Tiled array-based sequencing identifies enrichment of loss-of-function variants in the highly homologous filaggrin gene in African American children with severe atopic dermatitis

Abstract

Filaggrin (FLG) loss of function (LOF) variants are a major risk factor for the common inflammatory skin disease, atopic dermatitis (AD), and are often population-specific. African American (AA) children are disproportionately affected with AD, often later developing asthma and/or allergic rhinitis and comprise an atopy health disparity group for which the role of FLG LOF is not well known. Discovery of FLG LOF using exome sequencing is challenging given the known difficulties for accurate short-read alignment to FLG’s high homology repeat variation. Here we employed an array-based sequencing approach to tile across each FLG repeat and discover FLG LOF in a well-characterized cohort of AA children with moderate to severe AD. Five FLG LOF were identified in 23% of our cohort. Two novel FLG LOF singletons, c.488delG and p.S3101*, were discovered as well as p.R501*, p.R826*, and p.S3316* previously reported for AD. p.S3316* (rs149484917) is likely an African ancestral FLG LOF, reported in African individuals in ExAC (Exome Aggregation Consortium), Exome Variant Server (ESP), and 4 African 1000G population databases (ESN, MSL, ASW, and ACB). The proportion of FLG LOF (11.5%) among the total FLG alleles in our cohort was significantly higher in comparisons to FLG LOF reported for African individuals in ExAC (2.5%; P = 4.3×10⁻⁴) and ESP (1.7%; P = 3.5×10⁻⁵) suggesting a disease-enrichment effect for FLG LOF. Our results demonstrate the utility of array-based sequencing in discovering FLG LOF, including novel and population-specific, that are of higher prevalence in our AA severe AD group than previously reported.

INTRODUCTION

Atopic dermatitis (AD) is a common and chronic inflammatory skin disease with dry skin and pruritus [1]. AD greatly impacts 15–20% of the pediatric population and 1–3% of adults worldwide [2]. As the incidence of AD worldwide has almost tripled in the past few decades [3], the prevalence also continues to increase in young children and in low-income countries [2]. In the U.S. alone, up to 16% of African American children are disproportionately affected with AD in comparison to 9.8% reported for Caucasian American children and for which the risk factors are not clear [4]. Loss-of-function (LOF) variants in the skin barrier gene filaggrin (FLG), initially discovered as causative in ichthyosis vulgaris (dry, scaly skin) in Europeans [5], are a major risk factor for AD and for other associative atopic diseases involved in the atopic march including asthma, allergic rhinitis, and food allergies [1]. FLG is primarily expressed in differentiated epidermal keratinocytes with FLG LOF variants resulting in filaggrin haploinsufficiency associated with skin barrier defects. Copy number variations (CNV) in the number of intragenic filaggrin monomer repeats comprise 3 different FLG alleles (10-, 11-, or 12-repeats) [6]. The FLG 10-repeat allele harbors 10 repeats whereas the 11-repeat allele contains either a duplication of the 8^th repeat or the 10^th repeat and the 12-repeat allele possess both the 8^th and 10^th duplicated repeats. FLG CNV has also been linked to disease risk with the addition of each FLG repeat reducing the odds of developing AD by 0.88 per repeat [1, 7, 8]. To date, 110 population-specific FLG LOF variants have been reported in AD (Supplemental Table 1). Specifically, multiple and rare FLG LOF have been observed in Asian populations in contrast to more common FLG LOF observed in European populations [9]. Together, the data suggests recent and parallel emergences of these variants in each population that are also not well understood.

The determination of the prevalence of pathogenic FLG LOF variants in African Americans (AA) with AD has been a decade-long, active area of investigation. A previous study in search of four common European FLG LOF (R501*, 2282del4, R2447*, and S3247*) found 5.8% of AA AD with FLG LOF [10]. A higher prevalence of the same European FLG LOF (22%) was identified among 11 AA children with both AD and ichthyosis vulgaris despite only 11% of FLG open reading frame being Sanger-sequenced [11]. Together, these studies identified lower percentages of known pathogenic FLG LOF in AA children in contrast to the ~50% frequency reported in moderate-severe AD European cohorts [1] but were limited in detecting additional FLG LOF.

Discovery of FLG LOF using exome and whole genome sequencing poses its challenges for the following reasons. FLG is one of the known high homology genes to determine variants owing to the difficulties in the accurate alignments of short read sequencing data for FLG [12]. Furthermore, careful consideration of the appropriate reference allele for FLG for either 10-, 11-, or 12-repeats is needed to properly determine variants, namely in the duplicated 8^th and 10^th repeats found in either of the 11-repeat and in both for the 12-repeat alleles. The conventional human reference allele for FLG, ENSG00000143631, to which most LOF mutations are mapped, is the 10-repeat allele. A recent exome sequencing study in African American children with AD with high read depth (185X) reported a 6.3% proportion of FLG LOF alleles, including 6 newly reported and 3 previously described for AD to suggest multiple and perhaps rare FLG LOF in this ethnic group [13] (Supplemental Table 1). In support of this finding, the allele frequency for the total number of FLG LOF is 2.5% for African individuals in the Exome Aggregation Consortium (ExAC) (whereby each LOF is <1%) and 1.7% in the Exome Variant Server (ESP) for African Americans with no documented knowledges of AD in these datasets [14] (Supplementary Tables 2, 3). Moreover, the prevalence of FLG LOF is higher in African individuals compared to those of European and South and Central American but lower than South and East Asian populations (1000 Genomes Phase 3) (Figure 2 of Eaaswarkhanth et al. [15]). Together, the population genetics data further supports the global existence of FLG LOF variants in individuals of African ancestry that warrants further investigation with respect to AD disease.

We previously identified that fewer filaggrin monomer repeats correlate with more severe AD in our AA pediatric patients whose AD was well-characterized according to the widely accepted AD criteria by the United Kingdom Working Party [UKWP] [7]. We sought to discover FLG LOF variants in our AA AD patients (n=39) using our array-based targeted sequencing approach for FLG with alignment to the human 12-repeat allele [16] in contrast to the human reference 10-repeat allele [ExAC and EPS]. Our method enables concurrent discoveries for both intragenic FLG CNV and FLG variant detection that were validated by Sanger sequencing and long-range PCR, respectively, as previously described [16].

METHODS

Patients

African American children with atopic dermatitis according to the UK Working Party criteria for AD [17] were recruited to the study as previously described [7]. The study was approved by the Washington University in St. Louis Institutional Review Board and conforms to the US Federal Policy for the Protection of Human Subjects. All persons gave informed consent prior to his or her inclusion in the study. Disease severity was previously assessed using SCORAD [18] with moderate AD ( >25) and severe AD (>50).

FLG Array-based Sequencing

Targeted sequencing for FLG was performed as previously described [16]. Briefly, a set of 48 amplicons that overlaps and tiles across the entire FLG coding gene from each patient was generated using the Fluidigm Access Array 48/48 chip and sequenced using Illumina MiSeq 2×250 bp reads mapped to the FLG 12-repeat allele with at least 100X coverage and confirmed by Sanger sequencing.

RESULTS

FLG LOF variants were identified in 23% of our AA AD patients (9 out of 39 recruited; eligible age range 3 months to 18 years). A total of five FLG LOF were identified in 9 heterozygous AA AD patients (c.488delG, p.R501*, p.R826*, p.S3101*, and p.S3316*; Table 1). c.488delG and p.S3101* are novel FLG LOF as they have not been reported in dbSNP, ExAC, and ESP variant databases. p.S3316* (also known as rs149484917) was found in 3 unrelated patients and had been previously reported in a separate AA AD group [13]. The higher proportion of FLG LOF in our AA AD group (11.5%) compared to the 2.5% African ExAC frequency indicated an enrichment for FLG LOF for AA AD and was statistically significant (p=4.3E-04, Fisher’s Exact Test). Similarly, the significance was further demonstrated in comparison to the 1.7% FLG LOF in ESP for African Americans (p=3.5E-05, Fisher’s Exact Test).

Table 1:

FLG LOF variants and FLG repeat alleles in African American pediatric patients with atopic dermatitis (n=39) reported with respect to human reference FLG 10-repeat allele.

FLG LOF Variant	hg19 coordinate/ RSID (if known)	Location of variant to FLG repeat	FLG repeat allele 1	FLG repeat allele 2	Total # FLG repeats	Age	SCORAD	Asthma	Allergic rhinitis	Food Allergies	Ichthyosis Vulgaris (IV)/ Xerosis (X)
c.488delG¶	chr1:152,286,874	0	10	10	20	7	73	Yes	Yes	Yes1,2,3	IV
p.R501*	rs61816761	1	11	11	22	11	83	Yes	No	None	X
p.R501*	rs61816761	1	10	11	21	13	38	Yes	Yes	None	X
p.R826*	rs115746363	2	10	10	20	5	51	Yes	No	Yes1	X
p.R826*	rs115746363	2	10	10	20	1	65	Yes	No	None	X
p.S3101*¶	chr1:152,278,060	9	10	10	20	10	41	Yes	Yes	None	X
p.S3316*	rs149484917	9	10	11	21	4	50	Yes	No	Yes4	X
p.S3316*	rs149484917	9	10	11	21	2	94	No	No	Yes1	X
p.S3316*	rs149484917	9	10	10	20	0.5	85	No	No	None	X

^¶ denotes novel FLG LOF

¹Peanut

²Shellfish

³Seafood

⁴Eggs

Reports of asthma, allergic rhinitis, and food allergies are in response to our study questionnaire and obtained during medical exam history.

All patients with FLG LOF (age≥1 year) reported having at least one additional allergic disease: asthma, allergic rhinitis, or food allergies and either xerosis (n=8) or ichthyosis vulgaris (n=1, c.488delG) and a family history of atopy in at least one first-degree relative. Seven of the nine FLG LOF patients (77%) exhibited severe AD (SCORAD >50). Patients with FLG LOF exhibited a trend toward a greater percentage of the body affected with the disease (SCORAD component A mean, 56) compared to non-FLG LOF patients (mean, 40) but this was not significant (p=0.13, two-tailed t-test, data not shown) and warrants further investigation to test this emerging hypothesis.

We sought to further investigate the distribution and frequency of rs149484917 that results in FLG LOF S3316* using ExAC, ESP, and 1000G population datasets given the concurrent findings for AA AD in our cohort and a previous study [13]. rs149484917 has been reported primarily in African individuals with 96% and 100% of the minor allele reported in Africans in ExAC and ESP, respectively. A closer investigation into the 1000G populations revealed that rs149484917 was not found in any European, American, East Asian, or South Asian populations. However, of the seven 1000G African (AFR) ancestral populations, rs149484917 was reported as a less common variant in 4 AFR populations; ESN (Esan in Nigeria) (Minor allele frequency, [MAF], 0.02), MSL (Mende in Sierra Leone) (0.012), ASW (African ancestry in Southwest US) (0.016), and ACB (African Caribbean in Barbados) (0.010). Taken together, the population genetics data identifies rs149484917 (S3316*) as a population-specific FLG LOF unique to several populations of African ancestry.

We next sought to determine the genetic background from which our newly discovered FLG LOF mutations arose in the context of the 10-, 11-, or 12-repeat FLG alleles [1]. The 10-repeat allele is the major allele in Africans (73%) and is interpreted to be ancestral [15]. c.488delG was found in the incomplete, N-terminal repeat (repeat 0) [7] and inferred to be on the 10-repeat allele as the patient was homozygous for the 10-repeat allele (Table 1). The other novel FLG LOF mutation (p.S3101*) was identified on the 9^th repeat and also on the 10-repeat allele. Thus, the existences of c.488delG, p.S3101*, and p.S3316* on the 10-repeat allele suggest that these variants most likely arose independently and after the separation of the 11- and 12- expanded repeat, derived alleles.

DISCUSSION

In summary, our targeted sequencing strategy to identify FLG LOF variants in a well-characterized UKWP-defined AD group of African American children identifies a significant enrichment of pathogenic FLG LOF variant alleles (11.5% versus 2.5% [Africans; ExAC] and also 1.7% [African American; ESP]). We discovered 2 pathogenic FLG LOF not previously reported for AD (c.488delG and p.S3101*). The overlap of our p.S3316* findings in our cohort, in an independent AA AD study [13], and in African individuals (ExAC, ESP, and 1000G) supports p.S3316* as a population-specific FLG LOF variant for African ancestry. Our discovery of five pathogenic FLG LOF variants (including 2 newly discovered) in African American pediatric patients with severe AD has brought forth a hypothesis for the wider research community to test, FLG LOF variants among the African American population are enriched in AD, and thus justifies more precision medicine efforts for this health disparity group.