This mechanistic study of two independent European birth cohorts, the COPSAC2010 and the EMIL cohorts with detailed ambient air pollution exposure assessment of PM2.5, PM10 and NO2 throughout pregnancy and infancy and with a similar set of inflammatory proteins between the cohorts, demonstrated strong associations between an air pollution proteomic fingerprints and the subsequent risk of developing both upper and lower respiratory tract infections as well as other common infections in early childhood. Further, significant alterations in the maternal inflammation-related proteomic profile from prenatal air pollution were characterized by downregulation of several proteins, including AXIN1, which was also associated with a protective effect on risk of infection and asthma development.
We have investigated the impact of pregnancy and early childhood air pollution exposure on the inflammatory blood proteomic profile in relation to risk of respiratory infections in early childhood. Previously, a large cohort study12 from MoBA (n = 17,533) investigated the association between NO2 exposure in pregnancy and LRTIs until age 18 months in childhood and found no associations. However, they did not assess PM2.5 and PM10 exposures, concentrations were characterized as low, and diagnoses were based on questionnaires retrospectively at age 6 and 18 months without assessment of daily symptom load as in our study. We found an increased risk of gastric infections from increased PM10 exposure, although there were no associations with upper or lower respiratory tract infections directly from air pollution exposure in either the COPSAC2010 or the EMIL cohort. However, other larger cohort studies10,11,18,19,20 (n = 1510, n = 2568, n = 3515, n = 2199 and n = 1263, respectively) have previously identified both pre- and postnatal air pollution exposure as an independent risk factor for LRTI development in young children, which could suggest that our study sample size for detecting these direct associations was too small. The ESCAPE project, utilizing data from 10 birth cohorts (BAMSE, GASPII, GINIplus, LISAplus, MAAS, PIAMA, and four INMA cohorts) (n = 16,059), found that ambient air pollution exposure from birth increased the risk of pneumonia until age 2 years13. A recent (2024) large observational study (n = 224,214) found ambient PM2.5 exposure in early childhood to be associated with increased risk of respiratory infections in pre-school aged children as well14.
The mechanisms of disease development have mostly been speculative, including suggestions of a skewed immune system development and negative changes to the fetal and newborn lung development, and we are the first to demonstrate changes in both the maternal and newborn child blood inflammatory proteomic profiles, which were directly related to childhood infection proneness. Current literature describes air pollutants in general to be linked with low-grade systemic inflammation, an altered gut microbiota, and oxidative stress, resulting in an increased risk of cardiovascular disease development in particular21, but also leading to pathophysiological changes in the respiratory system12 and increased risk of gastrointestinal disorders22. By using mechanistic data layers to understand previously reported associations in other large cohorts between air pollution and respiratory infections, we are now able to characterize a high air pollution inflammatory proteomic profile based on blood plasma samples. We found a downregulation of IL8, confirming the findings of a previously conducted study15. Interestingly, the protein AXIN1 seemed to play a vital role in our study as it was significantly downregulated from both PM2.5 and PM10 exposure and contributed alone to the PM10 fingerprint, yielding the highest AUC, and to the PM2.5 and NO2 fingerprints with negative loadings as well. Interestingly, AXIN1 has been demonstrated to be highly expressed in the nasopharynx, bronchus, and lung tissue23, and an experimental mouse study demonstrated reduction of AXIN1 levels in the lungs at early stages of influenza pneumonia. Further, they found AXIN1 to inhibit both RSV and influenza virus replication as well as boost interferon (IFN) response through increased mRNA expression of IFNβ1 and the IFN-targeted anti-viral gene OAS1 in mice24. The authors even suggested the protein as a therapeutic target to prevent airway infections.
Another experimental study of mice recently demonstrated that AXIN1 acted as a regulator of antiviral innate immunity against virus infections by stabilizing interferon regulatory factor 3 (IRF3) and boosting IFN production, also suggesting this protein as an effective antiviral agent25.
This is in line with the findings from our study suggesting AXIN1 to play a key protective role against risk of infections in young children as we found higher levels of the protein to associate with a decreased risk of the overall infection burden, cold, pneumonia, tonsillitis and fever episodes and asthma diagnosis until age 10 years. Interestingly, similar results showing a decreased risk of respiratory infections from higher AXIN1 levels at age 1 year in the EMIL cohort were demonstrated, supporting the findings from COPSAC2010.
In addition, the protein seemed to be directly modified by air pollution exposure, with a significant downregulation in our cohort, highlighting this as a crucial protein in the link between air pollution exposure and increased infection risk found in other studies. These findings in two childhood cohorts indicate what others have previously speculated from experimental studies in animals, that AXIN1 could serve as a potential therapeutic target in the prevention of respiratory infectious diseases, which should be further tested in experimental studies using human cells.
This study was strengthened by the findings from two independent cohorts showing similar associations between the air pollution-derived proteomic fingerprint against mainly respiratory infection risk. This procedure allows for validation of the initial findings in COPSAC2010 where prenatal air pollution caused alterations in the maternal proteome relating to childhood infection risk, by later demonstrating a similar pattern in infancy of the air pollution-derived proteomic fingerprint against later infection risk in EMIL. For the COPSAC2010 cohort, the close longitudinal follow-up and unique daily parental registration of infection episodes from birth until age 3 years in combination with a comprehensive amount of information on environmental and social circumstances that could potentially act as confounders and are adjusted for, is a major strength of the study. Together with the proteomic mechanistic data layer, this allows for deep clinical phenotyping of the children and interpretation of the independent impact of air pollution on the mother’s proteomic profile that affects the newborn child’s infection risk. Further, our infection data in COPSAC2010 with daily registrations of symptoms of common infections during the first 3 years of life are unique and different from other cohorts. Another strength is the individual air pollution exposure assessment with high temporal and spatial resolution that takes into account full address history in both cohorts and assessment of ambient air pollution during both pregnancy and infancy whereas most previous large studies have focused on exposure during childhood alone and relied on modeling methods with less fine temporal resolution that did not consider meteorological factors and only had a single address available. The proteomic profiling of the same set of 92 inflammatory proteins in both cohorts followed the same procedure through the Olink Proteomics platform, mitigating the risk of systematic analyses errors. Finally, both cohorts were population-based, including both Danish pregnant women and children, and Swedish children from the greater Copenhagen and Stockholm areas, allowing for generalization of the findings.
Although the similar findings in an independent cohort strengthens our findings, the study was limited by the observational study design with the risk of residual confounding from other potentially associated factors e.g., noise pollution, co-pollutant exposures and lifestyle factors that could influence what the fingerprints also reflects other than air pollution and the overall risk of infections, and even though we adjusted for all available potential confounders from our data we cannot draw causal conclusions from this study. Another limitation is that air pollution assessment was not assessing specific periods of pregnancy. Our study sample size was limited compared with other observational studies linking air pollution to risk of respiratory infections and the lack of overall association between air pollution exposure in pregnancy and childhood infection risk may be explained by our limited sample size, however, we still had enough participants to demonstrate mechanistic alterations in the blood proteomic profile in relation to infection risk, which was the aim of this study. Hence, we believe this study adds important knowledge to existing literature, trying to understand the underlying mechanisms of previously found associations in larger cohort studies. Finally, the two cohorts were not directly comparable in terms of data availability and should be viewed as complementary, where the COPSAC study investigates prenatal air pollution exposure and the EMIL study investigates infancy air pollution exposure.
We have demonstrated changes in the maternal and newborn child’s inflammatory proteomic profile from ambient air pollution exposure relating to both respiratory and common infection risk in early childhood based on data from two independent European birth cohorts. In addition, our findings suggest that the AXIN1 protein, which was modified by air pollution exposure to play a role in the downregulation of infection proneness in young children and could potentially serve as a target protein in respiratory infection prevention.
