What vaccines can tell us about bacterial ecology
When a vaccine eliminates half a bacterial species, who is first to refill the niche? What does this tell us about how interventions should be designed to minimise disease?
The paper is describing this work in Nature Ecology & Evolution is here: http://go.nature.com/2kUm96r
The work reported in this paper began in 2001, when the SPARC project started collecting Streptococcus pneumoniae strains from across Massachusetts. Despite being very common, this bacterium is not the most easily accessible of species. Its normal habitat is the nasopharynx, the hollow space between the nose and mouth. Sampling the overall population is traditionally done through inserting a long swab through the nasal passage to reach the mucus-covered surfaces inside the head. Over a fifth, and sometimes up to three-quarters, of children under five will yield at least one strain, and in some cases two or more can be recovered.
The timing was critical; the previous spring, the first vaccine targeted against S. pneumoniae effective in infants had been introduced in the USA. The vaccine contained seven distinct carbohydrate-based molecules, which afforded protection against approximately half of the S. pneumoniae population. Were the bacterium to confine itself to the nasopharynx, such measures would not be necessary. However, the ability of S. pneumoniae to descend into the lungs means it is the most common bacterial cause of community-acquired pneumonia. The bacterium is also a common cause of sepsis, through invading into the bloodstream, and meningitis, an infection of the membranes surrounding the brain.
Over the subsequent 15 years, during which a dramatic fall in S. pneumoniae disease in children was observed across the USA, the SPARC project sampled over 6,500 children. There was no decrease in the frequency with which swabs yielded S. pneumoniae strains, making it clear the bacterium had not become less common. Yet serotyping, which distinguishes between bacteria using antibodies, confirmed the selective loss of ‘vaccine type’ strains, which express the carbohydrate structures included in the vaccine. Therefore the fall in disease was attributed to the vaccine’s clearance of more virulent strains, and their replacement by less invasive variants. Genetic analysis of the strains using ‘multilocus sequence typing’, a method of identifying closely-related isolates, suggested this was not a simple process: in some cases, ‘vaccine type’ strains died out, whereas others survived in an altered form that had evaded the vaccine. Some strains not targeted by the vaccine remained at stable frequencies, whereas others sprung up once competition from vaccine types was removed. It was not clear whether these changes were the consequence of stochastic changes, or were guided by underlying ecological processes.
Simulations of the S. pneumoniae population: each line shows the frequencies of a strain in the post-vaccine population. The more red the line, the higher the proportion of the strain targeted by the vaccine; the more blue the line, the higher the proportion of the strain unaffected by the vaccine.
A clue to resolving this came from a study of the pneumococcal pilus, a surface structure used to adhere to host tissues, present in about a fifth of strains. Many strains expressing the pilus were eliminated by the vaccine, causing it to drop in frequency. However, it later ‘bounced back’ closer to its pre-vaccination prevalence. When it became feasible to use whole genome sequencing to characterise the SPARC collection, it was clear there were hundreds of genes, each present only in a fraction of the population, which maintained quite consistent frequencies. The exceptions were genes encoding for the production of the surface structures directly targeted by the vaccine. This indicated the reshaping of the population occurred in such a manner to maintain these intermediate-frequency genes at their equilibrium levels.
Comparisons with other systematically-sampled collections of S. pneumoniae revealed conservation of these gene frequencies was not unique to Massachusetts. Despite the presence of a very different set of strains, the population of S. pneumoniae from the Maela refugee camp near the Thai-Myanmar border (pictured at the top) shared very similar gene frequencies. An extensive carriage surveillance project in Southampton, following the introduction of the vaccine in the UK, reproduced the same pattern, as did comprehensive surveillance of invasive disease in Nijmegen, the Netherlands. The proposed mechanism was negative frequency-dependent selection: the situation in which genes are more advantageous to their host when they are rare. When such genes fall in frequency, the cells in which they are found become fitter and increase in number. This can result in balancing selection that stabilises the genes’ frequencies. Analysing the likely functions of the genes involved suggested there were multiple mechanisms through which this could happen. These include ‘rock-paper-scissors’ interactions in the way strains in the same nasopharynx secrete chemicals to kill each other, and ‘kill-the-winner’ effects where viruses targeting the most successful strains become common to the extent at which they force down the strains’ abundances.
This explains why the post-vaccine strain alterations did not involve simple neutral expansion of strains not targeted by the vaccine, nor like-for-like replacement of individual vaccine-targeted strains. Instead, the complex strain-level changes can be modelled as the consequence of gene frequencies being preserved. Using recently developed statistical software capable of rapidly parameterising simulations to replicate complex data, a model of negative frequency-dependent selection could reproduce the post-vaccination changes in the bacterial population structure, with similar estimates of the strength of selection across each dataset. Although the model undoubtedly requires further refinement, its ability to replicate the impact of vaccination on a bacterial population opens up opportunities to rationally design interventions that efficiently remove virulent strains from a bacterial population.
The SPARC project, which recently marked 15 years of surveillance, has pioneered the use of different techniques to improve our understanding of the consequences of vaccination against S. pneumoniae. The lessons learned can now be applied to the samples being generated from around the world by the S. pneumoniae Global Sequencing Project, which will bring genomic epidemiology to bear on very different collections from lower and middle-income countries. In these high burden settings, the consequences of getting vaccine formulations right is all the more important.