The role of epigenetics in multi‐generational transmission of asthma: An NIAID workshop report‐based narrative review

Abstract There is mounting evidence that environmental exposures can result in effects on health that can be transmitted across generations, without the need for a direct exposure to the original factor, for example, the effect of grandparental smoking on grandchildren. Hence, an individual's health should be investigated with the knowledge of cross‐generational influences. Epigenetic factors are molecular factors or processes that regulate genome activity and may impact cross‐generational effects. Epigenetic transgenerational inheritance has been demonstrated in plants and animals, but the presence and extent of this process in humans are currently being investigated. Experimental data in animals support transmission of asthma risk across generations from a single exposure to the deleterious factor and suggest that the nature of this transmission is in part due to changes in DNA methylation, the most studied epigenetic process. The association of father's prepuberty exposure with offspring risk of asthma and lung function deficit may also be mediated by epigenetic processes. Multi‐generational birth cohorts are ideal to investigate the presence and impact of transfer of disease susceptibility across generations and underlying mechanisms. However, multi‐generational studies require recruitment and assessment of participants over several decades. Investigation of adult multi‐generation cohorts is less resource intensive but run the risk of recall bias. Statistical analysis is challenging given varying degrees of longitudinal and hierarchical data but path analyses, structural equation modelling and multilevel modelling can be employed, and directed networks addressing longitudinal effects deserve exploration as an effort to study causal pathways.


| INTRODUC TI ON
This review article is based on a workshop organized by the National Institute of Allergy and Infectious Diseases to bring together researchers interested in transgenerational effects focusing primarily, but not exclusively, on epigenetic transfer of information across generations. The specific objectives were to review and summarize the current status of knowledge for transgenerational effects in asthma and lung function, consider if epigenetic mechanisms explain this effect and identify major knowledge gaps and obstacles in furthering this field leading to recommendations for future research. The workshop focused on asthma, but the phenomenon has wider implications to other heritable chronic diseases such as diabetes and obesity.

| THE INHERITAN CE OF ALLERG I C DISE A SE
The most frequently identified predictor of asthma is a family history of asthma or atopy. 1 However, multi-generational effects may reflect a number of mechanisms. 2 These include a shared environment, genetic inheritance including genetic effects on the epigenome such as methylation quantitative trait loci and epigenetic inheritance.
Exposures such as smoking, diet, occupation, microbiome, air pollution and farming environments are often shared among family members across generations, meaning that multi-generational disease may not necessarily mean inherited disease. 3,4 Genetic inheritance is often assumed rather than proved. However, polygenic risk scores such as that developed based on a published asthma Genome-Wide Association Study in the Dunedin (New Zealand) longitudinal development study [5][6][7] do show that the likelihood of life-course persistent asthma is, in part, related to inherited genetic factors. Epigenetic changes in cells/tissues represent the accumulated impacts of environmental exposures, life changes (developmental stages) and genetic variants. 8 The role of epigenetics in asthma transmission between generations is thus of great interest and the focus of this review.

| EPI G ENE TI C TR AN S FER OF INFORMATION
Epigenetic factors are molecular factors or processes that regulate genome activity, independent of DNA sequence, and that are mitotically and/or meiotically stable. 9 These factors include DNA methylation, histone modifications, non-coding RNAs and chromatin The Role of Epigenetics in Multi-generational Transmission of Asthma: an NIAID Workshop RThe Role of Epigenetics in Multi-generational Transmission of Asthma: an NIAID Workshop Report-based Narrative Review

G R A P H I C A L A B S T R A C T
Parental and grandparental exposures influence risk of asthma in their offspring, in part mediated by epigenetic transfer of information across generations. An improved understanding of these effects will help to better estimate the true morbidity and mortality resulting from environmental exposures, and subsequently to develop concepts for prevention.

Key messages
• The risk to health should consider not only parental but also grandparental inheritance and exposures.
• Epigenetic processes might explain transgenerational effects, persisting in the absence of a direct environmental exposure.
• Multi-generational studies are required to provide insights into transgenerational epigenetic effects in human.
structure which controls the accessibility of DNA regions or loops functional DNA domains into proximity to each other. Epigenetic processes preside over the maintenance and termination of gene expression, and the remodelling of chromatin architecture is critical to prepare genes for regulated transcription.
In this report, intergenerational inheritance refers to maternal exposure having a direct effect on the developing fetus (and potentially on the germ line of the fetus), while transgenerational inheritance refers to the effect on subsequent generations persisting in the absence of a direct environmental exposure. 10 If a pregnant female (F0 generation) is exposed to an environmental toxicant, then the embryo(s) she carries (generation F1), as well as the developing germ cells within those embryos that will form the F2 generation, are directly exposed.
Therefore, the first generation in which one can observe transgenerational effects after exposure of a pregnant female is the F3 generation.
Where a male parent or non-pregnant female is exposed to an environmental toxicant, only the germ cells that will form the F1 generation are directly exposed, and therefore the first generation in which one can observe transgenerational effects is the F2 generation ( Figure 1).

| Epigenetic transgenerational inheritance
For environmentally induced epigenetic changes to be inherited, an altered epigenome must be present in germ cells (sperm or eggs).
Germ cells normally go through two rounds of epigenetic DNA

| Animal models of epigenetic inheritance
In a rat model, observed transgenerational epigenetic increases in susceptibility to diseases include increases in the incidence of germ F I G U R E 1 Transgenerational inheritance of disease risk: evidence from animal models. Gestational exposure of mice to an environmental trigger, in this case, diesel exhaust particles (DEP) or concentrated urban air particles (CAP), leads to increased asthma risk in up to 3 generations of offspring. It is important to distinguish intergenerational inheritance from transgenerational because in a gestational exposure model not only the pregnant F0 ancestor is directly exposed to the trigger, but also the fetus; moreover inside the F1 fetus there are predecessors of gametes that will give rise to F2 progeny which are theoretically also directly exposed. In the paternal line only the predecessors of gametes that form the F1 are directly exposed and effects that persist to the F2 can be considered transgenerational cell apoptosis in testes, male infertility, kidney disease, prostate disease, polycystic ovarian disease, decreased ovarian follicles and increased rates of cancer. 16 Transgenerational increases in rates of obesity and changes in behaviour have also been observed. 17,18 Several interesting findings have arisen from epigenetic transgenerational research in rats. Differential DNA methylation regions (DMRs; regions of the genome that are differentially methylated) between ancestrally toxicant-exposed animals and controls are fewer in the F1 generation than in the transgenerational F3 generations, and the F1 generation DMRs are different than those of the F3 generation. 19 Another interesting finding is that the pattern of DMRs in the transgenerational animals is specific to the ancestral toxicant exposure, suggesting that it may be possible to predict ancestral exposures in humans by examining epigenomic patterns in individuals. 15 Also, most DMRs in rodent studies have been found in intergenic regions of low CpG density, rather than in CpG islands. 19

| Experimental transgenerational transmission of asthma risk after exposure to environmental particles
While studies in human cohorts are still underway to determine the association between epigenetic alterations and transmission of the phenotype, a major challenge is to determine whether aberrant DNA methylation can be causative in this transgenerational transmission.
Hence, an experimental in vivo approach to track the aberrant methylation in three generations of mice whose increased asthma risk arises from a single ancestral exposure to environmental particles has been developed.

| DE VELOPMENTAL PROG R AMMING AND E ARLY LIFE ORI G IN S OF D IS E A S E
The strongest evidence for the importance of developmental programming in allergic disease is the observation of marked phenotypic differences already apparent at birth between individuals who do, or do not, develop allergic conditions later in life. 28 For example, children who develop asthma have impaired lung function shortly after birth in comparison to healthy children. 29 Neonates with allergic predisposition also have recognized differences in many aspects of immune function at birth. 28 While some of these differences will be due to inherited genetic predisposition to allergic disease, there is a clear role for maternal environmental exposures during pregnancy influencing subsequent development of allergy in offspring. For example, maternal allergy -and hence presumably an altered in utero environment -has an additional effect on risk of allergy in the offspring. 30 Evidence for the in utero origins of asthma comes from associations of childhood asthma with maternal pre-natal conditions and exposures such as maternal age, diet, smoking, infectious illness, stress, weight gain and exposure to farm animals. 21,31,32 The detection of asthma-associated differential epigenetic programming in neonates also supports the possibility that asthma might originate in utero. 33-35

| Maternal influences on childhood asthma risk
The maternal cytokine profile during the third trimester of pregnancy (defined as the ratio of IFNγ and IL-13 secreted by mitogenstimulated maternal peripheral blood mononuclear cells) has an inverse association with childhood asthma risk. Interestingly, this relationship was independent of childhood allergy, was found for pre-natal but not post-natal maternal cytokine levels, was evident in children of non-asthmatic but not in children of asthmatic mothers and held for maternal but not paternal cytokines. 36 Ongoing work is currently investigating whether this relationship involves the neonatal epigenome. Unpublished data show that a substantial proportion of the neonatal differential methylation associated with the maternal cytokine ratio is also associated with childhood asthma, thereby pointing to an epigenetic trajectory that connects immune responses in the mother to asthma in her child.

| Epigenetic predictors of childhood asthma at birth
Epigenetic studies focus on how a cell or a tissue become poised to express a given set of genes -that is, on a trajectory to events that will occur at a later time, thanks to epigenetic programming. is a well-replicated asthma risk gene, 5 and a master regulator of TGFβ-dependent signalling. In this capacity, SMAD3 is uniquely positioned to affect the trajectory to, and the pathogenesis of, childhood asthma. 38 Additionally, SMAD3 and TGFβ are expressed at high levels during early lung development regulating branching morphogenesis, epithelial cell differentiation and maturation of surfactant synthesis. 39 The identification of SMAD3 as an epigenetic predictor of childhood asthma at birth suggests that the early epigenetic trajectory to asthma proceeds at the intersection between immune regulation and lung development. In Germany, children born to participants of the International Study of Asthma and Allergies in Childhood are being recruited in a multigenerational study with information also collected from the parents of the original cohort and DNA collected for epigenetic analysis (ACROSSOLAR study). 49 In the UK, the ALSPAC study is recruiting both the parents and the offspring of the original birth cohort, thus providing information on three generations, two being the birth cohorts. 50 Asthma, allergy and obesity are (among others) the disease foci and both genetic and epigenetic mechanisms are being studied.
In California, the Children's Health Study (CHS) originally recruited school-age children and their parents for a study of the respiratory health effects of air pollution. 51,52 CHS subjects, now grown, are being recontacted and their own children recruited into a follow-up study to evaluate epigenetics across three generations. Data from these studies will enhance our knowledge of the extent of transgenerational effects. Evidence for effects of grandparental smoking on child health is also emerging ( Table 1). 56 Recent publications suggest grandmother's smoking may increase risk of asthma in the child independent of whether either parent smoked. 57-60 Father's smoking at young ages may also increase risk of asthma and low lung function in their future children. 4,58,61 One mechanism that can help explain inherited risk across generations is an altered epigenome in response to pollutant exposures. 62 Several epigenetic models of inheritance have been proposed in support of this hypothesis and are discussed below though many aspects of these models are as yet untested. 2

| Multi-generation studies based on adult cohorts
There are very few three-generation epidemiological studies re-

| The importance of puberty in men on risk in offspring
Analyses of the RHINESSA/RHINE/ECRHS cohorts suggest, consistently, that father's exposures in early puberty may be of key importance for asthma and lung function in the offspring. 4,58,70,71 In the RHINESSA cohort, father's smoking before age 15 years was strongly and consistently associated with early-onset offspring's non-allergic asthma, while duration of smoking, but not time of quitting, was of some importance. Further, a doubled risk of asthma was found for offspring of fathers with an occupational history of welding for at least 10 years before conception. Asthma was further associated with paternal grandmother's smoking -which, curiously, significantly modified the associations with father's smoking, totally attenuating effects of father's early onset smoking, but enhancing effects of father's later onset smoking or welding.
Association of father's smoking <15 years was replicated in a multigeneration analysis of the ECRHS cohort. Statistical methods were developed to account for the complexity in multi-centre multigeneration data, including simulation models that found that the estimated effect of unaccounted-for confounding was low. 72 Using more advanced statistical models for causal inference from observational data, father's smoking before age 15 years was a causal factor also for low lung function in offspring. 61 This was found both for FEV 1 and for FVC, suggesting effects on lung growth as well as airways calibre.
Further support for an important susceptibility window in early male puberty was obtained by a study of parents' puberty onset overweight. Johannessen et al. found that father's becoming overweight before voice break but after age 8 years was causally related to asthma in future offspring. 70 This effect was not mediated by offspring's own weight, nor were there significant effects of father's being overweight after puberty or mother's being overweight before conception. A recent study found that father's prepuberty onset overweight also was associated with lower lung function in offspring, and with considerably lower adult height in sons. 73 These observations have been supported by findings in Tasmanian Longitudinal Health Study showing association between paternal BMI trajectories from childhood to adolescence and asthma in their future offspring. 74 These analyses consistently suggest that there are significant windows of environmental exposure, such as male puberty and prior to conception that may be important with regard to offspring respiratory health. It seems plausible that male puberty could be a time window of higher susceptibility to environmental exposures due to epigenetic reprogramming during that period. 10,75 In addition to these studies that point to a role of the father's preconception environment, in particular during prepuberty, there are studies pointing to the mother's preconception. 76 A recent study identified higher asthma risk in offspring associated with mother's occupational exposure to cleaning products and disinfectants. 27 However, some of these methods do not address inter-connections between CpG sites such as methods for DMRs and those to assess mediation effects of a set of CpG sites. 77 (iv) Future studies should also investigate the functional consequences of these epigenetic signatures, transferred across generations, which emphasized the unmet needs for novel epigenetic editing tools.

| DATA ANALYS IS IN EPI G ENE TIC RE S E ARCH AND CRITI C AL RE VIE W OF DATA
(v) A challenge in statistical methods to examine epigenetic effects in transgenerational or intergenerational studies is to properly incorporate underlying biological mechanisms into the design of analytical methods.
(vi) An important question to ask is whether it is possible to intervene to modify the epigenome with postnatal exposures (treatments) to prevent disease?
(vii) These knowledge gaps will be important to address in future studies as we seek to better understand the risk factors for asthma and allergic disease and devise optimal public health strategies to prevent these chronic conditions to improve the health of the population.
based on pairwise correlations between CpGs in DNA methylation.
Because of these, inferences may be incomplete, biased or even misleading. Directed networks, e.g., via directed acyclic graphs, among CpGs have the potential to assess causality between epigenetic sites, but methods with the ability to address epigenetic changes over time are in great need.
As with all genome-scale data, DNA methylation data need to go through pipelines of quality control and pre-processing. Although multiple methods have been proposed, all these approaches in general involve a combination of quantile normalization of raw DNA methylation data followed by approaches to remove batch effects and technical variations. [84][85][86][87] In some situations, these pre- Methods have been proposed to infer cell type proportions, 85,88,89 which are then included in data analyses to adjust for the effects of cell heterogeneity. However, these methods are sensitive to the selected reference database used as the standard. 90 Methods not relying on reference databases, mostly based on principal component analyses or factor analyses, have been proposed. 89 Novel methods with the ability to infer cell type proportions, which are robust against the effect related to the choice of reference databases, are greatly needed. This variation in epigenetic factors among cells and tissues as a consequence of cellular differentiation is a major limitation of current epigenetic epidemiological studies. Usually blood or buccal/saliva DNA is used for assessment of epigenetic profiles due to ease of sampling. However, it is still unclear to what extent preconceptional or early-life exposure associated differences in methylation in blood are also observable in other disease relevant tissue, e.g., lung.
To improve statistical power in genome-scale studies, screening of methylation sites is usually performed prior to data analyses.
Statistical approaches have been proposed to filter out irrelevant or non-informative CpG sites. 91,92 To avoid data double dipping, statistical models for screening should be different from the models to be applied in final analyses (Box 1).

| SUMMARY
Recent evidence indicates that grand parental health status, nutri-

ACK N OWLED G EM ENTS
The authors thank National Institute of Allergy and Infectious Diseases, National Institutes of Health for organizing and supporting the workshop. Dr. Wheatley's and Dr. Togias' co-authorship of this report does not constitute endorsement by the U.S. National

Institute of Allergy and Infectious Diseases or by any other United
States government agency."

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interest with regard to this publication.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.