Air Pollution’s Effect on Pneumococcal Disease Depends on the Bacterial Lineage Circulating

For nearly two decades, South African hospitals have quietly recorded every case of invasive pneumococcal disease that crossed their wards. Now, a team of researchers has mined that extraordinary archive and uncovered something unexpected: the danger posed by dirty air is not fixed. It shifts depending on which strain of Streptococcus pneumoniae happens to be circulating in the population at the time. The finding, published in Nature Microbiology, suggests that the microscopic ecology of a pathogen can reshape how the environment threatens human health — and vice versa.

The study, led by Sophie Belman of the Barcelona Supercomputing Center and the Wellcome Sanger Institute together with Anne von Gottberg of South Africa’s National Institute for Communicable Diseases and Rachel Lowe, also of the Barcelona Supercomputing Center, analysed 59,017 cases of invasive pneumococcal disease (IPD) collected from 531 hospitals across all nine South African provinces between 2005 and 2023. Crucially, the team also sequenced the genomes of 4,350 bacterial isolates, allowing them to link each case not just to a serotype — the sugar capsule that defines over 100 pneumococcal types — but to a genomic lineage, one of hundreds of Global Pneumococcal Sequence Clusters (GPSCs) that make up the species’ population structure.

South Africa offered the perfect laboratory. The country spans temperate and arid climates, from the subtropical highlands of Gauteng to the Mediterranean coast of the Western Cape, and carries an enormous burden of pneumococcal disease. Nasopharyngeal colonization estimates range from roughly 40 to 60 percent of the population. During the study period, two major public health interventions reshaped the disease landscape: the seven-valent pneumococcal conjugate vaccine (PCV7) arrived in 2009, the thirteen-valent PCV13 followed in 2011, and antiretroviral therapy (ART) for people living with HIV became widespread from 2005. Yet despite these advances, IPD continued to spike each year in the colder months, hinting that environmental forces were still at work.

To disentangle those forces, the researchers built Bayesian spatiotemporal models that accounted for population density, vaccination periods, and residual seasonal, spatial and interannual variation. They then layered on distributed lag nonlinear models (DLNMs), a statistical framework designed to capture effects that are both delayed and nonlinear — the kind of relationship where a pollutant spike this week might translate into disease three weeks later, and where the dose-response curve bends rather than runs straight. The models tracked environmental exposures for up to eight weeks after each event, a window chosen because cross-correlation analyses showed associations extending to five weeks and because the bacterium’s estimated 30-day generation time makes longer biological effects plausible.

The meteorological results were nuanced. Cumulatively, over an eight-week period, moderate relative humidity between 33 and 49 percent increased IPD risk by 5 percent, while cold minimum temperatures between 4 and 10 °C raised it by 4 percent. Warm maximum temperatures between 27 and 38 °C, by contrast, triggered an acute risk increase of up to 10 percent within a single week of exposure — a divergent lag structure suggesting that heat and cold act through different mechanisms. At very high relative humidity, around 74.8 percent, the team found a strong protective effect, reducing risk by roughly 23 percent. Increasing precipitation and reduced drought conditions also lowered IPD risk, while higher wind speed raised it.

But the air pollution findings were the most striking. Particulate matter smaller than 2.5 micrometres (PM2.5) — fine particles produced largely by human activity in South Africa — showed a clear positive association with IPD. With a weekly average of 50 micrograms per cubic metre, just above South Africa’s national air quality standard of 40, the eight-week cumulative risk of IPD rose by 3.4 percent. At 100 micrograms per cubic metre, a concentration that Johannesburg regularly exceeds — 32 weeks in 2023 alone — the risk jumped by 18 percent, and the effect was immediate, beginning in the same week as exposure and persisting for six weeks. PM10 and sulfur dioxide also raised IPD risk, though with different lag patterns: SO2 effects began only three weeks after exposure.

The population-level stakes are considerable. In 2019, Gauteng province recorded a mean PM2.5 concentration of 117 micrograms per cubic metre alongside 718 observed IPD cases. The team calculated a population attributable fraction of 2.19 percent, meaning that if PM2.5 had been held at or below the national threshold of 40 micrograms per cubic metre, roughly 16 of those 718 cases could theoretically have been averted. Meeting the more stringent World Health Organization daily threshold of 15 micrograms per cubic metre would have prevented about 25 cases, or 3.45 percent. Across South Africa, roughly 23 districts a year experience weekly PM2.5 concentrations of at least 50 micrograms per cubic metre, with an average of 14 weeks per district.

Yet the risk was not uniform across the population or the pathogen. Adults aged 65 and older faced the highest cumulative risk from PM2.5, with a relative risk of 1.16 at 50 micrograms per cubic metre — 12 percent higher than the population average — consistent with cumulative exposure and reduced lung function in older adults. People aged 15 to 64 showed a more acute response, with risk rising in the same week as high pollution, possibly reflecting the higher prevalence of HIV in this group. Children under five were not spared: their cumulative risk was similar to the population average, translating to roughly 3 percent more children with IPD at high pollution levels. Meningitis carried a higher pollution-associated risk than bacteraemia, and penicillin-resistant infections showed a 5 percent higher risk than the main model, suggesting that air pollution may interact differently with different disease presentations.

The serotype-level analysis revealed something remarkable. Only three serotypes — 4, 8 and 23F — conferred an immediate risk of IPD following PM2.5 exposure, while others showed delayed effects. Serotype 8, a non-vaccine type, and PCV13 serotypes 23F and 4 all showed acute responses, and some serotypes, including PCV7 serotype 14 and PCV13 serotype 19A, carried higher cumulative risk than the population average. The researchers suggest that serotypes with higher population prevalence may respond immediately to air pollution, whereas those with reduced post-vaccine carriage may exhibit delayed risk. Mechanistically, PM2.5 probably acts on both the host — increasing susceptibility to colonization and decreasing mucociliary clearance — and the bacterium, which can increase its transformation capacity and biofilm formation in response to pollutants.

The most unexpected finding came from the genomic lineages. When the team interacted the PM2.5 cross-basis with the proportion of each dominant GPSC circulating in a given week and province, they found that lineage diversity modified the timing of the pollution effect. GPSC21, which carries serotype 19F and became predominant in South Africa after PCV7 drove down the initially common GPSC1-19F lineage, shifted the lag structure entirely. In places and times with high GPSC21 prevalence, IPD risk rose acutely in the same week as high PM2.5 exposure; where GPSC21 was rare, risk increased only five weeks after exposure. The cumulative effect did not change, but the timing did — a subtle but biologically meaningful shift that suggests circulating lineage diversity plays a role in environment-disease relationships, even though the mechanism remains unclear.

Belman and colleagues argue that these results demonstrate something profound: pneumococcal population structure shapes air quality risk, and air quality in turn can shape the fitness landscape of microbial populations. High air pollution may favour certain serotypes and lineages, conferring environmental fitness advantages that could explain why some strains persist despite vaccination. The relationship is further complicated by the inextricable link between particulate matter and climate change, whereby PM influences atmospheric chemistry while climatic shifts affect PM formation and distribution. In high-exposure regions like Gauteng, these pressures are particularly acute.

The study has limitations. The PM2.5 data come from global reanalysis models — the Copernicus Atmosphere Monitoring Service (CAMS) performed best against South African observational data, but correlations were imperfect, particularly in the Gert Sibande region. The team used weekly means rather than 24-hour means, which may underestimate risks relative to more stringent WHO standards. Household air pollution data were unavailable, and the analysis focused on invasive disease, excluding non-bacteraemic pneumonia. Ascertainment bias from access to care and specimen processing may also skew results, and cases were attributed to hospital locations rather than patients’ homes.

Still, the implications are clear. Meeting established air pollution policy thresholds remains a critical priority to protect human health, and more stringent goals could directly mitigate IPD. Beyond pneumococcal disease, the approach — integrating microbial genomic data into environmental epidemiological models — could be leveraged for other infectious diseases, offering a template for understanding how pathogens respond non-uniformly to the environments they inhabit. As climate change and urbanization continue to reshape both air quality and microbial populations, studies like this one suggest that the two cannot be understood in isolation.

Subject of Research: Interactions between air pollution, meteorological factors and pneumococcal population structure in shaping invasive pneumococcal disease risk in South Africa.

Article Title: Pneumococcal population structure influences the effects of air pollution on invasive disease risk in South Africa

Article References: Belman, S., Lekhuleni, C., Kleynhans, J., Moirano, G., Lührsen, D., Carnerero, C., Meiring, S., Lo, S. W., du Plessis, M., von Gottberg, A., & Lowe, R. (2026). Pneumococcal population structure influences the effects of air pollution on invasive disease risk in South Africa. Nature Microbiology. https://doi.org/10.1038/s41564-026-02458-5

Image Credits: AI Generated

DOI: 10.1038/s41564-026-02458-5

Keywords: Streptococcus pneumoniae, invasive pneumococcal disease, air pollution, PM2.5, genomic lineages, South Africa, Bayesian spatiotemporal models, pneumococcal conjugate vaccines, Pneumococcal, population, structure, influences