Selective advantages, selective mismatches, chronic inflammation, obesity, and type 2 diabetes: a new hypothesis to explain the effects of natural selection on genetic variation at the ectodysplasin A receptor locus

Abstract

Development of ectodermal tissues in humans is a complex process regulated by interacting signaling pathways. This review adopts an evolutionary medicine framework to examine how genetic variation in components of the ectodysplasin A (EDA) signaling pathway may have conferred human evolutionary fitness during the last ice age, but today may present an evolutionary mismatch, contributing to modern chronic diseases. A missense substitution at amino acid 370 in the human EDA receptor (EDAR) gene (EDARV370A), which is very common in East Asian, Native American, and Latino populations, produces pleiotropic effects in ectodermal tissues such as hair, teeth, mammary glands, and skin appendages. Several hypotheses have been offered to explain the high frequency of EDARV370A, proposing selective advantages related to thermal regulation, mother-infant vitamin D transmission, and/or tooth shape. However, new insight suggests that the selective advantage could have involved EDAR’s systemic and epithelial immune effects rather than embryonic effects alone. We explore the possibility that inflammatory effects associated with EDARV370A were the target of selection approximately 20 000 years ago, but that this selective advantage could now represent an evolutionary mismatch in the context of abundant caloric supply, resulting in chronic inflammation and a higher risk of metabolic syndrome. Finally, we outline directions for future research.

Introduction

Development of ectodermal tissues in humans is a complex process regulated by several interacting signaling pathways, including wingless-related integration site (WNT), ectodysplasin A (EDA), retinoic acid, Notch, Sonic Hedgehog (SHH), bone morphogenetic protein, and fibroblast growth factor (FGF).^{1 2} This review takes a novel evolutionary medicine approach to explain how genetic variation in components of the EDA signaling pathway, specifically the autosomal EDA receptor (EDAR) that binds EDA1, may affect selective fitness and, as an evolutionary mismatch, contribute to modern chronic diseases. The pathways involved in EDA signaling were first characterized through their involvement in hypohidrotic ectodermal dysplasias (HED),^3–9 characterized by various degrees of underdevelopment of skin and skin appendages. Tissues that are affected include sweat glands, sebaceous glands, and hair follicles, dentition, facial bone structure, and sometimes immune deficiencies.^{3 6}

Research into the underlying genetic variants involved in ectodermal dysplasias revealed that the majority are X-linked due to the location of two genes on the X chromosome: the EDA gene and one of the receptors (EDA 2 receptor (EDA2R)).^3–11 It is remarkable that as early as 1875, Charles Darwin is credited with the first description of a family with HED in which the pattern of inheritance was clearly X-linked: “I may give an analogous case, communicated to me by Mr. W. Wedderburn, of a Hindoo family in Scinde, in which ten men, in the course of four generations, were furnished, in both jaws taken together, with only four small and weak incisor teeth and with eight posterior molars. The men thus affected have very little hair on the body and become bald early in life. They also suffer much during hot weather from excessive dryness of the skin. It is remarkable that no instance has occurred of a daughter being thus affected; and this fact reminds us how much more liable men are in England to become bald than women. Though the daughters in the above family are never affected, they transmit the tendency to their sons; and no case has occurred of a son transmitting it to his sons. The affection thus appears only in alternate generations, or after longer intervals. There is a similar connection between hair and teeth, according to Mr. Sedgwick”.¹² This early observation came before the discovery of X-linked EDA signaling and revealed how closely hair, teeth, and skin development are linked. Autosomal forms of HED have also been reported,^{3 13 14} and some are due to variation in the autosomal EDAR on chromosome 2 that is functionally linked to X-linked EDA through their signaling pathway. We will focus our attention on EDAR because it contains a common, non-pathological functional variant that was the target of strong positive selection, as discussed below.

EDA1/EDAR signaling

Ectodysplasin signaling is initiated via two ligands, ectodysplasin (EDA) 1 and ectodysplasin 2, which bind in a highly specific manner to two receptors, the EDAR and the X-linked EDA2R. EDA1 and EDA2 are isoforms of the same X-linked EDA gene, differing by only two amino acids but having complete identity elsewhere in their sequences. Despite this seemingly small difference, EDA1 is highly specific for EDAR, and EDA2 is specific for EDA2R. EDAR and EDA2R are both members of the tumor necrosis factor (TNF) superfamily of receptors, and EDA1/2 are homologous to tumor necrosis factor-alpha (TNFA). Like the TNFA signaling system, these receptors trimerize in the plasma membrane and bind to EDA1/2 trimers via three extracellular cysteine-rich domains to initiate signaling cascades that primarily lead to activation of NF-κB signaling and other pathways.^{4 7 9 15}

We focus here on the EDA1-EDAR signaling network because a specific variant of EDAR occurred approximately 35 000 years ago in a human population living in northeastern Asia that was then positively selected during the last ice age, approximately 20 000 years ago.^16–19 Our goal is to explore the phenotypic effects that led to this evolutionary advantage, but before we dive into this, we need to consider this specific signaling network in more detail.

The EDA1-EDAR signaling network (figure 1) regulates ectodermal tissue development and influences the formation and pattern of skin and skin appendages such as hair follicles, sweat and sebaceous glands, teeth, and mammary glands.^{4 7 9 15} The EDAR pathway is characterized by unique receptor/adaptor interactions. EDA1 is present as a full-length transmembrane protein and as a soluble form that is released from the plasma membrane by cleavage at a furin site. EDA1 contains the TNF-homology domain, assembles as a homotrimer, and specifically binds to EDAR, a transmembrane receptor belonging to the TNF receptor superfamily. Either soluble EDA1 or transmembrane EDA1 can bind to the extracellular cysteine-rich domains of EDAR, allowing for paracrine or ‘juxtacrine’ signaling.^{2 5 9 20} EDAR trimerization occurs on EDA1 binding, which allows EDAR to recruit the adaptor protein EDAR-associated death domain (EDARADD). EDARADD acts as a bridge to cytoplasmic signaling, in this case the TNF receptor associated factor 6 (TRAF6)-mitogen-activated protein kinase kinase kinase 7 (TAK1) complex, which initiates downstream cascades. The engagement of EDAR with EDARADD and TRAF6 activates TAK1 (MAP3K7), a serine/threonine kinase that phosphorylates the inhibitor of nuclear factor-κB (IκB) kinase (IKK) complex (IKKα, IKKβ, and NEMO). This event triggers phosphorylation and degradation of IκBα, releasing NF-κB dimers (p65/p50) for nuclear translocation. Once translocated to the nucleus, NF-κB regulates transcriptional programs involved in ectodermal development, epithelial signaling, and inflammatory responses. These include expression of morphogenetic regulators such as WNT10A and SHH, as well as cytokine and antimicrobial peptide genes in skin epithelium (keratinocytes in the epidermis). Parallel signaling occurs through mitogen-activated protein kinases (c-Jun N-terminal kinase and p38), influencing keratinocyte differentiation and apoptosis resistance. EDA1/EDAR also interacts with the WNT, SHH, TGFB, and FGF signaling pathways to form an integrated developmental network that regulates ectodermal tissues.^{1 2 5 9 21 22}

Figure 1. Schematic of EDAR signaling. EDA1, ectodysplasin A1; EDAR, ectodysplasin A receptor; EDARADD, EDAR associated via death domain; IkB, nuclear factor kappa B inhibitor alpha; IKK1, component of inhibitor of nuclear factor kappa B kinase complex; IKK2, inhibitor of nuclear factor kappa B kinase subunit beta; NEMO, inhibitor of nuclear factor kappa B kinase regulatory subunit gamma; p50, nuclear factor kappa B subunit 1; p65, RELA proto-oncogene; TAB2, transforming growth factor-beta activated kinase 1 (mitogen-activated protein kinase kinase kinase) binding protein 2; TAK1, mitogen-activated protein kinase kinase kinase 7; TRAF6, tumor necrosis factor receptor associated factor 6.

Non-pathological variation in the EDAR gene

Although loss of function mutations in the EDAR gene is responsible for some forms of autosomally inherited HED, for example,¹³ a missense mutation that substitutes an alanine for a valine at position 370 in the human amino acid sequence of EDAR (often termed EDARV370A) is very common in East Asia and approaches fixation in many Native American populations.²³ Genetic analyses indicate that the ‘V’ allele is the ancestral allele and predominates in European and African populations. The EDARV370A allele arose in Northeast Asia around 35 000 years ago and was subject to strong positive selection that may have taken place between 26 500 and 19 000 years ago in the Arctic, specifically on or near the landmass that is known as Beringia, lying between what is now Siberia and Alaska.^{16 18 23} As the climate moderated, people living in Beringia migrated southward into the Americas and into Asia, giving rise to the high frequency of EDARV370A in those populations. This variant has proven to be a gain-of-function mutation in experiments in which the derived EDARV370A and ancestral isoforms were ectopically expressed in HEK293 cells along with an Nf-κB promoter/firefly luciferase reporter construct; the EDARV370A isoform activated NF-κB activity to a greater degree.^{7 18 24} While these findings support increased EDAR signaling capacity, NF-κB reporter assays provide only an indirect measure of biological function, and the physiological consequences of EDARV370A in vivo remain incompletely understood.

Phenotypic effects of variation in EDA signaling

The EDA signaling pathway is highly conserved across vertebrates, reflecting its essential role in the proper formation of ectodermally derived organs during embryogenesis,⁹ with numerous examples from humans and mice,^25–27 stickleback fish,²⁸ zebrafish,²⁹ sculpins,³⁰ and marine mammals.³¹

Transgenic mouse studies of the human-specific V370A variant, the target of strong natural selection approximately 20 000 years ago, reveal increased hair thickness, increased eccrine gland number, reduced mammary fat pad, increased mammary gland ductal branching,¹⁸ as well as enlarged sebaceous and Meibomian glands, and increased branching in salivary and mammary glands.³² Considerable experimental mouse developmental research has focused on understanding the role of EDA signaling in tooth development. Loss of function in two genes essential to the pathway’s activity, EDA or EDARADD, results in the development of a smaller tooth enamel knot^33–36 that is rope-like and elongated,^{36 37} and leads to fewer and smaller cusps and teeth.^{36 38 39}

Association studies in human populations demonstrate that EDARV370A influences a variety of phenotypes, including earlobe and chin shape, hair straightness, beard thickness, eccrine sweat gland density, lower breast density, and multiple dental traits: incisor shoveling, mesiodistal tooth diameter of anterior teeth, mandibular molar hypoconulid expression, and root morphology.^{18 40–49} Previously published analyses from the Arizona Insulin Resistance Registry (AIR) and Sangre Por Salud (SPS) cohorts identified associations between EDARV370A and several metabolic syndrome-related traits, including glycemic, lipid, and breast density traits (table 1).^{40 50 51}

Table 1. Previously reported associations between EDARV370A and selected metabolic and pleiotropic traits in Latino cohorts.

Cohort	Trait	Val/Val	Val/Ala	Ala/Ala	Adjusted effect size (95% CI)	Adjusted p value
AIR	Triglycerides (mg/dL)	120.0±63.9	136.5±74.0	157.8±97.7	β=16.5 (7.5 to 25.6)	0.00039
AIR	Very Low Density Lipoprotein (mg/dL)	20.1±10.7	22.2±11.0	23.4±10.6	β=1.57 (0.27 to 2.87)	0.018
AIR	Alanine aminotransferase (IU/L)	24.0±14.9	26.2±16.7	30.7±20.9	β=2.75 (0.72 to 4.78)	0.00807
AIR	2-hour glucose (mg/dL)	128.1±44.9	136.8±48.0	148.6±54.4	β=9.9 (3.98 to 15.86)	0.001137
AIR	Prediabetes (%)	29.2	38.0	42.7	aOR=1.36 (1.00 to 1.85)	0.05
AIR	Diabetes (%)	12.4	13.3	21.9	aOR=1.67 (1.12 to 2.48)	0.0119
SPS	Hemoglobin A1c (%)	5.98±1.11	6.16±1.35	6.20±1.45	β=0.13 (0.02 to 0.24)	0.0167
SPS	Breast density value	2.46±0.61	2.43±0.61	2.30±0.53	β=−0.09 (−0.16 to −0.03)	0.0066

Values are presented as mean±SD. Regression coefficients (β) and aORs represent additive effects per copy of the rs3827760 derived allele (EDARV370A). Continuous traits were analyzed using linear regression, whereas prediabetes and diabetes were analyzed using logistic regression. All analyses were adjusted for age, sex, and BMI, except breast density analyses, which were adjusted for age and BMI only because all participants were women. AIR and SPS results are reproduced from previously published analyses. Breast density values were obtained from mammography reports and categorized as: (1) almost entirely fatty; (2) scattered fibroglandular density; (3) heterogeneously dense and (4) extremely dense. Breast density categories were modeled as an ordinal numeric variable for exploratory analysis. El Banco Por Salud results are presented separately in tables 2 and 3.

AIR, Arizona Insulin Resistance Registry; aOR, adjusted ORs; BMI, body mass index; SPS, Sangre Por Salud Biobank.

Because type 2 diabetes and metabolic syndrome are highly prevalent in populations with substantial Native American ancestry, and because EDARV370A underwent strong positive selection in ancestral Native American and East Asian populations, interpretation of these associations is challenging due to the strong correlation between EDAR genotype and ancestry. Nevertheless, broadly consistent directions of effect were observed across these independent Latino cohorts despite relatively modest sample sizes (table 1).

To further investigate these observations, analyses were performed in El Banco Por Salud (EBPS),⁵² a cohort in which ancestry estimates were available, enabling adjustment for Native American ancestry. The methods for these analyses can be found in the online supplemental methods section. EBPS included the Centers for Disease Control and Prevention/Agency for Toxic Substances and Disease Registry Social Vulnerability Index (SVI) data, allowing adjustment for social vulnerability in the analyses. Analyses also evaluated whether associations differed according to diabetes status (table 2; online supplemental table 1). Because previously reported associations in AIR and SPS involved glycemic phenotypes, including hemoglobin A1c (HbA1c) (table 1), additional EBPS analyses were conducted to explore potential physiological mechanisms underlying these observations. Because HbA1c reflects the integrated effects of both insulin secretion and insulin action, homeostatic model assessment of insulin resistance (HOMA-IR) and fasting C-peptide were examined as surrogate markers of insulin resistance and endogenous insulin secretion, respectively.

Table 2. Sequential ancestry-adjusted and social vulnerability-adjusted analyses of metabolic phenotypes in El Banco Por Salud (n=844).

Trait	Model	Adjusted β (95% CI)	Adjusted p value
log(Hemoglobin A1c)	Age + sex + BMI	0.028 (0.003 to 0.053)	0.029
	Age + sex + BMI + Native American ancestry	0.017 (−0.010 to 0.043)	0.216
	Age + sex + BMI + Native American ancestry + SVI	0.017 (−0.010 to 0.043)	0.213
log(HOMA-IR)	Age + sex + BMI	0.043 (−0.035 to 0.121)	0.277
	Age + sex + BMI + Native American ancestry	0.023 (−0.059 to 0.105)	0.584
	Age + sex + BMI + Native American ancestry+SVI	0.024 (−0.058 to 0.106)	0.565
log(C peptide)	Age + sex + BMI	−0.045 (−0.112 to 0.022)	0.188
	Age + sex + BMI + Native American ancestry	−0.053 (−0.123 to 0.018)	0.142
	Age + sex + BMI + Native American ancestry + SVI	−0.051 (−0.121 to 0.019)	0.156

Regression coefficients (β) represent additive effects per copy of the rs3827760 derived allele (EDARV370A). Model 1 was adjusted for age, sex, and BMI. Model 2 additionally adjusted for Native American ancestry proportion. Model 3 additionally incorporated the Centers for Disease Control and Prevention/Agency for Toxic Substances and Disease Registry SVI overall ranked percentile metric (RPL_THEME), which reflects cumulative social vulnerability across multiple domains. HbA1c, HOMA-IR, and C peptide were log-transformed prior to analysis. Analyses were performed in participants with complete data for genotype, HbA1c, HOMA-IR, C peptide, age, sex, BMI, Native American ancestry, and SVI (n=844). Because EDARV370A is highly correlated with Native American ancestry due to historical positive selection, ancestry-adjusted analyses should be interpreted cautiously because of structural collinearity between genotype and ancestry.

BMI, body mass index; HbA1c, hemoglobin A1c; HOMA-IR, homeostatic model assessment of insulin resistance; SVI, Social Vulnerability Index.

Additional analyses in EBPS suggested that the relationship between EDARV370A and metabolic phenotypes may be influenced by both ancestry and diabetes status. In the full cohort, the association between EDARV370A and HbA1c was attenuated after adjustment for Native American ancestry (table 2). No significant associations were observed with HOMA-IR. Similarly, no significant association was observed between EDARV370A and C-peptide in the full cohort (table 2), although effect estimates remained directionally consistent across adjustment models.

Because diabetes prevalence differed across genotype groups in the EBPS cohort (online supplemental table 1), diabetes-stratified analyses were also performed. These analyses revealed distinct patterns, with associations involving HbA1c most evident among participants without diabetes. In contrast, among participants with diabetes, each copy of the EDARV370A allele was associated with lower fasting C-peptide levels after adjustment for age, sex, BMI, Native American ancestry, and SVI (table 3). These findings suggest that any metabolic effects of EDARV370A may be context-dependent and influenced by disease state, as associations with glycemic control and endogenous insulin secretion differed according to diabetes status. Adjustment for SVI produced little change in the estimated associations despite the uniformly high levels of social vulnerability observed across the cohort (online supplemental table 1), likely reflecting the relatively homogeneous socioeconomic characteristics of participants recruited from federally qualified health centers. However, given the modest sample sizes and exploratory nature of these analyses, these observations should be considered hypothesis-generating and require replication in independent cohorts.

Table 3. Diabetes status-stratified analyses of metabolic phenotypes in the El Banco Por Salud cohort.

Trait	Diabetes status	Adjusted β (95% CI)	Adjusted p value
log(HbA1c)	All participants	0.017 (−0.010 to 0.043)	0.213
	Without diabetes	0.014 (0.005 to 0.023)	0.003
	With diabetes	0.010 (−0.018 to 0.037)	0.489
log(HOMA-IR)	All participants	0.024 (−0.058 to 0.106)	0.565
	Without diabetes	0.022 (−0.083 to 0.127)	0.682
	With diabetes	0.009 (−0.107 to 0.125)	0.878
log(C peptide)	All participants	−0.051 (−0.121 to 0.019)	0.156
	Without diabetes	0.023 (−0.063 to 0.108)	0.601
	With diabetes	−0.114 (−0.219 to −0.010)	0.032

Regression coefficients (β) represent additive effects per copy of the rs3827760 derived allele (EDARV370A). All models were adjusted for age, sex, BMI, Native American ancestry proportion, and the Centers for Disease Control and Prevention/Agency for Toxic Substances and Disease Registry SVI overall ranked percentile metric (RPL_THEME). HbA1c, HOMA-IR, and C peptide were log-transformed prior to analysis. Diabetes-stratified analyses were performed in the same analytic dataset used for table 2 (n=844), including 483 participants with diabetes and 361 participants without diabetes.

BMI, body mass index; HbA1c, hemoglobin A1c; HOMA-IR, homeostatic model assessment of insulin resistance; SVI, Social Vulnerability Index.

The diabetes-stratified findings observed in EBPS were also consistent with analyses performed across the combined AIR, SPS, and EBPS cohorts (online supplemental table 2). In the pooled dataset, EDARV370A was associated with HbA1c after adjustment for age, sex, and BMI, and these associations were driven primarily by participants without diabetes. In contrast, no significant associations with HbA1c were observed among participants with diabetes. Although exploratory, these findings provide additional support for the possibility that the metabolic effects of EDARV370A may differ according to disease state.

Although EDARV370A also occurs at high frequency in East Asian populations, relatively few studies have examined its relationship with metabolic traits in these populations. Some large biobank studies, including Biobank Japan, KoGES, and FinnGen, have reported modest associations with metabolic and cardiometabolic phenotypes (online supplemental table 3). Taken together, the available evidence suggests that any metabolic effects of EDARV370A are likely modest and context-dependent. We do not propose that EDARV370A universally increases metabolic disease risk across populations; rather, any phenotypic effects of EDAR variation are likely modified by environmental, dietary, inflammatory, and socioeconomic factors.

Importantly, the strong correlation between EDARV370A and Native American ancestry creates structural collinearity between genotype and ancestry, limiting the ability of epidemiologic analyses alone to fully disentangle their independent effects. These findings therefore highlight the need for mechanistic studies directly examining the effects of EDAR variation on inflammatory pathways and their potential downstream metabolic consequences.

Natural selection acting on the EDARV370A allele

Because of the genetic signature of natural selection having operated on this allele around 20 000–25 000 years ago, several investigators have proposed hypothetical advantages for this allele in late Pleistocene environments. Chang et al³² proposed that increased glandular size could have been beneficial in providing greater humidification during respiration in the drier environment of the last ice age. Kamberov et al¹⁸ proposed two hypotheses. The first suggested that a higher density of eccrine glands may have advantaged evaporative cooling in hunter-gatherers during humid ice age summers in eastern Asia, and the second focused on the mammary gland effects and the potential functional consequences or effects on mate choice. Another suggestion is that the dentognathic effects were under positive selection for supposed advantages in masticating wild rice and millet.⁵³

Several of the coauthors of this paper relied on the effects of EDAR V370A on teeth and more specifically on incisor shoveling to survey the archeological record to identify more specifically the location of the population that experienced selection on this allele.²³ Our study indicates that the selection most likely occurred on a population living in or near Beringia, in the Siberian Arctic. Isolated by the environmental changes associated with the aridification of the last ice age, the V370A allele appears to have been brought to near fixation. At such a high latitude, Hlusko et al²³ concluded that the very low level of exposure to ultraviolet radiation, and the subsequent inability to biosynthesize vitamin D, would have been a powerful selective agent. Consequently, nursing infants could only obtain vitamin D through mothers’ milk, leading to the hypothesis that the increased ductal branching of the mammary glands arising from EDAR’s effects during embryogenesis may have facilitated nutrient transfer from mother to infant in a vitamin D-challenged environment (an environment that would also have compromised the function of the surrounding breast adipose tissue). However, it also is conceivable that more than one selective pressure operated on EDARV370A.

As the geographic origin of selection for EDARV370A has become clearer, so too has our understanding of its present-day frequency. The variant is common in East Asian and Native American populations, reflecting shared ancestry from ancient Beringian populations that were among the first to expand into these regions as the ice age ended.⁵⁴ But was the selective pressure that confronted that ancient population selecting for EDAR’s effects on embryonic development or on its more pervasive systemic effects? Prior hypotheses have all focused on the former; here, we explore the latter.

A new hypothesis: higher signaling activity of EDARV370A predicts enhanced inflammatory responses in skin to environmental pathogens

Recently, Hlusko and McNellis hypothesized that because the EDAR and the TNFA signaling pathways are highly homologous, both resulting in NF-κB-mediated gene transcription events, increased EDA1/EDARV370A signaling might confer a higher immune/inflammatory response.⁵⁵ This expands the range of phenotypes that could have contributed to the recent selective sweep. Importantly, any immune or inflammatory effects associated with EDARV370A may represent pleiotropic consequences of selection on the allele rather than the primary target of positive selection. Regardless of their evolutionary origin, such effects could still have important implications for inflammatory and metabolic phenotypes in present-day populations. Because skin is the largest immunologically active organ in the body and is central to the EDARV370A phenotype, altered epithelial immune signaling associated with EDAR activity could confer a selective advantage in pathogen-rich archaic environments, including the Arctic, in which modern hygiene and antibiotics were not present.^56–59

There is anecdotal evidence this could have happened. In recent history, before the advent of antibiotics around 1920, infections of the skin could be and often were fatal. For example, erysipelas, a streptococcal skin infection most often due to Streptococcus pyogenes and often related to scarlet fever, had a high mortality rate before the 1940s and the introduction of penicillin.⁶⁰ Although over time the reporting rate of erysipelas did not change in Norway between 1880 and the 1960s (about 10 cases per 10 000), the case fatality rate was about 25–40 per 1000 cases until the 1940s, after which it fell to near zero.⁶⁰

In modern populations living under relatively traditional conditions with lower levels of hygiene and poor access to medical care and antibiotics, there still is an extensive burden of pathogenic skin diseases on the ability of individuals to function normally. These diseases include, for example, tungiasis (Tunga penetrans), tinea capitis (dermatophyte scalp infection), onchocerciasis (onchocercal skin disease, or river blindness), yaws (endemic treponematosis), Buruli ulcer (Mycobacterium ulcerans), and leishmaniasis (leishmania parasites). These diseases can be severe and debilitating under the living conditions where they are common.^61–64 Even in modern, industrialized societies, fungal, bacterial, parasitic, and viral skin diseases present a greater health burden where the Social Deprivation Index (SDI) and SVI are high.^{62 64} Although mortality from such diseases is low on a per-case basis, these diseases can have significant effects on individuals, often preventing them from living a normal life, at least temporarily, as evidenced by 119 129 deaths and 41.9 million disability-adjusted life years due to skin and subcutaneous diseases in 2021.⁶⁴ Thus, there is abundant evidence for high morbidity in pathogenic skin disease prior to about 80–100 years ago, and this is still especially evident in populations living under conditions of low availability of modern medical care and lower standards of hygiene than those that have become common in economically advantaged societies.

In this context, the availability of SVI measures in the EBPS cohort is notable. Although adjustment for SVI produced little change in the estimated associations, social vulnerability was uniformly high across the cohort, limiting the ability to fully assess the contribution of socioeconomic and environmental factors. Nevertheless, broader environmental and social conditions may modify the phenotypic consequences of EDAR-related inflammatory signaling. Such interactions may be particularly relevant when considering evolutionarily selected alleles in modern obesogenic and socioeconomically heterogeneous environments.

EDAR variation and inflammation

The skin is a first-line defense against external pathogens. For most of prehistory and history, humans lived in pathogen-rich environments in which the skin was continually challenged to prevent infection. Thus, EDARV370A, which is associated with altered EDA1-EDAR signaling and downstream transcriptional responses, may have been subjected to strong positive selection. Because TNFA receptor signaling and EDAR signaling share downstream NF-κB-mediated signaling components, there may be downstream convergence between these signaling systems that could lead to synergistic interactions and enhanced immune or inflammatory responses. However, NF-κB signaling is activated by many upstream pathways, and shared downstream signaling does not necessarily imply equivalent biological effects across receptor systems.

This issue is complex and will require both in vitro and in vivo approaches to gain a better view of the biology, but a possible relationship between metabolic syndrome (obesity, hyperglycemia, insulin resistance, and hypertension) and skin diseases, mediated by inflammation, has been suggested.⁶⁵ There is substantial evidence that TNFA receptor signaling in keratinocytes is involved in skin immune function,^{66 67} but EDAR’s well-described role is in its participation in the development of ectodermal tissues and programming of subsequent adult phenotypes. For example, EDAR (and EDA signaling in general) plays a central role in dental development, as evidenced by major changes in the dentition in patients with HED. Tooth formation and the pattern of traits of the dentition are mainly programmed during development and change little during the life of an individual, except for wear and tooth loss.

However, to understand selective forces that may have been exerted on this locus, it is necessary to examine the expression of the EDA1/EDAR axis during adult life (figure 2). Are phenotypes produced during development the only substrate for natural selection in the adult, or is there sufficient expression of EDAR in adult tissues to alter ectodermal tissue phenotypes, such as potential effects on immunity and inflammation in the adult? In the Genotype-Tissue Expression Atlas portal,⁶⁸ EDA is expressed across many human tissues, with the highest transcript abundance observed in heart muscle and adrenal gland (~7–9 nTPM). Reproductive, epithelial, and endocrine tissues, including fallopian tube, ovary, thyroid, cervix, adipose tissue, skin, and pancreas, show intermediate levels (~3–4 nTPM). Expression is lower in metabolic tissues such as kidney, liver, and skeletal muscle (~1.5–3 nTPM), and minimal in the brain (<1 nTPM). However, these data do not distinguish between EDA1 and EDA2, both products of the EDA gene.

Figure 2. mRNA expression of elements of EDAR signaling (top panel) and TNFRSF1A signaling (bottom panel) using bulk RNA expression data for skin, breast, and whole blood derived from the GTEx Portal. EDAR, ectodysplasin A receptor; GTEx, Genotype-Tissue Expression Atlas; TPM, transcripts per million.

EDAR expression is enriched in epithelial barrier tissues, with the highest levels in the esophagus (~3.5 nTPM) followed by the urinary bladder, vagina, and skin (all ~1.4 nTPM), and then cervix (~0.9 nTPM). Expression is much lower in other tissues and undetectable in the brain regions.

EDARADD shows the highest transcript levels in stomach and urinary bladder (~3.5–3.8 nTPM), followed by skin and esophagus (~2.5–3.0 nTPM). Moderate expression is observed in the testes, thyroid, prostate, pancreas, and breast (~1.7–2.3 nTPM). Expression in metabolic tissues including kidney, liver, and adipose tissue is relatively low (<1.5 nTPM), and most brain regions show minimal or nearly undetectable expression (<0.5 nTPM). Overall, EDARADD is enriched in epithelial and mucosal barrier tissues, consistent with its role as an adaptor protein in the EDA-EDAR signaling pathway. Taken together, these data indicate that the three proximal elements of EDAR signaling are expressed in adult tissues. However, in comparison with homologous signaling elements of the TNFA signaling pathway, EDAR and EDARADD expression levels are much lower than TNFRSF1A or TNF Receptor-Associated Death Domain (TRADD, note the scale), suggesting that EDAR signaling in adult tissues may be too low to have much activity (figure 2). In this case, phenotypes related to EDARV370A likely are programmed during development and are the result of this programming interacting with the developmental and adult environment. However, this does not rule out the possibility that EDAR signaling could be highly compartmentalized and still have some activity in specialized cells in the adult, especially if these genes can be activated, perhaps by demethylation, under particular circumstances, such as during wound healing or response to pathogens. Parenthetically, TNFA expression in skin and breast is lower than EDA expression. So, it is possible that EDA-mediated signaling may be more paracrine or juxtacrine in nature, but TNF may signal in a more remote endocrine manner.

These expression patterns also clarify an important limitation of our hypothesis. We do not propose that EDARV370A directly alters glucose metabolism through high expression in classical metabolic tissues such as skeletal muscle, adipose tissue, and liver. Rather, we propose that EDAR variation may influence metabolic phenotypes indirectly through effects on epithelial and barrier tissue immune biology. The skin is an active immunologic organ capable of producing cytokines and other inflammatory mediators that may influence both local and systemic inflammatory responses. In the setting of obesity or other chronic environmental stressors, persistent epithelial immune activation may contribute to low-grade inflammation associated with insulin resistance and cardiometabolic dysfunction. Such inflammatory pathways, including TNFA and interleukin-6 (IL-6) mediated signaling, are well-recognized contributors to obesity-associated metabolic dysfunction. Thus, any relationship between EDARV370A and metabolic disease is likely indirect and context-dependent rather than the result of direct EDAR activity in metabolic organs. These possibilities remain speculative and require direct mechanistic testing.

A selective mismatch?

Considering the possibility that EDARV370A was subject to a recent selective sweep related to a higher immune response in people bearing the EDARV370A allele, what might be the consequences of a higher state of skin inflammation, particularly if such inflammation extended systemically? Substantial evidence indicates that chronic, systemic inflammation may contribute to the pathogenesis of chronic diseases that plague people living in Western societies, such as obesity, type 2 diabetes, and related cardiometabolic risk.^{69 70} Even though modern medical care can prevent most of the debilitating effects of skin inflammation, humans are still challenged by many pathogenic skin microbes that regularly produce immune responses, and thus people bearing the EDARV370A allele may still have higher levels of skin inflammation, even if antibiotics prevent many of the worst outcomes. If such inflammatory signaling contributed to broader systemic inflammation, it could potentially influence pathways relevant to insulin resistance, insulin secretion, obesity, and type 2 diabetes. There also could be local inflammatory effects on subcutaneous adipose tissue. And in our current society, with chronic oversupply of calories, all of this could be exacerbated, leading to a vicious cycle. Such effects may differ substantially across populations and environments and may not be detectable under all genetic or environmental contexts. Research into the relationships between evolutionary mismatches, aging, altered glucose metabolism through effects on insulin action or secretion, and metabolic disease is expanding with the advent of precision and genomic medicine.⁷¹

Natural selection, evolutionary mismatches, and social determinants of health in the context of ecological niches

Modern societies are not homogeneous with respect to access to medical care, exposure to environmental insults (including pathogens), and modern hygiene practices. Additionally, the extent of variability in such disease risk factors is not evenly distributed throughout society. Such societal factors that affect disease risk are commonly referred to as social determinants of health (SDOH), quantified by measures including the SVI or the SDI, and are the subject of extensive research. We argue here that it is possible for a community to have both limited access to healthcare (low SDOH measures) and an abundant caloric supply (in the context of overall reduced diet quality), and we contend that the coexistence of these conditions may produce an evolutionary mismatch. Our ideas build on the concept of the thrifty genotype hypothesis first proposed over 60 years ago to similarly explain the uneven distribution of type 2 diabetes among the global population,^{72 73} but with updated biological knowledge and concepts, as advocated by Brassington et al.⁷⁴

Although SDOH factors producing different living conditions often are considered in the context of social or income inequality, they also have a biological context. This variation essentially characterizes distinct ecological niches within our society, that is, environments that can exert different selective pressures. For example, the EDARV370A variant may produce a phenotype where obesity and type 2 diabetes present a greater risk in one niche compared with another within the same society. It has not escaped notice that it may be useful to view these differences in a biological rather than, or in addition to, a social context. When considered as biology, it might help society avoid judgments regarding the origin of such differences and allow policymakers to focus on the factors within an environment that could be targeted to improve human health most effectively and efficiently. On the other hand, we should guard against a biologically deterministic point of view, recognizing phenotypes, in this case health outcomes, as the result of differences in both heredity and environment.

Directions for investigating connections between EDAR variation and inflammation

It is plausible that EDARV370A influences immune and inflammatory responses and that any selective advantage associated with this variant may have been manifested in the skin. Evidence from loss-of-function mutations in components of the EDA signaling pathway provides a rationale for this hypothesis. Individuals with HED frequently exhibit abnormalities of the skin and skin appendages and may experience an increased susceptibility to cutaneous infections. Although many papers describing skin manifestations do not distinguish among pathogenic variants affecting different components of the EDA signaling pathway, some studies do.⁷⁵ Increased susceptibility to infection may result from an impaired barrier function associated with reduced numbers of sweat and sebaceous glands, as well as altered immune and inflammatory responses. In contrast, the physiological consequences of a gain-of-function mutation such as EDARV370A are less clear. If EDARV370A is associated with a more robust immune response, it could have conferred a selective advantage in preantibiotic environments and might be expected to reduce the frequency or severity of certain skin infections. However, there is very little evidence that such enhanced innate or adaptive immunity occurs in the skin or other tissues of individuals carrying at least one EDARV370A allele. The pathway to gaining such evidence is reasonably clear. Latinos represent a particularly informative population in which to study these questions because their intermediate frequency of the A allele means that the frequencies of the two homozygous and heterozygous genotypes are all high, such that large numbers of participants are available for study.

An important limitation of the currently available human association data is the absence of direct inflammatory biomarker measurements, including C-reactive protein, TNFA, and IL-6. Future studies incorporating inflammatory biomarkers together with more detailed metabolic phenotyping will be necessary to evaluate the proposed relationship more rigorously between EDAR variation, inflammation, and metabolic disease. Regarding immune function of the skin, several approaches are feasible and would provide relevant data on the effects of EDARV370A. First, the presence of immune and immune-related cells could be compared among the genotypes using bulk and single-cell RNA-Seq of small punch skin biopsies. These transcriptomic analyses also provide information about other cell types that might be affected by inflammation, such as dermal fibroblasts or vascular cells. These studies could be complemented by immunocytochemistry applied to thin sections of skin biopsies. Larger skin specimens obtained from patients of different genotypes undergoing elective surgeries could also allow studies using flow cytometry to directly quantify immune cell presence and permit proteomic analysis that would supplement the transcriptomics results. Second, it would be possible to use skin explants in culture to gain experimental evidence of altered cytokine responses to pathogenic stimuli such as lipopolysaccharides or other products of pathogens that produce inflammatory responses. Third, the expression of inflammatory and immune-related genes could easily be addressed using transcriptomic analysis of circulating immune cells to determine if evidence for high skin inflammation is accompanied by systemic inflammation. Fourth, genotype-specific induced pluripotent stem cells could be generated from peripheral blood mononuclear cells and reprogrammed into keratinocytes or other relevant cell types to study these mechanisms in greater detail. Parenthetically, because mouse skin differs substantially from human skin, including distinct genomic responses to inflammatory diseases,⁷⁶ and because studies using human specimens are feasible and allow direct linkage to in vivo phenotypes, humans may represent the most appropriate experimental model for these questions. Finally, associations between EDARV370A and skin infectious diseases should be assessed in large-scale studies or biobanks that have an intermediate frequency of the V370A allele and in which skin infections remain prevalent.

Conclusions

The EDAR signaling system, including EDAR and EDA1, plays a central role in the development and perhaps maintenance of ectodermal tissues. In Asian-derived populations, a missense gain-of-function mutation, EDARV370A, has had an outsized role due to its pleiotropic effects on skin and skin appendages and other ductal epithelial tissues. Because of its effects on the morphology of the dentition, this variant has been of primary importance in informing our understanding of how and when the Americas were initially settled by humans. EDARV370A was the target of an intense selective sweep during the past 20 000 years; the phenotypic basis for this selection remains uncertain, but altered epithelial immune responses represent one plausible hypothesis. Although this type of immune response could have been advantageous in the past, an increase in systemic inflammation could now be an evolutionary mismatch in our modern world of caloric oversupply, in which bearers of this allele may, in some environmental contexts, have a higher risk of obesity, diabetes, and cardiovascular diseases. Substantial investigation remains necessary before the biological and biomedical effects of EDARV370A are fully understood, but the evolutionary history of this gene raises novel hypotheses regarding population-level variation in chronic disease risk.

The views expressed are solely those of the authors and do not necessarily reflect those of the European Union, the European Research Council, or the University of Arizona Health Sciences.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.