Applying Predictive Modeling to Pneumococcal Vaccine Design Could Halve Disease Occurrence

Applying Predictive Modeling to Pneumococcal Vaccine Design Could Halve Disease Occurrence
Pneumococcal pneumonia. Computer illustration of Streptococcus pneumoniae (pneumococci) bacteria inside the alveoli of the lungs, causing pneumonia.

Studies by an international research team suggest that rates of disease caused by the bacterium Streptococcus pneumoniae could be substantially reduced by changing how vaccines are designed. The scientists, at the Wellcome Sanger Institute, Simon Fraser University in Canada, and Imperial College London combined genomic data, models of bacterial evolution, and predictive modeling to identify how vaccines could be optimized for specific age groups, geographic regions, and communities of bacteria.

The study, published in Nature Microbiology, simulated the performance of vaccines over time to assess the risk of vaccine-targeted strains being replaced by other, potentially dangerous strains of bacteria. Through this predictive modeling approach, the researchers identified new vaccine designs that could help reduce overall rates of disease. “Our research shows that the best vaccine designs strongly depend on the bacterial strains present in the population, which vary considerably between countries,” commented Nicholas Croucher, PhD, senior lecturer, bacterial genomics, at the MRC Centre for Global Infectious Disease Analysis, Imperial College London. “The best vaccine designs also depend on the age group being vaccinated. These ideas will be critical for applying lessons learned from introducing vaccines in high-income countries to combating the disease where the burden is highest.” Croucher and colleagues reported on their studies in a paper titled, “Designing ecologically optimized pneumococcal vaccines using population genetics.”

S. pneumoniae is commonly found at the back of the nasal cavity, where it is normally harmless. “The asymptomatic carriage prevalence of S. pneumoniae peaks in the first five years of life, reaching 25–50% in high-income countries and 20–90% in low- and middle-income countries,” the authors wrote. However, when the bacterium migrates to other parts of the body, it can cause serious bacterial infections such as pneumonia, sepsis, and meningitis—known collectively as invasive pneumococcal disease (IPD). IPD was estimated to cause around 1.6 million deaths per year worldwide prior to the introduction of widespread vaccination, with higher rates of disease in many low- or middle-income countries. Infants and the elderly are most at risk.

While vaccines against pneumococcus have prevented millions of infections, the bacteria are difficult to target precisely, because infection can be caused by different serotypes. “Each component of current protein–polysaccharide conjugate vaccines (PCVs) generally induces immunity specific to one of the approximately 100 pneumococcal serotypes …,” the researchers explained.

The most complex pneumococcal conjugate vaccine (PCV13) today targets 13 serotypes, but because there are approximately 100 S. pneumoniae serotypes around the world, vaccine effectiveness varies between countries depending on which serotypes are present. When serotypes are removed from circulation by a particular vaccine, other serotypes of S. pneumoniae take their place. This means that S. pneumoniae bacteria aren’t completely eradicated. “The resulting serotype replacement process typically eliminates vaccine types without any reduction in the overall S. pneumoniae carriage prevalence.”

The first pneumococcal PCV contained seven serotypes, and a subsequent updated formulation contained 10 serotypes. The current PCV13 vaccine contains three serotypes in addition to those in PCV10. “These updated formulations are now administered to millions of children across hundreds of countries, and recent modeling predicted the global use of PCV13 would avert almost 400,000 infant deaths annually,” the investigators commented.

Serotype replacement in response to vaccination is not without issues. In France, for example, the incidence of penicillin-resistant meningitis increased subsequent to the introduction of PCV13. More generally, the effects of PCV on the proportion of IPD caused by antimicrobial-resistant serotypes replacing the vaccine-targeted serotypes are variable. It may also take years for any impact to become evident after a new vaccine has been introduced. Unfortunately, manufacturing PCVs is complex and expensive, so it’s no simple matter to continually expand the number of serotypes covered.

For their reported study, the researchers at the Wellcome Sanger Institute, Simon Fraser University, and Imperial College London optimized a computer model to approximate the effect of vaccines targeting different serotype combinations. “…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,” they explained. Analysis of vaccine effectiveness was then carried out on S. pneumoniae genomic datasets from two geographic regions. One was from Massachusetts, comprising 616 bacterial genomes isolated from nasopharyngeal samples from children following the introduction of PCV7, “and representing a typical western S. pneumoniaepopulation.” The second dataset was from the Mae La refugee camp on the Thailand-Myanmar border, comprising 2,336 genomes carried by unvaccinated individuals.

The complexity of S. pneumoniae vaccines means that many designs are possible, each with different effects on disease. Based on the number of detected serotypes, approximately 3.47 x 109 and 1.05 x 1013 13-valent PCVs were possible for the Massachusetts, and Mae La, serotypes, respectively.

So, in Mae La, for example, the presence of 64 S. pneumoniae serotypes meant that around 100 trillion vaccine designs would be possible. But it would take 19,000 years to simulate them all, and most would be sub-optimal. The researchers developed a more efficient method that made it feasible to identify the best-performing designs from the trillions of possibilities. “We compared optimal formulations for reducing infant-only or population-wide IPD, and identified potential benefits to including non-conserved pneumococcal carrier proteins,” they stated. “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.”

The team discovered that rates of infant IPD in Mae La could actually be reduced by omitting components from the PCV13 vaccine to keep certain serotypes in place, and remove the possibility that they would be replaced by highly invasive serotypes. “In the Mae La dataset, optimized 15- and 20-valent PCV formulations were predicted to lower infant IPD substantially more than PCV13,” the authors stated. “The best performing ‘subset’ vaccine did not contain the pediatric serotypes 6A, 6B, 19F, and 23F, and reduced infant IPD partly by maintaining high levels of these serotypes in the post-vaccine population …Were these serotypes to be removed by PCV13, the mixture of serotypes replacing them would include some that are potentially highly invasive (for example, 40 and 46).” In Massachusetts, a vaccine targeting 20 serotypes was found to be more effective than the current PCV13. “In the Massachusetts dataset, 15-valent formulations could only slightly outperform PCV13 in terms of forecast infant IPD, whereas 20-valent formulations offered more opportunities for reducing disease.”

The results highlight the need for vaccine programs to be tailored to specific communities of bacteria and to consider vaccination at different ages. Vaccination of infants also impacts IPD in adults. However, trends in IPD can differ between infants and the elderly in the same country. Older adults can also suffer from increased incidence of IPD, but they don’t carry S. pneumonia at the same, high levels observed in children, the team noted. In many places, older adults already receive an S. pneumoniae vaccine, which was designed before the infant vaccine. “ … post-vaccine trends in infant and adult IPD can diverge,” the scientists stated. In the United Kingdom, for instance, there has been a 4% increase in adult IPD as infant IPD has declined ost-PCV13. “This highlights the risks attendant on reshaping the bacterial population through PCV associated strain replacement because the post-vaccine population can have an increased propensity to cause IPD relative to that preceding the immunization campaign.”

The scientists’ reported study suggested that adult disease rates could be reduced by almost 50% by redesigning adult vaccines to complement those administered to infants. “This approach to optimizing vaccines will help to address several problems, such as invasive disease among infants or adults and minimizing antibiotic resistance in the post-vaccine population,” noted Caroline Colijn, PhD, professor, Simon Fraser University and the Wellcome Sanger Institute. “Such an approach also enables public health policy-makers to assess the likely effectiveness of an existing vaccine for a local population based on genomic surveillance data.”

The findings coincide with growing concern about the threat of antimicrobial resistance (AMR) to common drugs. S. pneumoniae infections are sometimes resistant to multiple antibiotics and have been highlighted as a priority threat by the World Health Organization. The study by Croucher, Colijn, and co-author Jukka Corander, PhD, professor at the University of Oslo, University of Helsinki, and the Wellcome Sanger Institute, highlighted how vaccines can be designed to reduce the risk that an S. pneumoniaeserotype carried by an individual would be resistant to common treatments. “This analysis identified a set of pneumococcal vaccines, each of which was designed to be optimal for a defined starting population, a design constraint, and an optimization criterion specifying the type of IPD to be minimized,” the authors concluded. “ … it would appear to be beneficial to broaden the portfolio of licensed formulations, rather than globally optimize a single formulation.”

Corander noted, “With the power of the latest DNA sequencing technology we are heading towards a future where large-scale genomic surveillance of major bacterial pathogens is feasible. The approach we describe in this study will play an important role in accelerating future vaccine discovery and design to help reduce rates of disease.”