JA
    6. van Engelsdorp Gastelaars J
    7. Bruin JP
    8. Bogaert D
    9. Rots NY
    10. Sanders EAM
    11. Trzciński K

    (2016) Molecular surveillance of nasopharyngeal carriage of Streptococcus pneumoniae in children vaccinated with conjugated polysaccharide pneumococcal vaccines

    Scientific Reports 6:23809.

    https://doi.org/10.1038/srep23809

    • PubMed
    • Google Scholar
63. 1. Wyllie AL
    2. Warren JL
    3. Regev-Yochay G
    4. Givon-Lavi N
    5. Dagan R
    6. Weinberger DM

    (2020) Serotype patterns of pneumococcal disease in adults are correlated with carriage patterns in older children

    Clinical Infectious Diseases: An Official Publication of the Infectious Diseases Society of America 72:ciaa1480.

    https://doi.org/10.1093/cid/ciaa1480

    • Google Scholar
64. 1. Young BC
    2. Golubchik T
    3. Batty EM
    4. Fung R
    5. Larner-Svensson H
    6. Votintseva AA
    7. Miller RR
    8. Godwin H
    9. Knox K
    10. Everitt RG
    11. Iqbal Z
    12. Rimmer AJ
    13. Cule M
    14. Ip CLC
    15. Didelot X
    16. Harding RM
    17. Donnelly P
    18. Peto TE
    19. Crook DW
    20. Bowden R
    21. Wilson DJ

    (2012) Evolutionary dynamics of Staphylococcus aureus during progression from carriage to disease

    PNAS 109:4550–4555.

    https://doi.org/10.1073/pnas.1113219109

    • PubMed
    • Google Scholar
65. 1. Zafar MA
    2. Hammond AJ
    3. Hamaguchi S
    4. Wu W
    5. Kono M
    6. Zhao L
    7. Weiser JN

    (2019) Identification of pneumococcal factors affecting pneumococcal shedding shows that the dlt locus promotes inflammation and transmission

    MBio 10:e01032-19.

    https://doi.org/10.1128/mBio.01032-19

    • PubMed
    • Google Scholar

Decision letter after peer review:

Thank you for submitting your article "Pneumococcal genetic variability influences age-dependent bacterial carriage" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by Bavesh Kana as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Amelieke Cremers (Reviewer #2).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

Major Comments:

1. The Maela set is that these are mother-child samples. These seem like excellent sets to establish genomic features associated with age. Can the data be plotted for paired mother and child? At least for the serotype and sequence types that appear age-related.

2. The work is performed to inform future pneumococcal vaccine formulations. The authors looked for genome-wide pneumococcal traits that made them more likely to colonize a human host of particular age category. Although the bioinformatics analyses are extensive and of high quality, the sampling strategy needs clarification before the conclusions can be evaluated.

3. The manuscript may benefit from a hypothesis on what was expected. Were parents expected to acquire pneumococcal carriage from sources outside of their children? Or did the authors expect differences in viability or detectability of particular pneumococcal variants in adult carriage – i.e. a subset? What question does this study set up answer.

4. Is it possible that the major finding is explained by sampling rather than a biological phenomenon? The eigenvalue decomposed distance in the kinship matrix is a measure to express the distance between isolates within a group. And the authors used this measure to signal overrepresentation of (undefined) genomic variants in either category. For infants, in the preceding serotype-based comparison only a first isolate of a carriage episode was included, discarding 75% of the (duplicate?) isolates. It is unclear whether repetitive isolates from adults and infants were included in the 'heritability analyses'. Could the authors clarify what samples of genomes were included in the kinship matrices, and how such choices may affect the representativity of results?

5. Presentation of the results could be improved. While it is understandable that the sequencing data need to be compressed to a presentable format, essential information needs to be logically displayed for the sake of readers' understanding. As examples, the first section of the result section mentioned total numbers of isolates and serotypes from each cohort, but did not say how many of them were from children/adults. Figure 1 does have age information, but it is difficult to evaluate due to data transformation. Sequence clusters were mentioned without elaboration on what they mean. This style of data presentation may be readily comprehensive for sequencing gurus, but is hard to digest to the experimentalists like myself.

Other comments that should be addressed:

1. Supplementary Tables S8 and 9 would benefit from an annotations (or representative gene ID) associated with each of the regions.

2. Paragraph starting line 214 and Figure 4: Given that genes were mapped to D39V, it would be very useful to provide an ID. Similarly, TableS10 has a column for Marker Name, but the IDs are not present only sequence segments.

3. The description of the performed 'heritability analyses' requires more detail to enable the readership to repeat the analyses.

4. The title of the manuscript may be reconsidered given the cross-sectional nature of the study performed.

5. The discussion should contain a comparison to previous studies that performed pneumococcal GWAS on host age (f.e. doi.org/10.1093/cid/ciy417). The observed geographical differences in pneumococcal elements being associated with phenotype could also be related to previous studies reporting on this phenomenon (f.e. doi: 10.1038/s41467-018-07997-y).

6. The authors may also want to discuss potential biases or alternative explanations for observed differences between infants and adults (f.e. detectability – while pneumococcal carriage density is lower in adults compared to children, serotype 3 generally grows with high load and is relatively more likely to be detected if present. Viability – Survival of trpB defective pneumococci may depend on age-dependent host factors like diet).

7. Why do the authors present epidemiological arguments to prefer either serotypes or proteins as vaccine targets, while serotypes strongly co-occur with specific proteins across the pneumococcal population that could evenly be targeted? If they choose to do so, however, the report would benefit from concrete numbers to support the claim that based on this study serotypes would be a preferential vaccine target. Which vaccine-formulations would the authors recommend for the infant and parent populations studied here?

Essential revisions:

Major Comments:

1. The Maela set is that these are mother-child samples. These seem like excellent sets to establish genomic features associated with age. Can the data be plotted for paired mother and child? At least for the serotype and sequence types that appear age-related.

Our understanding is that the suggestion here is to look at mother-child pairs where direct transmission had been identified. Indeed, this would be an excellent set to identify age-related features, as genetic background would no longer be a major confounder, and variants occurring across the bottleneck over multiple events could be pooled (as is often done in within-host studies), especially if direction of loss/gain could also be inferred.

However, reconstructing transmission pairs with single colony-pick samples alone has proved to be highly uncertain, such that we cannot infer them in this dataset (despite trying various methods previously). They are also relatively rare events to observe – many of the mother/child series have clearly non-overlapping carriage episodes.

A newer analysis of swabs from this cohort was designed specifically to detect these events by sequencing within-host diversity to help improve power to reconstruct these events (https://doi.org/10.1101/2022.02.20.480002), and targeted those swabs from timepoints likely to be involved in transmission events. Despite this, those authors found only 40 direct mother-infant transmissions. This is a low number to pool variants, and would be the best case scenario here. So, there is insufficient power in this study to (a) find these pairs and (b) find common variation between them. We also note that the genome samples do not overlap with this newer study so we cannot simply lift their analysis over.

The advantage of doing a GWAS over well-controlled pairs is that one can pool samples from a variety cases and controls, and after accounting for biases as much as possible, allow the larger sample sizes to give the power to find associations.

So, while we welcome this suggestion, due to sampling design of the cohort we do not think it possible here. However, we have added this suggestion while explaining our study rationale in the introduction, and how it differs from what we did, and why it is difficult to do with these samples.

2. The work is performed to inform future pneumococcal vaccine formulations. The authors looked for genome-wide pneumococcal traits that made them more likely to colonize a human host of particular age category. Although the bioinformatics analyses are extensive and of high quality, the sampling strategy needs clarification before the conclusions can be evaluated.

We have made significant edits to the introduction to better motivate and describe our study, also to address point three below. Our sampling strategy is basically an opportunistic one: we collected and collated as many population-matched adult and child samples as possible to maximize statistical power, and use established techniques from GWAS to account for genetic background confounding results (as well as other covariates, where appropriate).

Furthermore, we have clarified the sampling strategy in the methods section of the manuscript. On page 16, line 382 “Cohorts were selected based on availability.”. On line 677 we added “parent-child paired isolates of”. On page 16, line 395 – 396:

“This cohort consists of mother – child paired samples, some of which were sampled from the same mother or child over multiple time points, and unpaired samples from mothers and children. A subset from this cohort was made to reflect the first sampled isolates for each mother – child pair and unpaired samples, to obtain isolates belonging to unique individuals.” And in the section on “Sequencing characteristics and quality control”, page 17, line 441 – 442: “In a subset from the Maela cohort, there were 762 isolates from unique hosts, of which 380 were paired isolates (190 from children and 190 from their mothers).”

Isolates in this subset are listed in Supplementary File 4 and this table is now referenced clearly in the Results section.

We also felt, while addressing these concerns, we had given too much weight to implications on vaccine design, where fully modelling these effects are outside the scope of this study. We have therefore reworded some of the introduction and abstract to be more specific about the contribution this study actually makes in vaccine design, and what further work would be required before recommendations could be made.

3. The manuscript may benefit from a hypothesis on what was expected. Were parents expected to acquire pneumococcal carriage from sources outside of their children? Or did the authors expect differences in viability or detectability of particular pneumococcal variants in adult carriage – i.e. a subset? What question does this study set up answer.

Reading our introduction again we can see that the motivation for a genome-wide association study was not necessarily clear, and we appreciate the suggestion of framing in terms of hypotheses. We do not look for direct transmission events between adults and children, as noted above (and in answer to the question, broadly we expect parents will acquire S. pneumoniae both from their children and from others in the community). Previous studies have observed differences in carriage duration, shedding and strain makeup between adults and children, which leads to the question of which factors cause these differences. If there are genetic factors, these should be accounted for in any models of carriage (particularly those which underlie transmission models used to propose vaccine design).

We therefore test for three things using our ‘opportunistic’ sample collection:

1. Whether carriage rates of individual strains or serotypes varies substantially between infants and adults in the same contact networks.

2. Whether this variation is attributable at least in part due to pathogen genetic adaptation to either the infant or adult nasopharynx, which are immunologically different niches.

3. Finally, whether this adaptation is due both to serotype and genetic background, and that some of the genetic effects are resolvable to individual genes, alleles or regulatory variants which arise on multiple different genetic backgrounds due to a selective advantage.

If we find clear associations, this would support proposals for age-specific vaccine design, and may also suggest specific protein components which more broadly suppress carriage in the target age group than multivalent PCV alone.

Please see the rewritten introduction for full details.

4. Is it possible that the major finding is explained by sampling rather than a biological phenomenon? The eigenvalue decomposed distance in the kinship matrix is a measure to express the distance between isolates within a group. And the authors used this measure to signal overrepresentation of (undefined) genomic variants in either category. For infants, in the preceding serotype-based comparison only a first isolate of a carriage episode was included, discarding 75% of the (duplicate?) isolates. It is unclear whether repetitive isolates from adults and infants were included in the 'heritability analyses'. Could the authors clarify what samples of genomes were included in the kinship matrices, and how such choices may affect the representativity of results?

Thank you for this important remark. Only the Maela cohort had multiple samples isolated longitudinally for individual children, but using estimates from a previous study we were able to find those samples which are from the same carriage episode. These samples were removed so only one sample is taken from any given carriage episode. The heritability analyses for this cohort have now been repeated with unique samples only. This did not significantly change our conclusions. For the genome-wide association study, we believe it was best to retain all samples, because in this case the kinship matrix adjusts for the repeated sampling, but retains any diversity accumulated across the carriage episode. Furthermore, the meta-analysis with the second cohort will remove any spurious findings. We have now mentioned more clearly what samples were analyzed for each experiment, e.g. page 7, line 173 – 174:

“We also looked at a subset of the Maela cohort, which included only the earliest obtained samples from unique individuals (mothers and children). This subset consisted of 762 isolates, including 380 from mother – child pairs.”

5. Presentation of the results could be improved. While it is understandable that the sequencing data need to be compressed to a presentable format, essential information needs to be logically displayed for the sake of readers' understanding. As examples, the first section of the result section mentioned total numbers of isolates and serotypes from each cohort, but did not say how many of them were from children/adults. Figure 1 does have age information, but it is difficult to evaluate due to data transformation. Sequence clusters were mentioned without elaboration on what they mean. This style of data presentation may be readily comprehensive for sequencing gurus, but is hard to digest to the experimentalists like myself.

Thank you for pointing this out. We have gone through the results, particularly the first section, trying harder to think about viewing them from the perspectives of experts from other disciplines.

The section titled “Serotypes and strains are variably carried between age groups, and between cohorts” (page 7, line 162) was intended as a global summary of population composition for both cohorts. There are many serotypes and strains, which are difficult to summarize in text and further adding the distribution over age categories would add more to this (we have tried this presentation, but it is poor in prose). To compare between cohorts and across strains/serotypes the transform from counts to frequencies (as in figure 1) is necessary to see the information on the correct scale. However, we appreciate that a lack of age split is not helpful, and have added this essential information to the start of this section. We have also added the requested age splits of each serotype as a table, where they can be more easily read.

The reference to Figure 1 here was incorrect and has been removed. Also, the section has been simplified and information has been moved to supplementary tables to increase readability. Figure 1 is now referenced in the 3rd paragraph of the Results section (page 8, line 190), as this section presents the differences in colonization frequency in relation to age groups.

Other comments that should be addressed:

1. Supplementary Tables S8 and 9 would benefit from an annotations (or representative gene ID) associated with each of the regions.

These have now been added.

2. Paragraph starting line 214 and Figure 4: Given that genes were mapped to D39V, it would be very useful to provide an ID. Similarly, TableS10 has a column for Marker Name, but the IDs are not present only sequence segments.

In Supplementary File 10 we have inserted a first column named ‘Sequence’. This column now provides the sequence segments. The second column, ‘MarkerName’ now has ‘intergenic’ for each segment sequence as these sequences do not map to an open reading frame (or any other sequence) in the D39V reference (or any other available reference), and therefore have not been allocated a gene ID. These sequences do map to the intergenic region between open reading frames encoding the accessory Sec-dependent serine-rich glycoprotein adhesen and a MarR like regulator in a subset of pneumococcal isolates from this study (but none of the available reference sequences).

3. The description of the performed 'heritability analyses' requires more detail to enable the readership to repeat the analyses.

The code used to perform the analyses can now be found at https://github.com/philipkremer123/carriage_pneumo_heritability/blob/main/code (page 18, line 474)

4. The title of the manuscript may be reconsidered given the cross-sectional nature of the study performed.

We have changed the title of the manuscript to “pneumococcal genetic variability in age-dependent bacterial carriage”.

5. The discussion should contain a comparison to previous studies that performed pneumococcal GWAS on host age (f.e. doi.org/10.1093/cid/ciy417). The observed geographical differences in pneumococcal elements being associated with phenotype could also be related to previous studies reporting on this phenomenon (f.e. doi: 10.1038/s41467-018-07997-y).

Thank you for these suggestions. We have added a comparison to the reference on genetic variation to IPD disease severity to the discussion (page 13, line 324 – 325 “Polygenic effects were also seen in another study on the contribution of genetic variation to disease severity of invasive pneumococcal disease”). Geographical differences, with a reference to the 2019 Nature Communications paper were also added (page 14, line 363: “geographical differences”)

6. The authors may also want to discuss potential biases or alternative explanations for observed differences between infants and adults (f.e. detectability – while pneumococcal carriage density is lower in adults compared to children, serotype 3 generally grows with high load and is relatively more likely to be detected if present. Viability – Survival of trpB defective pneumococci may depend on age-dependent host factors like diet).

We have added wording to the discussion on differences in carriage density, on page 14, line 364 – 365: “Pneumococcal factors such as differences in detectability due to carriage density could also influence results”. Thanks for these interesting examples, which we have also included in the text.

7. Why do the authors present epidemiological arguments to prefer either serotypes or proteins as vaccine targets, while serotypes strongly co-occur with specific proteins across the pneumococcal population that could evenly be targeted? If they choose to do so, however, the report would benefit from concrete numbers to support the claim that based on this study serotypes would be a preferential vaccine target. Which vaccine-formulations would the authors recommend for the infant and parent populations studied here?

Regarding co-occurrence of serotypes and proteins: previous work has shown that many proteins are present in every member of the population, and that proteins which are variably present (and often associated with specific serotypes) are selected to favor a frequency which is independent of vaccine driven serotype dynamics [Azarian et al., PLoS Biol. 2020; Corander et al., Nat Eco Evol 2017.]. Protein targets may therefore be desirable as they could target the entire population in a serotype-independent manner [Colijn et al., Nat Microbiol 2020]. Modifying the carrier protein used in the conjugate vaccine is one way to include a protein target using existing vaccine technology. We therefore wished to explore whether an alternative carrier protein found more commonly in infant carriage would be viable (given that reducing infant carriage has the greatest effect on reducing burden of disease), but we did not find a suitable candidate from our hypothesis-generating study.

While we appreciate the interesting suggestion to propose a particular formulation, pneumococcal population and vaccine dynamics are too complex for us to model in the scope of our paper. This is due to needing to model frequency dependent selection, vaccines which only target part of the pneumococcal population, and challenges quantifying disease burden from noisy observational study. This has previously been done by Colijn et al., Nat Microbiol 2020, who modelled this using optimization criteria, and included using part of the pilus as the carrier protein, and administering complementary adult vaccines.

Rather than simply citing this previous work on vaccine modelling, we have now added a more in-depth description of its relation to our work.

Author details

1. Philip HC Kremer

   Department of Neurology, Amsterdam UMC, University of Amsterdam, Meibergdreef, Netherlands

   Contribution

   Conceptualization, Formal analysis, Methodology, Writing - original draft, Project administration, Writing – review and editing

   For correspondence

   philip_kremer@hotmail.com

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-0483-841X

2. Bart Ferwerda

   1. Department of Neurology, Amsterdam UMC, University of Amsterdam, Meibergdreef, Netherlands
   2. Department of Clinical Epidemiology, Biostatistics and Bioinformatics, University of Amsterdam, Amsterdam, Netherlands

   Contribution

   Formal analysis, Supervision, Methodology, Writing – review and editing

   Competing interests

   No competing interests declared

3. Hester J Bootsma

   Centre for Infectious Disease Control, National Institute for Public Health and the Environment, Bilthoven, Netherlands

   Contribution

   Data curation, Writing – review and editing

   Competing interests

   No competing interests declared

4. Nienke Y Rots

   Centre for Infectious Disease Control, National Institute for Public Health and the Environment, Bilthoven, Netherlands

   Contribution

   Data curation, Writing – review and editing

   Competing interests

   No competing interests declared

5. Alienke J Wijmenga-Monsuur

   Centre for Infectious Disease Control, National Institute for Public Health and the Environment, Bilthoven, Netherlands

   Contribution

   Data curation, Writing – review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-5663-860X

6. Elisabeth AM Sanders

   1. Centre for Infectious Disease Control, National Institute for Public Health and the Environment, Bilthoven, Netherlands
   2. Department of Pediatric Immunology and Infectious D, Wilhelmina Children's Hospital, Utrecht, Netherlands

   Contribution

   Data curation, Writing – review and editing

   Competing interests

   No competing interests declared

7. Krzysztof Trzciński

   Department of Pediatric Immunology and Infectious D, Wilhelmina Children's Hospital, Utrecht, Netherlands

   Contribution

   Data curation, Writing – review and editing

   Competing interests

   No competing interests declared

8. Anne L Wyllie

   1. Department of Pediatric Immunology and Infectious D, Wilhelmina Children's Hospital, Utrecht, Netherlands
   2. Epidemiology of Microbial Diseases, Yale School of Public Health, New Haven, United States

   Contribution

   Data curation, Writing – review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-6015-0279

9. Paul Turner

   1. Cambodia Oxford Medical Research Unit, Angkor Hospital for Children, Siem Reap, Cambodia
   2. Centre for Tropical Medicine and Global Health, Nuffield Department of Medicine, University of Oxford, Oxford, United Kingdom

   Contribution

   Data curation, Writing – review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-1013-7815

10. Arie van der Ende

    1. Department of Medical Microbiology and Infection Prevention, Amsterdam UMC, Amsterdam, Netherlands
    2. The Netherlands Reference Laboratory for Bacterial Meningitis, Amsterdam, Netherlands

    Contribution

    Supervision, Methodology, Writing – review and editing

    Competing interests

    No competing interests declared

11. Matthijs C Brouwer

    Department of Neurology, Amsterdam UMC, University of Amsterdam, Meibergdreef, Netherlands

    Contribution

    Supervision, Writing – review and editing

    Competing interests

    No competing interests declared

12. Stephen D Bentley

    Parasites and Microbes, Wellcome Sanger Institute, Cambridge, United Kingdom

    Contribution

    Supervision, Funding acquisition, Methodology, Writing – review and editing

    Competing interests

    No competing interests declared

13. Diederik van de Beek

    Department of Neurology, Amsterdam UMC, University of Amsterdam, Meibergdreef, Netherlands

    Contribution

    Supervision, Funding acquisition, Methodology, Writing – review and editing

    Competing interests

    No competing interests declared

14. John A Lees

    1. European Molecular Biology Laboratory–European Bioinformatics Institute, Cambridge, United Kingdom
    2. MRC Centre for Global Infectious Disease Analysis, Department of Infectious Disease Epidemiology, Imperial College London, London, United Kingdom

    Contribution

    Conceptualization, Formal analysis, Supervision, Methodology, Writing – review and editing

    For correspondence

    jlees@ebi.ac.uk

    Competing interests

    No competing interests declared

    "This ORCID iD identifies the author of this article:" 0000-0001-5360-1254

• 666

  Page views

• 181

  Downloads

• 0

  Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.