Gene-environment interactions within a precision environmental health framework

Summary

Understanding the complex interplay of genetic and environmental factors in disease etiology and the role of gene-environment interactions (GEIs) across human development stages is important. We review the state of GEI research, including challenges in measuring environmental factors and advantages of GEI analysis in understanding disease mechanisms. We discuss the evolution of GEI studies from candidate gene-environment studies to genome-wide interaction studies (GWISs) and the role of multi-omics in mediating GEI effects. We review advancements in GEI analysis methods and the importance of large-scale datasets. We also address the translation of GEI findings into precision environmental health (PEH), showcasing real-world applications in healthcare and disease prevention. Additionally, we highlight societal considerations in GEI research, including environmental justice, the return of results to participants, and data privacy. Overall, we underscore the significance of GEI for disease prediction and prevention and advocate for integrating the exposome into PEH omics studies.

Graphical abstract

Motsinger-Reif et al. review the state of research investigating interactions between genetics and environmental factors, highlighting the translation of findings into precision environmental health. They discuss the need for new detection techniques due to the use of large-scale datasets comprising data from multiple sources; outline challenges in the field related to measuring environmental factors, environmental justice, and data privacy; and deliberate on the advantages of gene-environment interactions for understanding disease mechanisms.

Introduction

Multiple, often interacting, genetic and environmental risk factors contribute to the development of diseases, with an estimated 50%–80% of overall disease etiology based on environmental factors.¹ Gene-environment interactions (GEIs) vary based on a person’s genetic makeup and the nature of their environmental exposures. The involvement of GEI in disease etiology also varies across time and development stages, with significantly different effects on gene regulation over a lifetime, especially prenatally and during early childhood development.¹ An increasing body of evidence demonstrates the significance of GEI, with a growing number of replicated interactions, including BRCA-1 associated protein-1 (BAP1) mutations and asbestos exposure for mesothelioma,² chromodomain helicase DNA-binding protein 8 (CHD8) and pesticide exposure for autism spectrum disorder,³ fat mass and obesity-associated gene (FTO) and physical activity for obesity,⁴ and dopamine receptor D4 (DRD4) and parenting style for attention-deficit/hyperactivity disorder.⁵ In the context of precision health, several noteworthy GEIs that are relevant in various aspects of healthcare have been identified. These include the impact of gene polymorphisms in metabolic enzymes on health outcomes,⁶ the relationship between genetic factors and therapeutic response in drug efficacy,⁷ the influence of GEI on drug-drug interactions,⁸ and the association of GEI and drug toxicity.⁹

As reviewed in Baccarelli et al.,¹⁰ precision environmental health (PEH) employs genetic, epigenetic, environmental, and system-level data to elucidate the underlying environmental causes of disease, identify biomarkers of exposure, and develop new prevention and intervention strategies. Incorporating personalized information on genetic susceptibilities and environmental exposures enables improvements in the accuracy and efficacy of risk assessment, disease prevention, and interventions, with the ultimate goal of reducing the burden of environmentally induced diseases and health disparities. Figure 1 outlines the role of omic and exposure data in PEH. Advancements in PEH are anticipated to lead to more targeted and effective public health initiatives, healthcare strategies, and environmental policies that will foster healthier populations and environments.

Figure 1.

PEH

PEH uses available biological and exposure data to assess individual risk and tailor interventions. These data consist of omic data (blue box); data on internal exposures from diet, lifestyle factors, stress, and pharmacological agents; as well as data on prototypical external exposures from home, air, water, and social determinates of health (yellow boxes). Rigorous statistical machine-learning methods are used to winnow down the features that contribute to explaining disease likelihood (green boxes). This in turn enables differential treatment based on individual omic and exposure differences.

GEIs are foundational to PEH because they highlight the intricate interplay between genetic factors and environmental exposures in determining human health and disease risk. Individual reactions to environmental factors vary: what is harmless to one person may be detrimental to another due to genetic differences. Understanding GEI helps accurately predict individual susceptibilities, thus improving risk assessments. Unveiling the underlying biological mechanisms of diseases paves the way for health recommendations and effective interventions tailored to the individual.

Additionally, scrutinizing GEI engenders a holistic view of health determinants that does not oversimplify outcomes as the result of genetics or environmental exposures alone. Taking a comprehensive view aids in developing treatment approaches and prevention strategies, especially for those with genetic vulnerabilities to environmental hazards. This is a vital step in addressing health disparities as some populations have increased susceptibility due to the disproportionate number of exposures they experience and may have different susceptibilities due to genetics. Understanding GEI comprehensively ensures a more inclusive and personalized approach to environmental health.

In this manuscript, we review the current state of the field regarding GEI. We discuss the opportunities and obstacles involved in measuring environmental data; the advantages and challenges associated with interrogating interactions; and the interplay of the environment and genomics, transcriptomics, proteomics, lipidomics, metabolomics, and epigenetics. We outline approaches to GEI analysis and recent advancements, including emerging methods and sources to incorporate biological knowledge into analyses. We review known interactions and discuss how results are being translated into clinical practice. Further, we discuss the social impacts of GEI research regarding environmental justice, the return of results to cohort participants, and concerns surrounding data privacy in light of the increasing use and availability of geospatial data. Finally, we place the reviewed material into a PEH framework.

Gene-environment interactions

For this manuscript, we use a broad, inclusive definition of GEI and acknowledge that statistical definition of GEI is an area of active debate. A GEI occurs when exposure to an environmental factor affects the risk of developing a disease due to differences in an individual’s genetic makeup. Conversely, the effect of a genetic variant on a phenotype can also be influenced by exposures. The statistical definition determines the existence or absence of a GEI based on either an additive or multiplicative scale of measurement within linear modeling frameworks. There are several considerations when determining a suitable measurement scale for GEI, such as the primary goal of the inquiry (e.g., identifying causes of disease, predicting public health trends) and the proposed pathophysiologic model. Figure 2 demonstrates a general framework for GEI.

Figure 2.

Gene-environment contributions to trait variance

Given a genetic locus of interest (x axis) and an exposure of interest (exposed = blue), the ways in which they can be related to a trait of interest (y axis) are illustrated.

(A) Neither the genetic region of interest nor the exposure explain variance in the trait of interest.

(B) The genotype contributes to trait variance while the exposure does not.

(C) The exposure contributes to trait variance while the genotype does not.

(D) In the absence of exposure information, the genotype does not contribute to trait variance, but there is a gene-by-exposure interaction.

(E) Both the genotype and exposure contribute to trait variance as well as their interaction.

(F) Both the genotype and exposure contribute to trait variance but without a gene-environment interaction.

In traditional statistical frameworks, there is intricate discussion on GEI, with some statistical geneticists expressing uncertainty about the importance of these interactions. One issue is the seemingly minor impacts of GEI compared to the impacts of genetics or environmental exposures alone. Recent research has attempted to quantify interactions, disputing the idea that the impact of GEI on phenotypic variation is insignificant.¹¹ Ottman discusses traditional statistical frameworks in detail.¹²

Understanding the significance of GEI requires a broad view that emphasizes the intricate nature of biological systems and considers the combined effects of genetic and environmental influences on trait expression and disease susceptibility, even in the absence of statistically significant GEI. Further, it is essential to recognize that the occurrence of GEI does not necessarily indicate molecular interactions inside cells. While it is commonly held that a GEI is significant only if it has a direct impact on molecular mechanisms, GEI may influence phenotypic or disease outcomes without direct molecular interactions. For instance, expression quantitative trait loci (eQTLs) have similar effects across many contexts, but phenotypic results may differ due to threshold responses that are more prominent in a particular environment.

In a recent review, Westerman and Sofer described various relationships between genetics, environmental exposures, and outcome variables that could yield statistically significant GEI.¹³ The authors discuss intracellular molecular interactions but also highlight several statistical arrangements of genetics, environmental exposures, and outcomes that can produce GEI. These arrangements depart from the often default assumption that GEI must involve a molecular interaction and are generally less intuitive and obvious.

An underlying cause of potentially spurious detectable GEI is a non-linear relationship between a subset of modeled variables (environmental exposure and outcome or a mediator and outcome). Due to this non-linear relationship, an environmental exposure and outcome may have a different slope for different genotypes, indicating a GEI. For example, there is a non-linear relationship between body mass index (BMI) and low-density lipoprotein (LDL) cholesterol, so it is possible that a putative GEI for LDL between genotype and BMI is a product of this non-linearity and not an actual GEI. Heterogeneous variance can also be an underlying cause of detectable GEI.¹⁴ If there is additional variance in an outcome across genotypes (e.g., a variance QTL [vQTL]) or bias in the measurement error of an outcome that increases variance across genotypes (e.g., skin pigmentation and pulse-oximeter measurements from wearable technology),¹⁵ the measured effect of a constant environmental exposure could change across genotypes, which would be measured as a GEI. These underlying causes range from true instances of GEI, such as non-linearity or vQTLs, to instances due to measurement error.¹³

Environmental exposures affect health and illness not just at the cellular level but also via behavioral and systemic consequences. Toxic exposures can affect cellular function, while lifestyle factors can modify phenotypes by affecting brain networks or hormone levels. A diverse range of environmental factors affect biological systems from the cellular to organism level and interact with genetics in intricate ways.

Exposure measurement

As the scale of genotyping rapidly increases, opportunities to collect high-dimensional environmental data are expanding. The exposome describes the totality of environmental exposures over the course of an individual’s life, including external (e.g., pollution, radiation, social determinants of health) and internal (e.g., microbiome, metabolism, oxidative stress) factors.¹⁶ The individual factors that constitute the exposome are measured in numerous ways.

Mass spectrometry (MS) assays quantify environmental exposures in biological samples such as blood, urine, and tissues.¹⁷ Untargeted MS represents a significant advancement in exposome research and provides a robust method for the high-throughput identification and quantification of a vast array of small molecules in biological and environmental samples.¹⁸ Untargeted MS can comprehensively screen for chemical exposures without prior knowledge of the analytes present. It can detect environmental pollutants (e.g., air and water contaminants, soil pollutants), dietary exposures (e.g., food additives and natural toxins), lifestyle factors (e.g., components of tobacco smoke and alcohol metabolites), endogenous metabolites indicative of internal biological processes, and microbial metabolites from the gut microbiota.¹⁹ Untargeted MS offers extensive chemical coverage and captures a broad range of organic compounds, metals, and metabolites, including volatile and semi-volatile organic compounds; non-volatile organic compounds such as lipids, proteins, peptides, and nucleotides; and even inorganic ions and metals under certain conditions.²⁰ The technique’s high-throughput capability enables the detection and quantification of hundreds to several thousand unique analytes in a single run, depending on the sample matrix, technology employed, and analytical conditions. However, the sheer volume and complexity of data generated pose significant challenges in data analysis and compound identification, necessitating sophisticated bioinformatics tools and comprehensive reference databases for accurate analysis.²¹ The sensitivity and specificity of detection also vary among classes of chemicals and are influenced by factors such as a compound’s ionization efficiency and the presence of matrix effects.²¹

Geographic information systems (GISs) are used in epidemiology studies for environmental exposure assessment that integrates data from diverse sources.²² GIS linkages enable mapping that provides insights into the spatial and temporal distribution of exposures and facilitates the assessment of health risks associated with environmental factors. Analysis of air pollution is a primary application of GIS.²³ Particulate matter (PM2.5 and PM10) concentrations can be modeled using satellite data, data from air-quality-monitoring stations, and land-use information. The distribution of gases such as nitrogen dioxide (NO₂), sulfur dioxide (SO₂), and ozone (O₃) can be assessed using data on traffic, industrial activities, and meteorological conditions. Water-quality assessment is another prominent application of GIS in epidemiology studies.²⁴ GIS can assess the risk of exposure to contaminants in drinking water supplies, including heavy metals, nitrates, and microbial pathogens, by linking residential addresses to water-source data. In the context of soil and agriculture, GIS linkages are used to estimate pesticide exposure in agricultural areas and exposure risks from contaminated land due to historical industrial use, proximity to waste disposal sites, and natural deposits of harmful substances. Additionally, GIS applications are used in studies estimating the impact of socioeconomic status; access to healthcare, food, and recreational facilities; and factors of the built environment such as access to green space²⁵ and blue space, urban heat islands, and noise pollution on health outcomes. GISs can also be used to estimate environmental factors such as ultraviolet (UV) radiation exposure and the risk of extreme weather events such as floods and hurricanes.

Smart sensors in wearable devices that provide real-time data on individual exposure levels have transformed environmental exposure monitoring and assessment. Real-time data on air quality, noise levels, UV exposure and weather conditions, and physical activity can be paired with GIS data to reveal how environmental influences on human health differ within small geographic regions and across time.²⁶ Real-time monitoring is essential for understanding the air quality individuals are exposed to during daily activities and identifying air-pollution-related health hazards.²⁷ Combining GIS data with data from sensors in wearable devices can enhance contextual knowledge of exposures and how they vary across contexts. Geolocating sensor data can help mitigate environmental exposure hazards by correlating exposures to particular locations.²⁸ Wearable sensors can monitor small and coarse particulate matter (PM2.5 and PM10), volatile organic compounds (VOCs), and hazardous gases including CO and NO₂.²⁹ These sensors can also monitor noise levels, an exposure with a rapidly evolving body of evidence.³⁰ Data on environmental sound levels are used to determine exposure to potentially dangerous noise levels in urban areas and workplaces that could cause long-term hearing impairment or stress-related health concerns. Additionally, sensors in wearable devices can measure UV exposure, which is crucial for monitoring skin cancer risk and other sun-related health issues³¹ and regulating sun exposure. Sensors can also measure temperature and humidity, which affect comfort and health, particularly during harsh weather. Wearable sensors can detect steps, heart rate, and calories burned to measure lifestyle factors such as physical activity that interact with environmental exposures to affect health. Related technology can identify chemical exposures including heavy metals and organic contaminants.³² Due to the intricacy of detection, consumer wearables with these features are rare, although research and development is expanding availability.

Questionnaires administered to collect detailed data on lifestyle factors, environmental exposures, and social determinants of health are a pivotal source of information in environmental health research. While comprehensive data collection is instrumental in quantifying individual-level exposures and lifestyle factors, there are significant challenges, particularly regarding the standardization of questionnaires. The use of non-standard questionnaires introduces variability into the collected data, complicating comparisons and data synthesis across research efforts.³³ Subjective responses and recall bias further exacerbate these challenges, potentially skewing the data.³⁴ Moreover, cultural and linguistic differences can influence respondents’ understanding of questions, necessitating careful adaptation and validation across diverse populations to ensure accuracy and relevance.³⁴ Data quality and completeness are additional concerns, with the reliability of collected information dependent on respondents’ willingness and ability to provide accurate answers.^34,35 Environmental exposures can change over time, presenting additional difficulties with the use of questionnaires to assess historic and future exposures. To mitigate these challenges, there are ongoing efforts to develop standardized, validated questionnaires and implement advanced statistical techniques to adjust for bias and data variability.^33,36 Further, integrating questionnaire data with objective exposure measurements such as biomonitoring results and sensor data is increasingly recognized as a valuable approach.³⁷

Challenges with standardization of the exposome

Environmental exposures are responsible for 70%–90% of disease risk. While genetic studies commonly estimate heritability, defined as the proportion of phenotypic variation due to genetic variability, estimating the proportion of variability due to the exposome is challenging because of issues with direct exposure measurements. In contrast to genomics, where the genome is measured in a structured system with well-known estimates, and there are opportunities for data sharing through comprehensive databases with reference data, there is a lack of standardization and data sharing in exposome measurement. Accordingly, the proportion of disease risk due to exposome variability is underestimated. Table 1 details current estimates of the relative contribution of established exposures, underlining the importance of exposures in understanding overall disease etiology. While there is growing evidence in the GEI literature on specific diseases, an important open question is estimating the proportion of variation explained by GEI. Estimates of disease variation due to genetics and environmental factors can exceed 100% because they are often derived from different studies using different methods, leading to overlap. For example, twin studies may provide heritability estimates, while separate environmental studies offer risk estimates. When added together, the sum can erroneously exceed 100% due to this overlap and not accounting for GEI, which could either amplify or reduce the effects of each other. This highlights the complex interplay between genetics and the environment in disease causation.

Table 1.

Estimates of disease risk due to the exposome and heritability

Disease	Environmental factors	Exposome contribution	Heritability estimate
CVD	air pollution (PM2.5), leading to atherosclerosis, hypertension, and myocardial infarction	over 80% of CVD risk is attributable to environmental and lifestyle factors38	40%–60%11
T2D	dietary patterns, obesity, physical inactivity, exposure to chemicals (phthalates, bisphenol A)	70%–90% of T2D predisposition is explained by environmental exposures, lifestyle factors, and genetic predisposition39	25%–75%40
Asthma and respiratory diseases	air pollutants (NO2, O3), indoor allergens (dust mites, mold), tobacco smoke, occupational exposures	60%–70% of asthma incidence is attributed to environmental factors, especially early-life exposures41	35%–95%42
Cancer	tobacco smoking, dietary factors, alcohol consumption, sun exposure, environmental pollutants	70%–90% of most cancers are attributed to environmental factors; lifestyle also plays a dominant role43	varies widely, from low (∼15%) to high (>80%) depending on the type44
Neurodegenerative diseases (e.g., Alzheimer’s disease)	physical inactivity, diet, exposure to air pollution, occupational exposures	up to 60% of neurodegenerative disease risk is explained by environmental factors45	25%–80%, depending on familial or sporadic cases46

CVD, cardiovascular disease; T2D, type 2 diabetes.

There are several challenges involved with standardizing exposure measures across cohorts and studies, primarily due to the heterogeneity of data collection methods used. The approaches used for individual studies vary, resulting in inconsistencies in the type, quality, and granularity of data. Non-standardized data complicate comparisons and meta-analyses. Moreover, due to the dynamic nature of the exposome, which changes across time and location, it is difficult to establish standardized measures that accurately capture the variability of exposures.

The complexity of the exposome adds another layer of challenge, as the vast array of exposures requires unique measurement techniques. These challenges are further exacerbated by the differing methodologies and units of measurement used in studies. Additionally, the lack of structured data and opportunities for data sharing makes it challenging to standardize measures and compare findings across studies. Table 2 summarizes key challenges for types of environmental exposure estimates.

Table 2.

Key challenges involved in environmental exposure estimates

Methodology	Challenge	Description
GIS estimates of exposure	spatial resolution variability	the resolution of GIS data can vary significantly, affecting the accuracy of exposure estimates
	temporal dynamics	Environmental factors can change over time, making it challenging to estimate exposures accurately at specific time points
	lack of personalization	GIS estimates often rely on area-level data, which may not accurately reflect individual exposures
High-throughput MS exposomics technologies	standardization of analytical methods	standardized protocols and reference materials are needed to ensure comparability of results across laboratories and studies
	data interpretation and integration	interpretation of MS data can be complex, requiring sophisticated bioinformatics tools for data integration
	sensitivity and specificity	variations in sensitivity and specificity across compounds can affect the accuracy of exposure assessments
Survey assessments of exposure	subjectivity and recall bias	self-reported data are subject to biases, which can affect the accuracy of exposure assessments
	questionnaire variability	non-standardized questionnaire design can lead to inconsistencies in collected data
	cultural and language differences	surveys developed in one cultural or linguistic context may not be directly applicable in another

Advantages of GEI analysis

For complex traits influenced by both genotypes and environmental factors, GEI analyses have several advantages over genome-only models because they incorporate environmental exposures known to influence human health. GEI analyses can provide insights into disease mechanisms, identify sources of heterogeneity across individuals, and account for what some refer to as missing heritability, which is the proportion of heritability not explained by genetics alone.⁴⁷ A growing number of well-established GEIs have been replicated across studies in various populations. Examples include associations between air pollution and cardiovascular disease, alcohol and liver disease, smoking and lung cancer, diet and obesity, and pesticide exposure and Parkinson’s disease.

GEI analyses can also help identify novel pathways that influence human health.⁴⁸ Compared to genome-only approaches, a joint GEI model can explain significantly more phenotypic variance and substantially improve phenotypic prediction accuracy. Including environmental exposures in an analysis using the exposome framework allows additional phenotypic variance to be captured, including additive and non-additive effects such as GEI and environment by environment interactions (ExEs).⁴⁹ In addition, accounting for the influence of environmental exposures on health phenotypes can boost power.⁵⁰ Polygenic scores (PGSs), which provide a measure of an individual’s risk of a disease due to their genetics, may have low predictive power when applied to populations with different genetic ancestry or socioeconomic status or those that live in different environments.⁵¹ Accordingly, including the exposome in polygenic models can improve the portability of PGSs.⁵²

Interplay of environmental epidemiology and genetics

Epidemiology provides a framework for investigating the distribution and determinants of health and disease in populations, and genetics provides insight into the underlying biological mechanisms that contribute to disease. Genetic analysis examining the proportion of variation in a trait due to genetic factors is an increasingly important tool for estimating disease heritability and has substantially influenced study design in epidemiology. Family studies help distinguish genetic and environmental risk, and genetics help explain population heterogeneity. Because genetic markers provide an objective measure of ancestry that is not influenced by environmental factors, including this type of data can improve study design by identifying and adjusting for population stratification resulting from selection bias. This selection bias can lead to spurious associations in traditional epidemiological studies.

Epidemiology plays a crucial role in the collection of adequately phenotyped samples for genetic studies. High-quality, standardized measures of disease outcomes and environmental exposures are also necessary to interpret genetic associations. Epidemiological studies provide a framework for investigating interactions between genetic and environmental factors in disease development. Additionally, epidemiological research is essential for elucidating the clinical and therapeutic implications of genetics and establishing guidelines and policies for the appropriate use of genetic information in clinical practice and public health. Finally, including genetics in epidemiology studies enhances the ability to identify environmental effects, especially those found in subgroups that are not as apparent or strong in the overall population. In some cases, genetic effects are evident only with stratification by exposure level, or exposure effects are detected only in combination with genetic effects.

Candidate genes

While the selection of candidate genes ultimately depends on the context of a study and what is already known, some classes of genes have been established as environmentally responsive. Many genes play a direct role in an individual’s response to external factors, encompassing detoxification, DNA repair, and maintenance of cellular homeostasis. Key gene classes involved in responses to external factors include cytochrome P450 enzymes (CYPs) for metabolizing drugs and detoxifying xenobiotics, DNA repair genes for correcting damage from agents such as UV light and chemicals, and metabolic genes other than CYPs that transform substances for easier excretion. Antioxidant genes combat oxidative stress from pollutants, and heat shock proteins (HSPs) protect against environmental stressors by ensuring proper protein folding. Xenobiotic receptors regulate the metabolism and excretion of foreign compounds, inflammatory response genes mediate reactions to environmental agents, and transporter genes facilitate the movement of substances across cell membranes, including efflux pumps that export toxins. Table 3 summarizes genes that are highly responsive to environmental factors and outlines their function in the context of common exposures.

Table 3.

Gene classes highly responsive to environmental factors and their function

Class of Genes	Function
CYPs	metabolize drugs and detoxify xenobiotics; involved in the activation and deactivation of carcinogens, drugs, and pollutants
DNA repair genes	repair DNA damage caused by environmental agents such as UV light, radiation, and chemical mutagens; includes the nucleotide excision repair, base excision repair, DNA mismatch repair, and DNA double-strand break repair pathways
Metabolic genes	involved in converting substances into water-soluble compounds for excretion; includes the glutathione transferases and uridine diphosphate-glucuronosyltransferases enzymes
Antioxidant genes	code for enzymes that protect cells from oxidative stress by neutralizing reactive oxygen species; includes superoxide dismutase, catalase, and glutathione peroxidase
HSPs	upregulated in response to environmental stress to help fold proteins correctly and prevent aggregation of misfolded proteins
Xenobiotic receptors	sense the presence of xenobiotics and regulate the expression of metabolism and excretion-related enzymes and transporters; includes the aryl hydrocarbon receptor and pregnane X receptor
Inflammatory response genes	trigger and mediate inflammation in response to environmental agents; include genes coding for cytokines and chemokines
Transporter genes	encode for proteins involved in the transport of substances across cellular membranes, including efflux pumps such as P-glycoprotein

A growing number of resources catalog known or predicted genes with evidence of GEI. The Comparative Toxicogenomics Database (CTD) is a publicly available, manually curated resource that outlines interactions between chemicals and genes, chemicals and diseases, and genes and diseases⁵³ to elucidate a broad network of these components. GeneComps and ChemComps identify genes and chemicals with shared toxicogenomic profiles and provide enriched Gene Ontology terms and Venn diagram tools that identify common and distinct characteristics among chemicals, genes, and diseases, improving gene pathway data.⁵⁴ CTD is integrated into several databases, including PubChem, PharmGKB, UniProt, T3DB, GAD, ChemID, and TOXNET.⁵⁵ Additionally, microarray and proteomics datasets for various species can be accessed via the Chemical Effects in Biological Systems (CEBS) knowledgebase (http://cebs.niehs.nih.gov/cebs).⁵⁶

Evolution from candidate GEI studies to genome-wide interaction studies

The availability of high-throughput genotype data and large-scale phenotype and exposome data from large consortia has motivated advances in GEI methodology from candidate gene-environment interaction (cGEI) studies⁵⁷ to genome-wide GEI scans for millions of single nucleotide polymorphisms (SNPs), referred to as genome-wide environment interaction studies (GWEISs)⁵⁸ or genome-wide interaction studies (GWISs). cGEI studies consider pre-identified genes based on prior scientific knowledge or hypotheses, along with selected exposures, to determine the joint effects of the genotype and environment with multiplicative or additive models. cGEI studies have found several interesting GEI associations,⁵⁹ but this approach is limited by prior knowledge and low replicability.^57,59 Additionally, cGEI studies do not allow for hypothesis-free scans of the genome to identify novel GEIs. Further, the proportion of phenotypic variation explained by cGEI is anticipated to be low.⁶⁰

Genome-wide GEI scans do not require a priori hypotheses and are agnostic in nature and thus enable researchers to find novel main and interaction effects. GWISs have been used to identify GEIs with various environmental exposures such as toxic substances, diet, and smoking.⁶¹ While GWISs allow for hypothesis-free scans, they are often severely underpowered due to an extreme multiple testing burden. Further challenges for GWISs are related to exposure assessment, data harmonization in large consortia, reliance on electronic health records for phenotyping, multiple testing due to the large numbers of gene-exposure combinations tested, genetic heterogeneity, and the need for replication in other cohorts.⁶²

Epigenetics

Epigenetics involves the study of non-sequence modifications to DNA and modifications to chromatin that affect gene-DNA accessibility and subsequent gene regulation. The epigenome comprises the DNA base changes, chromatin protein tails, and non-coding RNAs that regulate gene expression and mRNA alterations that influence protein production. Epigenetics bridges an individual’s genetic makeup and environmental exposures to provide insights into how GEIs contribute to disease development. Cell-cell communication and tissue and cell growth depend on epigenetic markers, and environmental exposures can cause disease by altering these epigenetic markers. DNA CpG methylation was the first and easiest epigenetic signature to assess. Investigations of other epigenomic modifications, such as hetero-chromatic histone modification, are limited by technology costs and the difficulty of large-scale measurements, in part due to the requirement for high-quality antibodies. There are a growing number of replicated epigenetic-environment interactions, including smoking and DNA methylation,⁶³ maternal stress and methylation,⁶⁴ and prenatal famine and histone modification.⁶⁵

Epigenome-wide association studies (EWASs) are key to understanding GEI mechanisms, and commercial, high-density BeadChips⁶⁶ are making EWASs more affordable. The availability of this technology, lower per-sample cost, and the proliferation of replicated results in numerous study populations are driving EWAS adoption. Despite these advances, EWASs remain too expensive for very large population studies.

There is growing interest in studying the accumulation of epigenetic changes, known as epigenetic drift, in individuals with sustained generational environmental stressors such as stress and insufficient or improper nutrition. This is paramount for understanding the consequences of racism, poverty, and other multigenerational disparities. For example, historic and ongoing experiences of discrimination, poverty, and toxic exposures can contribute to chronic stress and other environmental exposures that can affect gene expression and increase the risk of disease. The weathering hypothesis posits that cumulative stress-related exposures due to discrimination, lower socioeconomic status, and racism result in increased disease risk in marginalized groups.⁶⁷ Additionally, the allosteric load theory, when considered in the context of GEI, highlights the sophisticated mechanisms through which living organisms adapt to their environments by modulating enzyme activity. This theory suggests that allosteric regulation, which occurs when molecules bind to enzymes at sites other than the active site to change their activity, plays a pivotal role in adjusting metabolic pathways in response to environmental changes.⁶⁸

Because epigenomics alone is insufficient for a comprehensive understanding of disease pathology, several approaches have been used to examine the collective genome and epigenome. One approach expands the quantitative trait analysis used in transcriptomics (e.g., eQTL analyses) to examine methylation-eQTLs (methyl-eQTLs). Methyl-eQTL analyses identify CpGs associated with gene expression and are often used to follow up top hits/findings from EWASs and genome-wide association studies (GWASs). This approach can reveal insights into relevant biological pathways. More general pathway analysis approaches can also be used.⁶⁹

Possible mechanisms of epigenetic inheritance include DNA methylation, heterochromatin, polycomb proteins, and non-coding RNAs.⁷⁰ Aligning with the Developmental Origins of Health and Disease (DOHaD) theory, early life, prenatal, and pre-conception settings have been shown to affect long-term health and chronic illness risk. For example, research has connected low birth weight and maternal habits including smoking and malnutrition with adult cardiovascular disease and type 2 diabetes.⁷¹ DOHaD also emphasizes the vulnerability of fetuses and neonates to environmental contaminants due to poor defensive mechanisms.⁷² For example, infants born to malnourished mothers and those exposed to endocrine disruptors are more likely to develop metabolic illnesses,⁷² and smoking during pregnancy is linked to low birth weight and obesity for offspring.⁷² Additionally, events such as famines provide evidence for intergenerational inheritance with long-term impacts such as elevated risk of obesity, diabetes, and mental health disorders seen in the offspring of those who experienced famine.⁷³ The DOHaD theory highlights early-life events’ impact on long-term health and has major implications for public health efforts to prevent chronic disease.

Multi-omics approaches for PEH

Multi-omics approaches, including genomics, transcriptomics, proteomics, epigenomics, lipidomics, and metabolomics, have been independently and jointly utilized to identify molecular features associated with environmental exposures that affect disease progression. Traditional omics-environment (OxE) interaction studies often analyze omics datasets independently. For example, transcriptomics has unveiled gene-expression signatures associated with exposure to various environmental chemicals that affect pathways involved in inflammation, oxidative stress, DNA damage, and apoptosis.⁷⁴ MS-based proteomics has been employed to investigate the effects of environmental exposures on protein expression and post-translational modifications such as phosphorylation, acetylation, and ubiquitination.⁷⁵ Additionally, MS has been used to examine the effects of biodiesel exhaust exposure on bioactive lipid mediators in human airways using lipidomic profiling methods.⁷⁶ Further, a metabolomics study⁷⁷ detected trichloroethylene metabolites in human plasma, linking these metabolites to changes in endogenous metabolites involved in immunosuppression, hepatotoxicity, and nephrotoxicity.

Integrative omics integrates and simultaneously analyzes multiple omics datasets for a comprehensive understanding of complex biological systems. This approach can unravel complex interactions between genetic variations, gene-expression patterns, protein profiles, metabolic changes, and epigenetic modifications in response to environmental factors. For example, a multi-level omics (proteomics, metabolomics, lipidomics, one-carbon metabolism, and neuroinflammation) analysis in Alzheimer’s disease pathology identified interacting pathway alterations, revealing dimensions of heterogeneity and potential novel disease mechanisms.⁷⁸ Another study of Alzheimer’s disease⁷⁹ integrated GWAS data, multi-omics data, drug-target networks, and protein-protein interactions to identify high-confidence Alzheimer’s disease-risk genes and candidate drugs for repurposing. Additionally, to explore the mechanisms of cognitive impairment in individuals with type 2 diabetes, researchers conducted a multi-omics analysis (transcriptome, plasma metabolome, and gut microbiota) in db/db mice with cognitive decline.⁸⁰ Further, to understand the etiology of coronary artery disease through an integrative approach, a large-scale study⁸¹ combined GWAS variants, eQTLs, protein quantitative trait loci (pQTLs), and cell-type-specific gene-expression datasets. This framework enabled the identification of cell-type-specific OxE effects from bulk expression data, which is confounded by single-cell heterogeneity.

Advancements and challenges in studying OxE interactions using multi-omics

Advancements in single-cell omics technologies, such as single-cell RNA sequencing (scRNA-seq)⁸² and single-cell assay for transposase accessible chromatin sequencing (scATAC-seq),⁸³ have enabled the identification of associations between traits and cell types from high-resolution profiling of individual cells. Single-cell integrative omics methods such as cellular indexing of transcriptomes and epitopes by sequencing (CITE-seq)⁸⁴ merge RNA and protein measurements, providing concurrent insights into transcriptomes and cell-surface proteins. However, integrative OxE analyses face significant challenges involving data harmonization, the integration of multiple types of environmental exposure data from diverse sources, the consideration of cell type-specific responses, how to capture dynamic interactions, and sample-size limitations. To address these challenges, promising tools are being developed to increase the precision of OxE analyses. For example, scQTLbase⁸⁵ and scGWAS⁸² facilitate the identification of loci that play regulatory roles as well as associations between traits and cell types in single-cell integrative omics analyses.

Methods for GEI analysis

In traditional quantitative genetic studies, linear additive models offer a straightforward framework for testing GEI by incorporating both genetic and environmental variables, along with their interaction terms, into a regression equation. In these models, the dependent variable typically represents a phenotype of interest, such as disease status or a quantitative trait, and the independent variables include a genetic factor (e.g., genotype or allele count), an environmental exposure, and an interaction term that represents the product of the genetic and environmental variables.^1,86 This approach is based on the premise that, if the interaction term significantly improves a model’s fit to the data, it suggests the presence of a GEI. This approach has been crucial in pinpointing essential genes that interact with the environment, but it does not show how the collective effects of genes are altered in response to environmental stimuli.⁸⁷ With full-genome sequences, it is now possible to conduct GWEISs, which integrate GEI analyses into GWASs.⁵⁹ This is anticipated to become a standard method for mapping complex traits,⁸⁸ and methodological work is rapidly advancing. In addition, there is growing appreciation for an omnigenic hypothesis that posits that complex features or illnesses are influenced by all genes, which work together in a network inside an organism.⁸⁹

Machine learning and artificial intelligence (AI) methods are being increasingly used in algorithms that include random forests, support vector machines, and neural networks. Deep-learning models excel in capturing complex, non-linear interactions, enhancing the understanding of gene-environment dynamics. Additionally, AI aids in constructing interaction networks and enables cross-validation and ensemble methods for robust model evaluation. Bayesian networks facilitate causal inference, and natural language processing can extract valuable information from scientific literature and clinical records, facilitating the identification of potential GEI. While it is beyond the scope of this review to cover all study designs and methods for detecting GEI, Table 4 summarizes current and emerging approaches and their advantages and disadvantages.

Table 4.

Study designs and methods for detecting GEI

Approach	Description	Advantages and limitations
Candidate gene-environment analysis57,59,90	pre-identified genes or variants and selected exposures are used to determine joint genotype and environmental effects on a multiplicative or additive scale	specific genes and exposures can be investigated; limited by prior knowledge and low replicability; does not allow for hypothesis-free scans
GWISs or GWEISs91	genome-wide scans for GEI interactions enable the identification of novel main and interaction effects without a priori hypotheses	does not require a priori hypotheses; severely underpowered due to multiple testing burden
Variable selection92,93	the number of features is reduced using methods such as lasso, ridge regression, and Bayesian variable selection	enhances power for GEI scans
Tree-based methods94	tree-based methods are applied to interaction models	provides easily interpretable models; may not capture complex interactions
Combinatorial approaches94,95	all possible variable combinations are searched for the best predictive model; multifactor dimensionality reduction is popular for detecting gene-gene interactions and GEI	non-parametric and model-free approach; computationally intensive and may suffer from the curse of dimensionality
Case-only design96	detects GEI interactions under the G-E independence assumption	greater power than traditional case-control studies when G-E independence assumption holds; increased type I error if the G-E independence assumption is violated
Multi-stage analysis97	SNPs are filtered in a multi-stage process before testing GEI interactions using SNPs selected based on marginal effects from GWAS results, gene-environment correlation tests, functional annotation, or external pathway knowledge	filters and prioritizes variants; may miss important interactions
Filtering and prioritization87,98,99	biological knowledge from variant annotations, molecular phenotypes, pathway-based annotations, and associations from external databases, variants, and exposures are filtered and prioritized for GEI analyses; vQTLs enriched for GEI can be used for prioritizing genetic variants	incorporates biological knowledge into analysis; quality and completeness of knowledge base affect results
Set-based methods for rare variants100	burden tests or variance component tests are used to group rare variants and test interactions between the set of rare variants and an exposure	increased statistical power and reduced multiple testing burden compared to single variant tests; loss of information for individual variants because variants are grouped
Polygenic score by environment (PGSxE) interactions101	interactions between individual polygenic scores and environmental exposures are tested using methods that include polygenic gene-environment interactions (PIGEONs)	reduces multiple testing burden; accounts for the polygenic nature of complex traits; quality and portability of PGS can lead to biases and limited generalizability
Multi-exposure analysis92,102,103	the effects of the genome and exposome are jointly modeled using linear mixed models such as StructLMM, GxEMMs, and IGE to investigate the effects of multiple exposures on complex traits	accounts for genome-exposome correlations and correlated exposome variables; computationally intensive; difficult to interpret and determine causality
Mixtures analysis104,105	risk from complex mixtures such as air pollution, diet, or cigarette smoke is assessed by analyzing effects of mixture components and examining metabolic pathways for specific exposures	realistically models exposure mixtures; increased risk of confounding factors; population-specific exposure patterns can reduce the generalizability of results
Longitudinal analysis106,107,108	gene x longitudinal-exposure interactions are modeled by incorporating cumulative and time-varying exposures from longitudinal studies into functional logistic regression modeling	increased power with repeated measurements time and resource intensive
Machine learning109	GEI analyses can be conducted using advanced machine-learning models such as neural networks; evolutionary algorithms such as genetic programming neural network and grammatical evolution neural network can optimize the parameters of a neural network	models complex, non-linear relationships; requires large amounts of data; requires cross-validation to avoid overfitting and assess predictive ability of the model; tends to produce black-box models, which are challenging to interpret

G-E, gene-environment.

Other advances in the field include set-based methods that use burden or variance component tests to group rare variants, which are SNPs with minor allele frequency less than 0.01, to test for interactions between a set of variants and an exposure.¹¹⁰ Other methods estimate and partition heritability by modeling the genome and exposome with linear mixed models. The integrative analysis of genomic and exposomic data (IGE)⁴⁹ and gene-environment mixed models (GxEMMs),¹⁰² a mixed model for GEI, are whole-genome methods for estimating additive and interaction effect heritability and variance. Structure linear mixed models (StructLMMs)¹¹¹ examine SNPs with multiple environmental variables to identify GEI-interacting SNPs. vQTLs enriched for GEI can be used to identify and prioritize genetic variants associated with phenotype variance.⁹⁹ Additionally, PGSs are powerful tools for assessing interactions, particularly for highly polygenic traits.¹¹² GEIs for rare diseases or genotypes can be explored using meta-analyses of exposure data and study results.

Methods development is also focused on assessing risk from complex mixtures of exposures such as air pollution, diet, or the carcinogen mixtures in cigarette smoke or pesticides.¹⁰⁴ Effects of the components of a complex mixture can be dissected by examining metabolic pathways in individuals with different genetic backgrounds.⁵⁹ For example, studies have identified the interaction between N-acetyltransferase 2 (NAT2) and heterocyclic amines found in red meat as a potential risk factor for colorectal cancer.¹¹³ Longitudinal measures of environmental exposure, such as those in Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE),¹⁰⁶ can be leveraged to incorporate cumulative and time-varying exposures for complex disease-risk assessments.¹⁰⁷

As reviewed in McAllister et al.,¹¹⁴ detecting GEI presents many statistical challenges. High dimensionality complicates analyses and requires multiple comparison adjustments, increasing the risk of type II error. Additionally, many studies lack power to detect interactions due to small effect sizes and insufficient sample sizes due to logistics and costs. However, the availability of data from mega-sized observational studies such as the UK Biobank¹¹⁵ and All of Us¹¹⁶ is helping address these limits. Further, heterogeneity in genetic and environmental factors across populations can obscure true interactions and hinder replicability. Model complexity is another issue because complex models are needed to accurately explore the many potential interactions and confounders.

Measurement errors and misclassifications can introduce bias into estimates and reduce power to detect true interactions, complicating GEI detection. Due to confounding and collinearity, disentangling the independent effects of correlated exposures is difficult and often inconclusive. Differences in interaction scale can also affect detection and interpretation. True interactions may also be missed due to multiple testing. Additionally, the difficulty of collecting high-quality data on genetic variants and environmental exposures can bias detected interactions. Finally, ethical and privacy concerns about genetic and environmental data can limit its use. Innovative methods and interdisciplinary collaboration are needed to overcome these challenges and gain insights into disease etiology and prevention. Although machine learning and AI are emerging methods for interaction detection, generalized linear models are still widely used.⁹²

The lack of a standardized roadmap for quantifying the exposome and harmonizing environmental measurements across studies causes data collection and measurement protocol inconsistencies in GEI analyses. Global cooperation and funding agency enforcement are needed to solve these problems. Researchers are aware of these challenges, and efforts such as those from the Human Health Exposure Analysis Resource (HHEAR) network can help expand standardized exposure analysis in human health studies.¹¹⁷ hhearprogram.org, led by NIEHS, was recently issued to create the Center for Exposome Research Coordination to Accelerate Precision Environmental Health, which is tasked with helping resolve the challenges associated with exposomics studies.

Methodological advances are needed to address heterogeneity in exposure measurements, variable exposure distributions across populations, confounding factors, and population stratification when integrating exposure data from diverse studies for larger sample sizes. Additionally, methods to improve GEI detection in diverse populations are needed to ensure findings are broadly applicable. Finally, large population-based studies using geospatial estimates of environmental exposures are promising but require new methods for integrating data from diverse sources to explore complex relationships between health and the environment.

Translation for PEH

In the context of PEH, GEI studies aim to identify exposure-response relationships to predict disease risk and improve health. Analyzing exposure data across the course of life reveals inter-individual differences in susceptibility and informs the development of biomarkers of exposure/disease risk. Determining which genetic variations increase susceptibility to the toxic effects of environmental pollutants can help identify individuals at risk of adverse health outcomes and inform tailored interventions.¹⁰ GEI studies provide insights into how the complex interplay between genetic and environmental factors affects disease development and outcomes and elicits the underlying biological mechanisms of disease. We highlight early exemplars of how GEI can be translated into PEH practice. We then discuss what is needed to translate basic research from model organisms and new approach methods (NAMs) into PEH practice.¹¹⁸

Exemplars of PEH translation

GEI research has a direct impact on individual-level healthcare and preventive medicine in a growing number of real-world scenarios. Interactions have been found between genetic variation in CYP2D6 and pesticide exposure for Parkinson’s disease, NAT2 and smoking for bladder cancer, ALDH2∗2 and alcohol intake for esophageal cancer, and AS3MT and arsenic for skin lesions.¹¹⁹ As recently reviewed in Kabbani et al.,¹²⁰ pivotal real-world pharmacogenomics applications showcase how knowledge of GEI can be incorporated into clinical practice, enhancing personalized medical treatment. For example, personalized warfarin dosing, guided by variations in genes that include VKORC1 and CYP2C9, optimizes dosage to balance efficacy and bleeding risk.¹²¹ Further, genetic testing for the HLA-B∗57:01 allele before prescribing abacavir for HIV has become standard practice to prevent reactions due to severe hypersensitivity.¹²² In another example, due to severe toxicity risks at standard doses in individuals with variants in the TPMT gene, dose adjustments can be made or alternative therapies suggested when prescribing thiopurine drugs such as azathioprine and mercaptopurine.¹²³ Additionally, CYP2D6 gene variants help guide the selection of antidepressants and dosages to optimize treatment efficacy and minimize side effects.¹²³ Identifying individuals with BRCA1/2 mutations and HER2 expression enables personalized cancer therapy with PARP inhibitors and trastuzumab, respectively, ensuring targeted and effective treatment strategies for patients.¹²⁴

In addition to drug-response outcomes, there are a growing number of translational implementations for disease prevention in PEH. An example is GEI related to susceptibility to asbestos exposure. BAP1, a tumor suppressor gene, when mutated, increases susceptibility to several cancers, notably mesothelioma, in the context of asbestos exposure.¹²⁵ These mutations not only predispose individuals to disease but may also modulate clinical outcomes, suggesting varied prognoses for affected individuals. This underscores the potential for targeted screening and prevention strategies, especially for those with known BAP1 mutations and a history of asbestos exposure. Investigating how these mutations interact with asbestos to promote carcinogenesis is vital for developing targeted therapies, as this helps with understanding the affected molecular pathways.¹²⁶ Table 5 lists examples of GEI applications in PEH. Baccarelli et al.¹⁰ provide additional examples and context.

Table 5.

Exemplar GEI applications in PEH

Scenario	Genetic and environmental factors	Impact
Personalized air-quality alerts	genetic susceptibilities to respiratory conditions and air quality	manage symptoms of respiratory conditions and prevent exacerbation by reducing exposure to pollutants
Lead exposure and genetic susceptibility	genetic variations and lead exposure	implement interventions to reduce the risk of cognitive and developmental delays associated with lead exposure in susceptible children
BAP1 and asbestos exposure	BAP1 mutations and asbestos exposure	implement targeted surveillance and early detection strategies for improved outcomes in individuals with an elevated risk of asbestos-related diseases
UV exposure and skin cancer risk	MC1R gene variants and UV radiation	implement skin cancer screening for early detection and treatment of skin cancer and provide personalized sun-protection advice to reduce the incidence of melanoma in susceptible individuals
Personalized nutrition plans	genetic variations and nutrient intake	manage chronic conditions through dietary modifications tailored to individual genetic makeup to reduce the risk of diet-related diseases such as type 2 diabetes and cardiovascular disease
Pesticide exposure and Parkinson’s disease	genetic susceptibilities and pesticide exposure	implement stringent protective measures and monitoring to manage and possibly prevent the onset of Parkinson’s disease in susceptible individuals

Translation from model organisms and NAMs for PEH

Discovery and validation of GEI in human epidemiology studies are limited by low power compared to main-effect studies, the difficulty of harmonizing environmental risk factors across consortia, the complexity of measuring environmental exposures (especially mixtures) due to temporality, genetic and phenotypic heterogeneity, linkage disequilibrium patterns and genetic modifiers that can reduce power, and challenges associated with the location of most genetic variants (SNPs) in non-coding regions. GEI studies with population-based in vivo and in vitro experimental models have advanced the understanding of population variability. Especially for GEI studies that are unethical or intractable to conduct in humans (e.g., toxicity studies), testing and genetic analysis can be conducted in model organism population models or NAM experiments.¹²⁷ Studies have demonstrated the utility of population-based human in vitro cells and population-based in vivo mouse models in hazard identification, exposure-response assessment, and mechanistic understanding.¹²⁸ Hypotheses arising from these studies can then be tested using human genetic data, improving detection power and enabling the analysis of subpopulations.

Several population-based model organism resources have advantages compared to (and can complement) human GEI studies. Rodent models are heavily used in the field, and each has respective advantages and disadvantages. Among these are the genetically diverse Collaborative Cross,¹²⁹ Diversity Outbred lines,¹³⁰ Complementary CC-RIX mice,¹³¹ Mouse Hybrid Diversity Panel,¹³² Rat Hybrid Diversity Panel,¹³³ and genetically heterogeneous rat stocks. These models help identify individual quantitative trait locus (QTL) and candidate genes involved in GEI interplay with complex human diseases due to the high sequence homology between mice and humans.

In addition to rodent resources, other population-based model organism systems have potential for GEI discovery, including Drosophila,¹³⁴ Caenorhabditis elegans, and fish panels.¹³⁵ An advantage of these alternative resources is broad exposome coverage due to their amenability to high-throughput data generation.¹²⁷ Population-level studies utilizing complementary resources of cells and derived cell lines also enable in vitro population-level studies.¹³⁶

For GEI studies, NAMs have historically been used during early discovery or in a screening context because of their amenability to high-throughput scaling. High-throughput screening (HTS) assays incorporating genetic data and environmental and chemical exposures can prioritize candidate exposures for in-depth hypothesis testing, including GEI analyses.¹³⁷ Tox21¹³⁸ and ToxCast¹³⁹ enable HTS through in vitro assays produced by inserting chemical exposure-sensitive reporters into human cell lines. Using computational methods, the toxicity and risk of chemicals are ranked from in vitro screening data for genotoxic agents,¹⁴⁰ endocrine-disrupting chemicals,¹⁴¹ thyroid receptor-interfering compounds,¹⁴² and environmental, manufacturing, and agrochemical compounds in large chemical libraries.¹⁴³ HTS assays were used to assess the toxicity and endocrine disruption activity of dispersants in the DeepWater Horizon oil spill¹⁴⁴ and evaluate the toxicity of chemicals in the West Virginia Elk River chemical spill.¹⁴⁵ High-throughput transcriptomics has recently been incorporated into screening assays¹⁴⁶ to identify the effects of genes, signaling pathways, and biological networks on environmental and chemical exposures, identify endocrine-disrupting environmental contaminants,¹⁴⁷ and determine hazard value estimates for long-term exposures using short-term assays.¹⁴⁸

Adverse outcome pathways (AOPs) are a framework for organizing scientific knowledge regarding the linkage between perturbation of a specific biological target by stressors and a resultant adverse outcome relevant to risk assessment and regulatory decision making. This framework provides approaches for assessing how biological processes such as gene expression, signaling pathways, and cellular functions respond to molecular events initiated by environmental exposures. Aggregate exposure pathways (AEPs)¹⁴⁹ have been proposed as a geospatial environmental exposure complement to AOPs. Environmental exposure assessment ends where AOPs begin, at a key biological or molecular event. Eccles et al.¹⁵⁰ demonstrated how AEPs and AOPs can be integrated for geospatially resolved mapping of molecular perturbations. To increase the accuracy and geographic coverage of geospatial exposure models and the human relevance of AOPs, an improved NAM-based approach for characterizing GEI-related outcomes and risk is needed.

Recent advances in functional genomics tools and technologies, in vitro approaches, and biological knowledge have enabled novel approaches for functionally validating, experimentally characterizing, and interpreting both newly identified and existing GEI findings relevant to human disease outcomes and better inferring biological plausibility after the discovery stage. The use of embryonic stem cells and induced pluripotent stem cells (iPSCs) from relevant cell types is an example of such an approach. Combined with genome/epigenome editing tools, these cell populations help identify and validate genetic variants responsible for environmental sensitivity. Additionally, organoid culture models such as tissue-chip platforms and other culture systems and microfluidics are emerging. Tissue chips or “organs on chips” derived from patient iPSC lines have enabled accurate in vitro human models of human diseases. Genome/epigenome editing tools such as CRISPR-Cas9 can generate robust high-content genome-wide screens that can be used to dissect functional non-coding genetic variants that drive susceptibility to environmental exposures.

Social and societal considerations in GEI research

Environmental justice

PEH combines large-scale individual-level biological data with individual-, community-, and population-level environmental data to enhance population-level health.¹⁵¹ For instance, GEI analysis can identify genetic variations that increase an individual’s susceptibility to pollutants. In the context of environmental justice, these results enhance understanding of how environmental factors affect health outcomes for different populations. This enables targeted population-level interventions to reduce hazardous exposures for vulnerable populations with the goal of addressing health disparities. As climate change accelerates, the impact of such GEI is likely to magnify, and genomics is posited as one mechanism for understanding climate change response.¹⁵²

PEH also plays a crucial role in addressing environmental justice issues such as unequal pollutant exposure in low-income and marginalized communities.¹⁵³ Historically, genetic studies have focused on populations of European descent and thus have not accurately represented the genetic and environmental diversity of populations underrepresented in clinical and scientific studies. Including more diverse populations in studies can expand understanding of the genetic and environmental factors that contribute to health disparities and help develop more effective interventions.¹⁵⁴ In addition to addressing health disparities, integrating historically underrepresented populations into large cohorts improves the generalizability of research findings.¹⁵⁵

The impact of exposures on gene expression varies between individuals and populations and can contribute to differences in disease risk and health outcomes. Accordingly, it is important to consider the potential consequences of racism and discrimination on individual and intergenerational epigenetics to understand the underlying mechanisms of disease and develop interventions. This includes research to identify populations at greater risk for transgenerational epigenetic changes due to generational exposure to toxicants or stress and interventions to mitigate the harm in the current generation and prevent harm in future generations.

Return of results

As the research community begins to report findings from GEI studies to research participants, there are many ethical, legal, and social implications. Nearly all participants in genetic and environmental studies want to know the results of their chemical exposures and GEI, even if the data have uncertain health implications and may reveal risks that cannot be addressed.¹⁵⁶ Report-back of findings to communities and individuals should be routine as a matter of reciprocity, respect, transparency, and trust.¹⁵⁶ The concerns of communities and their members can inform and guide the report-back of results. For findings involving high environmental exposures or high genetic susceptibility to an environmental exposure, researchers can help participants access resources to help them take appropriate actions, particularly in the context of understanding how GEI findings could affect their health. Report-back of findings should include information on choices to reduce individual-level exposures. Findings can aid decision-making in clinical settings and guide policy changes. Additional community engagement, guidelines, educational resources, training, and tool development are needed to support the effective communication of findings.

Data privacy

Data privacy and human subject protection are essential when collecting and analyzing environmental exposure data. Geospatial and sensor-based exposure data lack well-developed policies and protections, unlike identifiable genomics data. The Genetic Information Non-discrimination Act (GINA) protects genetic data, but there is no equivalent for environmental data.

As geospatial data are integrated into PEH studies, data privacy must evolve with technology. To assess environmental risk factors, participants’ residential addresses during critical susceptibility periods (e.g., infancy, childhood, adolescence, adulthood) are collected. Spatial exposure data that track location can reveal routines, habits, health, and pollutant exposures, making them vulnerable to identity theft and malicious activity. Accordingly, geospatial exposure data collection and use require clear guidelines and informed consent.¹⁵⁷ Data collection, storage, and sharing should incorporate encryption, access controls, and data minimization.

With the growing demand for data sharing, individualized exposure assessment, and expanding repositories of public and commercial data, the risk of re-identification in environmental health research is increasing. As reviewed in Onsrud,¹⁵⁸ location-specific environmental exposure measurements from wearable sensors capturing location, exposures, and biometrics can expose individuals to risk.

Breaches in data privacy can have severe consequences, including higher insurance costs, employability impacts, stigma at the individual and community levels, effects on home prices, and legal repercussions.¹⁵⁸ Insurers and employers subject to environmental regulations could re-identify environmental data and deny coverage or employment based on exposures. This information can potentially lower property values and stigmatize communities. Privacy breaches and re-identification can also damage public confidence in environmental health and harm litigants and research. Importantly, privacy breaches may exacerbate negative effects on vulnerable populations because environmental research often targets populations with high levels of exposures. These issues and specific examples are reviewed in Onsrud.¹⁵⁸

Conclusion

Understanding environmental exposures and GEI is crucial for predicting and preventing diseases, and the expansion of GEI research must include the exposome. Through the integration of multiple levels of genetic and environmental data, PEH can help identify populations at high risk due to genetic variants, exposure to specific environmental stressors, and combinations of both genetic and environmental risk factors. Integrating the exposome into emerging and existing studies focused on genetics/genomics will enable precise risk assessment and enhance regulatory decision-making. In PEH, PGSs and environmental risk scores can be merged to create individual-level metrics that are more clinically translatable and accurate than PGS alone. This could allow the generation of data-driven predictive risk scores that accurately predict an individual’s risk (or resilience) based on their unique genetic makeup and lifetime history of exposure to environmental risk factors, including those related to the built environment and socioeconomic status. In this comprehensive concept, emerging knowledge about inter-individual genetic/epigenetic diversity and its influence on susceptibility to environmental exposures could be used to predict and explain disease pathology more comprehensively and accurately. Building on findings at the individual level, PEH can identify novel points of intervention at the population level. For instance, changes can be made to the built environment to increase physical activity or access to smoking cessation programs for pregnant people can be expanded. Further, PEH results can be leveraged to prioritize the implementation and enforcement of environmental regulations to reduce health disparities by identifying populations with elevated risk due to environmental exposures.