Association of FLG single nucleotide variations with clinical phenotypes of atopic dermatitis

Background FLG encodes a large protein called profilaggrin, which plays a key role in maintaining an effective skin barrier against the environment. In this study, we identified FLG single nucleotide variations (FLG-SNVs) and evaluated the association of FLG-SNVs with clinical phenotypes including atopic dermatitis (AD)-associated minor clinical features, presence of specific allergic sensitization, and serum parameters. Methods Eighty-one Korean patients with AD were enrolled. AD-associated minor clinical features as well as allergic rhinitis and asthma were diagnosed by specialists. FLG-SNVs were identified by Sanger sequencing of entire exons through long-range PCR. Allergic sensitization to a specific allergen was evaluated by multiple allergen simultaneous test. Serologic parameters such as serum eosinophil cationic protein (ECP) and eosinophil derived neurotoxin (EDN) were measured. Results A total of seventy-three SNVs and 4 LOF mutations were successfully genotyped. rs71626704 and rs76413899 were significantly associated with a history of asthma and cheilitis (P = 0.002 and P = 0.033, respectively), however, the associations were not found statistically significant after adjustment by multiple comparisons. In addition, we detected haplotype blocks which were correlated with non-specific hand or foot dermatitis and scalp scale. We identified FLG-SNVs which were associated with sensitization to environmental allergens; rs62623409 and rs71625199 (P = 0.038 and P = 0.008, respectively). Patients with FLG P478S TT and history of allergic rhinitis showed a higher EDN level, and among those patients, the ones with asthma showed a higher ECP level. Conclusion This study revealed the association of FLG-SNVs with AD-associated minor clinical features. We firstly identified rs71625199 which was associated with higher environmental allergic sensitization. We also suggest that FLG P478S is a kind of disease modifier which affects serologic parameters such as EDN and ECP.


Methods
Eighty-one Korean patients with AD were enrolled. AD-associated minor clinical features as well as allergic rhinitis and asthma were diagnosed by specialists. FLG-SNVs were identified by Sanger sequencing of entire exons through long-range PCR. Allergic sensitization to a specific allergen was evaluated by multiple allergen simultaneous test. Serologic parameters such as serum eosinophil cationic protein (ECP) and eosinophil derived neurotoxin (EDN) were measured.

Results
A total of seventy-three SNVs and 4 LOF mutations were successfully genotyped. rs71626704 and rs76413899 were significantly associated with a history of asthma and cheilitis (P = 0.002 and P = 0.033, respectively), however, the associations were not found statistically significant after adjustment by multiple comparisons. In addition, we detected haplotype blocks which were correlated with non-specific hand or foot dermatitis and scalp scale. We identified FLG-SNVs which were associated with sensitization to environmental allergens; rs62623409 and rs71625199 (P = 0.038 and P = 0.008, respectively). Patients with FLG P478S TT and history of allergic rhinitis showed a higher EDN level, and among those patients, the ones with asthma showed a higher ECP level. PLOS

Introduction
Atopic dermatitis (AD) is a pruritic, inflammatory skin disease with a chronic relapsing course. This disease has been known to be influenced by environmental and genetic factors [1]. Profilaggrin is cleaved to produce multiple copies of the filaggrin protein, which plays a key role in maintaining an effective skin barrier against the environment [2][3][4][5]. The filaggrin gene loss-offunction (FLG-LOF) mutations lead to reduce protein expression, making the skin more permeable to environmental allergens and increasing trans-epidermal water loss [6][7][8]. FLG-LOF mutations have been identified as a risk factor for allergic sensitization, and allergic rhinitis and asthma [9][10][11]. Not only FLG-LOF mutations but also mutations associated with epidermal barriers and immune response have also been linked to AD [12,13]. In addition, one FLG single nucleotide variation (FLG-SNV) corresponding to P478S has been reported to have a clinical implication in AD patients [14] and to increase the risk of asthma combined with allergic sensitization [15]. FLG-LOF has been recognized as a disease-modifying factor in AD strongly associated with early onset and severe disease in European populations [8].
In this study, we analyzed FLG-SNVs by Sanger sequencing of entire exons through longrange PCR and evaluated the association of FLG-SNVs with clinical phenotypes such as ADassociated minor clinical features, presence of specific allergic sensitization, and serum parameters in AD patients.

Materials and methods Patients
Eighty-one Korean patients with AD were recruited. The mean age of the study group was 15.25 ± 8.07 years and 56.8% of patients were male. Family history of AD was found in 53.1% of patients and 18.5% of patients had a history of early onset disease. These patients were outpatients referred to the Department of Dermatology, Seoul St Mary's Hospital. Detailed phenotypic data, especially AD-associated minor clinical features, including xerosis, pityriasis alba, cheilitis, tendency towards non-specific hand or foot dermatitis (NS-HFD), scalp scale, perifollicular accentuation, keratosis pilaris, palmar hyperlinearity, and ichthyosis, were gathered and examined by experienced dermatologists. Allergic rhinitis and asthma were diagnosed based on the questionnaire answers and the diagnosis made by pediatricians or otolaryngologists. Subjects were informed about the study design and purpose according to the Declaration of Helsinki. All subjects of the study provided informed consent, and the protocol of the study was approved by the institutional review board (KC10TISI0582) of The Catholic University of Korea.

Computational prediction tools
Several prediction tools were used to evaluate the FLG mutations identified in Korean AD patients. The following widely established computational prediction methods were applied: the evolution-based prediction tools including Sorting Intolerant From Tolerant (SIFT, http:// sift.jcvi.org/), and protein structure and function were assessed by using straightforward comparative physical and evolutionary analyses and sequence-and structure-based bioinformatics tools, such as Polymorphism Phenotyping v2 (PolyPhen-2, http://genetics.bwh.harvard.edu/ pph2/) [19,20].

Eosinophil cationic protein and eosinophil derived neurotoxin
Serum eosinophil cationic protein (ECP) and eosinophil derived neurotoxin (EDN) levels were measured. Both were secreted by activated eosinophils; therefore, they were considered as sensitive markers for disease activity in AD [21,22]. We measured the ECP level using a solid-phase, 2-site chemiluminescent immunometric assay on an Immulite 2000 analyzer (Siemens Healthcare Diagnostics, Tarrytown, N.Y., USA). The cutoff limit for a positive serum ECP result was set at >19 μg/L on the basis of the manufacturer's instructions. The EDN level was analyzed using The BioTracer™ K1 EDN ELISA Kit according to the manufacturer's instructions. After reaction, absorbance was measured at 450 nm by a Micro Plate Reader Infinite 200 PRO (TECAN, Mannedorf, Switzerland). The concentration obtained from the standard curve after calculating the average absorbance values for each set of duplicate standards and duplicate samples. The analytical measurement range of the EDN assay is 6.0 to 400 ng/ mL. AD patients included in this study showed significantly higher levels of both ECP and EDN compared to healthy controls (59.12 μg/L vs. 17.04 μg/L, P<0.001, 110.51 ng/mL vs. 43.44 ng/mL, P<0.001, respectively).

Specific allergic sensitization
Multiple allergen simultaneous test (MAST) was performed to analyze the serum-specific IgE using Advansure Alloscreen (AS: LG Life Science, Seoul, Korea). A total of 20 allergens commonly present in AD patients were included; two mite allergens (Dermatophagoides farinae (Df), Dermatophagoides pteronyssinus (Dp)), 8 food allergens (egg, milk, soy bean, wheat flour, shrimp, chicken, peanut, and pork), and 10 environmental allergens (cat epithelium, dog dander, cockroach, birch-alder mix, oak white, ragweed short, mugwort, Alternaria alternata, grass mix, and Japanese hop). Briefly, 250 μL of serum was reacted to the allergen on a test strip. After washing, antihuman IgE antibody coupled with biotin was added and incubated. Then, unbound antibody was removed, which was followed by incubation with streptavidin conjugated to alkaline phosphatase. After incubation with luminescent solution, test strips were completely dried and read using AdvanSure AlloScan. Specific IgE concentration was classified from 0 to 6, and results greater than the levels in class 2 (sIgE ! 0.7 kU/L) were considered positive.

Statistical analysis
Chi-square test was used to compare the allele frequencies of FLG-SNVs according to ADassociated minor clinical features. Odds ratios (OR) and 95% confidence intervals (95% CI) were calculated from logistic regression analysis. Haploview was used to generate a linkage disequilibrium (LD) map by using Gabriel's rule. Tests for associations using multi-marker haplotypes were performed using the statistics software ''R" (http://www.R-project.org), package 'haplo.stat'. To assess the population impact of the variants we identified, we performed a casecontrol analysis of each FLG-SNVs. The genotype and allele frequencies for patients and control subjects were used to calculate the odds ratios (ORs). We used the allele data of East Asian population in Exome Aggregation Consortium for the control group. Multiple comparisons were performed by one-way ANOVA and Bonferroni correction. A P value of less than 0.05 was considered statistically significant. PLINK software program (http://pngu.mgh.harvard. edu/~purcell/plink/), Haploview program (http://www.broad.mit.edu/mpg/haploview), and SAS program (version 9.2; SAS Institute Inc., Cary, NC, USA) were used for statistical analysis.

FLG-SNV and AD-associated minor clinical features
A total of seventy-three SNVs and 4 LOF mutations were successfully genotyped in this study. The data set are provided in Supplementary table (S1 Table). The results comparing their frequencies to normal East Asian population and in silico prediction are listed in S2 Table. We analyzed the interaction of SNVs with AD-associated minor clinical features including NS-HFD, asthma, cheilitis, and scalp scale. The A allele frequency of rs71626704 was significantly higher in patients with asthma (37.5%, 3/8) than in patients without asthma (4.8%, 3/ 62) (P = 0.002). The allele A of rs76413899 was only identified in patients with cheilitis (16.3%, 7/43) (P = 0.033). NS-HFD was associated with FLG-SNVs including P478S (S3 Table). A Bonferroni correction was applied for multiple comparisons with all FLG-SNV associations, with a correction factor derived from the number of SNVs tested (for 73 comparisons, critical P value is 0.000685). However, no significant associations were found between any single FLG-SNVs and AD-associated minor clinical features.
https://doi.org/10.1371/journal.pone.0190077.t001 rs71625199 showed a higher risk of sensitization to environmental allergens (P = 0.008, OR 4.173, 95% CI 1.454-11.972). We found eight FLG-SNVs which were significantly associated with dog dander allergic sensitization (Table 2). Interestingly, patients with any of these SNVs showed a higher risk of sensitization to dog dander allergen compared to patients without these eight FLG-SNVs (P = 0.002, OR 9.091. 95% CI 1.893-43.657). In our study, we did not find any association between FLG-LOF mutations and the presence of specific allergic sensitization.

FLG-SNV and serum parameters
Scoring atopic dermatitis (SCORAD) is a clinical tool used to assess the extent and severity of eczema. Dermatologists may use this tool before and after treatment to determine whether the treatment has been effective [23]. In this study, serum IgE, ECP, and EDN levels, and eosinophil count were significantly different according to SCORAD (P = 0.028, P = 0.001, P = 0.022 and P < 0.001, respectively). EDN level was higher in the moderate group compared to the mild group (P = 0.018), while ECP level was higher in the severe group compared to the mild and moderate group (P = 0.001 and P = 0.028, respectively). Eosinophil count was higher in severe group compared to the mild and moderate group (P < 0.001 and P = 0.037) (Fig 2). Patients with FLG P478S TT and history of allergic rhinitis (n = 11) showed a higher EDN level (154.5 ± 34.7 ng/mL vs. 99.1 ± 9.3 ng/mL, P = 0.041). Serum IgE and ECP levels were significantly different according to FLG-SNV and history of allergic rhinitis and asthma (P = 0.009 and P = 0.013, respectively). Patients with FLG P478S TT and history of allergic rhinitis and asthma showed a higher IgE and ECP level compared to those without asthma (P = 0.037 and P = 0.022) and patients without P478S TT (P = 0.008 and P = 0.010).

Discussion
AD is a complex disease and genetic and environmental factors are implicated in the heterogeneity of AD. Because filaggrin is a multifunctional protein that can alter the skin barrier function and inflammatory response [24], FLG-SNVs as well as FLG-LOF may have an influence on the clinical phenotype of AD. Here, we performed FLG sequencing including all exons and flanking introns and identified most of the known SNVs (n = 73). We searched public databases of genetic variants including the KRGDB and performed in silico analyses of the identified missense variants. To the best of our knowledge, this is the first study to investigate the correlation of FLG-SNVs with clinical phenotypes including AD-associated minor clinical features, specific allergic sensitization, and serum parameters. rs71626704 (D2015E), rs76413899 (G837S), and rs11584340 (P478S) were significantly associated with the development of asthma, cheilitis, and NS-HFD, respectively. The linkage analysis showed a few meaningful haplotype blocks associated with NS-HFD and scalp scale. The reduction or deterioration of filaggrin influences the skin barrier formation and makes the patient's skin susceptible to environmental allergens [25] because allergic sensitization occurs through the damaged skin barrier [26]. A previous study found associations between the FLG-LOF mutations and sensitizations to grass, house dust mite and cat dander [27]. Few studies have demonstrated that FLG-SNV was associated with allergic sensitization in AD patients. We identified FLG-SNVs including rs62623409 (Q1873K) which were associated with sensitization to environmental allergens. In addition, rs71625199 was a significant FLG-SNV associated with sensitization to environmental allergens. SIFT and polyphen-2 predicted rs62623409 as tolerate and benign, respectively. Even rs71625199 does not induce amino acid change (synonymous). Another in silico prediction tool predicted that rs71625199 induce the alteration of splicing through the activation of an exonic cryptic acceptor site [28,29]. Filaggrin is highly polymorphic and filaggrin itself breaks down, the resulting products contribute to the formation of natural moisturizing factor, which is important for epidermal hydration and barrier function [30]. Therefore, the functional change of the FLG-SNV cannot be directly expected via in silico analysis.
Currently, there is no reliable serum parameter that can distinguish AD from other disease entities [31]. Although the parameter has little diagnostic significance, several characteristic laboratory changes occur in AD patients. Eosinophils play a role in the inflammatory process of AD by which the activated eosinophils release granule proteins such as ECP and EDN. ECP is a cytotoxic agent secreted by activated eosinophils during allergic and inflammatory processes and it is elevated in AD patients [32]. EDN may also be useful to determine eosinophil activity in allergic diseases [22]. This study demonstrated that serum parameters including ECP and EDN were significantly different according to SCORAD. EDN level was higher in the moderate and severe group compared to the mild group, while ECP level was higher in the severe group compared to the mild and moderate group. Patients with FLG P478S TT and history of allergic rhinitis showed a higher EDN level, and among them, patients with asthma showed a higher ECP level. These results strengthen the findings of the previous study which demonstrated that EDN reflects strongly disease severity in AD [33] and increased ECP levels correspond to asthma symptom onset [34]. FLG P478S was suggested to hamper protease cleavage and it was associated with the risk of developing asthma or allergic rhinitis [15,35]. We consider that P478S may act as a modifier of AD especially associated with allergic rhinitis and asthma represented by serologic parameters suggesting disease severity.
There are acknowledged limitations of this study. First, although we found a statistically significant association between FLG-SNVs and clinical phenotypes, it was not individually supported by multiple comparisons, previous investigation or accepted physiologic models. Therefore, it remained to be supplemented with an experimental evidence of the functional role of the SNVs. In addition, the pathomechanisms of filaggrin or filaggrin-modifying natural moisturizing factors in atopic dermatitis are complex and more works are required to establish influences FLG-SNVs on epidermal barrier defects [36].
In conclusion, this study revealed the association of FLG-SNVs with AD-associated minor clinical features such as asthma, cheilitis, NS-HFD, and scalp scale. There is a possibility that FLG-SNV rs71625199 may be associated with higher environmental allergic sensitization. We also suggest that FLG P478S is a kind of disease modifier which affects serologic parameters such as EDN and ECP.
Supporting information S1