Designing ecologically optimized pneumococcal vaccines using population genomics

Abstract

Streptococcus pneumoniae (the pneumococcus) is a common nasopharyngeal commensal that can cause invasive pneumococcal disease (IPD). Each component of current protein–polysaccharide conjugate vaccines (PCVs) generally induces immunity specific to one of the approximately 100 pneumococcal serotypes, and typically eliminates it from carriage and IPD through herd immunity. Overall carriage rates remain stable owing to replacement by non-PCV serotypes. Consequently, the net change in IPD incidence is determined by the relative invasiveness of the pre- and post-PCV-carried pneumococcal populations. In the present study, we identified PCVs expected to minimize the post-vaccine IPD burden by applying Bayesian optimization to an ecological model of serotype replacement that integrated epidemiological and genomic data. We compared optimal formulations for reducing infant-only or population-wide IPD, and identified potential benefits to including non-conserved pneumococcal carrier proteins. Vaccines were also devised to minimize IPD resistant to antibiotic treatment, despite the ecological model assuming that resistance levels in the carried population would be preserved. We found that expanding infant-administered PCV valency is likely to result in diminishing returns, and that complementary pairs of infant- and adult-administered vaccines could be a superior strategy. PCV performance was highly dependent on the circulating pneumococcal population, further highlighting the advantages of a diversity of anti-pneumococcal vaccination strategies.

Access options

Rent or Buy article

Get time limited or full article access on ReadCube.

from$8.99

All prices are NET prices.

Fig. 1: Variation in the invasiveness of pneumococcal serotypes.
Fig. 2: Optimizing conjugate vaccines to minimize infant IPD.
Fig. 3: Vaccine strategies for minimizing population-wide IPD.
Fig. 4: Optimizing multi-protein vaccines to minimize IPD.
Fig. 5: Optimizing conjugate vaccines to minimize AMR IPD.
Fig. 6: Comparing the design and effectiveness of different vaccination strategies.

Data availability

The original sequence datasets underlying this analysis are in the public sequence databases with the accession codes given in Supplementary Dataset 3 of Corander et al.12. The epidemiological and phylogenetic data are available at https://microreact.org/project/multilocusNFDS. The input matrix G, the serotype for each isolate, the equilibrium frequencies for each locus and other input data are available from https://github.com/carolinecolijn/optimvaccine.

Code availability

The model code is available at https://github.com/carolinecolijn/optimvaccine.

References

  1. 1.

    Croucher, N. J., Løchen, A. & Bentley, S. D. Pneumococcal vaccines: host interactions, population dynamics, and design principles. Annu. Rev. Microbiol. 72, 521–549 (2018).

  2. 2.

    Turner, P. et al. Improved detection of nasopharyngeal cocolonization by multiple pneumococcal serotypes by use of latex agglutination or molecular serotyping by microarray. J. Clin. Microbiol. 49, 1784–1789 (2011).

  3. 3.

    Cobey, S. & Lipsitch, M. Niche and neutral effects of acquired immunity permit coexistence of pneumococcal serotypes. Science 335, 1376–1380 (2012).

  4. 4.

    Weinberger, D. M., Malley, R. & Lipsitch, M. Serotype replacement in disease after pneumococcal vaccination. Lancet 378, 1962–1973 (2011).

  5. 5.

    Johnson, H. L. et al. Systematic evaluation of serotypes causing invasive pneumococcal disease among children under five: the pneumococcal global serotype project. PLoS Med. 7, e1000348 (2010).

  6. 6.

    Flasche, S. et al. Effect of pneumococcal conjugate vaccination on serotype-specific carriage and invasive disease in England: a cross-sectional study. PLoS Med. 8, e1001017 (2011).

  7. 7.

    Huang, S. S. et al. Continued impact of pneumococcal conjugate vaccine on carriage in young children. Pediatrics 124, e1–e11 (2009).

  8. 8.

    Masala, G. L., Lipsitch, M., Bottomley, C. & Flasche, S. Exploring the role of competition induced by non-vaccine serotypes for herd protection following pneumococcal vaccination. J. R. Soc. Interface 14, 20170620 (2017).

  9. 9.

    Gjini, E., Valente, C., Sá-Leão, R. & Gomes, M. G. M. How direct competition shapes coexistence and vaccine effects in multi-strain pathogen systems. J. Theor. Biol. 388, 50–60 (2016).

  10. 10.

    Croucher, N. J. et al. Population genomics of post-vaccine changes in pneumococcal epidemiology. Nat. Genet. 45, 656–663 (2013).

  11. 11.

    Chewapreecha, C. et al. Dense genomic sampling identifies highways of pneumococcal recombination. Nat. Genet. 46, 305–309 (2014).

  12. 12.

    Corander, J. et al. Frequency-dependent selection in vaccine-associated pneumococcal population dynamics. Nat. Ecol. Evol. 1, 1950–1960 (2017).

  13. 13.

    McNally, A. et al. Signatures of negative frequency dependent selection in colonisation factors and the evolution of a multi-drug resistant lineage of Escherichia coli. mbio 10, e00644-19 (2019).

  14. 14.

    Azarian, T. et al. Prediction of post-vaccine population structure of Streptococcus pneumoniae using accessory gene frequencies. Preprint at https://doi.org/10.1101/420315 (2018).

  15. 15.

    Hausdorff, W. P., Bryant, J., Paradiso, P. R. & Siber, G. R. Which pneumococcal serogroups cause the most invasive disease: implications for conjugate vaccine formulation and use, part I. Clin. Infect. Dis. 30, 100–121 (2002).

  16. 16.

    Hausdorff, W. P., Feikin, D. R. & Klugman, K. P. Epidemiological differences among pneumococcal serotypes. Lancet Infect. Dis. 5, 83–93 (2005).

  17. 17.

    Feikin, D. R. et al. Serotype-specific changes in invasive pneumococcal disease after pneumococcal conjugate vaccine introduction: a pooled analysis of multiple surveillance sites. PLoS Med. 10, e1001517 (2013).

  18. 18.

    Nurhonen, M. & Auranen, K. Optimal serotype compositions for pneumococcal conjugate vaccination under serotype replacement. PLoS Comput. Biol. 10, e1003477 (2014).

  19. 19.

    Chen, C. et al. Effect and cost-effectiveness of pneumococcal conjugate vaccination: a global modelling analysis. Lancet Glob. Heal. 7, e58–e67 (2019).

  20. 20.

    Ouldali, N. et al. Incidence of paediatric pneumococcal meningitis and emergence of new serotypes: a time-series analysis of a 16-year French national survey. Lancet Infect. Dis. 18, 983–991 (2018).

  21. 21.

    Kyaw, M. H. et al. Effect of introduction of the pneumococcal conjugate vaccine on drug-resistant Streptococcus pneumoniae. N. Engl. J. Med. 354, 1455–1463 (2006).

  22. 22.

    Lee, G. M. et al. Immunization, antibiotic use, and pneumococcal colonization over a 15-year period. Pediatrics 140, e20170001 (2017).

  23. 23.

    Tomczyk, S. et al. Prevention of antibiotic-nonsusceptible invasive pneumococcal disease with the 13-valent pneumococcal conjugate vaccine. Clin. Infect. Dis. 62, 1119–1125 (2016).

  24. 24.

    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).

  25. 25.

    van Hoek, A. J., Choi, Y. H., Trotter, C., Miller, E. & Jit, M. The cost-effectiveness of a 13-valent pneumococcal conjugate vaccination for infants in England. Vaccine 30, 7205–7213 (2012).

  26. 26.

    Centers for Disease Control and Prevention. CDC Vaccine Price List (2019); https://www.cdc.gov/vaccines/programs/vfc/awardees/vaccine-management/price-list/index.html

  27. 27.

    Mackenzie, G. A. et al. Effect of the introduction of pneumococcal conjugate vaccination on invasive pneumococcal disease in The Gambia: a population-based surveillance study. Lancet Infect. Dis. 16, 703–711 (2016).

  28. 28.

    Ladhani, S. N. et al. Rapid increase in non-vaccine serotypes causing invasive pneumococcal disease in England and Wales, 2000–17: a prospective national observational cohort study. Lancet Infect. Dis. 18, 441–451 (2018).

  29. 29.

    Weinberger, D. M. et al. Relating pneumococcal carriage among children to disease rates among adults before and after the introduction of conjugate vaccines. Am. J. Epidemiol. 183, 1055–1062 (2016).

  30. 30.

    Hanage, W. P. et al. Evidence that pneumococcal serotype replacement in Massachusetts following conjugate vaccination is now complete. Epidemics 2, 80–84 (2010).

  31. 31.

    Ubukata, K. et al. Serotype changes and drug resistance in invasive pneumococcal diseases in adults after vaccinations in children, Japan, 2010–2013. Emerg. Infect. Dis. 24, 2010–2020 (2015).

  32. 32.

    Kavalari, I. D., Fuursted, K., Krogfelt, K. A. & Slotved, H. C. Molecular characterization and epidemiology of Streptococcus pneumoniae serotype 24F in Denmark. Sci. Rep. 9, 5481 (2019).

  33. 33.

    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).

  34. 34.

    Park, I. H. et al. Differential effects of pneumococcal vaccines against serotypes 6A and 6C. J. Infect. Dis. 198, 1818–1822 (2008).

  35. 35.

    Croucher, N. J. et al. Diverse evolutionary patterns of pneumococcal antigens identified by pangenome-wide immunological screening. Proc. Natl Acad. Sci. USA 114, E357–E366 (2017).

  36. 36.

    Campo, J. J. et al. Panproteome-wide analysis of antibody responses to whole cell pneumococcal vaccination. eLife 7, e37015 (2018).

  37. 37.

    Tleyjeh, I. M., Tlaygeh, H. M., Hejal, R., Montori, V. M. & Baddour, L. M. The Impact of penicillin resistance on short-term mortality in hospitalized adults with pneumococcal pneumonia: a systematic review and meta-analysis. Clin. Infect. Dis. 42, 788–797 (2006).

  38. 38.

    Navarro-Torné, A. et al. Risk factors for death from invasive pneumococcal disease, Europe, 2010. Emerg. Infect. Dis. 21, 417–425 (2015).

  39. 39.

    Atkins, K. E. & Lipsitch, M. Can antibiotic resistance be reduced by vaccinating against respiratory disease? Lancet Respir. Med. 6, 820–821 (2018).

  40. 40.

    Finkelstein, J. A. et al. Impact of a 16-community trial to promote judicious antibiotic use in Massachusetts. Pediatrics 121, e15–e23 (2008).

  41. 41.

    Wroe, P. C. et al. Pneumococcal carriage and antibiotic resistance in young children before 13-valent conjugate vaccine. Pediatr. Infect. Dis. J. 31, 249–254 (2012).

  42. 42.

    Davies, N. G., Flasche, S., Jit, M. & Atkins, K. E. Within-host dynamics shape antibiotic resistance in commensal bacteria. Nat. Ecol. Evol. 3, 440–449 (2019).

  43. 43.

    Ruczinski, I., Kooperberg, C. & Leblanc, M. Logic regression. J. Comput. Graph. Stat. 12, 475–511 (2003).

  44. 44.

    Kay, E., Cuccui, J. & Wren, B. W. Recent advances in the production of recombinant glycoconjugate vaccines. NPJ Vaccines 4, 16 (2019).

  45. 45.

    Andrews, N. J. et al. Serotype-specific effectiveness and correlates of protection for the 13-valent pneumococcal conjugate vaccine: a postlicensure indirect cohort study. Lancet Infect. Dis. 14, 839–846 (2014).

  46. 46.

    Gladstone, R. A. et al. International genomic definition of pneumococcal lineages, to contextualise disease, antibiotic resistance and vaccine impact. EBioMedicine 43, 338–346 (2019).

  47. 47.

    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.e1–1002.e8 (2016).

  48. 48.

    del Amo, E. et al. High invasiveness of pneumococcal serotypes included in the new generation of conjugate vaccines. Clin. Microbiol. Infect. 20, 684–689 (2014).

  49. 49.

    Parra, E. L. et al. Changes in Streptococcus pneumoniae serotype distribution in invasive disease and nasopharyngeal carriage after the heptavalent pneumococcal conjugate vaccine introduction in Bogotá, Colombia. Vaccine 31, 4033–4038 (2013).

  50. 50.

    Rivera-Olivero, I. A. et al. Carriage and invasive isolates of Streptococcus pneumoniae in Caracas, Venezuela: the relative invasiveness of serotypes and vaccine coverage. Eur. J. Clin. Microbiol. Infect. Dis. 30, 1489–1495 (2011).

  51. 51.

    Sá-Leao, R. et al. Analysis of invasiveness of pneumococcal serotypes and clones circulating in Portugal before widespread use of conjugate vaccines reveals heterogeneous behavior of clones expressing the same serotype. J. Clin. Microbiol. 49, 1369–1375 (2011).

  52. 52.

    Sandgren, A. et al. Effect of clonal and serotype‐specific properties on the invasive capacity of Streptococcus pneumoniae. J. Infect. Dis. 189, 785–796 (2004).

  53. 53.

    Scott, J. et al. Serotype distribution and prevalence of resistance to benzylpenicillin in three representative populations of Streptococcus pneumoniae isolates from the coast of Kenya. Clin Infect Dis 27, 1442–1450 (1998).

  54. 54.

    Sharma, D. et al. Pneumococcal carriage and invasive disease in children before introduction of the 13-valent conjugate vaccine: comparison with the era before 7-valent conjugate vaccine. Pediatr. Infect. Dis. J. 32, e45–e53 (2013).

  55. 55.

    Smith, T. et al. Acquisition and invasiveness of different serotypes of Streptococcus pneumoniae in young children. Epidemiol. Infect. 111, 27–39 (1993).

  56. 56.

    Trotter, C. L. et al. Epidemiology of invasive pneumococcal disease in the pre-conjugate vaccine era: England and Wales, 1996–2006. J. Infect. 60, 200–208 (2010).

  57. 57.

    Varon, E., Cohen, R., Béchet, S., Doit, C. & Levy, C. Invasive disease potential of pneumococci before and after the 13-valent pneumococcal conjugate vaccine implementation in children. Vaccine 33, 6178–6185 (2015).

  58. 58.

    Zemlickova, H. et al. Serotype-specific invasive disease potential of Streptococcus pneumoniae in Czech children. J. Med. Microbiol. 59, 1079–1083 (2010).

  59. 59.

    Browall, S. et al. Clinical manifestations of invasive pneumococcal disease by vaccine and non-vaccine types. Eur. Respir. J. 44, 1646–1657 (2014).

  60. 60.

    Yildirim, I. et al. Serotype specific invasive capacity and persistent reduction in invasive pneumococcal disease. Vaccine 29, 283–288 (2010).

  61. 61.

    Brueggemann, A. B. et al. Clonal relationships between invasive and carriage Streptococcus pneumoniae and serotype‐ and clone‐specific differences in invasive disease potential. J. Infect. Dis. 187, 1424–1432 (2003).

  62. 62.

    Brueggemann, A. B. et al. Temporal and geographic stability of the serogroup‐specific invasive disease potential of Streptococcus pneumoniae in children. J. Infect. Dis. 190, 1203–1211 (2004).

  63. 63.

    Gray, B. M., Converse, G. M. & Dillon, H. C. Serotypes of Streptococcus pneumoniae causing disease. J. Infect. Dis. 140, 979–983 (1979).

  64. 64.

    Hanage, W. P. et al. Invasiveness of serotypes and clones of Streptococcus pneumoniae among children in Finland. Infect. Immun. 73, 431–435 (2005).

  65. 65.

    Jroundi, I. et al. Streptococcus pneumoniae carriage among healthy and sick pediatric patients before the generalized implementation of the 13-valent pneumococcal vaccine in Morocco from 2010 to 2011. J. Infect. Public Health 10, 165–170 (2017).

  66. 66.

    Kellner, J. D. et al. The use of Streptococcus pneumoniae nasopharyngeal isolates from healthy children to predict features of invasive disease. Pediatr. Infect. Dis. J. 17, 279–286 (1998).

  67. 67.

    Levidiotou, S. et al. Serotype distribution of Streptococcus pneumoniae in north-western Greece and implications for a vaccination programme. FEMS Immunol. Med. Microbiol. 48, 179–182 (2006).

  68. 68.

    Viechtbauer, W. Conducting meta-analyses in R with the metafor package. J. Stat. Softw. https://doi.org/10.18637/jss.v036.i03 (2015).

  69. 69.

    Mostowy, R. et al. Heterogeneity in the frequency and characteristics of homologous recombination in pneumococcal evolution. PLoS Genet. 10, e1004300 (2014).

  70. 70.

    Croucher, N. J. et al. Diversification of bacterial genome content through distinct mechanisms over different timescales. Nat. Commun. 5, 5471 (2014).

  71. 71.

    Lees, J. A. et al. Fast and flexible bacterial genomic epidemiology with PopPUNK. Genome Res. 29, 304–316 (2019).

  72. 72.

    Croucher, N. J. et al. Evidence for soft selective sweeps in the evolution of pneumococcal multidrug resistance and vaccine escape. Genome Biol. Evol. 6, 1589–1602 (2014).

  73. 73.

    Croucher, N. J. et al. Rapid pneumococcal evolution in response to clinical interventions. Science 331, 430–434 (2011).

  74. 74.

    Croucher, N. J. et al. Variable recombination dynamics during the emergence, transmission and ‘disarming’ of a multidrug-resistant pneumococcal clone. BMC Biol. 12, 49 (2014).

  75. 75.

    Fahrmeir, L. & Tutz, G. Multivariate Statistical Modelling Based on Generalized Linear Models 2nd edn (Springer, 2013).

  76. 76.

    Flasche, S. The scope for pneumococcal vaccines that do not prevent transmission. Vaccine 35, 6043–6046 (2017).

  77. 77.

    Mrkvan, T., Pelton, S. I., Ruiz-Guiñazú, J., Palmu, A. A. & Borys, D. Effectiveness and impact of the 10-valent pneumococcal conjugate vaccine, PHiD-CV: review of clinical trials and post-marketing experience. Expert Rev. Vaccines 17, 797–818 (2018).

Download references

Acknowledgements

We thank C. Levy for sharing epidemiological data. C.C. was supported by the Engineering and Physical Sciences Research Council of the United Kingdom (grant nos. EP/K026003/1 and EP/N014529/1) and the Government of Canada’s Canada 150 Research Chair programme. J.C. was supported by European Research Council grant no. 742158. N.J.C. was supported by a Sir Henry Dale Fellowship, jointly funded by Wellcome and the Royal Society (grant no. 104169/Z/14/A), and by the UK Medical Research Council and Department for International Development (grant no. MR/R015600/1).

Author information

C.C. developed and fitted the updated version of the model, and performed the optimization of vaccine designs and most statistical analyses. J.C. advised on statistical analyses and optimization. N.J.C. performed the epidemiological meta-analysis and other statistical analyses of vaccine designs.

Correspondence to Caroline Colijn.

Ethics declarations

Competing interests

C.C. and N.J.C. have protected the formulations identified in this work. N.J.C. has consulted for Antigen Discovery Inc. and been awarded an investigator-initiated award by GlaxoSmithKline.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data

Extended Data Fig. 1 Comparison of the Massachusetts and Maela S. pneumoniae populations.

Comparison of the Massachusetts and Maela populations. a, Frequencies of serotypes across the two studied populations; serotypes 15B and 15C, which rapidly interchange but were resolved separately in the Maela dataset, are merged into 15B/C for comparability in this plot. b, Frequencies of sequence clusters, groupings analogous to strains defined by Corander et al, across the two populations. Both plots demonstrate the dissimilarity of these two S. pneumoniae populations, despite them being isolated from nasopharyngeal carriage almost contemporaneously. c, Distribution of resistance scores relative to the population structure within each serotype across the Massachusetts and Maela populations. Each colour represents a different sequence cluster, such that they can be distinguished within a serotype.

Extended Data Fig. 2 Validation of the ODE negative frequency-dependent selection model.

a, Parameterising the ordinary differential equation (ODE) model. The solid lines (stochastic model output) and dashed lines (ODE model output) show the post-vaccine rate of serotype replacement. They demonstrate that the deterministic ODE model is appropriately parameterised to replicate the post-PCV7 temporal dynamics of the stochastic model, which was directly fitted to genomic surveillance data. b, Serotype frequencies observed in the genomic data (horizontal axis) and in simulations (vertical axis). The outputs of the NFDS model (blue) were compared to those from a neutral ‘proportional replacement’ model (red), in which each non-vaccine serotype expanded to replace the eliminated vaccine serotypes in proportion to its original carriage prevalence. The best-fitting linear relationships are shown by the corresponding coloured lines, and the surrounding shaded regions represent the 95% confidence intervals for each. The NFDS model correlates more strongly with the observed data (Pearson correlation, n = 32, R2 = 0.90) than the neutral model (Pearson correlation, n = 32, R2 = 0.78). c, This plot compares the predicted frequencies of pneumococcal sequence clusters 10 years post-PCV7 when using the ODE and stochastic NFDS models (n = 100 replicates for both versions of the stochastic model). The differences between the ODE and stochastic models are smaller than the differences between alternative mechanisms of strain migration implemented in the stochastic NFDS model (see Methods for details), demonstrating the ODE implementation to accurately replicate the population dynamics of the stochastic implementation.

Extended Data Fig. 3 Comparison of optimisation criteria at the 10 and 25 year timepoints.

These scatterplots compare the IPD burden measures used for vaccine optimisation at 10 and 25 years post-vaccination in Massachusetts (n = 480 optimised formulations in each plot) and Maela (n = 440 optimised formulations in each plot). Each plot also displays the simulated effect of introducing PCV13 into a vaccine-naïve population. Plots are separated by population and IPD burden measure. Points are coloured by the constraint on the formulation, and the criterion used for optimisation. The line of identity is marked in black. The IPD measures are very similar at the two timepoints, indicating that while the model dynamics have long transient behaviour driven by drift among similar genotypes, the IPD burden criteria converge towards a feasible-time value relatively early.

Extended Data Fig. 4 Invasiveness of pneumococcal serotypes in infant and adults.

Variation in invasiveness between serotypes in infants and adults. Each bar represents the logarithmic invasiveness odds ratios for a serotype, estimated from the meta-analyses of IPD and carriage isolates (Supplementary Tables 13) using a random effects model. The 95% confidence intervals associated with these estimates are shown by the error bars. The number of studies contributing to the estimates for each age group for each serotype are enumerated at the top of the plot, with the individual study estimates overlaid as individual points. Results are coloured according to the currently-available vaccines in which the serotype is found, if any. a, Invasiveness in infants relative to carriage in infants. b, Invasiveness in adults relative to carriage in infants. Fewer serotypes are present in this panel, as there were fewer datasets available to estimate these values.

Extended Data Fig. 5 Post-vaccine populations forecast following optimised 20-valent PCV introductions.

Differences in serotype prevalences, forecast 10 years after vaccine introduction, between the best-performing 20-valent strategies optimised under different criteria in a, Massachusetts, and b, Maela. Bars are coloured according to whether they represent the frequency of a vaccine or non-vaccine serotype in the corresponding formulation. In Massachusetts, serotypes 6C, 11A, 15B/C and 35B are typically prevalent in the post-vaccine population regardless of the optimisation criterion, owing to their low infant invasiveness. Serotypes 15A and 23A are higher when minimising infant IPD, whereas serotypes 6A and 23B are higher when minimising overall IPD, in accordance with their age-specific invasiveness (Extended Data Fig. 4). Minimising AMR IPD results in higher prevalence of serotype 10A, which is pansusceptible in Massachusetts. In Maela, all optimal formulations result in high post-vaccine prevalences of serotypes 6A, 6C, 11A, 15F, 19B, as well as non-typeables. Serotypes 19F and 23F remain prevalent when optimising for overall and infant IPD, respectively; both are suppressed when optimising for AMR IPD, owing to their antibiotic resistance profiles (Fig. 5). These are partially replaced by serotypes 6B and 34, which have a weaker association with resistance.

Extended Data Fig. 6 Distribution of protein antigens across serotypes.

This barchart shows the prevalences of the intermediate-frequency protein antigens within each serotype with at least 10 representatives across the Massachusetts and Maela populations.

Extended Data Fig. 7 Alternative strategies for minimising IPD.

a-b, These plots summarise the formulations of PCVs optimised with pneumococcal carrier proteins in a, Massachusetts and b, Maela. Results are displayed as in Fig. 2c,d, except that the first column denotes the carrier protein on which the design was based. Rows are ordered first by the featured pneumococcal carrier protein, and then by the predicted post-vaccine infant IPD burden, which the formulations were designed to minimise. c-d, These plots summarise the compositions of complementary adult vaccines (CAVs) designed to minimise adult IPD following introduction of infant vaccines to minimise AMR IPD (corresponding to the vaccines in Fig. 5) in c, Massachusetts and d, Maela. On each row, the light blue cells define the infant formulation, and the dark blue cells define the adult formulation. Rows are ordered by the overall IPD burden estimated following the implementation of the combined vaccination strategy.

Extended Data Fig. 8 Comparing formulations’ effects on IPD using different criteria.

Performance of vaccination strategies judged by different criteria: a, minimising infant IPD; b, minimising overall IPD; c, minimising AMR IPD. Each violin plot is labelled with the constraint on formulation design, and coloured according to the criterion optimisation was intended to minimise. The overlaid points show the estimated effects of each individual optimised formulation (n = 20 for each combination of constraint and optimisation criterion in each population). The purple point in each panel shows the corresponding estimates for PCV13. For the Maela population, no optimisation was performed for two proteins (RrgB2 and ZmpC) that were below the threshold frequency of 0.05 in the starting population (Supplementary Fig. 6), and therefore not included in the multi-locus NFDS simulations. The diminishing returns of expanding infant vaccine valency can be inferred from the predicted effects of the 10-, 15- and 20-valent vaccines relative to the horizontal dashed line, which marks the pre-vaccine value of the optimisation criterion.

Extended Data Fig. 9 Comparing formulations’ effects on IPD in different populations.

Performance of vaccine strategies in the alternative population to that for which they were designed. Simulations of each strategy were run in the alternative population, and their performance compared to that in the intended recipient population using different criteria: minimising infant IPD; minimising overall IPD, and minimising AMR IPD. Panels are labelled to indicate the population for which the formulation was designed. For those panels in which the intended target population was Massachusetts, results are shown for 480 optimised formulations. For those panels in which the intended target population was Maela, results are shown for 440 optimised formulations. Each plot also displays the estimated effect of introducing PCV13 into a vaccine-naïve population. Notably, those vaccines designed to reduce infant and overall IPD in Massachusetts are predicted to perform poorly in Maela.

Extended Data Fig. 10 Comparing formulations’ effects on IPD using different ecological models.

These scatterplots compare the simulated effectiveness of vaccine formulations in the original multi-locus NFDS model and an otherwise equivalent ‘proportional replacement’ neutral model (Extended Data Fig. 2). Each plot compares the expected post-vaccine IPD burdens expected under NFDS and neutral evolution. Points (n = 480 for the Massachusetts population; n = 440 for the Maela population) are coloured by optimisation constraint and criterion, and the line of identity is marked in black. For the formulations identified by optimization, the results are very similar under both ecological models, with vaccine compositions that we predict to perform better than PCV13 using the NFDS model also tending to do so in the neutral model. This indicates that the formulations we have identified perform well despite the predicted effects of NFDS, rather than because of them.

Supplementary information

Supplementary Information

Supplementary Figs. 1–11 and Supplementary Tables 4–6.

Reporting Summary

Supplementary Tables

Supplementary Tables 1–3.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Colijn, C., Corander, J. & Croucher, N.J. Designing ecologically optimized pneumococcal vaccines using population genomics. Nat Microbiol 5, 473–485 (2020). https://doi.org/10.1038/s41564-019-0651-y

Download citation