Abstract
Streptococcus pneumoniae causes substantial mortality among children under 5-years-old worldwide. Polysaccharide conjugate vaccines (PCVs) are highly effective at reducing vaccine serotype disease, but emergence of non-vaccine serotypes and persistent nasopharyngeal carriage threaten this success. We investigated the hypothesis that following vaccine, adapted pneumococcal genotypes emerge with the potential for vaccine escape. We genome sequenced 2804 penumococcal isolates, collected 4-8 years after introduction of PCV13 in Blantyre, Malawi. We developed a pipeline to cluster the pneumococcal population based on metabolic core genes into “Metabolic genotypes” (MTs). We show that S. pneumoniae population genetics are characterised by emergence of MTs with distinct virulence and antimicrobial resistance (AMR) profiles. Preliminary in vitro and murine experiments revealed that representative isolates from emerging MTs differed in growth, haemolytic, epithelial infection, and murine colonisation characteristics. Our results suggest that in the context of PCV13 introduction, pneumococcal population dynamics had shifted, a phenomenon that could further undermine vaccine control and promote spread of AMR.
Similar content being viewed by others
Introduction
Streptococcus pneumoniae (the pneumococcus), a common commensal of the upper respiratory tract, is responsible for a high burden of severe pneumonia, septicaemia and meningitis1,2, mainly affecting children under 5-years-old, the immunocompromised and the elderly. Pneumococcal disease causes almost 300,000 deaths annually in under-5’s worldwide, and disproportionally affects people in resource-poor settings, where 57% of the total pneumococcal deaths occur3. This high burden of pneumococcal disease is largely vaccine preventable1,4. Indeed, reports from The Gambia, South Africa, Kenya and Malawi5,6,7,8,9 show that pneumococcal conjugate vaccines (PCV), targeting the polysaccharide capsule of the most common disease-causing pneumococcal serotypes (currently PCV10 or PCV13), have been highly effective in reducing invasive pneumococcal disease (IPD). However, the highly adaptable nature of S. pneumoniae has the potential to undermine this success. Through the expression of approximately 100 capsule serotypes, the pneumococcus has the ability to adapt to change, by mutation and by the capacity to readily acquire new traits through horizontal gene transfer from other pneumococci and related streptococci occupying the same niche10,11.
In the context of PCV introduction, vaccine type (VT) replacement IPD is widespread in resource-rich settings12,13 and is being increasingly reported on the African continent14,15. In addition, we and others have shown that despite excellent vaccine uptake in many African countries, there is considerable residual VT nasopharyngeal carriage in both PCV-vaccinated and PCV-unvaccinated populations several years after PCV introduction6,7,16. Routine PCV programmes have also been shown to reduce the burden of antimicrobial resistant (AMR) S. pneumoniae VT disease and carriage in some settings17,18. However, the impact has not been uniform, with the emergence of certain AMR serotypes, most notably 7F and 19A19,20,21. Furthermore, our modelling suggests that under vaccine pressure, antimicrobial resistant non-VT (NVT) strains could replace susceptible NVT strains through the removal of competition from vaccine-susceptible VTs22.
Emerging evidence from Africa, Europe and the USA suggests that PCV introduction has driven genomic alterations in S. pneumoniae populations, contributing to capsular switching, selection of specific lineages and genetic recombination21,23,24,25,26. We have previously described a theoretical framework whereby pneumococcal vaccines targeting particular serotypes drive the emergence of NVT S. pneumoniae strains with metabolic and virulence-associated characteristics similar to the VTs commonly circulating prior to vaccine introduction27,28.
Here, to explore the genotypic and phenotypic basis for this framework, we have used a large, well-characterised pneumococcal carriage strain collection from an urban population in Blantyre, Malawi (2015–2019), established starting 4 years after the introduction of PCV13 into the national immunisation programme. PCV13 was introduced 12 November 2011 using a 3 + 0 schedule, with primary doses given at 6, 10 and 14 weeks of age. Field studies in Malawi have reported high PCV13 uptake of 90%–95%29, similar to the 92% PCV13 coverage recently reported by WHO/UNICEF (WHO and UNICEF estimates of immunisation coverage, 2018. https://www.who.int/immunisation/monitoring_surveillance/routine/coverage/WUENIC_notes.pdf). We hypothesised that in this population with a high force of infection30 and a high frequency of multiple serotype carriage31,32, there would be an increase in pneumococcal diversity, adaptability and potentially AMR. We have therefore developed a metabolic core genome allelic profiling method, which has allowed us to analyse 2804S. pneumoniae carriage isolates collected over 4 years of surveillance. We show that discreet “metabolic genotypes” have expanded amongst both VT and NVT serotypes up to 7 years after vaccine introduction. These expanding metabolic genotypes (MT) are characterised by accessory genes linked to virulence and AMR. Strains selected from the emerging MT have in vitro and in-vivo phenotypes that suggest adaptation. Together these data highlight the genetic changes in pneumococcal population structure and biology beyond serotype replacement that, if not addressed, may undermine current and future vaccine strategies.
Results
Homogeneity of the Blantyre pneumococcal carriage strain collection across age-cohorts using classical serotyping and genotyping methods
We have previously reported an analysis of the first 7 surveys from our population-based carriage surveillance in Blantyre, Malawi (2015–2018), which showed high persistent residual VT carriage among PCV-vaccinated children 3–5-year-old (16·7%) and PCV-unvaccinated children 6–8-year-old (15·7%) and HIV-infected adults 18–40 years old on antiretroviral therapy (ART; 8·9%)16. Including swabs that were collected from an 8th survey conducted in 2019, overall (NT + NVT) S. pneumoniae carriage prevalence was 5115/8413 (60.8%). Using this Blantyre pneumococcal carriage strain collection (Table 1), WGS libraries were developed using a random subset of isolates from each of the eight surveys (Fig. 1).
The upper row shows prevalence (fraction) of vaccine serotypes (1–23F). The following rows show non-vaccine serotypes. Non-typeable strains are excluded from the analysis. The dashed blue line separates early-stages of carriage surveys (June 2015–April 2017, survey 1–4) from late stages of carriage surveys (May 2017–June 2019, survey 5–8). Serotypes are determined by WGS using the PneumoCaT tool. Error bars show standard error of the mean. The number of isolates for each point is given in Table S3.
Among the final dataset of 2804 sequenced carriage isolates, we observed a decline of VT serotype 1 and 6B carriage isolates, and an increase in frequency amongst several NVTs over time, in particular 23B and 38 (Fig. 1). However, the majority of VTs and NVTs persisted over time and although there were some differences between age groups, each cohort contributed a similar proportion of the dataset at each timepoint (Figs. 1, S1). Two hundred sequence types (ST) were identified on the basis of the allelic profiles of a standard set of seven housekeeping genes (see “Methods”). Most STs were present in each cohort (Fig. S1a), with the exception of the STs circulating at very low frequencies, including ST347, ST4084, ST10880 and ST10992, which were not identified among the isolates from the HIV-infected adults. Analogously, identified serotypes and global pneumococcal sequencing clusters (GPSCs) contributed a similar proportion of the dataset at each timepoint (Fig. S1b, c). Given this spread of serotyping and genotyping characteristics throughout the collection, all isolates were aggregated into a single dataset for the subsequent analyses.
Serotype switching amongst both vaccine and non-vaccine serotypes
On the basis of the sequence-typing, we identified 44 separate serotype switching events in the Blantyre collection, described by a total of 22 STs which corresponded to more than one serotype (Table S1). Directionality of these switch events was determined by considering the genetic relatedness of the strains and assuming that the most common serotype (capsule type in a dominant lineage) within an ST was the serotype from which the switch originated33. Rather than a clear shift from VTs towards NVTs, with only a few statistically significant trends found (Tables S2 and S3), low-frequency VT or NVT variants were identified to undergo switching throughout the carriage surveys in several ST and GPSC lineages (Fig. S2a, b). These data support the emerging evidence that even in the context of vaccine introduction, the process of serotype switching is not limited to VTs (e.g., ST7653 6C/6D or ST3214 35A/11A). We suggest that although a serotype switch may be directly vaccine-induced, switched serotypes may have occurred stochastically and then emerged alongside other vaccine-associated changes in population structure. Indeed, it seems likely that selection forces that are more complex than serotype-directed vaccine pressure alone facilitate many of such switches such as environmental and antimicrobial pressures34.
Shifts in metabolic gene profiles of the pneumococcal population
Bacterial metabolism is intimately linked to virulence27,35,36. To further identify dynamic changes in the pneumococcal population structure beyond serotype switching and genotypes defined by evolutionary history alone, we defined metabolic genotypes (MTs) based on our previously described theoretical framework27. Each MT was classified on the basis of differences in core genes involved in metabolism and energy production, following functional annotation of the core genome of the pneumococcal population. This typing method therefore relies on sequence differences between genes involved in bacterial metabolism (i.e., function), rather than lineages defined exclusively by bacterial evolutionary history.
We grouped the Blantyre pneumococcal carriage isolates into 148 discrete MTs (Fig. 2). As shown in the clustering tree, these MTs were neither independent of serotype nor did they completely overlap with serotype. For instance, the 14 different serotype 19F MTs were spread across the clustering tree. As might be anticipated, overlap between MTs and GPSCs has been also observed (Supplementary Data 3). Our previous study of PCV13 uptake in rural Malawi in the 1st year after vaccine introduction29 and our estimates of vaccine uptake in the Blantyre population 4 years after introduction16, show that vaccine uptake has been rapid and sustained. Our statistical modelling of the carriage survey data estimated that the population-level half-life of VT vs. NVT carriage was similar among PCV-vaccinated (3.34 vs. 9.46 years) and PCV-unvaccinated (3.26 vs. 9.83 years) children16. Therefore, to assess whether there had been a change in the S. pneumoniae genotypic profile over the surveys with sufficient precision, we compared surveys 1–4 (19 June 2015–12 April 2017) with 5–8 (3 May 2017–9 August 2019). Overall, the majority of MTs did not significantly change in frequency when comparing early and later surveillance isolates (p > 0·05; Fisher’s exact test), highlighting a high residual diversity within the bacterial population (Fig. S3a, b). Nonetheless, our analysis highlighted a shift in the pattern of several MTs: for instance, with the NVTs 10A, 17F, 34, 38 and 23B which showed a significant increase in MTs 74, 109, 33, 120 and 93 (Fig. S3a, b, p < 0·0001 for MTs 109, 120, and 93; p = 0.026 MT33; p = 0.0051 MT74 Fisher’s exact test).
The dendrogram is based on of the hamming distance between the different allelic profiles of the metabolic genes. Coloured strips represent the Metabolic type, the serotype, and the carriage survey group (early = survey 1 to 4, late = survey 5 to 8). MTs are defined by curtailing the dendrogram at half the maximum distance, indicated by the red line (n = 148 discrete MTs, containing only MTs with more than 15 observations). The inset show a magnification of the dendrogram around serotype 17F/MT109.
In total, we observed 32 instances in which VTs switch capsular serotype to NVTs, or vice versa, while maintaining the same metabolic profile, which is in line with a vaccine-induced metabolic shift as described in ref. 27.
Metabolic genotypes increasing in frequency are characterised by increased AMR
To address the hypothesis that vaccination drives changes in the pattern of AMR by removing the competition of non-AMR VTs22, we assessed whether the emerging MTs amongst the NVTs 10 A, 17 F, 23B, 34 and 38 were also associated with increased AMR. We determined resistance genetically to penicillin, an important first-line antibiotic for pneumococcal disease, together with chloramphenicol, erythromycin and tetracycline, which are also commonly used antibiotics in this population. Higher penicillin MICs (minimum inhibitory concentration) were observed in the later surveillance period compared to the early surveillance period (Wilcoxon rank sum test, p < 0·0001) (Fig. S3e). The majority of the serotypes and MTs identified in this dataset (n = 39 and n = 79, respectively) were statistically significantly associated, either positively or negatively, association with penicillin resistance (i.e., MIC > 0·06 µg/ml, p < 0·05; Fisher’s exact test), rather than being independent of resistance. This indicates that the phenotypic characteristic of AMR is likely linked to the isolate’s genotype23 (Fig. S3c, d).
We observed several instances in which emerging metabolic genotypes were associated with higher AMR. Among NVTs 17F, 10A, 23B, 34 and 38 we show that MTs 109, 74, 93, 22 and 120 (respectively, Fig. 2) were more frequently identified in the later surveillance period (Fig. S3), indicating a shift in the dominant genotype observed within each serotype. In four of these NVTs (17F, 10A, 23B and 38), the emergence of MTs in later surveillance was associated with increased AMR: either in terms of an increased MIC for penicillin (genetically assessed, Fig. 3) or in terms of presence of other antibiotic resistance genes such as tetM and ermB (Fig. S4). In serotypes 17F, 23B and 38, there was an increase in the frequencies of MT114, MT97 and MT122, respectively. In each case, the MT that emerged showed a higher penicillin MIC, compared to MTs common in the bacterial population in the earlier surveys. Interestingly, the increased frequency of MT46 of serotype 34 in later surveillance surveys without exhibiting an increase in AMR suggests a variety of forces that drive the selection and restructuring of bacterial populations.
Boxes show the frequency of isolation of each metabolic profile per survey (“MT distribution in time”) and penicillin MIC of each isolate (box plot representing the lower and upper quartiles as the box edges, a horizontal line as the median, and whiskers are the minimum between 1.5-fold the interquartile range and the farthest observation from the quartiles; with points representing isolates; and numbers on top representing the number of isolates in each survey) for to serotypes 17F, 10A, 23B, 34 and 38. Vertical blue lines separate the early-late survey periods. Presence of AMR genes per serotype is shown in Fig. S4.
There were no significant changes (p > 0·05, Fisher’s exact test) in carriage frequency of the majority of the VTs over the duration of the surveillance period - i.e., they did not increase nor decrease when comparing the early and later periods of surveillance. These included, 3, 19F and 23F (Fig. 1 and S3). Serotypes 3 and 23F in particular were found at relatively high frequencies after vaccine introduction in other settings37,38. Contrary to what was described with NVTs, the persistent VTs in this dataset were characterised by the presence of a dominant MT (MT9 for serotype 3, MT3 for serotype 23F) consistently present during the survey period, showing increased penicillin MIC and the presence of antimicrobial resistance genes (Fig. S5).
Within those serotypes where we identified genotype switch events (namely, 17F, 10A, 23B, 38 and 23B), we investigated genetic relatedness amongst MTs and time-calibrated phylogeny, to reconstruct the emergence of those recent dominant MTs. We assessed the genetic relatedness between MTs, hence their likelihood to have recently expanded clonally, by calculating the core genome SNP (single nucleotide polymorphism) divergence between MTs (Fig. S6). For serotypes 38, 17F and 34, recently identified MTs were characterised by lower genotypic difference (lower SNP variance, serotype 38 MT120-28 p = 0·016, serotype 17F MT109-66 p < 0·0001, serotype 34 MT33-53 p < 0·0001, F-test for variances, Fig. S6), indicating that these isolates expanded clonally. This recent clonal expansion was confirmed by the temporal signal identified in the phylogeny (Fig. S7). Root-to-tip regression highlighted that the emergence of the recently identified lineages (MT109 and MT120) in serotype 17F and 38 was likely to have happened before the year 2011 (17F MT114: 2000·9 [95% CI 1988·35–2008·1] node 95, 2009·7 [2005·7–2012·3] node 94; for 38 MT122: 2009·3 [2000·3–2013·3] node 85, 2007·5 [1996·7–2012·2] node 84, 2004·1 [1985·2–2011·4] node 83). Thus, none of these MTs are likely to have originated after PCV13 introduction, but rather we propose that they existed as variants at low frequency before vaccine introduction that underwent clonal expansion. It was not possible to temporally calibrate the phylogeny of serotypes 23B, 10A and 34, hence we could not infer the time of emergence of those lineages. This may be due to the smaller sample size of some lineages (i.e., in serotype 10A) or to lineages that originated much earlier than their isolation in this study.
Emerging metabolic genotypes show multiple adaptations
Having demonstrated a shift in MTs in serotypes 10A, 17F, 23B, 34 and 38 associated with increased AMR during the post-vaccine introduction observation period, we then assessed whether these emerging MTs had associated virulence characteristics that could convey a competitive advantage. To do this, we investigated the accessory genome using a pangenome wide association study approach (pan-GWAS39). We identified 93, 73, 51, 30 and 175 genes present exclusively in the most common emerging metabolic profiles of serotypes 10A, 17F, 23B, 34 and 38, respectively (Fig. S8; full gene list is reported in Supplementary Data 1; examples of genes associated to each MT analysed are shown in Fig. 4). Functional annotation showed that these included genes responsible for metabolism and carbohydrate transport. In serotype 38 MT120, for instance, mannose- and sorbose-specific transporter genes, absent in other MTs, were identified. MT93 serotype 23B was characterised by lactose and galactiol transporters (Fig. 4) which were absent in other MTs. The emerging MTs were also characterised by the presence of bacterial defence systems, such as multidrug resistance ABC transporters, restriction enzymes and antimicrobial resistance genes (Fig. 4).
The table shows examples of genes specific of MTs that expanded at late stages of carriage survey identified via pan-GWAS. Columns show the annotation of the genes or the gene system, the MTs and serotypes in which the gene was found, the sensitivity and specificity related to its presence in the MT reported in the table (and its absence in the other MTs). Full table is shown in supplementary material (Supplementary Data 1). All p-values (adjusted using a Bonferroni correction) were <0·0001, as reported by Scoary. The barplot graph shows clusters of orthologous genes enriched in Serotype 10A (MT74), 17F (MT109), 23B (MT93), 34 (MT33), and 38 (MT120). Each barplot shows the number of genes in each functional class for the dominant metabolic type in the 5 serotypes, as reported by the pan-GWAS analysis. The columns refer to the functional classes “Metabolism” (MET), “Cellular processes and signalling” (CELL), “Information storage and processing” (INFO), or poorly characterised genes (UNKNOWN).
The MT93 and MT120 of serotypes 23B and 38 accessory genomes encode several Toxin-Antitoxin (TA) systems. Variants of the pezAT system was present in MT93 and MT120 and a yefM-yoeB system was present in MT120. This particular TA-system is located in a previously described integrative-conjugative region (ICE)40, which is absent from the non-MT120 strains. The ICE region is present in several non-pneumococcal Streptococci, such as S. dysgalactiae and S. suis41, suggesting that its acquisition may have occurred via horizontal gene transfer (HGT). This ICE also carries specific metal transporters, phage resistance and competence systems, as well as a previously described agglutinin receptor involved in colonisation40.
To assess the generalisability of our findings, we undertook a wider comparison using pneumococcal genomes sequences from other African countries21,42. This revealed the presence of analogous MTs outside Malawi. In particular, MT93 and MT120 (associated with serotypes 23B and 38 respectively) have been identified after vaccine introduction in invasive pneumococcal disease isolates in both South Africa and The Gambia42. Rather than originating from a common source, differences in the core genome SNPs among MT120 strains isolated in South Africa and in Malawi (~6000 SNPs, Fig. S9) suggest that these different lineages originate from different ancestors that have subsequently converged to the same MT.
Together, these data indicate that a range of potentially evolutionary relevant genes are associated with the emerging MTs (Fig. 4). In populations with a high force of infection and host vulnerabilities such as malnutrition and HIV, this potential for competitive advantage in nutrient transport, cellular metabolism, nasopharyngeal epithelial colonisation and AMR could lead to a resurgence in invasive NVT pneumococcal disease.
Phenotypic characterisation of emerging MT reveals adaptations that may confer competitive advantage
To further explore this potential competitive advantage amongst emerging NVT metabolic genotypes with increased AMR, we compared MT shifts within serotype 38 and serotype 23B in several in vitro phenotypic assays. Rather than use a very large panel of strains that captures the variation within the accessory genome, we used representative strains of these MTs. To support this approach, we have assessed within MT genetic variation of virulence factors implicated in colonisation43 (Supplementary Data 4). We assessed amino acid sequence similarity rather than nucleotide similarity, because although nucleotide sequences would likely result in more variance, the impact of synonymous single nucleotide polymorphism on virulence is uncertain. To avoid missing potential changes in virulence, a conservative 95% amino acid similarity cutoff point was chosen. This analysis revealed that the variance in amino acid sequences of virulence genes ranged from 0 (540/865, 62.4%) to 2.54 (MT55, virulence gene pce/cbpE), with a mean variance of 0.03. Generally, we observed little variance in the amino acid sequences across virulent genes in metabolic types, with 10 out of 16 virulence genes having a mean average variance of less than 0.01. Together this indicates that there is little variance in the virulence gene repertoire of isolates belonging to the same MTs.
Our phenotypic experiments aimed at representing some of the challenges of bacterial lifestyle in the nasopharyngeal niche. Firstly, we cultured combinations of different MTs in standard growth medium. The growth parameters and the in vitro fitness were calculated by Curveball44 for MTs 28 and 120 for serotype 38 and MTs 23 and 93 for serotype 23B (Fig. 5). This software fits the growth curves of the different bacterial strains to several growth models and attempts to predict which one would have a competitive advantage over the other while sharing resources.
The growth curves of the MTs comprising serotype 23B (a) and 38 (b) separately (blue and red curves) and in an in vitro co-culture (green curve). For each curve the MT or the combination of MTs is reported in parentheses. The table shows the growth parameters as predicted by Curveball (nu-deceleration parameter, q0-initial physiological state, r-specific growth rate in low density). These experiments were performed on at least three independent occasions with 16 replicates.
Within serotype 23B, the emergent strain MT93 reached a higher maximum optical density (OD) than MT23 (Fig. 5). Using the Baranyi-Roberts model to approximate co-culture growth, we found that MT93 is expected to have a higher in vitro fitness than MT24. Within serotype 38, the emergent strain MT120 grew at a lower rate and to a lower maximum OD than MT28 (Fig. 5) and had a lower predicted fitness than MT28. Given the opposing findings, we expanded the growth curve experiments to include dominant and non-dominant MTs within serotype 3 and 23F. We observed nearly identical growth patterns for serotype 23F strains, with a higher stationary-phase autolysis rate for serotype 23F’s dominant MT3 compared to the non-dominant MT123. Moreover, we observed a slightly higher growth rate for the non-dominant MT60 (ST5435) compared to the dominant MT9 (ST700) for serotype 3 strains (Fig. S10). Thus, while some emergent MTs have a growth advantage in vitro, others do not. However, all emergent and dominant MTs within serotype 23B, 38, 3 and 23F had a higher MIC to penicillin compared to their non-dominant counterparts (Figs. 3, S5).
The epithelial colonisation process is a prerequisite for both disease and onward transmission to another host45. We therefore then tested representative S. pneumoniae strains from these MT pairs in a well-standardised Detroit 562 human nasopharyngeal epithelial infection model46. Within serotype 38, we observed no difference in the ability to colonise the epithelial cells between MTs 28 and 120 (Fig. 6a). In contrast, serotype 23B, MT93 was internalised into epithelial cells at a higher frequency than MT23 (Fig. 6a, b) even though both genotypes were characterised by the same ability to associate with epithelial cells. We have previously suggested that differences in this so-called epithelial “microinvasion” may be fundamental to the outcome of colonisation, explaining strain differences in the potential of S. pneumoniae to cause invasive disease, persist during carriage, and be transmitted from person-to-person46.
a Number of bacteria associated with epithelial cells, recovered after 3 h of incubation, b Number of intracellular bacteria after 3 h incubation and 1 h with gentamicin treatment, c IL-8 produced by epithelial cells, incubated with bacterial cells for 6 h. NI non-infected control. d haemolysis promoted by the different bacterial strains (lower OD corresponds to higher haemolysis), for different bacterial dilutions (neat~1·5 × 106 bacterial cells. Results for 1/2, 1/5 and 1/10 dilutions are shown). Representative strains from each MT were used in these experiments. Each experiment was repeated at least 3 times independently. In a, b, and c, dots on the bar plots show the result of each repetition. All p-values derived using Wilcoxon ranked-sum test: a serotypes 38 and 23, p-values = 0.31 and 0.34, respectively; b serotypes 38 and 23 p-values = 1 and 0.03, respectively; c NI vs MT120, p-value = 0.032; NI vs MT28, p-value = 0.16; MT28 vs MT120, p-value = 0.42; NI vs 93, p-value = 0.1; NI vs 23, p-value = 0.016; 93 vs 23, p-value = 1; d MT120 vs MT28: neat, p-value = 0.11; 1-in-2, p-value = 0.33; 1-in-5, p-value = 0.4; 1-in-10, p-value = 0.3; MT93 vs MT23: neat, p-value = 0.03; 1-in-2, p-value = 0.33; 1-in-5, p-value = 0.2; 1-in-10, p-value = 0.3. Data are represented as mean values and error bars are the standard errors of the mean. All p-values reported are for two-sided statistical tests.
Other phenotypic characteristics in different MTs were assessed, such as their ability to stimulate the production of the phagocyte chemoattractant IL-8 in epithelial cells (Fig. 6c) and their haemolytic potential, a measure of functional pneumolysin production (Fig. 6d). Serotype 38’s MT120 stimulated lower IL-8 production, compared to MT28, while the two MTs in serotype 23B did not stimulate a higher IL-8 production when compared to the non-infected control (Fig. 6c). The haemolytic potential of serotype 23B’s MT23 and serotype 3’s MT9 were higher than their serotype matched counterparts, while no significant difference was identified between serotype 38 and 23F strains (Figs. 6d, S10).
Finally, we employed a 7-day murine colonisation model to investigate the ability of the MTs to colonise the nasopharynx in competition47. Despite similar levels of epithelial adherence and internalisation in vitro, serotype 38 MT28 had a tendency to colonise the murine nasopharynx to a greater extent than the emerging genotype MT120, when inoculated singly (Wilcoxon, p = 0.06), and had a significant advantage in a competition (Wilcoxon, p-value < 0.001, Fig. 7b, c). In contrast, there were no differences in the number of CFUs recovered from the murine nasopharynx for serotype 23B MT93 and MT23 (Fig. 7a).
CFUs recovered from nasopharyngeal washes of CD1 outbred mice inoculated intranasally singly with serotype 23B MT93 or MT23 (a), singly with serotype 38 MT120 or MT28 (b). In c CFUs from a competition experiment with a 1:1 ratio of serotype 38 MT120 and MT28, seven days post inoculation, are transformed into the competitive index. Six mice were used for each condition. All p-values derived using Wilcoxon ranked-sum test: a p-value = 0.7. b MTs 120 and 28 p-value = 0.06; c MTs 93 and 23 p-value < 0.001. All p-values reported are for two-sided statistical tests.
Together these data suggest that the emerging MTs have not arrived at a single adaptive profile but have undergone a variety of changes (growth; epithelial invasion; inflammation; and pneumolysin production) that may have enabled them to become more prominent in the context of vaccine introduction. However, while these experiments provide proof of concept for our approach to the phenotypic characterisation of emerging MTs in a vaccine-exposed population, further validation is needed through similar analyses in other populations.
Discussion
How a pneumococcal population evolves both under natural or vaccine-induced pressures remains an open question. It has been long known that S. pneumoniae is characterised by a high level of genome plasticity, and this species diverges more quickly than many other mucosal bacterial species associated with vaccine preventable disease25,48. Here, using a large post-PCV introduction pneumococcal carriage surveillance dataset from an urban population in Malawi, we describe the emergence of pneumococcal lineages with unique virulence, AMR and metabolic characteristics. Rather than using the populational structural methods of the Pneumococcal Molecular Epidemiology Network (PMEN)49 or the Global Pneumococcal Sequencing (GPS) consortium42, our genotyping approach defined pneumococcal lineages based on metabolic genotypes (MT) that we have proposed27 may give a wider understanding of the biological drivers to changes in pneumococcal population structure.
We show that in addition to serotype replacement, there has been a shift in the profile of AMR MTs among NVTs characterised by exclusive genes responsible for metabolism and carbohydrate transport, and toxin-antitoxin systems located in an integrative-conjugative region suggestive of horizontal gene transfer. Phenotypically, representative strains from these emergent genotypes were found to have differential growth, haemolytic, inflammatory or epithelial association/ invasion traits that may confer advantage in the nasopharyngeal niche. However, the genotype-phenotype relationship is multidimensional. For example, our finding that while similar single strain colonisation levels were achieved in a murine model of pneumococcal carriage, when inoculated together, serotype 38’s MT28 outcompeted the emerging MT120 genotype is contrary what might be anticipated for an emerging genotype. We maintain that the emergence of such genotypes results from multiple phenotypic adaptations that may impact on both colonisation efficiency and transmission. Indeed, together these findings support theoretical models that predict that PCV introduction disrupts complex competitive interactions between colonising pneumococci, with the potential to promote replacement with fitter, more virulent, and more antimicrobial resistant existing pneumococcal lineages22,27. The serotypes contained within PCVs are selected primarily on the basis of invasive potential in infants21. We postulate that if these emergent AMR NVT MTs become fixed in the population, out-competing NVTs with less invasive/ disease-causing potential, they may lead to lower vaccine effectiveness against pneumococcal disease overall. We also observed stable MTs among VTs and although our observations are limited to the post-PCV13 era, we speculate that these genotypes may confer a competitive advantage leading to inadequate PCV13 effectiveness in the control of VT carriage, including serotypes 3, 19F and 23F. While the increased penicillin MICs that were observed amongst emerging or persistent MTs do not cross the thresholds that would likely lead to treatment failure for diseases such as pneumonia or meningitis, in this heavily antibiotic exposed population50, we speculate that if this upward drift in MIC continues, and it may confer a competitive advantage in the nasopharyngeal niche.
The overall effect of the shift in MTs that we have observed is dependent on their colonisation potential, transmissibility and their relative invasiveness21. These MTs have emerged within serotypes that are known to be important contributors to IPD and are commonly carried. Serotype 38 has high heterogeneity in its invasive potential in South Africa and the USA;21 serotype 23F ST2059/MT3 was reported to have a higher risk for invasiveness in strains isolated in South Africa and in the USA; serotype 23B ST4423/MT93 was identified in IPD in South Africa and Israel; and serotype 35B, which also expanded in frequency in this study, was earlier highlighted as an expansion serotype in Malawi IPD21,42. Moreover, serotypes 23B and 38 are among the top 10% of serotypes causing IPD after vaccine introduction in Europe, Asia and US51.
The emergence and expansion of particular NVTs in the period after PCV introduction has been widely reported21. The serotype switching processes and dynamics highlighted here have also been reported by others23,42. For instance, the serotype switch event involving ST8672 (serotypes 3 and 7F) was also identified in a post-vaccination carriage study in northern Malawi23. Similarly, the Global Pneumococcal Sequencing study, identified ST10599 and ST989 as part of switch events, highlighting how these mechanisms are common in pneumococcal genotype and serotype dynamics21. Our finding of a shift in MTs in Malawi appears to have also been seen in other African countries, leading us to hypothesise that these MTs have become more prevalent across the continent following PCV introduction.
A theme that has emerged from our analysis is that although the emergent MTs have successful metabolic, virulence and AMR properties that may confer a competitive advantage over other MTs, we did not identify genetic or phenotypic traits that suggest a single adaptive profile amongst the emergent MTs. In contrast, we identified a complex set of multiple different adaptations that have the potential to provide an advantage in distinct ecological niches. For instance, some MTs are characterised by sugar transporters, TA-systems and bacteriocins production systems, which would confer competitive advantage while sharing resources. Carbohydrate metabolism is an increasingly recognised determinant of pneumococcal niche adaption and progression to disease52. In addition, a characteristic of the MTs that appeared to have emerged in recent years in our study population is the presence of genes that encode particular toxin/antitoxin systems. These systems are present in a variety of bacterial species both at a plasmid and chromosomal level, and involved in a variety of cellular processes–including virulence, environmental stress response and niche adaptation53,54. The pan-GWAS analysis highlights the presence of virulence factors in expanding MTs: in MT33 for instance the presence of a ydcP protease, a collagenase which facilitates breaking of extracellular structures and a known virulence factor in other bacterial species55, identified to be associated with pneumococcal IPD;56 and in MT109 and MT74 of a haemolysin (pneumolysin) translocation ATP-binding protein gene, known virulence factors for S. pneumoniae. Together, these observations indicate that the expanding metabolic types accumulate characteristics that may affect niche competition and virulence.
Our phenotypic analysis of representative strains further highlights the complex and sometimes conflicting effects of the genotypic shifts identified. Recent studies have described bacterial fitness in different terms, adapting its definition to the available data. For instance, the description of negative frequency dependent selection in S. pneumoniae populations, whereby bacterial fitness is affected by the frequency of rare accessory genes57,58. We have used growth curve analysis to infer relative, mixed-culture bacterial fitness44. In combination with previous modelling analyses, the results described in this work are consistent with a complex and diverse behaviour of a S. pneumoniae population under vaccine pressure. Serotype 23B MTs for instance, behaves similarly to the VIMS model description whereby the fitter genotype which was suppressed beforehand expands after vaccine introduction due to niche clearance (Fig. 3). On the other hand, serotype 38’s MTs behave in accordance with the hypothesis of an increase in frequency of AMR NVTs, driven by vaccination22. An antibiotic susceptible MT of serotype 38 exhibited a higher intrinsic fitness, as determined by the growth curve analysis. However, we postulate that this MT was overtaken by another, antibiotic resistant MT of serotype 38, when vaccination alleviated the competition pressure from VTs.
We also found that the most common recent strain of serotype 23B, namely MT93, is less haemolytic (pneumolysin mediated) than MT23 and more successful in invading nasopharyngeal cells in vitro but has similar colonisation levels in a murine colonisation model. On the other hand, for serotype 38, the recently dominant MT120 stimulated less IL-8 production but was less successful in the murine colonisation model, when compared to the less common metabolic genotype MT28. Pro-inflammatory interleukin-8 attracts neutrophils to the site of infection59 and therefore by evading the immune system MT120 may have a competitive advantage. While the similar or lower murine colonisation densities seen with the emerging genotypes may seem disadvantageous, we propose that features such as AMR, together with measured and unmeasured virulence traits, may outweigh these effects during human carriage.
Importantly, our approach to metabolic genotyping should not be viewed as a proposed alternative to established pneumococcal strain-typing systems. Instead, the genetic and phenotypic analyses, and their varying results, highlight that pneumococcal evasion of the immune system, or more successful colonisation of different ecological niches, is both complex and multifactorial. This is often the case in evolutionary landscapes60, where a simple explanation of complex phenomena is unattainable. We have not identified a common phenotype characteristic between all the recently expanded genotypes, but we highlighted several differences that could help the success of each genotype in different environments and at different time points27. However, although a very large undertaking requiring analyses from different populations under vaccine pressure, if phenotypic characteristics of the multiple genotypes that have emerged since PCV13 introduction are characterised across both in vitro and in vivo platforms, common profiles may emerge.
Our study has several limitations. First, the lack of genomic data prior to the introduction of PCV in Blantyre. Pneumococcal carriage surveys have been undertaken in the pre-PCV era in Malawi but not with the same systematic approach, and largely not in the same populations geographically or in people living with HIV31,61,62,63,64. To avoid these biases, we have therefore not undertaken a direct comparative analysis. However, we have been able to leverage this large dataset to capture ongoing changes in a population with high vaccine coverage16,29. Because of the cross-sectional nature of the surveillance contributing to the Blantyre collection we were not able to track the emergence of MTs within individual hosts over time. Therefore, as is common in observational studies, we have not been able to attribute causality to routine vaccine introduction.
Second, the MTs identified in this study in Malawi may not be fully generalisable to other pneumococcal populations. For example, the core genome underlying our analysis is likely to change with varying number of isolates and genetic heterogeneity. Hence, population-specific analyses will be required to address particular hypotheses in other populations.
Third, our experimental observations are limited by the use of a single representative strain from each MT and although we saw little variation in virulence genes within MTs, there is the potential that our model systems may not fully capture the competitive advantage of emerging MTs. Rather than analysing the large number of strains required to capture the full extent of the genetic variation within the accessory genome, to further validate our findings, we advocate for similar analyses in other populations under vaccine pressure. We used an endpoint-outcome of murine colonisation to assess differences in carriage potential. However, a kinetic model, as described by Zafar et al.65, may have provide greater insights into competitive advantage. Nonetheless, our experiments have highlighted the complex, multidimensional nature of pneumococcal dynamics. Comprehensive future work will be needed to pinpoint the effects of each dimension of the fitness landscapes of different populations.
In conclusion, we postulate that in this high burden vaccine-exposed population, AMR and a range of genotypic and phenotypic traits may facilitate the expansion of MTs in the community. Vaccine strategies that increase PCV valency may be insufficient to interrupt this process66. The multiple adaptations by these MTs extend beyond simple serotype replacement, and successful adaptation could further undermine vaccine control and promote the spread of AMR.
Methods
Bacterial isolates and carriage surveys
The strains analysed in this study were isolated as part of a prospective observational study using stratified random sampling to assess pneumococcal nasopharyngeal carriage in Blantyre, Malawi16. In brief, sampling consisted of twice-annual rolling cross-sectional surveys over the course of 4 years amongst PCV13-vaccinated children 2–7 years old, PCV13-unvaccinated children 5–10 years old, and HIV-infected adults 18–40 years old on ART. Nasopharyngeal sample collection, storage and testing were all performed using standard approaches16. As commonly done for pneumococcal carriage surveys, one colony was selected for purification and subsequent serotyping from each sample, selected by its ease of sampling and minimisation of contamination risk. 2804 carriage isolates were thus selected at random from the 8 surveys to develop the WGS libraries. Isolates from the HIV-infected adults were sequenced from surveys 1, 2, 5 and 6 only. Particularly in children, multiple serotype carriage as determined by DNA microarray, is relatively high in this population31,32. However, our comparative analysis shows good concordance between a single S. pneumoniae colony selected for phenotypic serotyping and the dominant serotype determined by microarray molecular serotyping32.
DNA sequencing and assembly
NP swabs were stored in skim-milk-tryptone-glucose-glycerol (STGG) at −80 °C. After being thawed and vortexed, 30 µl of STGG was plated on gentamicin-sheep blood agar (SBG; 7% sheep blood agar, 5 µl gentamicin/mL) and incubated overnight at 37 °C in 5% CO2. Plates showing no S. pneumoniae growth were incubated overnight a second time before being reported as negative. S. pneumoniae was identified by colony morphology and optochin disc (Oxoid, Basingstoke, UK) susceptibility. The bile solubility test was used on isolates with no or intermediate optochin susceptibility (zone diameter < 14 mm). A single colony of confirmed pneumococcus was selected and grown on a new SBG plate as before. Growth from this second plate was used for serotyping by latex agglutination (ImmuLex™ 7-10-13-valent Pneumotest; Statens Serum Institute, Denmark)16,32. DNA was extracted from an overnight lawn plate culture from isolates archived after serotyping, using DNAeasy blood and tissue kit (Qiagen) following the manufacturer’s guidelines and sequenced using HiSeq4000 (paired-end library 2 × 150) platform at Oxford Genomics Centre UK.
Raw DNA reads were trimmed of low-quality ends and cleaned of adaptors using Trimmomatic software (ver. 0·32)67. De novo assembly was performed with SPAdes software (ver 3·8·0)68, using a sample of 1,400,000 reads and k-mer values of 21, 33, 55, and 77. De novo assemblies were checked for plausible length (between 1,900,000 and 2,200,000 bp), annotated using Prokka (ver. 1·12)69, and checked for low-level contamination using Kraken (ver. 0·10·5)70 and the module “rMLST species id” on the pubmlst database (pubmlst.org)71. In cases for which more than 5% of the contigs belonged to a species different from Streptococcus pneumoniae, the genome sequence was not included in any further analysis. Resulting assemblies are available on pubmlst.org (https://pubmlst.org/organisms/streptococcus-pneumoniae). Metadata of the strains used in this work are reported in Supplementary Data 2.
Definition of serotype, multilocus sequence type, global pneumococcal sequencing cluster and AMR
Serotypes were determined via DNA sequence, using the PneumoCaT software (Ver. 1·2·1, https://github.com/phe-bioinformatics/PneumoCaT)72. The tool reports serogroup 15 as serotype 15 A, 15B, 15 C and 15B/C. Phenotypic serotype definition by latex agglutination was used as validation of the genetic serotype definition, observing a concordance of over 90%, in alignment with previous estimations32. For the analyses in this work, serotype determined by DNA sequencing was used.
Multilocus sequence types (MLST, STs) were assigned using the databases from the allelic profiles of 7 housekeeping genes (adhP, pheS, atr, glnA, sdhA, glcK, and tkt). BLASTn (ver. 2·10·0 + ) was used to align the DNA assemblies to the DNA fragments typical of each allele for each housekeeping gene (parameters: E-value 1e − 10, minimum 95% identity, minimum 90% query coverage). This grouped strains into 200 unique STs. Strains which did not show a full set of housekeeping gene alleles or were not assigned to any previously described ST (n = 783) were double-checked for sequence contamination and assigned to a new sequence type definition. The database of MLST housekeeping fragments was downloaded from pubmlst.com/Spneumoniae in October 2019. The GPSC genetic typing method (Global Pneumococcal Sequence Clusters) was described in several recent publications21,23,42. We collected the dataset to assign the GPSCs from https://www.pneumogen.net/gps/assigningGPSCs.html and used PopPUNK (ver. 2·0·2)73 to cluster the isolates in the pre-defined GPSCs as described in Gladstone and colleagues in 201942. AMR gene presence (for genes cat, tetM, ermB, and mefA) was detected via nucleotide-BLAST (E-value < 0.001, sequence coverage and identity >80%)74. For penicillin resistance, MIC was genetically assessed using the pbp genes allelic profile75.
Pangenome, definition of metabolic profiles and phylogenetic analyses
In order to capture the total variability of the population, a pangenome was generated from the combined isolates from surveys 1 to 8 using Roary (ver. 3·8·0)76. Parameters for each run were: 90% of minimum BLASTp identity; MLC inflation value 1.5; with 99% of strains in which a gene must be present to be considered “core”. Saturation of the pangenome was assessed. The core genome of this bacterial population included 1061 genes which was consistent with previous estimates from Malawi77. The metabolic genes were selected via functional annotation of the core genome (n = 386 core metabolic genes)27 through the eggNOG database (http://eggnogdb.embl.de). This was done according to the KEGG Orthology (KO) groupings of the KEGG database27. Functional classes selected to define the core metabolic genome were: D-Energy production and conversion, E-Amino acid transport and metabolism, F-Nucleotide transport and metabolism, G-Carbohydrate transport and metabolism, H-Coenzyme transport and metabolism, I-Lipid transport and metabolism, P-Inorganic ion transport and metabolism, Q-Secondary metabolites biosynthesis, transport, and catabolism.
Metabolic profiles were defined using the following pipeline: (i) for each genome, the allelic profile of the metabolic core genes was defined using the Genome Comparator module of Bigsdb (bigsdb.com);71 (ii) the hamming distance between each allelic profile was calculated (hamming. dist function of base R 3·6, https://www.R-project.org/); (iii) the hamming distance between each isolate was used to hierarchically cluster the population (hclust); (iv) the hierarchical clustering dendrogram was cut at midpoint to define the discrete metabolic profiles.
Maximum-Likelihood, recombination censored phylogeny was calculated with ClonalframeML (ver. 1·12)78, under a generalised time-reversible model with 100 bootstrap replicates. Core genomes used to produce ML phylogeny were calculated separately for the isolates belonging to the serotypes of interest (17 F, 10 A, 38, 34 and 23B) with Roary as described above. SNPsites (Ver 1·04)76 was used to identify the variable part of the core genome alignment, which was then used for phylogeny reconstruction. Root-to-tip regression and phylogeny dating were calculated with the BactDating R package (https://github.com/xavierdidelot/BactDating, 700000 permutations)79.
The pan-GWAS analyses were carried out using Scoary (https://github.com/AdmiralenOla/Scoary)80 as described by Gori and colleagues in 202039.
To ascertain the expected properties of the metabolic genes, an analysis of linkage disequilibrium (LD) using the approach taken by Watkins et al.27. The analysis used the ATCC 700669 reference genome and 100 randomly selected loci pairs to compare LD D’ values in metabolic and non-metabolic genes, resulting in an average D’ values of 0.93 and 0.9, respectively (Fig. S11). We note that our larger dataset forces the core genome to be more conserved, leading to higher D’ values than in Watkins et al.
Co-culture fitness and epithelial association/ invasion experiments
Strains BVY5TE, BVY11B (serotype 23B, MT23 and MT93 respectively), BVY2DJ and BVY123 (serotype 38, MT120 and MT28 respectively) were selected for phenotypic evaluation (Supplementary Data 2).
In vitro broth culture was carried out in order to apply the pipeline described by Ram and colleagues in 201944. Isolates belonging to different MTs were harvested from a THY-10% glycerol stock and grown overnight on a Columbia agar plate with 5% horse blood (CBA) plate. The plate was used to seed 10 ml of Todd-Hewitt broth with 2·5% yeast extract (THY) to an optical density at 600 nm (OD600) of 0·5–1, which was then diluted to an OD600 0·05 and used directly as starter in at least 16 replicate 200ul wells of a 96-well plate. Each experiment also contained a mixture of 100 μl each of the two strains. Each plate was incubated statically for 18–20 h, at 37 °C, with 5% carbon dioxide in a plate reader (Tecan Spark M20). OD600 was measured every 15 min. In order to explore the possibility that ill-defined components of THY may bias the differences seen, as well as to subject the cells to an environment more similar to the one found in the nasopharynx, chemically defined minimal medium proposed by Aprianto and colleagues81 was tested, but the isolates failed to grow in this medium.
Curveball (ver. 0·2·5) was used to fit each growth curve to a bacterial growth model and calculate each strain-relative fitness by modelling the behaviour of each strain in co-culture. As curveball was designed to calculate fitness in E. coli growth experiments, the growth curves were censored after each strain reached the maximum OD600, to correct for the characteristic S. pneumoniae autolysis.
Association and invasion experiments on human respiratory tract epithelial cells (Detroit 562) were carried out as reported by Weight and colleagues in 201946. Briefly, human pharyngeal carcinoma Detroit 562 epithelial cells (ATCC_CCL-138) were grown in 10% FCS in alpha MEM media (Gibco). Confluent Detroit 562 (typically day 8 post-plating) were co-cultured with S. pneumoniae for 3 h in 1% FCS alpha MEM (MOI~1 cell:10 pneumococci). The medium was removed, and cells washed three times in Hanks Buffer Saline Solution (HBSS +/+, with calcium and magnesium, Gibco). Cells were incubated in 1% saponin for 10 min at 37 °C and lysed by repetitive pipetting. Dilutions of bacteria were plated on blood agar and colonies counted after 16 h. To quantify internalised bacteria, 200 µg/ml gentamicin was added for 1 h to the cells, which were washed another three times, before incubating with 1% Saponin and plating on blood agar plates. CFUs were counted after 16 h incubation at 37 °C, 5% CO2. For analysis of IL-8 secretion, Detroit 562 cells were infected with S. pneumoniae for 6 h and supernatant was collected. The protocol for the DuoSet® ELISA kit (R&D Systems) was followed according to manufacturers’ instructions.
Finally, for testing the haemolytic activity, bacteria were suspended in phenol free RPMI (Invitrogen) and incubated with 2% red blood cells (EO labs) in a U-bottom 96 well plate for 30 min. at 37 °C / 5% CO2. Saponin was used as a positive control for cell lysis. The plate was centrifuged at 1500 g for three minutes and supernatant was transferred to a new plate to read absorbance at 540 nm. All the abovementioned experiments were performed on at least three independent occasions.
Variance between MTs
To determine the variance between isolates of the same MT virulence genes’ amino acid sequences, all protein-coding genes were first identified and translated into amino acid sequences from their assembled WGS using Prodigal82. The amino acid sequences were then compared to VFDB (Virulence Factor Database) amino acid sequences of known Pneumococcal virulent genes using BLASTP (particularly those involved in colonisation43), filtering out any hits that do not have a full alignment with the reference sequence, have gaps in the alignment, or have less than 95% sequence identity to that of the reference virulence genes’ amino acid sequence83.
Murine colonisation experiments
Five-week-old outbred female CD1 mice (Charles River Laboratories) were inoculated intranasally with 5 × 106 CFU S. pneumoniae under anaesthetic (isoflurane) for colonisation experiments. For competition experiments, 1:1 mix of serotype 38 MT28 and MT120 was used. Six mice were used for each experiment. Nasopharyngeal washes were obtained 7 days post inoculation and CFU enumeration performed using Columbia blood agar (CBA) plates supplemented with 4 µg/ml gentamicin. For competition experiments, plates were additionally supplemented with 10 µg/ml tetracycline, and the competitive index was calculated84.
All animal procedures were approved by the local ethical review process and conducted in accordance with the relevant UK Home Office-approved project license (PPL70/6510). Mice were housed for at least one week under standard conditions before use. Randomisation or blinding was not performed for these experiments.
Throughout the manuscript and supporting material, box plots represent the lower and upper quartiles as the box edges, a horizontal line represents the median, and whiskers are the minimum between 1.5-fold the interquartile range and the farthest observation from the quartiles.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
Data availability
All required data are found in the Supplementary Material or archived online, as described in the Methods. All isolate sequences are available from NCBI (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA1011974).
Code availability
All data analysis protocols and software used are explicitly detailed in the manuscript and the supplementary files.
References
O’Brien, K. L. et al. Burden of disease caused by Streptococcus pneumoniae in children younger than 5 years: global estimates. Lancet 374, 893–902 (2009).
Antimicrobial Resistance C. Global burden of bacterial antimicrobial resistance in 2019: a systematic analysis. Lancet 399, 629–655 (2022).
Wahl, B. et al. 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. Lancet Glob. Health 6, e744–e757 (2018).
Ngocho, J. S. et al. Effectiveness of pneumococcal conjugate vaccines against invasive pneumococcal disease among children under five years of age in Africa: a systematic review. PLoS ONE 14, e0212295 (2019).
Roca, A. et al. Effects of community-wide vaccination with PCV-7 on pneumococcal nasopharyngeal carriage in the Gambia: a cluster-randomized trial. PLoS Med. 8, e1001107 (2011).
Nzenze, S. A. et al. Temporal changes in pneumococcal colonization in a rural African community with high HIV prevalence following routine infant pneumococcal immunization. Pediatr. Infect. Dis. J. 32, 1270–1278 (2013).
Cohen, C. et al. Effectiveness of the 13-valent pneumococcal conjugate vaccine against invasive pneumococcal disease in South African children: a case-control study. Lancet Glob. Health 5, e359–e369 (2017).
Hammitt, L. L. et al. Effect of ten-valent pneumococcal conjugate vaccine on invasive pneumococcal disease and nasopharyngeal carriage in Kenya: a longitudinal surveillance study. Lancet 393, 2146–2154 (2019).
Bar-Zeev, N. et al. Impact and effectiveness of 13-valent pneumococcal conjugate vaccine on population incidence of vaccine and non-vaccine serotype invasive pneumococcal disease in Blantyre, Malawi, 2006–18: prospective observational time-series and case-control studies. Lancet Glob. Health 9, e989–e998 (2021).
Salvadori, G., Junges, R., Morrison, D. A. & Petersen, F. C. Competence in Streptococcus pneumoniae and close commensal relatives: mechanisms and implications. Front. Cell. Infect. Microbiol. 9, 94 (2019).
Kalizang’oma, A. et al. Streptococcus pneumoniae serotypes that frequently colonise the human nasopharynx are common recipients of penicillin-binding protein gene fragments from Streptococcus mitis. Micro. Genom. 7, 000622 (2021).
Sheppard, C. L. et al. The genomics of Streptococcus pneumoniae carriage isolates from UK children and their household contacts, pre-PCV7 to post-PCV13. Genes 10, 687 (2019).
Vestrheim, D. F., Hoiby, E. A., Aaberge, I. S. & Caugant, D. A. Impact of a pneumococcal conjugate vaccination program on carriage among children in Norway. Clin. Vaccin. Immunol. 17, 325–334 (2010).
Kwambana-Adams, B. et al. Rapid replacement by non-vaccine pneumococcal serotypes may mitigate the impact of the pneumococcal conjugate vaccine on nasopharyngeal bacterial ecology. Sci. Rep. 7, 8127 (2017).
Vadlamudi, N. K., Chen, A. & Marra, F. Impact of the 13-valent pneumococcal conjugate vaccine among adults: a systematic review and meta-analysis. Clin. Infect. Dis. 69, 34–49 (2019).
Swarthout, T. D. et al. High residual carriage of vaccine-serotype Streptococcus pneumoniae after introduction of pneumococcal conjugate vaccine in Malawi. Nat. Commun. 11, 2222 (2020).
Ben-Shimol, S. et al. Impact of pneumococcal conjugate vaccines introduction on antibiotic resistance of Streptococcus pneumoniae meningitis in children aged 5 years or younger, Israel, 2004 to 2016. Eur. Surveill. 23, 1800081 (2018).
Tin Tin Htar, M. et al. The impact of routine childhood immunization with higher-valent pneumococcal conjugate vaccines on antimicrobial-resistant pneumococcal diseases and carriage: a systematic literature review. Expert Rev. Vaccines 18, 1069–1089 (2019).
Lee, J. K. et al. Changes in the serotype distribution among antibiotic resistant carriage Streptococcus pneumoniae isolates in children after the introduction of the extended-valency pneumococcal conjugate vaccine. J. Korean Med. Sci. 32, 1431–1439 (2017).
Habibi Ghahfarokhi, S. et al. Serotype distribution and antibiotic susceptibility of Streptococcus pneumoniae isolates in Tehran, Iran: a surveillance study. Infect. Drug Resist. 13, 333–340 (2020).
Lo, S. W. et al. Pneumococcal lineages associated with serotype replacement and antibiotic resistance in childhood invasive pneumococcal disease in the post-PCV13 era: an international whole-genome sequencing study. Lancet Infect. Dis. 19, 759–769 (2019).
Obolski, U. et al. Vaccination can drive an increase in frequencies of antibiotic resistance among nonvaccine serotypes of Streptococcus pneumoniae. Proc. Natl Acad. Sci. USA 115, 3102–3107 (2018).
Chaguza, C. et al. Early signals of vaccine-driven perturbation seen in pneumococcal carriage population genomic data. Clin. Infect. Dis. 70, 1294–1303 (2020).
Azarian, T. et al. Global emergence and population dynamics of divergent serotype 3 CC180 pneumococci. PLoS Pathog. 14, e1007438 (2018).
Croucher, N. J. et al. Diversification of bacterial genome content through distinct mechanisms over different timescales. Nat. Commun. 5, 5471 (2014).
Croucher, N. J. et al. Population genomic datasets describing the post-vaccine evolutionary epidemiology of Streptococcus pneumoniae. Sci. Data 2, 150058 (2015).
Watkins, E. R. et al. Vaccination drives changes in metabolic and virulence profiles of Streptococcus pneumoniae. PLoS Pathog. 11, e1005034 (2015).
Watkins, E. R., Kalizang’Oma, A., Gori, A., Gupta, S. & Heyderman, R. S. Factors affecting antimicrobial resistance in Streptococcus pneumoniae following vaccination introduction. Trends Microbiol. 30, 1135–1145 (2022).
Mvula, H. et al. Predictors of uptake and timeliness of newly introduced pneumococcal and rotavirus vaccines, and of measles vaccine in rural Malawi: a population cohort study. PLoS ONE 11, e0154997 (2016).
Lourenco, J. et al. Determinants of high residual post-PCV13 pneumococcal vaccine-type carriage in Blantyre, Malawi: a modelling study. BMC Med. 17, 219 (2019).
Kamng’ona, A. W. et al. High multiple carriage and emergence of Streptococcus pneumoniae vaccine serotype variants in Malawian children. BMC Infect. Dis. 15, 234 (2015).
Swarthout T. D., et al. Evaluation of pneumococcal serotyping in nasopharyngeal carriage isolates by latex agglutination, whole genome sequencing (PneumoCaT) and DNA microarray in a high pneumococcal carriage prevalence population in Malawi. J. Clin. Microbiol. https://doi.org/10.1128/JCM.02103-02120 (2020).
Chaguza, C. et al. Population genetic structure, antibiotic resistance, capsule switching and evolution of invasive pneumococci before conjugate vaccination in Malawi. Vaccine 35, 4594–4602 (2017).
Croucher, N. J. et al. Selective and genetic constraints on pneumococcal serotype switching. PLoS Genet. 11, e1005095 (2015).
Rohmer, L., Hocquet, D. & Miller, S. I. Are pathogenic bacteria just looking for food? Metabolism and microbial pathogenesis. Trends Microbiol. 19, 341–348 (2011).
Leonard, A. & Lalk, M. Infection and metabolism - Streptococcus pneumoniae metabolism facing the host environment. Cytokine 112, 75–86 (2018).
Kandasamy, R. et al. Persistent circulation of vaccine serotypes and serotype replacement After 5 Years of Infant Immunization With 13-Valent Pneumococcal Conjugate Vaccine in the United Kingdom. J. Infect. Dis. 221, 1361–1370 (2020).
Lewnard, J. A. & Hanage, W. P. Making sense of differences in pneumococcal serotype replacement. Lancet Infect. Dis. 19, e213–e220 (2019).
Gori, A. et al. Pan-GWAS of Streptococcus agalactiae highlights lineage-specific genes associated with virulence and niche adaptation. mBio 11, e00728–00720 (2020).
Brown, J. S., Gilliland, S. M., Spratt, B. G. & Holden, D. W. A locus contained within a variable region of pneumococcal pathogenicity island 1 contributes to virulence in mice. Infect. Immun. 72, 1587–1593 (2004).
Huang, J. et al. Evolution and diversity of the antimicrobial resistance associated mobilome in Streptococcus suis: a probable mobile genetic elements reservoir for other Streptococci. Front. Cell Infect. Microbiol. 6, 118 (2016).
Gladstone, R. A. et al. International genomic definition of pneumococcal lineages, to contextualise disease, antibiotic resistance and vaccine impact. EBioMedicine 43, 338–346 (2019).
Subramanian, K., Henriques-Normark, B. & Normark, S. Emerging concepts in the pathogenesis of the Streptococcus pneumoniae: from nasopharyngeal colonizer to intracellular pathogen. Cell Microbiol. 21, e13077 (2019).
Ram, Y. et al. Predicting microbial growth in a mixed culture from growth curve data. Proc. Natl Acad. Sci. USA 116, 14698–14707 (2019).
Weiser, J. N., Ferreira, D. M. & Paton, J. C. Streptococcus pneumoniae: transmission, colonization and invasion. Nat. Rev. Microbiol. 16, 355–367 (2018).
Weight, C. M. et al. Microinvasion by Streptococcus pneumoniae induces epithelial innate immunity during colonisation at the human mucosal surface. Nat. Commun. 10, 3060 (2019).
Cohen, J. M. et al. Contributions of capsule, lipoproteins and duration of colonisation towards the protective immunity of prior Streptococcus pneumoniae nasopharyngeal colonisation. Vaccine 30, 4453–4459 (2012).
Tettelin H., Hollingshead S. K. Comparative genomics of Streptococcus pneumoniae: intrastrain diversity and genome plasticity. In: The Pneumococcus (eds Tuomanen E. I., Mitchell T. J., Morrison D., Spratt B. G.) (ASM Press, 2004).
McGee, L. et al. Nomenclature of major antimicrobial-resistant clones of Streptococcus pneumoniae defined by the pneumococcal molecular epidemiology network. J. Clin. Microbiol. 39, 2565–2571 (2001).
Dixon, J. et al. Antibiotic stories: a mixed-methods, multi-country analysis of household antibiotic use in Malawi, Uganda and Zimbabwe. BMJ Glob. Health 6, e006920 (2021).
Balsells, E., Guillot, L., Nair, H. & Kyaw, M. H. Serotype distribution of Streptococcus pneumoniae causing invasive disease in children in the post-PCV era: a systematic review and meta-analysis. PLoS ONE 12, e0177113 (2017).
Minhas, V., Paton, J. C. & Trappetti, C. Sickly sweet - how sugar utilization impacts pneumococcal disease progression. Trends Microbiol. 29, 768–771 (2021).
Van Melderen, L. & Saavedra De Bast, M. Bacterial toxin-antitoxin systems: more than selfish entities?. PLoS Genet. 5, e1000437 (2009).
Wen, Y., Behiels, E. & Devreese, B. Toxin-antitoxin systems: their role in persistence, biofilm formation, and pathogenicity. Pathog. Dis. 70, 240–249 (2014).
Navais, R., Mendez, J., Perez-Pascual, D., Cascales, D. & Guijarro, J. A. The yrpAB operon of Yersinia ruckeri encoding two putative U32 peptidases is involved in virulence and induced under microaerobic conditions. Virulence 5, 619–624 (2014).
Obolski, U. et al. Identifying genes associated with invasive disease in S. pneumoniae by applying a machine learning approach to whole genome sequence typing data. Sci. Rep. 9, 4049 (2019).
Corander, J. et al. Frequency-dependent selection in vaccine-associated pneumococcal population dynamics. Nat. Ecol. Evol. 1, 1950–1960 (2017).
Azarian, T. et al. Frequency-dependent selection can forecast evolution in Streptococcus pneumoniae. PLoS Biol. 18, e3000878 (2020).
Cockeran, R., Durandt, C., Feldman, C., Mitchell, T. J. & Anderson, R. Pneumolysin activates the synthesis and release of interleukin-8 by human neutrophils in vitro. J. Infect. Dis. 186, 562–565 (2002).
Obolski, U., Ram, Y. & Hadany, L. Key issues review: evolution on rugged adaptive landscapes. Rep. Prog. Phys. 81, 012602 (2018).
Heinsbroek, E. et al. Pneumococcal acquisition among infants exposed to HIV in rural Malawi: a longitudinal household study. Am. J. Epidemiol. 183, 70–78 (2016).
Heinsbroek, E. et al. Persisting high prevalence of pneumococcal carriage among HIV-infected adults receiving antiretroviral therapy in Malawi: a cohort study. AIDS 29, 1837–1844 (2015).
Glennie, S. J. et al. Defective pneumococcal-specific Th1 responses in HIV-infected adults precedes a loss of control of pneumococcal colonization. Clin. Infect. Dis. 56, 291–299 (2012).
Iwajomo, O. H. et al. Deteriorating pneumococcal-specific B-cell memory in minimally symptomatic african children with HIV infection. J. Infect. Dis. 204, 534–543 (2011).
Zafar, M. A., Hamaguchi, S., Zangari, T., Cammer, M. & Weiser, J. N. Capsule type and amount affect shedding and transmission of Streptococcus pneumoniae. MBio 8, e00989–17 (2017).
Colijn, C., Corander, J. & Croucher, N. J. Designing ecologically optimized pneumococcal vaccines using population genomics. Nat. Microbiol. 5, 473–485 (2020).
Bolger, A. M., Lohse, M. & Usadel, B. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics 30, 2114–2120 (2014).
Bankevich, A. et al. SPAdes: a new genome assembly algorithm and its applications to single-cell sequencing. J. Comput. Biol. 19, 455–477 (2012).
Seemann, T. Prokka: rapid prokaryotic genome annotation. Bioinformatics 30, 2068–2069 (2014).
Wood, D. E. & Salzberg, S. L. Kraken: ultrafast metagenomic sequence classification using exact alignments. Genome Biol. 15, R46 (2014).
Jolley, K. A., Bray, J. E. & Maiden, M. C. J. Open-access bacterial population genomics: BIGSdb software, the PubMLST.org website and their applications. Wellcome Open Res. 3, 124 (2018).
Kapatai, G. et al. Whole genome sequencing of Streptococcus pneumoniae: development, evaluation and verification of targets for serogroup and serotype prediction using an automated pipeline. PeerJ 4, e2477 (2016).
Lees, J. A. et al. Fast and flexible bacterial genomic epidemiology with PopPUNK. Genome Res. 29, 304–316 (2019).
McCall, C. & Xagoraraki, I. Comparative study of sequence aligners for detecting antibiotic resistance in bacterial metagenomes. Lett. Appl. Microbiol. 66, 162–168 (2018).
Metcalf, B. J. et al. Using whole genome sequencing to identify resistance determinants and predict antimicrobial resistance phenotypes for year 2015 invasive pneumococcal disease isolates recovered in the United States. Clin. Microbiol. Infect. 22, 1002 e1001–1002 e1008 (2016).
Page, A. J. et al. SNP-sites: rapid efficient extraction of SNPs from multi-FASTA alignments. Micro. Genom. 2, e000056 (2016).
Kulohoma, B. W. et al. Comparative genomic analysis of meningitis- and bacteremia-causing pneumococci identifies a common core genome. Infect. Immun. 83, 4165–4173 (2015).
Didelot, X. & Wilson, D. J. ClonalFrameML: efficient inference of recombination in whole bacterial genomes. PLoS Comput. Biol. 11, e1004041 (2015).
Didelot, X., Croucher, N. J., Bentley, S. D., Harris, S. R. & Wilson, D. J. Bayesian inference of ancestral dates on bacterial phylogenetic trees. Nucleic Acids Res. 46, e134 (2018).
Brynildsrud, O., Bohlin, J., Scheffer, L. & Eldholm, V. Rapid scoring of genes in microbial pan-genome-wide association studies with Scoary. Genome Biol. 17, 238 (2016).
Aprianto, R., Slager, J., Holsappel, S. & Veening, J. W. Time-resolved dual RNA-seq reveals extensive rewiring of lung epithelial and pneumococcal transcriptomes during early infection. Genome Biol. 17, 198 (2016).
Hyatt, D. et al. Prodigal: prokaryotic gene recognition and translation initiation site identification. BMC Bioinforma. 11, 119 (2010).
McGinnis, S. & Madden, T. L. BLAST: at the core of a powerful and diverse set of sequence analysis tools. Nucleic Acids Res. 32, W20–W25 (2004).
Durmort, C. et al. Deletion of the zinc transporter lipoprotein AdcAII causes hyperencapsulation of Streptococcus pneumoniae associated with distinct alleles of the type I restriction-modification system. mBio 11, e00445–20 (2020).
Acknowledgements
We thank the individuals who participated in this study and the local schools and authorities for their support. We are grateful to the study field teams (supported by Farouck Bonomali and Roseline Nyirenda) and the MLW laboratory team. We are grateful to the hospitality of the QECH ART Clinic, led by Ken Malisita. Our thanks also extend to the MLW laboratory management team (led by Brigitte Denis) and the MLW data management team (led by Clemens Masesa). We thank Andrea Gori for his contribution to the analysis. R.S.H., N.F., A.G. and T.S. are supported by the National Institute for Health Research (NIHR) Global Health Research Unit on Mucosal Pathogens using UK aid from the UK Government. The views expressed in this publication are those of the author(s) and not necessarily those of the NIHR or the Department of Health and Social Care. We thank the High-Throughput Genomics Group at the Wellcome Trust Centre for Human Genetics (funded by Wellcome Trust grant reference 203141/Z/16/Z) for the generation of the Sequencing data. This work was funded by Bill & Melinda Gates Foundation (OPP1117653 to R.S.H.), Medical Research Council (MRC Grant Number: MR/N023129/1), a National Institute for Health Research (NIHR) Global Health Research Unit on Mucosal Pathogens using UK aid from the UK Government (16/136/46 to R.S.H.). The Malawi–Liverpool– Wellcome Clinical Research Programme is supported by a Strategic Award from the Wellcome Trust (206545/Z/17/Z).
Author information
Authors and Affiliations
Contributions
U.O., T.D.S., M.C.M., S.G., N.F., and R.S.H. designed the study and contributed to the development of the methodology. T.D.S., N.F., and R.S.H. oversaw the study, data collection, and data management. U.O., R.C., J.L. and S.G. developed the genomic analyses. C.M.W., J. M. C., and G. E. performed laboratory experiments. T.D.S., J. M. and C. B. oversaw the genome extraction and pneumococcal serotyping. U.O and T.D.S wrote the first draft of the paper. U.O., T.D.S., J.L., C.M.W., J.C., A.K., T.S.M., J.M., C.B., J.S.B., G.E., M.C.M., N.F., S.G., R.S.H. contributed to subsequent drafts and read and approved the final version of the report.
Corresponding authors
Ethics declarations
Competing interests
The authors declare no competing interests.
Ethics approval and consent to participate
The study protocol was approved by the College of Medicine Research and Ethics Committee, University of Malawi (P.02/15/1677) and the Liverpool School of Tropical Medicine Research Ethics Committee (14.056). Adult participants and parents/guardians of child participants provided written informed consent, children 8–10 years old provided informed assent. This included consent for publication.
Peer review
Peer review information
Nature Communications thanks Taj Azarian and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Obolski, U., Swarthout, T.D., Kalizang’oma, A. et al. The metabolic, virulence and antimicrobial resistance profiles of colonising Streptococcus pneumoniae shift after PCV13 introduction in urban Malawi. Nat Commun 14, 7477 (2023). https://doi.org/10.1038/s41467-023-43160-y
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41467-023-43160-y
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.
