Atopic dermatitis (AD) is a chronic, itchy inflammatory skin disorder associated with multiple genetic risk loci.[1–3] The most widely replicated genetic risk factor for AD is loss-of-function (LoF) variants in the filaggrin gene (FLG; NM_002016.2). FLG is translated as profilaggrin and post-translationally cleaved into functional FLG monomers.[4] Exon 3 contains 10, 11 or 12 near identical tandem repeats, determined by allelic variation producing copy number variants (CNV) that appear to be independent duplication events of repeat 8, repeat 10, or both.[1, 5] FLG CNV number has been reported as an independent risk factor for AD.[6, 7] AD is more prevalent and more severe in African-Americans.[8] However, FLG LoF variation is less frequent in African-American children compared to Caucasian children.[2, 3] The goal of our study was to evaluate FLG CNV as a major mechanism of skin barrier defect in African-American individuals with AD.
FLG sequencing from the Genetics of Atopic Dermatitis cohort identified 15 FLG LoF. Four were previously unreported and confirmed by Sanger sequencing (methods and table 1 in supplement). The odds of having FLG LoF and not having AD was 0.34 (0.15, 0.76; p=0.009).
The 10 CNV allele was detected in 65.5% of AD individuals and 64.1% of controls. The 11 CNV allele was detected in 72.1% of AD individuals and 63.2% of controls. The 12 CNV allele was detected in 28.4% of AD individuals and 42.7% of controls. A significant difference exists between the prevalence of 12 CNV between the cases and controls (p=0.01). The MVF for the AD and control groups are in Table 1.
Table 1:
Association of FLG copy number for those with (N=197) and without a history of atopic dermatitis (N=117). Odds ratios and 95% confidence intervals presented based on the presence of 10, 11, or 12 variant, based on total number of copies per person, and based on the allelic variant. Percentage presented with sample size (N) *As a measure of the prognostic usefulness, the aROC, a measure of the ability of a model to predict an outcome (i.e., AD), was 0.58 for total number of variants per person and 0.59 if those with FLG LoF were excluded. Dichotomizing the number variants at ≥ 22 maximized the sensitivity, specificity, and correct classification at 58.1%, 55.8% and 56.7%, respectively.
| Variant | Full cohort | Only those without a FLG LoF variant | CNV adjusted for FLG LoF composite | Percentage of AD group with variant (N) | Percentage of control group with variant (N) |
|---|---|---|---|---|---|
| CNV Variant Present | |||||
| 10 | 1.06 (0.66,1.71) | 1.17 (0.71,1.93) | 1.03 (0.63,1.67) | 65.5 (129) | 64.1 (75) |
| 11 | 1.50 (0.92,2.44) | 1.46 (0.88,2.43) | 1.48 (0.85,2.29) | 72.1 (142) | 63.2 (74) |
| 12 | 0.53 (0.34,0.86) | 0.51 (0.31,0.84) | 0.55 (0.34,0.90) | 28.4 (56) | 42.7 (50) |
| Total copy number* | |||||
| 20 | ref | ref | ref | 10.7 (21) | 3.4 (4) |
| 21 | 0.38 (0.12, 1.16) | 0.40 (0.13,1.26) | 0.34 (0.11,1.07) | 45.2 (89) | 38.5 (45) |
| 22 | 0.22 (0.07,0.68) | 0.19 (0.06,0.61) | 0.20 (0.06,0.64) | 25.4 (50) | 37.6 (44) |
| 23 | 0.38 (0.10,1.38) | 0.34 (0.09,1.28) | 0.36 (0.10,1.31) | 11.2 (22) | 9.4 (11) |
| 24 | 0.22 (0.06,0.81) | 0.24 (0.06,0.90) | 0.23 (0.06,0.86) | 7.6 (15) | 11.1 (13) |
| CNV Allelic Variant | |||||
| 10/10 | ref | ref | ref | 10.7 (21) | 3.4 (4) |
| 10/11 | 0.38 (0.12,1.16) | 0.40 (0.13,1.26) | 0.34 (0.11,1.07) | 45.2 (89) | 38.5 (45) |
| 10/12 | 0.14 (0.04,0.47) | 0.13 (0.04,0.44) | 0.13 (0.38,0.45) | 9.6 (19) | 22.2 (26) |
| 11/11 | 0.33 (0.10,1.11) | 0.28 (0.08,0.97) | 0.30 (0.10,1.32) | 15.7 (31) | 15.4 (18) |
| 11/12 | 0.38 (0.10,1.38) | 0.34 (0.09,1.28) | 0.36 (0.10,1.32) | 11.2 (22) | 9.4 (11) |
| 12/12 | 0.22 (0.06,0.81) | 0.24 (0.06,0.90) | 0.23 (0.06,0.86) | 7.6 (15) | 11.1 (13) |
A range of 20 to 24 FLG monomers can be present in the genome of an individual due to the CNV number (biallelic total). In this study, the mean CNV per person was 21.7 (sd: 1.06) for the full cohort, 21.9 (sd: 1.02) for the controls, and 21.6 (sd: 1.07) for those with AD (p=0.03). The average age at enrollment and the average age of onset did not vary by total number of variants per person (p=0.41, p= 0.11, respectively). CNV per person was not associated with asthma, seasonal allergies, food allergies, age of onset, or pet allergies among those with atopic dermatitis. CNV per person were nearly identical among those without FLG LoF; full cohort 21.7 (sd:1.09), controls 22.0 (sd:1.04) and those with AD 21.6 (sd:1.11) (p=0.027). The total number of CNV was associated with a decreased odds of having AD as compared to controls (0.79 (0.64, 0.98; p=0.033) and 0.78 (0.62, 0.97; p=0.028) for those without FLG LoF. As compared to 20 FLG monomers per person, those with 22 FLG monomers per person (0.22 (0.07, 0.68 p=0.009) and 24 FLG monomers per person (0.22 (0.06, 0.81; p=0.023) had the smallest odds of having AD. (Table 1). A linear trend (Kendall’s tau-b p>0.05) was not present.
The association between the allelic CNV and the odds of AD were similar to total number of FLG monomers. Those with the 10/12 and 12/12 alleles had the lowest odds of having AD (0.22 (0.07, 0.68); 0.22(0.06, 0.81), respectively) (Table 1). The presence of a twelve allele had the greatest influence on AD/control status. The absence of FLG LoF variant did not appreciably change the effect estimates. Epistasis between FLG LoF and individual CNV was not noted. Additional data were obtained from the St. Louis study of 39 African-Americans with AD[9] and 43 African Ancestry genomes from 1000 Genomes [6]. The baseline frequency of the CNV in these cohorts varied greatly (Table 2).
Table 2:
The number of individuals sequenced and copy number variation by dataset. 1000 Genomes (Eaaswarkhanth et al); St. Louis (Mathyer et al), and GAD controls and GAD cases from this investigation. Data presented as a number for a given variant with percentage in parentheses. 1000genomes was different from controls and cases with respect to total copy number and allelic variant (p<0.0001, chi-squared). St. Louis cases were different from GAD controls and GAD cases with respect to total copy number and allelic variant (p<0.0001, chi-squared). 1000 Genomes and St. Louis cases were not different (p=0.191) but allelic variant was different (p=0.042) as compared to each other.
| CNV Variant (by allele) | Total copy number | CNV Allelic Variant | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Source | Subjects/alleles | 10 | 11 | 12 | 20 | 21 | 22 | 23 | 24 | 10/10 | 10/11 | 10/12 | 11/11 | 11/12 | 12/12 |
| 1000 Genomes | 43/86 | 63 (73.3) | 15 (17.4) | 8 (9.3) | 28 (65.1) | 5 (11.6) | 7 (16.3) | 0 (0) | 3 (7.0) | 28 (65.1) | 5 (11.6) | 2 (4.7) | 5 (11.6) | 0 (0) | 3 (7.0) |
| GAD Controls | 117/234 | 78 (33.3) | 92 (39.3) | 63 (26.9) | 4 (3.4) | 45 (38.5) | 44 (37.6) | 11 (9.4) | 13 (11.1) | 4 (3.4) | 45 (37.9) | 26 (22.4) | 18 (15.5) | 11 (9.4) | 13 (11.1) |
| GAD Cases | 197/394 | 150 (38.1) | 173 (43.9) | 71 (18.0) | 21 (10.6) | 89 (45.2) | 50 (25.4) | 22 (11.2) | 15 (7.6) | 21 (10.6) | 89 (45.4) | 19 (9.7) | 31 (15.8) | 22 (11.2) | 15 (7.6) |
| St. Louis cases | 39/78 | 48 (61.5) | 16 (20.5) | 14 (17.9) | 16 (41.0) | 9 (23.1) | 9 (23.1) | 3 (7.7) | 2 (5.1) | 16 (41.0) | 9 (23.1) | 7 (17.9) | 2 (5.1) | 3 (7.7) | 2 (5.1) |
FLG CNV is an important risk factor for the development of AD, but there is a paucity of studies that have accurately described FLG CNV particularly in African-Americans patients. To address this, we present the largest case-control study in African-Americans. The frequency of FLG LoF are similar to a previous study of African-Americans with mild to moderate AD.[3] Our study demonstrates that CNV is associated with AD. The risk from FLG CNV is independent from FLG LoF, similar to reports in other ancestries.[5, 10] Similar to Irish children, the 11 CNV allele was most prevalent in association with AD.[5] However, in the Irish study FLG LoF variation modified the effect which did not occur here. Those with the fewest copy numbers (i.e., 20) have the highest risk for AD, but the risk is not linearly associated with increased number of copies. The lowest risk is associated with having at least one 12 CNV allele. In a study of Korean children, allelic CNV was also found to be a better predictor of AD than total CNV.[10]
Our results sharply contrast with a study using a sample from 1000 Genomes. Those of African ancestry seemed to overwhelmingly have 10 CNV.[6] There is reason to question the 1000 Genomes analysis. The 1000 Genomes data is a control population, yet the 10 CNV was the most frequent. In fact, it was more frequent than an African-American case-series with severe AD from St. Louis.[9]
There are possible limitations to our study. Unlike many case-control studies evaluating genetic variation, we enrolled controls within the same clinical settings (i.e. same “reference population”) thereby minimizing an important form of selection bias as well as information bias. Our study was limited to academic sites and might not generalize to all with AD. Also, as with other studies, we did not measure FLG protein changes resulting from genetic variation to provide a direct link between genetic variants and actual expression.
African-Americans with AD have CNV frequencies that are similar to those of European ancestry. This is in contrast with what is reported for FLG LoF variation. CNV and FLG LoF appear to have independent associations with AD. Individuals with 20 copies of FLG monomers are most prone to AD. However, CNV number is at best a moderate predictor of AD. The total number of FLG monomers and the specific CNV allelic variants, seem to both be similarly related to AD variation. Assessment of CNV adds value to our overall understanding of genetic association of AD but it does not account for the large difference between the frequency of FLG LoF seen in those of European and African ancestries.
Supplementary Material
Acknowledgement:
This study was funded by support from the NIH NIAMS R01-AR069062. J.E.C and X.F.C.C.W. are funded by Biomedical Research Council (BMRC) A*STAR, BMRC-A*STAR-EDB IAF-PP (H17/01/a0/004) for the Skin Research Institute of Singapore. None of the authors have a financial conflict of interest with respect to this investigation.
Conflict of interest:
D.M. had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. D.M. receives research funding as the principal investigator via the Trustees of the University of Pennsylvania (R01-AR060962, R01- AR070873, and R01-DK116199) and from the NIH and Valeant Pharmaceuticals (PEER study) and Sunovion Pharmaceuticals. Only R01-AR060962 funding was used for this study. He has consulting activities primarily as a member of data monitoring boards or scientific advisory boards with Leo, Johnson and Johnson, Pfizer, Sanofi, Kerecis, and Cell Constructs. None of these activities are associated with the outcomes of this study. With the exception of research funding from R01-AR060962, no other authors report conflicts of interest.
Abbreviations:
- AD
Atopic dermatitis
- CI
Confidence interval
- CNV
Copy number variation
- FLG
Filaggrin
- GAD
Genetics of Atopic Dermatitis
- LoF
Loss of function
- MVF
Minor variant frequency
- OR
Odds Ratio
- aROC
Area under the receiver operator characteristic curve
- sd
Standard deviation
References
- [1].Brown SJ, McLean WH, One remarkable molecule: filaggrin, J. Invest. Dermatol 132(3 Pt 2) (2012) 751–62. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [2].Margolis DJ, Apter AJ, Gupta J, Hoffstad O, Papadopoulos M, Campbell LE, Sandilands A, McLean WH, Rebbeck TR, Mitra N, The persistence of atopic dermatitis and filaggrin (FLG) mutations in a US longitudinal cohort, J Allergy Clin Immunol 130(4) (2012) 912–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [3].Margolis DJ, Mitra N, Wubbenhorst B, D’Andrea K, Kraya AA, Hoffstad O, Shah S, Nathanson KL, Association of Filaggrin Loss-of-Function Variants With Race in Children With Atopic Dermatitis, JAMA Dermatol (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [4].McAleer MA, Irvine AD, The multifunctional role of filaggrin in allergic skin disease, Journal of Allergy & Clinical Immunology 131(2) (2013) 280–291. [DOI] [PubMed] [Google Scholar]
- [5].Brown SJ, Kroboth K, Sandilands A, Campbell LE, Pohler E, Kezic S, Cordell HJ, McLean WH, Irvine AD, Intragenic copy number variation within filaggrin contributes to the risk of atopic dermatitis with a dose-dependent effect, Journal of Investigative Dermatology 132(1) (2012) 98–104. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [6].Eaaswarkhanth M, Xu D, Flanagan C, Rzhetskaya M, Hayes MG, Blekhman R, Jablonski NG, Gokcumen O, Atopic Dermatitis Susceptibility Variants in Filaggrin Hitchhike Hornerin Selective Sweep, Genome Biol. Evol 8(10) (2016) 3240–3255. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [7].Wong X, Denil S, Foo JN, Chen H, Tay ASL, Haines RL, Tang MBY, McLean WHI, Sandilands A, Smith FJD, Lane EB, Liu J, Common JEA, Array-based sequencing of filaggrin gene for comprehensive detection of disease-associated variants, J Allergy Clin Immunol 141(2) (2018) 814–816. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [8].Abuabara K, Hoffstad O, Troxel A, Gelfand JM, Margolis DJ, Atopic dermatitis disease control and age: A cohort study, J Allergy Clin Immunol 136(1) (2015) 190–192.e3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [9].Mathyer ME, Quiggle AM, Wong X, Denil S, Kumar MG, Ciliberto HM, Bayliss SJ, Common JE, Strong CD, 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, Experimental Dermatology 27(9) (2018) 989–992. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [10].Li K, Seok J, Park KY, Yoon Y, Kim KH, Seo SJ, Copy-number variation of the filaggrin gene in Korean patients with atopic dermatitis: what really matters, ‘number’ or ‘variation’?, British Journal of Dermatology 174(5) (2016) 1098–1100. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.