Filaggrin variation differs for European-Americans and African-Americans

To the Editor

Atopic dermatitis (AD) is a common chronic illness that often manifests in childhood as red, itchy, inflammatory, crusting skin lesions. However, the disease can occur at any age and waxes and wanes in severity. The most common genetic findings in children with AD are filaggrin (FLG; OMIM:135940) loss of function mutations (LOF) found on the gene’s third exon. . Europeans with the most severe and persistent forms of AD are more likely to have FLG LOF (Irvine et al., 2011, Margolis et al., 2012). Asians also have FLG LOF though many of their specific variants differ from those of Europeans (Wong et al., 2018). Past studies have claimed that FLG LOF variants are very rare, if not absent, in those of African ancestry (Thawer-Esmail et al., 2014, Winge et al., 2011, Margolis et al., 2014, Margolis et al., 2020a, Margolis et al., 2020b, Margolis et al., 2018). More recent studies of African-Americans indicate that while the prevalence of variants is much less than that of European ancestry, FLG LOF is not as rare as previously noted (Margolis et al., 2019). In fact, recent data from public access databases that used whole exome/genome approaches reveal more than 100 FLG LOF variants in those of African ancestry (https://gnomad.broadinstitute.org/). Oddly, gnomAD shows FLG to have many more distinct variants than other closely located genes in the epidermal differentiation complex found on chromosome 1 such as filaggrin-2, hornerin and trichohyalin (see supplement). The most common FLG mutations in those of African ancestry differ from those observed in European as well as Asian populations. Regardless of these newer findings, it has been suggested that AD outcomes in African Americans can be explained by population admixture between European and African ancestries. The goal of our study was to determine whether African Americans (AA) and European Americans (EA) are similar to each other and also to reference populations of Africans and Europeans, respectively, in terms of overall FLG variation.

Our primary data come from the Pediatric Elective Eczema Registry (PEER), which is a prospective 10-year observational cohort. Details regarding this study can be found in Margolis et al. (2012). We obtained FLG data on African and European ancestry participants from phase 3 of the 1000 Genomes Project of the International Genome Sample Resource (IGSR); we refer to this sample as 1000 Genomes (1000G). Data from PEER were initially analyzed to assess differences in the total number of FLG variants by ethnicity (European or African) using linear regression or negative binomial regression (IRR), as appropriate.

To explore genetic similarity between subjects within the FLG locus, we employed logistic principal component analysis (PCA) on FLG SNP observations using the approach by Landgraf & Lee (2015) in R (Landgraf and Lee, 2015). Their method extends the original PCA formulation of Pearson (1901) by constructing a projection of data that is closest to the original with respect to minimizing Bernoulli deviance (Pearson, 1901). This dimensionality reduction approach extracts information in the SNP data and summarizes that variation into a few principal components allowing for easier visualization of underlying patterns. Using the first and second principal components which account for a large portion of the variability in FLG SNPs, we assessed clustering of observations for the PEER population (both AA and EA) and combined with data from 1000G. We also conducted PCA on a separate cohort, Genetic of Atopic Dermatology (GAD), to verify that our results were not sample-specific.

The number of FLG synonymous and non-synonymous variants (excluding LOF variants) ranged from 2 to 109 per person in the PEER population with mean38.3 (sd: 33.4) and median of 49 (IQR: 5, 60). The number of FLG variants per person was greater in AA (mean: 54.7, sd: 31.8) than EA (mean: 24.7, sd: 24.7), IRR 2.0 (95% confidence interval: 1.7, 2.3; p=0.0009). For all FLG LOF variants, a composite was constructed (no variant, a single variant, two or more variants). In AA, the mean number of FLG SNPs was 55.1 for those with no LOF variants, 43.9 for those with one LOF variant, and 28.5 for those with two or more. The mean numbers for the EA group were 26.0, 23.0, and 18.2 for the three composite categories, respectively. Thus, there appears to be an inverse relationship between number of SNPs and presence of LOF variants.

Figure 1(a) presents the first two principal components based on 484 FLG SNPs for the 741 AD subjects. We distinguished between AA and EA to compare and contrast their respective genetic variations. For the most part, FLG variation is differs between AA and EA. To better understand the FLG variation of AA AD patients, we combined the PEER data with subjects who were African (population code: AFR) or European (population code: EUR) in 1000G (combined N=1905). 335 FLG SNPs noted in both PEER and 1000G were evaluated. Figure 1(b) shows that AA in PEER tend to cluster with Africans from 1000G, while EA in PEER tend to cluster with Europeans in 1000G. All humans originated from the same ancestral root, and as would be expected, some clusters include both ethnicities. The observed nonoverlap (clusters containing only AA or only EA in Figure 1(a)) suggests that FLG variation in AA is not explained by genetic admixture with EA in the United States. A comparison of patients with FLG LOF and those without is presented in the supplement.

Figure 1:

PCA plots showing clustering by ethnicity/ancestry for PEER cases based on FLG variation. (a) 484 FLG SNPs for subjects with AD in the PEER cohort and (b) 335 FLG SNPs in common to the PEER and 1000 Genomes data sets.

To replicate our results, a similar analysis was conducted for subjects with AD in the GAD cohort, which includes 412 subjects with 579 FLG SNPs. Figure 2(a) again indicates that there are differences in the genetic variation of FLG between AA and EA. Comparison of the GAD cohort to 1000G shows that AA from the GAD cohort and Africans from 1000G are more likely to cluster together (Figure 2(b)).

Figure 2:

PCA plots showing clustering by ethnicity/ancestry for GAD cases based on FLG variation. (a) 579 FLG SNPs for subjects with AD in the GAD cohort and (b) 205 FLG SNPs in common to the GAD and 1000 Genomes data sets.

The number of FLG variants per individual appears to be different between EA and AA. While FLG LOF variation is associated with AD, there appears to be an inverse association between total FLG variation and the presence of FLG LOF. FLG variation seen in those of African ancestry (includes AA) is different from those of European ancestry (includes EA). Our team as well as other investigators have shown that FLG variation differs by ancestry and may be shaped by genetic drift as well as selection (Eaaswarkhanth et al., 2016). Our results indicate that it is unlikely that the risk for or persistence and severity of atopic dermatitis in AA as compared to EA is due to admixture of the FLG gene from Europeans to Africans. With respect to FLG variation, it appears that AA map to individuals of African ancestry rather than individuals of European ancestry. Because AD is more common in AA than in EA, other genic or environmental differences may account for AD disease mechanisms.

Abbreviations:

FLG

Filaggrin

LOF

Loss of function

PEER

Pediatric Elective Eczema Registry

AA

African Americans

EA

European Americans

PCA

Principal Component Analysis