What Is the Exposome and How It Can Help Research on Air Pollution

Ultrafine particles (UFP) exposure has been less studied than exposure to larger particles so far. UFP have characteristics that may lead to higher toxicity, including their extremely small size (allowing them to reach deeper into the tissues) and higher surface-to-mass ratio (that facilitates adhesion of larger amounts of hazardous materials). Whether or not this translates into increased risks of respiratory or cardiovascular diseases in humans remains to be clarified.

In Exposomics, we conducted a study on adult onset asthma (AOA) (n = 73) with non-asthmatics as the reference group (n = 115). Assessment of exposure to UFP was based on multi-area land use regression (LUR) models derived from specific-measurement campaigns. We found a strong association of UFP exposure with AOA [8]. The odds ratios were 1.80 (95% CI 1.26, 2.55) for an increase in particle number concentration (PNC) by 5000 particles/cm³, and 1.73 (95% CI 1.27, 2.36) for an increase in lung-deposited surface area (LDSA) by 10 μm²/cm³. PM_2.5 and NO₂ did not show a significant association with AOA.

In a study on cardio- and cerebro-vascular disease (CCVD) in Exposomics, we aimed to investigate the involvement of oxidative stress and inflammation in the causal chain, and to identify intermediate biomarkers that are associated retrospectively with the exposure and prospectively with the disease. We designed a case-control study on CCVD nested in a cohort of 18,982 individuals from the EPIC-Italy study [9]. We measured air pollution, inflammatory biomarkers, and whole-genome DNA methylation in blood collected up to 17 years before the diagnosis. The study sample included all the incident CCVD cases among former- and never-smokers that arose in the cohort during the follow-up. We identified enrichment of altered DNA methylation in “ROS/glutathione/cytotoxic granules” and “cytokine signaling” pathways-related genes, associated with both air pollution and with CCVD risk. Interleukin-17 was associated with higher exposure to NO₂ (P = 0.0004), NO_x (P = 0.0005), and CCVD risk (OR = 1.79; CI 1.04–3.11; P = 0.04 comparing extreme tertiles). Our findings indicate that chronic exposure to air pollution can lead to oxidative stress, which in turn activates inflammatory responses mainly involving the “cytokine signaling” pathway, leading in turn to increased risk of CCVD. Our research suggests that inflammatory proteins and DNA methylation alterations can be detected several years before CCVD diagnosis in blood samples.

Epidemiologic evidence indicates in fact common risk factors, including air pollution exposure, for respiratory and cardiovascular diseases. This may imply the involvement of common perturbed molecular pathways. The goal of a further investigation within Exposomics was to find intermediate metabolites or metabolic pathways in two cohorts (EPIC and SAPALDIA), that could be associated with both air pollutants and health outcomes, thus shedding light on mechanisms and reinforcing causality (Jeong et al., 2018). We measured untargeted metabolomics in two independent case-control studies nested in cohorts on adult-onset asthma (AOA, same as before) and cardio-cerebrovascular diseases (CCVD, same as before). We have identified several metabolic pathways that potentially mediate the effect of air pollution on health outcomes. Among those, perturbation of Linoleate metabolism pathway was associated with all three—air pollution exposure, AOA, and CCVD. Our results suggest that common pathway perturbations may occur as a consequence of chronic exposure to air pollution leading to increased risk for both AOA and CCVD.

Air pollution is a mixture of compounds that can be in a gaseous or particle phase. One of the obvious practical questions of research on health is whether components in a mixture act separately (via different metabolic or molecular pathways) or they impact common pathways. Omics investigations from Exposomics have shown that both cases are realistic depending on the mixture we consider. First, we considered air pollutants from the Oxford Street randomized cross-over trial [10] in which volunteers were exposed to high (Oxford Street, London) or low (Hyde Park) levels of air pollution. Both for RNA (mRNA and miRNA, the latter involved in post-transcriptional gene expression modulation) and for metabolites from mass spectrometry the different components of air pollution gave rise to signals that did not overlap, suggesting that each pollutant follows a different metabolic or molecular pathway to exert its effects [10, 11]. An opposite example comes from water (Piscina study in Exposomics). In this case, we enrolled swimmers (volunteers) to swim in a normal Barcelona pool contaminated by chlorinated or brominated disinfection by-products ([11, 12]). In this example, there seems to be a broad overlap between miRNA and, respectively, MS metabolomic signals across five different disinfection by-products. This is biologically plausible since the five compounds had very similar chemical composition. Therefore, we can expect different health impacts of components of the air pollution mixtures, but similar effects of water contaminated by disinfection by-products (for statistical methods of mixture analysis see [13]).

There is no evidence that common exposures such as air pollution or secondhand smoking show a threshold in their action. For example, in the ESCAPE network, we found that mortality increases below the current “acceptable” limits for particulate set by WHO ([14]). The very establishment of low-dose effects requires very careful epidemiological designs, extremely accurate individual exposure assessment, good characterization of confounders, and very large sample sizes. One potential explanation of low or very low dose effects is the combination of different exposures conferring acquired susceptibility, plus genetic predisposition.

The existence of low-dose effects implies that we should be able to detect molecular changes at those levels of exposure. Support for this is starting to increase through the application of omics in investigations of low-dose exposures. For example, we have found that the levels of exposure to PM₁₀ experienced in utero from four different European areas (Italy, Belgium rural, Greece, Spain) influence cord blood metabolomic signals (unpublished). This can be perceived at low levels of exposure to air pollution, though we also noted that the effects were stronger in the areas with higher levels of pollution than in rural areas with lower levels.

In Exposomics, we had three categories of studies:

Table 1 shows the main results we have found so far in the published papers and some still unpublished findings. The purpose of this paragraph is to address consistency of the findings across omic platforms and across study designs.

There are several lessons that can be drawn. First, DNA methylation results do not seem to replicate across studies. This is not entirely surprising because—with exceptions—only few genes have been found to be consistently hypo- or hyper-methylated in relation to strong carcinogens such as tobacco smoke (notably AHRR, [19]). As Table 1 shows, signals that replicate across studies in Exposomics are IL-8 in three studies (including Linoleate that is involved in IL-8 activation), carnitine shuttle in 2 studies, and Lyso-PC in two studies (Table 1). As indicated above, these signals suggest that pathways commonly perturbed after exposure to air pollution involve inflammation and oxidative stress.

A research program within the exposome community is currently trying to reconcile the pathways that have been empirically identified with “hallmarks” of disease or key characteristics of toxicants. For example, López-Otín et al. [20] proposed nine ‘hallmarks of aging,’ including primary hallmarks (genome instability, telomere attrition, epigenetic alterations, loss of proteostasis), antagonistic hallmarks (deregulated nutrient sensing, mitochondrial dysfunction, and cellular senescence), and integrative hallmarks (stem cell exhaustion and altered intercellular communication). Similarly, hallmarks of cancer have been proposed, partly overlapping with those for aging [21]. Others have suggested to classify the key characteristics of carcinogens into 10 categories as described in Table 2 [22].

When in Exposomics (Table 1), we looked for signals corresponding to hallmarks of cancer or key characteristics of carcinogens, we detected significant perturbation of the following categories: miRNA expression; cytokine signaling (inflammation); reactive oxygen species (oxidative stress); glycosphingolipids (apoptosis, cell growth, senescence, cell cycle control); NOTCH1 (cell cycle control); carnitine shuttle; and acyl-carnitines (oxidative stress). Inflammation, oxidative stress, and miRNA are also involved in respiratory and cardiovascular diseases. Oxidative stress pathways are compatible with our observations on the oxidative potential of air pollutants as measured in filters [23].

Thus, though still very preliminary, this approach looks like a promising avenue to make sense of the multitude of signals emerging from omic investigations. Looking at pathways and pathway perturbation is a way to confer biological plausibility to observations and to increase statistical power.

Exposomics is one of the first investigations on the exposome in environmental research. Some results are promising, particularly in the investigation of mixtures, and in the identification of biological pathways perturbed by air pollution that lead to diseases in different organs. Much work still needs to be done for a more thoughtful interpretation of such pathways, for example, using the paradigm of “hallmarks” of disease as I propose here.