Genetics and Epigenetics of Atopic Dermatitis: An Updated Systematic Review.

Background: Atopic dermatitis is a common inflammatory skin disorder that affects up to 15–20% of the population and is characterized by recurrent eczematous lesions with intense itching. As a heterogeneous disease, multiple factors have been suggested to explain the nature of atopic dermatitis (AD), and its high prevalence makes it necessary to periodically compile and update the new information available. In this systematic review, the focus is set at the genetic and epigenetic studies carried out in the last years. Methods: A systematic literature review was conducted in three scientific publication databases (PubMed, Cochrane Library, and Scopus). The search was restricted to publications indexed from July 2016 to December 2019, and keywords related to atopic dermatitis genetics and epigenetics were used. Results: A total of 73 original papers met the inclusion criteria established, including 9 epigenetic studies. A total of 62 genes and 5 intergenic regions were described as associated with AD. Conclusion: Filaggrin (FLG) polymorphisms are confirmed as key genetic determinants for AD development, but also epigenetic regulation and other genes with functions mainly related to the immune system and extracellular matrix, reinforcing the notion of skin homeostasis breakage in AD.


Introduction
Atopic dermatitis (AD), also known as atopic eczema, is a common inflammatory skin disorder that affects up to 15-20% of children [1] and 7-10% of adults [2] in developed countries. AD typically develops during childhood and is characterized by recurrent eczematous lesions with intense itching [1]. It is considered the first step of the atopic march, associated with an increased risk of developing allergic rhinoconjunctivitis, asthma, or food allergy [3]. Worldwide, the prevalence of AD used to be higher in countries with higher incomes. However, due to the globalization process and a more westernized way of life, an increase of AD prevalence in low-income countries of Africa and East Asia has been reported, stressing the role of environment together with genetic and immunologic factors in the pathogenicity of AD [4].

Materials and Methods
This systematic review has been performed using the PRISMA guidelines for Systematic Reviews and Meta-Analysis 2009 checklist and GRADE recommendations [18].
Original articles and meta-analyses indexed from July 2016 to December 2019, describing genetic or epigenetic aspects of atopic dermatitis, were searched. We identified eligible studies using the following inclusion criteria: (1) primary study or meta-analysis, (2) written in English or Spanish, (3) human subjects, both children and adults, (4) patients suffering from atopic dermatitis or atopic eczema, and (5) studies describing mutations, single nucleotide polymorphisms (SNPs), or epigenetic modifications in association with disease onset, severity, or prevalence in the population. The exclusion criteria were: (1) animal, in vitro or in silico studies, (2) review articles, (3) proteomics or expression analysis without epigenetic/genotyping study, (4) articles focused in other diseases, such as psoriasis or ichthyosis vulgaris, in which AD was merely mentioned, and (5) articles whose full-text version was not available to us.
We performed the literature search between December 2019 and January 2020 in PubMed, Cochrane Library, and Scopus databases, using the following terms: "atopic dermatitis" OR "atopic eczema" AND "gene" OR "genetic" OR "mutation" OR "epigenetic" OR "DNA methylation" OR "sequencing" OR "microRNA" OR "polymorphism" OR "genome-wide association study" OR "microarray" OR "gene profiling".
Three authors individually reviewed the database search results, assessing titles, evaluating abstracts, and considering or not the study for full review. Any disagreements in either the title/abstract or the full manuscript review phases were resolved by consensus. All eligible studies were formally evaluated and included in this systematic review.
The authors independently evaluated the quality appraisal and graded the risk of bias of the included studies. The risk of bias was assessed by Rob2, the recommended tool to assess the risk of bias in randomized trials included in Cochrane Reviews [19], slightly adapted by the authors to fit the nature of the selected articles. Studies were classified as low, moderate, or high risk of bias.
Quality was assessed using the Newcastle-Ottawa scale [20]. Each study was awarded one point per positive item, according to the scale. Scores over 6 merited "high quality"; those below 4 were considered as "low quality"; the rest being classified as "moderate".
Epigenetic methodology has some peculiarities that prevent the application of bias and quality scoring by the commonly used scales. Some notes regarding this will be mention when discussing the selected epigenetic studies.

Selection, Bias and Quality of Articles
The database search yielded 914 articles ( Figure 1). Atopic eczema was used for the search engine as a synonymous term. After title and abstract review, 810 articles were rejected since they did not fulfill eligibility criteria, i.e., those describing animal or in vitro studies, literature reviews, analysis of protein or gene expression, and articles written in languages other than English or Spanish. Therefore, 104 articles qualified for full text review. Of those, we eliminated 13 studies that mentioned AD in comparison to other conditions but were not fully dedicated to it and 18 studies that did not perform any genetic association with the disease. As a result, 73 articles were evaluated. Out of 73 studies, 11 were related to epigenetics [13,[24][25][26][27][28][29][30][31][32][33], 39 were candidate gene studies , 5 were genome-wide association studies (GWAS) [5,13,[74][75][76], whole-exome sequencing (WES) was performed in 7 articles [77][78][79][80][81][82][83], and phenome-wide association sequencing was done in 1 article [84]. Four studies described results from next-generation sequencing (NGS) [85][86][87][88], and 2 showed analyses of copy number variations (CNV) [89,90]. Besides, 6 meta-analyses were also included [91][92][93][94][95][96]. A description of the 64 selected non-epigenetic studies is presented in Table 1. Epigenetic articles are summarized in Table 2. rs893051 is associated with development of AD in early life. rs9290927 rs9290929 rs893051 [46] Candidate gene Jordan To study the association between resistin gene polymorphisms and AD 162 AD patients, 161 HC RETN SNP +299 G>A (rs3745367) rs3745367 associated with AD in a genderand age-specific manner (male, less than 10 y) SNP +157 C>T (rs3219177) [ This SNP is a stronger risk factor for eczema than for hay fever or asthma.

RPTN-[ ]-HRNR rs12123821
This SNP is a stronger risk factor for eczema than for hay fever or asthma.
This SNP is a stronger risk factor for eczema than for hay fever or asthma.
This SNP is a stronger risk factor for hay fever than for eczema or asthma.

IL2RA
rs61839660 This SNP is a stronger risk factor for eczema than for hay fever or asthma.

GSDMB rs921650
This SNP is a stronger risk factor for eczema than for hay fever or asthma. MIF promoter polymorphisms in the −173 C allele and the MIF C/5-CATT and MIF C/7-CATT haplotypes were significantly associated with an increased risk for AD.

German
To identify disease association in the locus 11q13.5 using combination of sequencing and functional annotation.

A407T
Association of low-frequency and rare missense mutations within the LRRC32 gene with AD.
Meta-analysis French, French-Canadian and UK Rare FLG LoF variants in African American children are associated with AD and more persistent AD. In contrast to Europeans, no FLG LoF variants predominate in African American children. The most common variants were R501X, S3316X, and R826X. This frequency was lower than that described for other Asian populations (Chinese, Japanese, Singaporean) Candidate gene Korea Null alleles were associated with early onset of AD and higher risk of developing the disease by age 2 years.
3321delA EASI score was also higher in these patients

INPP5D
rs1057258-c Six classes based on temporal trajectories of rash were identified: persistent, early-onset/late resolving, early-onset/early-resolving, medium-onset/resolved, late-onset/resolved. FLG null mutations were strongly associated with early-onset and late-onset of AD (p < 0.00001: ALSPAC) and early-onset/late-resolving (p = 0.0006: To assess the genetic relationship between Il-10 polymorphisms and susceptibility to AD. 16 case-control studies IL-10 IL-10 -1082a/G These polymorphisms showed a weak association with AD susceptibility IL-10 -819T/C IL-10 -592a/C [53] Candidate gene Mutation S3296X only appeared in Japanese AD patients. R501X and R826X only appeared in IV patientsThe rest of the mutations were found in both Korean and Japanese patients.  AD lesioned skin Microarray miRNA expression (GSE31408) Hsa-let-7a-5p, has-miR-26a-5p and has-miR- 143-3p differentially expressed in lesioned tissues [32] Whole blood samples NLRP2 methylation NLRP2 methylation is associated both to the environment and SNP [33] AD lesioned skin Real-time PCR assay miR-124 downregulated in AD lesional tissue [25] As mentioned above, we followed the Cochrane guidelines to assess the risk of bias of the selected non-epigenetic studies, using the current version of the Rob2 tool [19]. As this tool has been developed for randomized trials, the authors decided to make some assumptions in order to adapt it to the specific nature of the genetic analysis. Taking into account that our main concern with respect to bias referred to the lack of appropriate controls or non-adequate genetic or epigenetic techniques for achieving the intended aim, we responded to questions about intervention or randomization consequently. Therefore, a study was classified as high risk when healthy controls were missing or the methodology to analyze the samples was not clearly explained in the text.
Under these conditions, 20.3% of the non-epigenetic studies were considered at a high risk of bias. Healthy controls were not included in 14 studies, some of them referring to very few patients. One of the studies had no reference to the sample size, and another one did not describe the used methodology. We found some concerns in 4 studies, mainly referring to the selection of subjects. The rest of the selected studies (71.9%) accomplished our criteria for low risk of bias ( Figure 2, Table S1).
Correspondingly, 71.2% of the articles merited high quality after running the NOS questionnaire (Table S2). Overall, the representativeness of the cases was the better-scored category. Thirteen articles were considered low-quality studies, mainly due to failed selection and definition of relevant controls.

Genetic Studies
A total of 62 genes and 5 intergenic regions were described as associated with AD in the selected studies; 32 of them were related to the disease for the first time during this period. Among them, filaggrin was the most widely reported gene, being over 90% of the other genes cited in only one article.
The interaction analysis performed by STRING showed some connectivity enrichment of the listed proteins (p-value < 1e-16). The network was clustered using the k-means method. Clustering results are shown in Figure 3 (non-linked proteins were removed from the graph). Three main clusters were found. The most populated included cytokines related to STAT3, with connections with ORM2 and RETN, and TNF.  With respect to the pathway analysis, a genome-wide overview from Reactome is shown as a Reacfoam in Figure 4 (for a higher resolution, a zoomable pdf version is available as Figure S1). Reacfoam shows a high-level pathway overview visualization based on Voronoi tessellation. Darker functions correspond to those that are over-represented in the list of genes identified in the selected studies, i.e., immune system, developmental biology, signal transduction, and extracellular matrix (ECM). Immune system functions stood up among the high hierarchy pathways (False Discovery Rate (FDR) 2.21e-6; p-value 2.22e-8). The most significant pathways were related to signaling by interleukins (FDR 1.7e-11; p-value!4.17e-11) or different variants, i.e., IL-4/IL-13 (FDR 7.02e-10; p-value 5.2e-10), IL-2 (FDR 3.89e-5; p-value 4.8e-7) or IL-12 signaling (FDR 1.42e-4; p-value 2.11e-6). Identifiers found in the former pathway were IL6R, IL10, TGFB1, TNF, STAT3, ADAM33, IL4, IL13, and MMP9. IL5RA, STAT3, IL2RA, INPP5D, IL9, and IL21 were associated with the IL-2 signaling pathway, while MIF, STAT3, IL10, and IL12RB1 were the IL-12 pathway entities found in the analysis. IL22 and INNP5D had not been related to AD before 2016. Moreover, enrichment was also found in transcriptional regulation of granulopoiesis (FDR 0.074; p-value 0.009), the process leading to the production of neutrophils, eosinophils, and basophils. Signaling cascades by MAPK1 (FDR 0.04; p-value 0.002) and FGFR (FDR 0.04; p-value 0.003) were also significantly enriched. The related entities found in the analysis were STAT3, FLG, IL2RA, IL5RA, IL6R, and MCM10. Seven out of the 62 genes described in the period covered by this review have been curated with functions in pathways related to extracellular matrix organization (FDR 4.58e-2; p-value 3.81e-3). Interestingly, 6 of these genes (COL5A3, COL6A6, KDR, MCM10, MMP9, and STS) had not been associated with AD yet, and only TGF-β1, involved in a broad spectrum of pathways, had been previously associated [97].
We also performed an analysis of disease-related genes using the FunRich software. The results are shown in Figure 5. The appearance of DOCK8 in all these biological processes stands out, taking into account that it has been described as related to AD only in one study reporting a single case [77].  Additionally, 17 of the new AD identified genes could not be ascribed to significant biological functions, i.e., CARD14, CRNN, TCHHL1, RPTN, PANX3, PHLDB1, LILRA6, NLRP2, MTF1, LTA, MAST2, DOCK8, CUX2, ADCY10, VSTM1, and RTEL1.

Filaggrin
During the period of this revision, we have identified 33 studies on the filaggrin (FLG) gene association with AD. It is remarkable that 16 novel mutations have been reported [56,81,85,88].

Filaggrin Mutations and Other Allergic Diseases
Eleven studies analyzed the association between FLG mutations and allergic sensitization, showing that FLG alleles conferred an increased risk, mainly in children with eczema [13,39,41,44,47,50,51,58,70,72,76]. In Polish children, Debinska et al. showed that several FLG mutations predisposed patients to eczema plus asthma, increasing more than 6-fold the risk of this complex phenotype (p-value 0.043) [72]. By contrast, such increased risk of asthma in FLG mutation was not confirmed in adult twins, although the risk of having AD was increased in those individuals with asthma, compared to individuals without asthma (27.6% vs. 18.5%; OR 1.68, 95% CI 1.12-2.52; p-value 0.012) [58]. Chan et al. showed a significant effect of FLG loss-of-function (LoF) mutations on both asthma and rhinitis at ages 1, 2, 4, 10. and 18 years, particularly at the age of 10 years (RR 1.96; 95% CI 1.70-2.26; p-value 0.003), early eczema being a requisite to suffer asthma at all ages [70]. Ferreira et al. carried out GWAS on individuals suffering from asthma, hay fever, and eczema to identify shared risk variants. The rs6181676[A] FLG variant was 1.32-fold more common in individuals suffering only from eczema when compared to individuals suffering only from hay fever (p-value 7.2e-8), and 1.26-fold comparing with asthma-only cases [13]. Two FLG single nucleotide polymorphisms (SNP), rs71626704 and rs76413899, were significantly associated with a history of asthma and cheilitis (p-value 0.002 and p-value 0.003, respectively) and rs62623409 and rs71625199 SNPs were associated with sensitization to environmental allergens (p-value 0.038 and p-value 0.008, respectively). Rs11584340 was associated with an increase of eosinophil-derived neurotoxin serum levels in allergic rhinitis patients and eosinophilic cationic protein serum levels in asthmatic patients [41]. Park et al. reported an association between FLG LoF mutation and early onset of asthma and AD [87]. The association between an FLG mutation and IgE sensitization to peanut at age 4 years (OD, 1.88; 95% CI 103-3.44), but not to other allergens was reported by Johansson et al. [39]. Equally FLG mutations were significantly associated with elevated IgE in a population of Korean patients with AD (>200KIU/L and/or MAST-CLA>+, p-value 0.005), palmar hyperlinearity (p < 0.001), and a family history of allergic disease (p-value 0.021) [50]. However, there was no significant difference in IgE levels between AD patients with non-mutated FLG and those carrying FLG LoF mutations (p-value 0.062) [44].

Filaggrin Mutations and Early Onset of AD
Patients who carried FLG mutation alleles are associated with early-onset AD [62,72]. Wan et al. found a dose-dependent association between the number of common FLG mutations and early onset [62]. In a Finish population, the combination of FLG mutations was shown to be significantly associated with early-onset of AD (<2 years) (OR 4.15, p-value 1.82e-10) and asthma (OR 2.76, p-value 1,57e-6) [47]. In two independent cohorts, FLG LoF mutations were associated with subphenotypes of AD. Thus, in a cohort study of 14,701 children from Avon (UK), the strongest association was detected with early-onset-persistent AD (OR 4.31; 95% CI 3.29-5.63; p-value 2e-26) and in a Dutch cohort of 3963 children, only the group of children with early-onset-late-resolving AD was associated with FLG LoF mutations (OR 5.63; 95% CI 2.65-11.95; p-value 7e-6) [76].
Additionally, it has been demonstrated that FLG expression in umbilical cord blood was associated with eczema development in infancy, being significantly lower in children with FLG variants when compared to children with wild-type FLG genotype (p-value 0.007) [75].

Filaggrin Mutations and other Skin Diseases
Andersen et al. studied the prevalence of FLG null mutations in adult patients with actinic keratosis (AK), premalignant intra-epidermal skin lesions that can progress into squamous cell carcinomas (SCCs). In their study, 7.5% AK patients had an FLG LoF mutation, of whom only the homozygous mutation carriers (0.8%), but not heterozygous, showed an increased risk of AK compared with wild-types (p-value 0.0017) [65]. Elhaji et al. found a significant association between the R501X mutation with polysensitivity in contact dermatitis when three or more positive patch test reaction occurred (8.5% patients vs. 4% controls; p-value 0.008) [91].
FLG mutations were significantly associated with palmar hyperlinearity in a population of Korean patients with AD (p < 0.001) [50], and also in a Finish population (OR 4.67, p-value 1.46e-5) [47].
The specific variant rs558269137 was exclusively detected in Italian children with AD and Molluscum contagiosum virus (MCV) infection, while rs374910442, rs138055273, rs113136594, and rs11584340 variants were found both in AD children and AD plus MCV-infected children [48].
In a population of African American (AA) children with AD and FLG LoF, 77% exhibited severe AD (SCORAD >50) [90]. A negative effect on the success of the immunosuppressive treatment was reported in FLG mutated patients when compared with those without FLG mutations [53]. When exploring the correlation between Eczema Area and Severity Index (EASI) and FLG-related AD in Korea, contradictory results were reported [50,87]. In addition, Wong et al. did not find any significant association between FLG LoF and the severity of AD [88].
Elbert et al. carried out a study to analyze the association of ethnic origin with FLG mutations and environmental risk factors in children from multiethnic origins but living in the Netherlands, showing that minority ethnicity children had a higher risk of eczema than Dutch children [73]. Gimalova et al. studied LoF variants in Russians and Tartars AD patients, reporting that c.2282del4 was the most prevalent mutation in both populations, whereas R501X and R2447X mutations were rare [36]. In India, a study of the association between FLG mutations and hand eczema showed that mutations in S2889X constituted 96.4% of all FLG mutations, while European mutations were not found [37].
An overview of the genetic map and geographic distribution of FLG mutations across East Asia found that 3321delA is a pan-Asian mutation [45]. K4022X, the most prevalent FLG mutation in Korea and northern China, showed a south-to-north distribution gradient. In contrast, c.6950del8 showed the reverse effect. On the other hand, S2554X, S2889x, S3296X, and Q1701X mutations were Japanese-specific. These FLG mutations were associated with an increased risk of AD but did not confer a risk of asthma [45].
Pigors et al. analyzed the genetic scheme of AD patients from the South Asian Bangladeshi community using WES combined with rare variant enrichment analysis [81], showing that FLG carried the highest number of enriched dominant (OR 12.1; p-value <0.0001) and recessive (OR 43.4; p-value <0.0001) LoF mutations. Three of the LoF mutations were previously unreported (S923Ffs*2, T1545Qfs*163, and S2352X). Furthermore, these genetic data revealed intrafamilial heterogeneity with multiple FLG variants often segregating within the Bangladeshi families with AD [81].
The common European FLG LoF R501X and 2282del4 were significantly associated with the risk of developing AD (OR 11.29, p-value 0.00022 and OR 2.66, p-value 0.00016, respectively) in Finnish patients [47]. In addition, having two 12-repeat alleles (rs12730241) was found to be significantly associated with a higher risk of AD (OR 1.96, 95% CI 1.36-2.81, p-value 0.00056) [47].
Using massively parallel sequencing, Margolis et al. identified nine FLG LoF variants in AA children, including 6 newly reported and 3 previously described, suggesting multiple and rare FLG LoF variants. Those children with FLG LoF variants had more persistent AD than wild-type children for FLG LoF [85]. These findings were supported by Mathyer et al., who identified five FLG LoF variants in 9 heterozygous AA AD patients (488delG, R501X, R826X, S3101X, and S3316X) [90].
New mutations were also found in Japanese (R826X) and Korean (S2889X) ichthyosis vulgaris (IV) patients [56]. Although the R826X mutation has not been detected in Japanese and Korean AD patients to date, it had been previously reported in Chinese and AA populations [95,99], suggesting that it is not a population-specific FLG mutation. Japanese and Korean patients shared 4 FLG mutations, Gly1109Glufs*13, Ser2889X, Ser3296X, and Lys4022X, being the latter more frequent in Korean than in Japanese AD or IV patients [56].
A robust and cost-effective high-throughput PCR-based method using microfluidics technology and NGS was applied to study the FLG coding region in cohorts of Chinese, Malaysian, and Indian AD patients living in Singapore. Thirty-three FLG LoF variants were identified in Chinese subjects, being 5 of them novel mutations. Unreported FLG LoF variants in Indian and Malaysian patients confirmed the diversity depending on the ethnic group [88].

Other Genes
Significant associations with AD have been reported for most of the analyzed genes and variants. Thus, ACTL9, C11orf30, IL6R, IL21, IL22, INPP5D, KIF3A, OVOL1, PRR5L, PPP2R3D, and STAT3 were investigated in two cohorts, ALSPAC and PIAMA [76]. ADCY10, CUX2, MAST2, MCM10, MTF1, ORM2, PHLDB1, and TCHHL1 were analyzed in Bangladeshi patients [81]. A CLDN1 polymorphism was positively associated with early onset of AD in Ethiopian patients [35], but no association was found in the Finnish variants [47]. CARD11-R30W has been associated with recurrent infections, autoimmunity, and severe atopy [78], and other dominant, negative mutations in CARD11, leading to dominantly inherited, severe atopy have been described in 4 unrelated USA families [79]. Within the same protein family, downregulation of CARD14 was reported to lead to severe AD and reduced skin protection against infection as well as dysregulated cutaneous inflammation pathways [80]. Rs199691576, a polymorphism of CP27A1, was also related to AD in Japan [82].
Other genes involved in immune functions, such as IL2RA [13,76], IL4 and ADAM33 [84], TGFB1 [57], and MIF promoter [42] have also been significantly associated to increased risk of AD. The relationship of TLRs polymorphisms with AD, i.e., TLR2 rs55743708(G>A) and TLR4 rs4986790(A>G) were reported to increase levels of IL-4 and IL-10 in Russian AD patients [61], while other authors found no association of TLR2 polymorphisms, i.e., rs5743708 and rs4696480, in Turkish children with AD [69]. Also, significant SNPs in TSLP, a lymphopoietin, have been reported in Chinese Han [38], Korean [43], and American [62,71] patients. Lymphotoxin α, a protein involved in IL-2 and IL-4 signaling events, more specifically, LTA rs2844484, was associated with AD in Greenland patients [67]. In a study performed in Jordan investigating the relationship between RETN gene polymorphisms and AD, rs3745367 was found significantly associated with AD in a gender-and age-specific manner [46].
Variants in extracellular matrix genes such as COL5A3 rs2287807 and MMP9 rs17575 were found as significantly related to AD in a meta-analysis performed in French, Canadian, and UK families [93]. COL6A6 minor allele (AA) in rs16830494 and the rs59021909 (TT) allele and the rs200963433 heterozygous (CT) showed higher frequency in patients than in controls, although no statistical significance was reached [83]. TMEM232 rs11357450 had the strongest association with the risk of AD among all the variants analyzed by Wu et al. [74]. Mutations in SHARPIN, a protein involved in epidermis development, that were exclusively present in patients, decreased its expression in AD lesions [55]. GSDMB rs921650 was characterized as a strong risk factor for eczema [13].

Epigenetic Studies
Over the period included in this revision, the studies of epigenetic modifications on atopic dermatitis were focused on DNA methylation and microRNAs. These epigenetic mechanisms have been shown to be crucial regulators in different allergic conditions [100,101], although histone modifications have also been studied in the context of allergic diseases and have been shown to play a role in their development [102].
With respect to DNA methylation, two articles studied such modification at a whole-genome level in blood samples. In the first one, Ferreira et al. analyzed the association of DNA methylation with different allergic risk factors. In this manner, they detected 36 genes with DNA methylation sites nearby that were associated with differences in gene expression between allergic patients and healthy controls. Additionally, they found an association between smoking and the methylation state of PITPNM2 [13], potentially involved in neutrophil function [103,104]. The second study found 490 CpGs differentially methylated between AD patients with eczema herpeticum and healthy controls and 6 CpGs differentially methylated when comparing AD patients without eczema herpeticum and healthy controls. Among these sites, they identified CpG methylation sites in IL4 and IL13, which suggested that there was a significant association between these methylations and the phenotype observed in the patients [28]. Another two studies were centered on the methylation state and its effect on gene expression of NLRP2 [33] and SIRL-1 [31]. These studies showed the influence of single nucleotide polymorphisms, a correlation of the gene expression level, and the presence or absence of AD condition.
Regarding the miRNA research, the different analyzed studies can be grouped by two main approaches to the function of miRNAs in the development of AD. On the one hand, the studies that assessed for the upregulation or downregulation of miRNAs in lesioned tissue of AD patients. In this way, several differentially expressed miRNAs were described in AD lesions. When comparing the lesional tissue of AD patients with normal skin samples of healthy controls, Yang et al. described miR-124 downregulation in AD lesional tissue [25] and in an in silico interaction analysis of differentially expressed miRNAs, Li et al. [32] postulated that downregulation of hsa-let-7a-5p would potentially upregulate CCR7, a chemokine receptor involved in the activation of T cells [105]. They also found a differential expression of miR-143, whose potential target is DENND1B, which is involved in the proliferation of T-cells [34]. In addition, the authors suggest that the downregulation of miR-26 would regulate hyaluronan synthase 3 (HAS3), which is upregulated in AD skin [32,106]. After comparing lesional skins samples with non-lesional skin samples in AD patients, Ding et al. proposed a regulatory network of differentially expressed genes that included 182 miRNAs, and among them, hsa-miR-148b, hsa-miR-152, and hsa-miR-324 [24]. Finally, using primary adult human keratocytes, several out of the 372 most common miRNAs were dysregulated when exposed to IL-4, which plays a key role in the development of AD [27].
On the other hand, there are some studies that look for miRNAs differentially expressed in sera from AD patients compared with healthy controls. Thus, the levels of miR-144 detected in umbilical cord serum were higher in those Japanese children that would develop AD at one year of age [26], levels of miR-151a and miR-409 were found to be in higher in sera from Chinese AD patients compared with healthy controls [30], and finally, miR-146a showed no difference in serum levels between patients with AD and healthy individuals [29], despite previously demonstrating a role in the regulation of the immune system and inflammatory responses pathways [107,108]. This second method of retrieving candidate miRNAs may serve as a feasible and less invasive way of obtaining new AD biomarkers, whereas the former procedure in lesioned tissue focused more on the mechanistic of action of such RNAs.

Discussion
Herein, we have systematically reviewed the literature related to genetics and epigenetics of AD published between June 2016 and December 2019. We have found 58 original articles and 6 meta-analyses and also 9 epigenetic studies. A total of 62 genes have been analyzed in the selected publications, 31 of which had not been reported as potentially associated with AD before June 2016.
One remarkable feature about allergic diseases is the diversity in the potential phenotypes sharing the same genotype, indicating that there appear to be additional components that increase the complexity of the regulation and development of such conditions, or at least shape its evolution over time [109,110]. Epigenetic regulation has emerged as a key factor that was missing to completely understand the molecular basis of allergic disease [111].

Filaggrin
Up to the revision date of this review, FLG LoF mutations were the most significantly associated genetic variants for AD. Filaggrin is a key protein in the differentiation of the epidermis and the formation of the skin barrier, which is necessary to prevent water loss through the epidermis and to avoid the entry of allergens, toxins, and pathogens [112]. Its precursor profilaggrin is encoded by the FLG gene, which is located on chromosome 1q21.3 [113] within a region known as the epidermal differentiation complex (EDC) comprising over 50 genes encoding proteins involved in terminal differentiation and cornification of keratinocytes [114]. LoF mutations in the exon 3 completely hinder FLG protein expression, increasing the risk of AD [113,[115][116][117][118][119]. A meta-analysis of 24 studies on FLG mutations determined a 3-fold increased risk of AD in those individuals carrying one or more FLG LoF, singling out the influence of one gene in such a heterogeneous disease [120]. More than 300 FLG LoF variants have been identified in the gnomAD browser, an international database of exome and genome sequencing data (https://gnomad.broadinstitute.org), more than 20 of them associated with susceptibility to AD [85].
The results showing that FLG null mutations conferred risk for allergic sensitization and susceptibility to ezcema-associated asthma are well aligned with previous studies [121][122][123][124][125][126][127]. All these findings support the idea that FLG mutations lead to functional epidermal barrier defects, increasing skin permeability and subsequent allergic sensitization, promoting the Th2 inflammatory response, and eventually leading to asthma [122]. The "outside-inside" theory of AD pathogenesis proposes that epidermal APCs in AD patients are overexposed to danger signals because of their impaired skin barrier, leading to APC maturation and T-cell-mediated inflammatory skin disease [128].
The most frequent FLG LoF mutations (R501X, 2282del4, S3247X, and R2447X) are present in 7-10% of Europeans [115,120] while these mutations are rare in Asian patients, who carry specific mutations [45,50,139,140]. Thus, K4022X has been reported as the most prevalent variant in Korean AD patients [50,51,87]. Interestingly, FLG mutations in Korean AD patients seem to be less frequent than in other East Asian countries, most likely due to genetic and environmental factors or mutations in other barrier genes [50,87]. Also, the analysis of FLG mutations in East Asia showed a geographic distribution in agreement with the history of human migrations [45].
On the other hand, FLG LoF mutations could be less common in patients with African descent than in those with European or Asian descent [139,140], although other studies have shown that AA children had an increased risk of AD compared with European children [140]. The prevalence of AD in the US was reported as the highest among AA patients, but this population remains largely understudied [141,142]. Recently, two studies using current sequencing methods were able to identify rare FLG LoF variants in AA children associated with more persistent AD [85,90]. The prevalence of common FLG variants in children of African ancestry is less frequent than those of European or Asian ancestry [85]. Moreover, hygiene habits, vitamin D level, sun exposure, microbiota, genetics, and skin barrier characteristics could also influence the association of ethnicity with AD [1,143].
The implementation of new technologies like NGS to analyze cohorts of understudied populations, as well as newer bioinformatic tools, will allow the identification of new FLG LoF mutations and confirm the variation among the different ethnicities.

Other Genes
Regarding research contributions in the period of this review, new components of the extracellular matrix have been described to be associated with AD [64,81,83,93,144]. These new associations highlight the importance of such structure in the development of AD and in the integrity of the skin barrier. Thus, COL5A3, COL6A6, and MMP9 are important for the collagen formation [145][146][147]; KDR, one of the two receptors of the VEGF, has been associated with integrin cell surface interactions with extracellular matrix [148]; mutations in STS (steroid sulfatase) have been associated to X-linked ichthyosis [148]; rare variants of MCM10, a key component of the pre-replication complex, have been described for the first time associated to AD [81]. Regarding TGF-β1 [57], besides its roles in other pathways, different works have shown its role in extracellular matrix assembly and disassembly [149,150].
Some genes associated to AD have been related to innate immune system pathways, providing solid evidences of the relationship of the innate immune system with the disease and its progression. On its behalf, most of the genes of the review associated to innate immune pathways (ADAM33, MIF, MMP9, ORM2, RETN, and TLR2) are related to neutrophil degranulation that contributes to the inflammation of the tissue in the AD [151,152]. In addition, a substantial number of publications [61,69,95,153] emphasize the importance of Toll-like receptor cascades on the development of AD and its link with other allergic diseases [154,155]. The TLR-2 rs4696480 polymorphism has also been associated with AD severity in adult patients in two different populations, and supported with functional studies [156,157]. However, one of the studies included in this review reported a lack of association of this polymorphism in Turkish children with AD [69], which could be due to the fact that differences in genetics and environmental factors appear to be relevant in the development of allergy. New pieces of evidence have been added to the previously reported association of genes such as ADAM33, CARD11, and DEFB1 with AD [64,67,78,79,84]. In this line, ADAM33 has also been associated to other allergic diseases like asthma [158] and allergic rhinitis [159]. CARD11 is required for B-and T-cell receptor signal transduction and activation of NF-κB transcription factor [160]. DEFB1 is an antimicrobial peptide implicated in the resistance of epithelial surfaces to microbial colonization. It is a member of the family of defensins, peptides made by neutrophils, and it has been proposed as a link between the innate and the adaptive immune systems [161,162].
It is noteworthy that several of the new genes listed in this revision do not fall into any of these functional groups. This may be due to different causes, as because of little knowledge about some genes or because they are not properly curated yet. This might be the cases of CARD14, with similar functions to CARD11 [163]; VSTM1, which behaves as a cytokine [164]; LILRA6, a member of the leukocyte immunoglobulin-like receptor family [165]; mutations in DOCK8 are responsible for an immunodeficiency syndrome [166]; NLRP2 is involved in inflammatory processes [167,168]; RTEL1, a helicase involved in telomere maintenance, has also been found associated to severe dyskeratosis congenita [169]; LT-α (also known as TNF-β), which is a cytokine produced by lymphocytes [170]. Additionally, ADCY10, CUX2, MAST2, MTF1, PANX3, PHLDB1, and SCAND3 have been associated to AD in a single study of exome sequencing [81]. The S100 fused type protein (SFTP) family includes genes which are mainly expressed in stratified epithelia and play a role in epithelial homeostasis [114]. SFTPs contain two calcium-binding domain EF-hand motifs and are associated with cytoplasmic intermediate filaments as well as minor components of the cornified envelope [114]. This family of proteins include 7 members, FLG being its most studied member and certainly showing an association with AD. Besides FLG, only FLG2 and HRNR were previously associated with AD [171][172][173]. Interestingly, over the last 5 years, other 3 members of the family have been associated with AD. The three members now associated with AD, CRNN, RPTN, and TCHHL1 were known to be involved in different epithelial disorders [174][175][176]. Taken together, SFTP family proteins pose as pivotal players in the proper skin cornification and in the development of AD.

AD Epigenetics
In recent years, the search for risk factors that help to understand how allergic diseases develop has become one of the main objectives of the research. The plasticity observed in the different phenotypes associated to an underlying genotype suggests that additional components may provide complexity to the processes that lead to the development of the disease, or, at least, influence its evolution [109,177]. In the last years, the focus has been set on epigenetic modifications, which can lead to the development of allergic diseases. Epigenetic regulation has emerged as a pivotal key in the comprehension of the molecular basis of allergic conditions [111,178].
Most of the research on AD epigenetic regulation has focused on the posttranscriptional regulation mediated by miRNAS. miRNAs constitute a class of small non-coding RNAs, with a size ranging from 17 to 25 nucleotides, and a sequence that allows them to bind to specific mRNAs. This key feature permits the posttranscriptional modulation of targeted genes by triggering mRNA degradation and/or inhibition of translation [179]. According to several functional studies, miRNAs are involved in virtually every cellular process [180]. miRNAs have also been related to immune system regulation, miR-21, miR-146a, and mIR-155 being the most extensively studied. A role in the regulation of the immune response and tissue inflammation in allergic diseases has been shown [101].
The analysis of lesional tissues has provided new miRNA molecules that could regulate different mechanisms and signalling pathways that are altered in AD lesions. As a general feature, a decrease of miRNAs involved in the regulation of the immune response and an increase of miRNAs involved in epidermis development is observed in these studies. Thus, downregulation of miR-124 in AD lesions has been shown to control NF-κB-dependent inflammatory responses in keratinocytes and chronic skin inflammation in atopic eczema [25]. Bioinformatics analyses from two studies suggest that miRNAs can influence the transcriptional regulation of signalling pathways related to the synthesis of extracellular matrix components, such as arachidonic acid and hyaluronic acid, as well as participate in processes such angiogenesis, lymphangiogenesis and apoptosis, all of which are involved in AD progression [24,27].
In addition, the use of miRNAs as biomarkers in allergic diseases is increasingly described [181,182]. In this sense, miR-151a and hsa-mir-144-3p have been proposed as potential biomarkers in AD, as they have been shown to be differentially expressed in serum and umbilical cord serum, respectively [26,30]. miR-151a would reduce IL12RB2 levels in T-cells, favouring the increase of Th2 cells, which are central in the pathogenesis of AD [30,183]. Also, an increased expression of miR-144 would reduce ABCA1 mRNA and protein levels and induce a proinflammatory response via NF-κβ [26]. Nevertheless, the differences in the expression of these small non coding RNAs observed in the lesional skin does not necessarily translate to changes in serum levels, which impairs their use as biomarkers. This is the case of miR-146a, which has been shown to be upregulated in AD lesional skin when compared to healthy controls [184] but showed no differences in serum levels [29].
Another extensively studied epigenetic mechanisms is DNA cytosine methylation [185]. This modification occurs in CpG dinucleotides which are grouped in the so called CpG islands, frequently located in intergenic regions, as well as in promoter region of genes [186]. CpG islands methylation of gene promoters is related to the repression of the transcription, i.e., CpG islands of genes that are being actively transcribed do not usually present methylation, while non-transcribed genes present CpG islands with high degrees of methylation [187]. Cytosine methylation suppresses gene transcription, as it causes chromatin condensation and prevents transcription factors from binding to their target sequences in promoters [185]. Different EWAS have shown differential methylation patterns associated with some pathologies [188,189] or even with the exposure to different agents [190][191][192]. Two studies showed gene expression modulation due to exposure to tobacco smoke in AD [13,33]. Thürmann et al. showed that NLRP2 was differentially methylated, due to both the effect of polymorphisms and tobacco smoke [33], and Ferreira et al. found differences in the methylation of PITPNM2, which were partly associated with environmental tobacco smoke [13]. Interestingly, both genes are involved in innate immune responses, with NLRP2 involved in inflammatory processes related to macrophages [193,194] and PITPNM2 related to neutrophils [103,104]. Other two studies found differences in DNA methylation patterns in AD patients. The first one found differences in methylation of the promotor region of the VSTM1 gene locus [31], which encodes the protein SIRL-1 that has been proposed to inhibit crucial pro-inflammatory functions in human myeloid cells [195,196]. The second one was a genome wide methylation study that found different patterns of methylation in over 490 sites in AD patients with eczema herpeticum, and where Boorgula et al. identified a significant association between IL4 and IL13 methylation and the AD phenotype, as well as with serum IgE levels [28]. However, this methylation patterns where shown to be highly influenced by the eosinophilic count [28].

Final Remarks
In the present article, we have evaluated the last 5 years of AD-related literature using systematic review methodology. We have focused on genetics and epigenetics aspects of the disease, monitoring the different polymorphisms and gene variations associated to the onset or severity of AD, comparing studies performed in different locations and including several ethnicities, therefore showing an up-to-date picture of current knowledge. Some of the retrieved articles used state-of-the-art technology when assessing their findings, including genome-wide sequencing of representative samples of patients. An exhaustive analysis of risk of bias and quality of the 64 selected articles have allowed us to ponder the validity of the reported associations. Another strong point is the inclusion of epigenetic studies.
Regarding limitations to the present review, it has to be point out that we have restricted our analysis to those genes included in the articles published in the last 5 years. Although the main genes related to disease onset and development, i.e. filaggrin, have been included, we are aware that other important genes, already reported elsewhere, may be missing here. Since our goal is to update the topic with new results, we highly recommend the interested reader to consult the previous reviews for more information [16,17].
In addition, we should remark that most genes have been described only once and for a limited number of patients. For instance, DOCK8 has been identified in a single case report. Larger clinical trials would be required to unambiguously link these genes to AD. The universalization of the whole genome techniques will allow the discovery of new mutations or confirm the already known ones in different populations.
New developments in genetics and epigenetics technology offer opportunities to improve the diagnosis of AD patients, ascribing them to specific genetic groups and allowing the tailoring of therapy with the best response to ensure the most convenient patient care.
Supplementary Materials: The following are available online at http://www.mdpi.com/2073-4425/11/4/442/s1, Figure S1: Reactfoam of pathways overview of the genes reported in the selected studies; Table S1: Analysis of the risk of bias for the selected genetic studies; Table S2: Quality assessment of the selected genetic studies.