How exposomic tools complement and enrich genomic research

Summary

Because genetics and the environment interact to drive gene expression, we propose that exposomics must now be incorporated into the multi-omics paradigm to complete the overall biological pathway. Exposomics’ groundbreaking tools and life-course framework holistically characterize non-genetic (environment) components of chronic diseases and integrate with multi-omics. This work brings forward the importance of the human exposome as a major driver of gene/protein expression across the life course. Exposome features are noteworthy for multi-omics as they (1) show where and when biodynamic trajectories of gene x environment interactions meet; (2) move beyond single-environmental-factor-centric views; (3) integrate exposomic measurements during and outside of critical windows of susceptibility; (4) provide agnostic discovery and hypothesis-generating studies; and (5) are biodynamic over time. Upon applying these unique features of the human exposome, future human studies are anticipated to revolutionize the integration of genetics and environmental health sciences.

Graphical abstract

Makris et al. present the case of human exposomic tools that are warranted to complement and enrich genomic research. This is supported by describing the four novel features of the human exposome concept, which would help humanity to advance its knowledge of chronic disease processes.

Background

The Human Genome Project and the missing heritability

Since the latter half of the 20^th century, the global burden of disease morbidity and mortality has been mainly driven by non-communicable diseases (NCDs), which account for >60% of healthcare costs.^1,2 Early in the 21^st century, groundbreaking research methods in genomics (Human Genome Project [HGP]) allowed the assessment of genetic risk factors for NCDs.³ Following the launch of the HGP, a slew of research was conducted to understand the role of genomics in chronic disease processes, with limited results for understanding how complex diseases arise. Specific disciplines cataloged these genetic risk factors. For example, the Cancer Genome Atlas project mapped cancer-associated genetic variants, gene expression, and DNA methylation using advanced bioinformatics algorithms.⁴ However, despite these technological advances, the HGP has produced modest societal benefits to improving population health, and the burden of NCDs has actually increased. While the HGP established the reference code for the DNA sequence of every known gene, only ∼2% of the human genome is actually used for protein coding, and few variants associated with the onset of the most common NCDs change the amino acid sequence of proteins.⁵ In parallel, non-genetic (environment, lifestyle, and metabolic) risk factors appear as major determinants of disease. During the 1990–2015 period, public health improvements driven by the human exposome accounted for the largest part of mortality improvement overall when compared with pharmaceuticals.⁶ In the US, the biggest advances in life expectancy and quality of life from 1900 to 1999 were driven by environmental factors in the human exposome.⁷

Causal associations for specific genetic loci have yielded few actionable findings to target new therapies, in part because genetic associations provide genomic locations of interest rather than identifying key genes driving the associations, and they reflect many variants with small overall effects instead of a small number of variants with larger effects.⁸ They may also reflect unmeasured gene x environment (GxE) interactions. Genetic loci with a small contribution to measured phenotypic variance can have a substantial effect if pharmacologically altered to address environmental factors, such as diet. For example, the 3-hydroxy-3-methylglutaryl (HMG) coenzyme A (CoA) reductase gene (HMGCR) has a modest effect size in association with coronary artery disease in genome-wide association study (GWAS) analysis,⁹ but major effects on disease risk result from lipid-reducing medications targeting this gene. Recent developments make use of polygenic risk scores that summarize the contribution of multiple genetic variants in predicting predisposition to a disease.¹⁰ However, the polygenic nature of heritability makes it difficult to decode the full biological process. Familial aggregation of disease is not solely dependent on genetics but is also controlled by both measured and unmeasured environmental factors that correlate across generations within families. Children not only inherit genes; they also inherit the wealth or poverty of their parents, as well as neighborhoods, educational attainment, diet, and the environment of their shared home.

Several examples highlight the importance of GxE interactions in disease prevention and control. Environmental effects have been usually treated in quantitative genetics research as randomly generated noise drawn from an underlying distribution of stochastic input, although they explain more than half of interindividual variation in disease risk.¹¹ Common traits, such as height, may have a high heritability, but the genetic component is composed of numerous genes with small contributions to the overall phenotype. Further, the environment still plays a role in height, as evidenced at the population level. The Dutch were the shortest European population at the start of the 20^th century but became the tallest European population over just 2–3 generations, a change far too rapid to ascribe to genetics alone. Only environmental factors such as wealth, hygiene, and diet can explain such a rapid increase in average height and likely operate in the context of GxE interactions. The multifactorial network of interactions between genetic and environmental components works at various levels of network hierarchy and at the biological level, i.e., in cells, tissues, organs, the body, and their combination.¹²

We propose that a greater understanding of the underlying causes of NCDs could be found using exposomics to study environmental (non-genetic) interactions with the genome using a life-course perspective. A recent definition proposed by the European Parliament study panel for the future of science and technology and adopted from the International Human Exposome Network defines human exposomics as “a field that studies the compilation of all physical, chemical, biological, and psychosocial factors that impact biological systems by integrating data from a variety of interdisciplinary methodologies and streams, to enable a discovery-based analysis of environmental influences on human health.”¹³ In recognition of the need to expand scientific horizons, we are now entering a new “multi-omics” era that integrates gene expression and protein data as well as epigenomics and environmental stressors.^14,15 This multi-omics era is holistic and considers that gene expression/translation, processing, and the relative levels of gene products are connected and dynamic and cannot be predicted by DNA sequence alone. Biological variability and complexity are driven by interactions with the too-often unmeasured environmental components of disease phenotype, i.e., the human exposome.^{16,17,18,19,20,21,22,23} The substantial gap between genetic variance attributed to identified genetic loci and total estimated genetic variance has been named the “missing heritability,” signaling an assumption that unmeasured genomic features in isolation explained the discrepancies rather than the far more likely scenario that GxE interactions (both biological and mathematical) provide major contributions to our limited understanding of complex diseases.

This work aims to (1) bring forward the importance of the human exposome in moving past the gene-centric view of genomics research and (2) highlight the special features of the human exposome concept and its tools that make it unique against the conventional practice applied so far in biomedical and environmental health sciences.

Bringing forward the value of the human exposome concept

Exposomics point #1: Where and when the biodynamic trajectories of GxE meet

In the 21^st century, the “nature versus nurture” dogma of the 20^th century that assumes environment and genes act independently can finally be transformed into the biologically fundamental question of how, where, and when the environment, genes (GxE), and their products interact in disease pathogenesis; these factors act in a developmental context, which is also influenced by stochastic factors (Figure 1). Exposome research is similar to genomics, as it can entail non-hypothesis-driven (i.e., discovery) approaches. Exposomics can help discover GxE interactions by offering a framework that can be incorporated into genomics and systems biology studies.²⁴ For example, the relative levels of air pollution components and their interactions with receptors will occur within and between biological layers of the lungs and circulatory system as they stochastically generate inflammation over time, creating biological responses in a series of feedback loops.²⁵ The concept of the “functional” exposome considers both the parent compounds from the environment and their metabolites that arise from the host and its microbes.^26,27 Such exposures are not limited to chemicals but can include hormonal changes from chronic stress, for example. This concept is not new, as it has been part of early definitions of the human exposome.^17,18,23 In fact, the integration of the internal environment with the external environment, including the socio-ecological and cultural environments of individuals, was discussed as part of the human exposome domains’ constant flux.^17,18

Figure 1.

Time-resolved monitoring of individual and joint effects of the exposome (exposures) and the genome as they interact in the disease process across multiple -omics platforms overlaid as biological layers of information that prospectively evolve over time

The concept of GxE interactions has existed for decades but has been understudied, primarily due to the lack of concerted, collaborative efforts between scientists working in the fields of genetics and environmental health sciences. Suggestive evidence from early GxE studies hints at promising results, however.^28,29,30 For example, genetic variants at the 10q24.32 locus near AS3MT were associated with inefficient arsenic metabolism and subsequent toxic inorganic arsenic exposure for affected populations in Bangladesh.³¹ In another study, variance quantitative trait loci (vQTLs) were found to be enriched for GxE effects in large cohort studies (UK Biobank) by associating select genetic variants with phenotypic variability (e.g., birth weight, height, etc.).³² Using data from the UK Biobank, Kim et al.³³ showed that cigarette smoking exposure interacted with a polygenic risk score for lung function to influence the risk for airflow obstruction. Zhu et al.³⁴ screened for interactions across the genome using an approach similar to the Mendelian randomization framework and identified/confirmed five loci that interacted with either cigarette smoking or alcohol consumption for serum lipid levels. Because all proteins arise from genes and some proteins (enzymes) use substrates that at some point entered the body from our environment, one could argue that GxE interactions are mandatory biologically. Epidemiologic research has generally ignored GxE interactions in favor of simpler main effect models in which one of the two components of the interaction remains unmeasured. This approach ignores biology.

Although GxE effects have been readily detected in model organisms, it has been challenging to reliably identify GxE effects in human populations. Potential explanations for this discordance could relate to small GxE effects in human populations, measurement errors in assessing environmental factors, and contributions from higher-order interactions. Creative methodological approaches to detect GxE will be required, such as using environmental exposure as the outcome and assessing for gene x disease interactions in a “reverse” test of linear regression.³⁵

Multi-omics molecular phenotyping using nontargeted chemical assays offers new avenues to explore GxE interactions, particularly in longitudinal settings, such as in pregnancy-birth cohort studies.^14,15 Epigenetic DNA marks regulate gene expression and gene products, including microRNA (miRNA), long non-coding RNA (lncRNA), and downstream proteomics and metabolomics signatures. Integrating the structural information and the changes between these -omics signatures could greatly facilitate the assessment of GxE interactions in pregnancy, childhood, or both windows of susceptibility using joint dimensionality reduction algorithms.³⁶

Exposomics point #2: Moving beyond the single-gene- and single-environmental-factor-centric views of disease processes

The HGP created a gene-centric view where distinct human traits could be related to a specific gene(s) by analyzing differences in single-nucleotide polymorphisms (SNPs). The grand challenge of sequencing three billion base pairs hid the fact that this is actually a reductionist approach, as it assumed these measures would be the main disease determinants. In parallel, the field of environmental health proved not to be immune to reductionism, as its epidemiological and toxicological analyses for decades relied on a “one-exposure-one-outcome at a time” approach that ignores the human exposome. The recent growth of “mixtures” studies of multiple biomarkers of exposure/effect shows that a sea change may be coming. An exposomics approach would also include the assessment of multiple exposures and consider them as mixtures in their effects on disease processes and would allow for the delineation of the main and interaction effects of both chemical and non-chemical stressors (infectious agents, stress, racism, noise pollution, etc.) on the endogenous response.²⁰ We provide some examples here.

Exposome-wide association studies

This is a data-driven approach that can be applied to different study designs (case-control, cohort, etc.).^26,27 In analogy to the GWAS approach, the exposome-wide association study (ExWAS) evaluates multiple associations between environmental agents and an outcome of interest and rank orders them to carry through to a replication stage. Because of the large number of comparisons, replication in an independent study population is critical, as well as the application of rigorous false discovery rate testing. Multi-stage sampling designs maximize variability in exposome data and maximize power to detect effects while also addressing false positives through replication and validation. It should be noted that biological function testing, either in vitro or in vivo, is also likely needed to complete a rigorous multi-stage design. However, with the rise of predictive toxicology and artificial intelligence, it may be possible to address future functional testing without experiments. Finally, given the exposome is time varying, prospective exposure data, or nested case control study designs within longitudinal settings, are often preferred over cross-sectional or case-control designs.

Multiple exposures

The multiple statistical testing challenges for large-scale environmental studies will likely be even greater than for genome-wide genetic studies, and the effect sizes of environmental factors are likely often quite small. New statistical algorithms can simultaneously assess multiple exposures for their joint/cumulative health effects.^37,38,39,40 The testing of mixtures can be applied to ExWAS discovery, although, to our knowledge, there are not yet examples of doing so. These statistical algorithms are gaining momentum and include the Bayesian kernel machine regression (BKMR), random forest, the least absolute shrinkage and selection operator (LASSO), weighted quantile sum regression, and quantile G computation.^37,38,39,40 These algorithms allow for the identification of important multiple exposures that are simultaneously examined. Several of these algorithms allow for the estimation of joint health effects and relative contributions of each mixture component, such as the weighted quantile sum regression and the BKMR. Data are split into training, validation, and testing subsets where multiple variable selection steps (e.g., ExWAS and LASSO) are employed to identify significant combined exposure factors associated with an outcome.

Polyexposure and polygenic risk scores

The assessment of individual genomic variants in GWASs by comparing groups with a certain disease to a group without the disease can be used to calculate a polygenic risk score, which is an estimate of a person’s relative risk to that of others with a different set of genetic variants. This is a time-invarying risk estimate, which does not account for age or ethnic differences or other environmental/exposomic factors that would provide us with a better disease risk estimate. Polyexposure risk scores (PXSs) are an analogous concept, still in an early stage of development, where a handful of well-known non-genetic risk factors (e.g., alcohol and smoking) with or without a larger putative group of environmental stressors may be included in the calculation of PXS for a specific disease outcome.⁴¹

Another important challenge to exposomics relates to the need to understand the biochemical pathway dynamics that impact the manifestations of environmental risk factors on well-being and disease processes.²² Methods to reconstruct the past exposome are critically needed. For example, endogenous response measures collected simultaneously with exogenous chemical measures do not solely arise from those chemical exposures that can be measured in a biosample collected cross-sectionally. They are far more likely to be due to past exposures, many of which are not reflected in a cross-sectional blood or urine specimen. This is further complicated by the fact that co-occurring social, psychological, mental, and physical exposures may influence the downstream biological response.

Exposomics point #3: Integrating different levels of exposure measurement at critical windows of susceptibility

The human exposome provides the opportunity to characterize the black box lying between the network of environmental stressors that comprises the external exposome and the endogenous response, which comprises the internal exposome, as it interacts with the genome.¹⁷ For example, air pollution is an external exposome component and has no specific internal biomarker of exposure. This necessitates the characterization of both external and internal exposomic domains and their groups of exposomic components and variables. The exposomic biological response may be characterized by looking at combinations of select targeted biomarkers of biological effect in biospecimens, such as urine, blood, hair, teeth, saliva, tissues, or cells, which often include genes and gene products, SNPs, epigenetic marks (such as methylated DNA bases), RNA transcripts, post-transcriptional modifications of RNA bases (epitranscriptomics), proteins (including post-translational modifications), and metabolites and, specifically, how exposome biomarkers and exposure models relate to them (Figure 2); the genetic and environmental determinants (as well as their interactions) of these targeted biomarkers will need to be carefully assessed. The most rigorous approach is to have prospective measures of exposure that track the past exposome to relate it with biomarkers of biological response. Nonetheless, the key concept here is the dynamic relationship between two or more of the abovementioned biological and environmental entities while accounting for their temporal or spatial continuity and evolution (Figure 2). These concepts are especially important during critical windows of susceptibility where the spatiotemporal dimensions of external environmental stimuli may affect such relationships, signaling whether to switch on or off key biological processes (Figure 2). Perturbation between and within the entities of multi-omics systems and their networks over time and space shall dictate the disease process continuum evolving either to resilience (homeostasis) or to a differentiated disease risk because of disrupted homeostasis. Changes in systems-based networks at critical windows of susceptibility between repeated longitudinal measures of external exposomic features and their associated endogenous response profile are likely the key to understanding how the environment can maintain or disrupt well-synchronized biological systems.⁴²

Figure 2.

GxE interactions using the human exposome concept in a life-course approach, i.e., integrating information on critical time windows of susceptibility

Presumably, GxE interactions shall be strongest when environmental exposures (x axis) occur during a critical time window of susceptibility (scenario A: GxE interaction present at life stage A). GxE interactions may even be null if the exact same combination of environmental exposures occurs outside such a critical time window (scenario B: GxE interaction present at life stage B). The only difference between scenarios A and B is actually the dose-response curve, as we have the exact same exposure(s) (magnitude/variance) at the exact same dose. The biological interfaces that shape the temporal profiling and evolution of the disease process as a result of exposomics temporal dynamics shall be well characterized and integrated with multi-omics datasets that comprise different layers of biological information combining the entities of environmental exposures with those of the metabolome, proteome, transcriptome, and genome as they dynamically evolve in time and space. Perturbations between and within the entities of multi-omics systems and their networks over time and space shall dictate the disease process continuum evolving either to resilience (homeostasis) or to a differentiated disease risk because of disrupted homeostasis.

The temporal dynamics linking external exposome components and the internal exposome profiling in time, dictating downstream sequelae of homeostatic biological events, is what underlies the relationship between exposomics and disease process. The homeostatic theory has historically advocated for the constancy of biological processes and functions, systematically ignoring the dynamical character of homeostasis. The human exposome fills this gap by monitoring the time-resolved behaviors of such structured dynamic systems. From single-protein molecules whose internal oscillatory dynamics stem from the mutual relationship among amino acid residues solicited by thermal noise⁴³ to ecological systems whose relative abundance of species oscillation is driven by their relative position in the food web,⁴⁴ many of the system components have dynamic changes, such as oscillation. Temporally synchronized oscillations may be subject to alterations due to external environmental perturbations, altering the harmonic character of the oscillatory behavior and disrupting their dynamic character.⁴⁵ For example, the time-resolved monitoring of early life metabolic dynamics of heavy-metal exposures predicted the emergence of autism spectrum disorder cases using the recurrence quantification analysis algorithm and its metrics.^45,46 As such, the study of temporally dynamic networks of systems could facilitate the detection of early warning signals of the pre-disease state.⁴⁷

Life stage critical windows of susceptibility

The exposomic changes during select critical windows of susceptibility are likely key drivers of health and disease trajectories. During the first 3 months of life, a dynamic yet stereotypic immune system development process operates for newborns that is driven by microbial interactions and environmental stimuli.⁴⁸ The Barker hypothesis, which evolved into the “developmental origins of health and disease,” was the first to prospectively report that fetal nutritional status was a key predictor of coronary heart disease and diabetes 50–60 years after birth. Measuring life-stage-specific exposomes and linking them with gene expression and genomic variation patterns are critical needs. For practical reasons, the human exposome typically studies specific windows of susceptibility rather than the whole life course, as it would take decades of follow-up to do the latter. While difficult to achieve, the integration of temporal processes and observations from different life stages would provide clues regarding the ways that exposomics shapes health.

Integrating multiple levels of exposomics measurements and analysis

The human exposome offers the inherent flexibility to integrate individual-level data on location (e.g., address or zip code), psychosocial, or sociodemographic data to small-area, neighborhood-, and community-level data, including regional health data, environmental monitoring, economic indicators, and climate data. As an example, we now approach cities as complex living systems of networks whose resilience to shocks and pressures is subject to a suite of local (e.g., spatial and population characteristics, infrastructure, local policies, and small-area effects), national, regional, and international trends. The urban exposome framework has emerged as a holistic study paradigm of urban health that captures the totality and spatiotemporal integration of exposomics and its contextual nature.^19,49,50 The urban exposome allows the integration of different methodologies in describing how urban health evolves over time and space, together with the integrated systems network of individual-level health trajectories. In a post-pandemic era, urban exposomics shall be comprehensively characterized in urban space and time if we are to better understand infectious disease transmission dynamics and to better identify high-risk groups for infection.²⁰

Exposomics point #4: Provides agnostic discovery mode and hypothesis-generating studies

The exposome concept moves beyond hypothesis testing and instead looks for the discovery of new health and environmental data. The agnostic nature offered by exposomic approaches such as high-resolution mass spectrometry (HRMS) to enable the detection of hundreds to thousands of biological or environmental signals in a single biospecimen facilitates biomarker discovery applications and hypothesis-generating studies. The exposomic trajectories can be integrated dynamically with multi-omics datasets that comprise different layers of biological information, combining the entities of environmental exposures with those of the metabolome, proteome, transcriptome, and genome. These advanced precision health technologies facilitate the generation of new hypotheses that can be rigorously tested in subsequent human or animal studies. The generation of such enriched yet high-dimensional datasets requires the implementation of advanced computational tools and machine learning algorithms.^51,52 The application of exposomics tools, such as HRMS and similar omics platforms, requires investments in data resources such as libraries, HMDB, PubChem, USEPA Comp Tox, NIST, etc.^51,52 Access to computational mass spectrometry workflows coupled with open mass spectral libraries and new identifications must be developed so that more epidemiological studies can employ exposomic tools. Tools must also be developed to integrate non-chemical stressors (climate, psychosocial stress, etc.) into ExWAS studies, as well.

Discussion

Applications of human exposomics

Albeit a relatively young concept, human exposomics has the potential to advance our understanding of chronic disease pathobiology beyond the HGP achievements alone. To this extent, exposomics is particularly relevant to precision public health initiatives that call for the integration of genetics, lifestyle, and environment toward disease prevention and control at the population level.⁴⁰ Pathogen sequencing in wastewater may provide insights into viral trajectory dynamics at the population level.⁴⁰ Other relevant applications are direct-to-consumer testing services for polygenic risk scores of multifactorial complex diseases or the integration of genomic data into electronic health records. Some of the most promising applications relate to the integration of environmental and social determinants of disease into GxE studies or DNA methylation studies of prevention or prediction of chronic disease risk; groups of lower socioeconomic position may be at higher risk of cardiovascular disease due to higher exposures to pro-inflammatory agents in the environment.^37,53

Differences in disease susceptibility, progression, and severity may be characterized and linked with environmental exposures throughout the lifetime or in critical windows of susceptibility; this might be highly relevant for precision health applications in cancer treatment and diagnosis.⁵⁴ As an example, depending on the body burden of various xenobiotics, ranging from per- and polyfluoroalkyl substances (PFASs) to air pollutants and pesticides or metals, cancer treatment efficacy may be substantially perturbed.⁵⁴ Human exposomics to various environmental pollutants may be key in disease prevention, control, and treatment patterns by ethnic groups or other sociodemographic parameters.⁵⁵ Epigenetics would advance the field by assigning molecular pathways to unexplained interindividual variation of within-generation phenotypic diversity observed in human studies.¹¹

Challenges of human exposomics data collection and processing

Several challenges to human exposomics data collection and processing exist. Obviously, this goes beyond the typical effort devoted to genomics data collection, but nevertheless, it is essential to better understand phenotypic diversity. Data collection for exposomics covering all domains of the human exposome methodological framework¹⁷ may be particularly troublesome because of the high interindividual variance in exposomic variables as they dynamically evolve in space and time. As such, a combination of environmental measurements is warranted together with measurements of biomarkers of exposure to environmental stressors and their downstream multi-omics profiling. Biomarkers of exposure to environmental stressors should be obtained from invasive (blood) or non-invasive (saliva, urine, hair, nails, etc.) matrices in a spatiotemporal fashion from reference and understudied populations covering various study designs (cohort, case-control, panel studies, etc.) and a wide span of ethnic, age, and socioeconomic position groups.

Policy and interventions

Both the European Commission and the US National Institutes of Health have highlighted the staggering burden of chronic diseases on their societies, calling for action.^56,57 For example, health technology assessments (HTAs) are designed to evaluate health technologies, as they provide a comprehensive analysis that considers multiple dimensions, such as technology effectiveness, safety, cost effectiveness, and organizational impact. Exposomics tools, such as multi-omics, biomarkers, or wearable sensors, would be invaluable toward developing, testing, and validating non-pharmacological and other health technologies for the clinical sector or for public health practice. Advances in data science are anticipated to support and facilitate the growth of exposomics.

Conclusion

This work highlights the most important features of human exposomics toward advancing our knowledge of chronic disease processes. A multi-omics view of biomedical science using exposomics has emerged, and while initially gene-centric, it has evolved to recognize that DNA sequence alone cannot explain chronic disease processes. Because the environment is a prime driver of gene expression, there are enormous opportunities to integrate the power of exposomics into the policy frameworks of precision health and precision public health. The time is right for biomedical sciences to consider incorporating the dynamic profiles of exposomic domains and their components into clinical practice and precision public health using data-driven methodologies. The human exposome is inherently dynamic in time and space, and exposomics tools, together with advanced biostatistical time-varying algorithms and biomedical technologies, will find great use in a new multi-omics era that considers genomics as the study of the drivers of gene expression and not just DNA sequence variation. Without this critical piece of the complex disease puzzle, a comprehensive understanding of human health will never be realized.

Acknowledgments

We would like to acknowledge funding support by the EPHOR and IHEN exposome projects. We would like to thank Dr. Georgia Soursou from the Cyprus International Institute for Environmental and Public Health for preparing Figure 1. We would also like to acknowledge funding from US National Institutes of Health P30ES023515, U2CES026561, UL1TR004419, and R01ES013744.

Author contributions

K.C.M. conceived and designed the study, contributed materials/analysis tools, and wrote the paper. R.O.W., E.K.S., and A.B. provided expert advice and co-wrote the paper.

Declaration of interests

In the past 3 years, E.K.S. has received institutional grant support from Bayer and Northpond Laboratories.