Filaggrin Genotype Determines Functional and Molecular Alterations in Skin of Patients with Atopic Dermatitis and Ichthyosis Vulgaris

Background Several common genetic and environmental disease mechanisms are important for the pathophysiology behind atopic dermatitis (AD). Filaggrin (FLG) loss-of-function is of great significance for barrier impairment in AD and ichthyosis vulgaris (IV), which is commonly associated with AD. The molecular background is, however, complex and various clusters of genes are altered, including inflammatory and epidermal-differentiation genes. Objective The objective was to study whether the functional and molecular alterations in AD and IV skin depend directly on FLG loss-of-function, and whether FLG genotype determines the type of downstream molecular pathway affected. Methods and Findings Patients with AD/IV (n = 43) and controls (n = 15) were recruited from two Swedish outpatient clinics and a Swedish AD family material with known FLG genotype. They were clinically examined and their medical history recorded using a standardized questionnaire. Blood samples and punch biopsies were taken and trans-epidermal water loss (TEWL) and skin pH was assessed with standard techniques. In addition to FLG genotyping, the STS gene was analyzed to exclude X-linked recessive ichthyosis (XLI). Microarrays and quantitative real-time PCR were used to compare differences in gene expression depending on FLG genotype. Several different signalling pathways were altered depending on FLG genotype in patients suffering from AD or AD/IV. Disease severity, TEWL and pH follow FLG deficiency in the skin; and the number of altered genes and pathways are correlated to FLG mRNA expression. Conclusions We emphasize further the role of FLG in skin-barrier integrity and the complex compensatory activation of signalling pathways. This involves inflammation, epidermal differentiation, lipid metabolism, cell signalling and adhesion in response to FLG-dependent skin-barrier dysfunction.


Introduction
Atopic dermatitis (AD; OMIM #605803) is a common chronic, non-contagious, inflammatory skin disorder. Clinical manifestations include early onset of dry skin, pruritus, eczema with typical age-dependent distribution, and personal or family history of atopic disease [1]. Knowledge of the pathophysiology behind the disease is emerging, several common genetic, environmental disease mechanisms and individual trigger factors being of importance [2]. Central in the pathogenesis are combinations of inherited and acquired insults thought to alter epidermal structure. These changes in the physiological skin barrier predispose to increased allergen presentation and are followed by immune activation, which in turn has negative consequences for skinbarrier homeostasis [3]. Impaired homeostasis of the skin leads to increased trans-epidermal water loss (TEWL) and changes in gene expression patterns [4] and enzymatic activity [5].
The most common monogenic disorder of keratinisation, ichthyosis vulgaris (IV; OMIM # 146700), is associated with AD and related atopic manifestations in up to 50% [6]. This contrasts with X-linked recessive ichthyosis (XLI; OMIM # 308100), which is due to mutations in the STS gene leading to accumulation of cholesterol sulphate in the stratum corneum. XLI occurs almost exclusively in males and may look almost indistinguishable from IV. However, skin histology and surface pH differ in the two conditions [7] and no association to AD has been reported in XLI. In 2006, it was found that mutations in the FLG gene resulting in filaggrin (FLG) dysfunction are the causative genetic factor for IV [8]. Following the frequent co-existence of IV and AD it was also discovered that 20-40% of European and Asian patients with moderate-to-severe AD carry FLG mutations. This is so far the most significant genetic finding associated with AD [9]. FLG is important for the structural integrity of the skin, and other functions are attributed to acidic degradation products of FLG, e.g. urocanic acid (UCA) and pyrrolidone carboxylic acid (PCA). These are components of natural moisturizing factors (NMFs) [10] and contributes to maintaining a low pH in the stratum corneum (SC) [11].
In addition to FLG dysfunction, it has previously been demonstrated that the molecular background to the pathogenesis of AD is complex, and that several clusters of genes, including inflammatory and epidermal differentiation [4,12] are altered in lesional AD skin. We set out to study whether the functional and molecular alterations in AD and IV skin depend directly on FLG loss-of-function variants, and whether the FLG genotype determine the type of downstream molecular pathways affected.

Patient material
Patients (n = 43) with AD (n = 35), AD and IV (n = 5) and IV (n = 3) together with controls (n = 15; subjects without past or present history of AD, dry skin or other atopic manifestations) were identified at the dermatology outpatient clinics at Karolinska University Hospital Solna, Sophiahemmet Stockholm and Uppsala University Hospital; or recruited from a Swedish family material with known FLG genotype as described previously [13]. All patients were investigated by a dermatologist performing clinical examination and recording medical history with a standardized questionnaire. Inclusion criteria were: age 18-65 years and diagnosed AD and/or IV. Exclusion criteria were pregnancy; other concomitant skin disease; recent UV-treatment; or recent use of topical or systemic corticosteroids, systemic immunosuppressives or systemic retinoids (,4 weeks). AD was diagnosed according to the UK Working Party's diagnostic criteria and the disease severity for AD was assessed using the scoring atopic dermatitis index (SCORAD) [14]. IV was diagnosed by clinical examination and genetic testing of the FLG gene, and in male patients with ichthyosis genetic testing of the steroid sulphatase (STS) gene to rule out XLI. Other atopic manifestations such as allergic asthma and allergic rhinoconjunctivitis were assessed through the questionnaire. Blood samples and punch biopsies were taken from all patients and controls. Two 3 mm punch biopsies were obtained from a non-lesional area on each patients forearm, after local anaesthetic with lidocain hydrochloride with adrenalin (Astra Zeneca, Södertä lje, Sweden). TEWL was assessed using a Tewameter TM 300 Multi Probe Adapter (Courage+Khazaka electronic GmbH, Köln, Germany) and skin pH was measured using a skin-pH-Meter PH 905 Multi Probe Adapter (Courage+Khazaka electronic GmbH). TEWL and pH were measured from the forearms of patients and controls. The patients were divided into three groups (AD FLG+/+; AD FLG+/2 and AD/IV FLG2/2) depending on genotype of the four most prevalent European FLG-mutations [6].
Subjects for microarray analysis. Five patients from each patient group (AD FLG+/+, AD FLG+/2 and AD/IV FLG2/2) were randomly selected for microarray analysis after removing outliers in TEWL and pH. In the AD FLG+/2 group four were heterozygous carriers of the 2282del4 mutation and one was a heterozygous carrier of the R501X mutation. In the AD/IV FLG2/2 group four were homozygous carriers of the 2282del4 mutation and one was a homozygous carrier of the R501X mutation. All selected patients had AD. The groups were compared to five healthy controls randomly selected after removing outliers compared to the rest of the control group regarding TEWL or pH. They carried no tested FLG mutations.
Ethics. The study was conducted according to Declaration of Helsinki principles and was approved by the regional ethics committees at Uppsala University and at Karolinska Institute. All study participants gave written informed consent.

Genotyping
Genomic DNA was isolated from peripheral blood using QIAampH DNA mini kit (Qiagen, Hilden, Germany).
FLG genotyping. FLG genotyping was performed with allelic discrimination in patients and controls for the prevalent European FLG mutations R501X, S3247X and R2447X. Genomic DNA was PCR-amplified in 384-well plates. Each well contained 5 ng genomic DNA, 2.5 ml TaqMan Universal PCR Master Mix, 0.125 ml specific Taqman assay solution and 2.375 ml H 2 O. Allelic discrimination was carried out with the ABI PRISMH 7900HT Sequence Detection System and the SDS 2.2.1 sequence detection system program (Applied Biosystems, Stockholm, Sweden).
Primers and PCR conditions for tested FLG mutations were as described previously [6].
STS genotyping. Multiplex Ligation-dependent Probe Amplification (MLPA) analysis was run for the STS gene using the P160 A2 kit (MRC-Holland, Amsterdam, the Netherlands), as previously described [15] with minor modifications. Typically 100 ng genomic DNA was amplified. The sample was analyzed on the ABI 3130xl Genetic Analyzer. In addition, exon 1-10 of the STS gene was sequenced using primers and PCR conditions previously described in male patients where no deletion was detected [16].
All primer pairs were confirmed specific by database queries (using BLAST and BLAT). The polyphen [17] and the Alamut mutation interpretation software (Interactive Biosoftware, Rouen, France) was used to predict pathogenicity of single nucleotide polymorphisms (SNPs) compared to reference sequence.

RNA extraction
Skin biopsies were trimmed of subcutaneous fat prior to homogenization. The biopsies were placed in 1 ml Trizol (Invitrogen) and subsequently homogenized using a Polytron homogenizer. Total RNA was isolated as described elsewhere [18]. Total RNA concentration was determined with spectrophotometric analysis and purity was analyzed by the 260:280 absorbance ratios.

Microarray analysis
Microarray hybridization and scanning. Trizol-extracted total RNA was purified using the RNeasy MiniKit (Qiagen, Valencia, CA). Samples were re-quantified with spectroscopy, and purity was re-analyzed through the 260:280 absorbance ratios. RNA quality and integrity were assessed and ensured using Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA) and RNA 6000 NanoAssay. Hybridization was performed with Human Gene 1.0 ST arrays (Affymetrix, Inc, Santa Clara, CA). Briefly, 100 ng of total RNA from each sample was reversetranscribed to complementary DNA (cDNA) using the Ambion WT Expression kit. The cDNA was subsequently converted to complementary RNA using in vitro transcription with an amplification kit. 10 mg purified complementary RNA was used as a template for another cycle of first-strand cDNA synthesis. Single-stranded cDNA samples were fragmented and end-labeled with the Gene Chip WT cDNA Synthesis Kit (Affymetrix). Approximately 25 ng/ml cDNA was added to the hybridization cocktail, followed by hybridization with the Human Gene 1.0 ST Array GeneChip at 45uC for 16 hours. This was then washed using the Affymetrix Fluidics Station 450. A final step was to measure probe intensities using the GeneChip Scanner 3000. The raw intensity data was normalized using Command Console Software (Affymetrix). The average fluorescence intensity of all annotated genes was calculated using the Robust Multiarray Analysis (RMA) algorithm [19], including a quartile normalization (all arrays are considered to have an equal intensity distribution) and using a background correction for GC-content.
Microarray gene expression, data processing, quality control and statistical analysis. The values of individual probes belonging to one probe set were averaged and normalized using Partek Genomics Suite 6.4 (Partek Inc., St. Louis, MO, USA, www.partek.com), from which probes with lowest available p-value and a known GenBank accession ID correspondence were selected for functional analysis. The distribution of the intensity values on the individual arrays was visualized in a signal histogram. One sample was removed due to deviating intensity values compared to the other samples. No other obvious outliers were detected. The intensity values of probe sets specific for the pre-labeled hybridization controls were analyzed and corresponded with the expected values. To check overall data quality, the array contained probe sets for exonic and intronic regions of reference genes (genes thought to be constitutively expressed in many different samples). Their probe set intensities were used to calculate the difference between the area under the curve of the positive and negative probe sets according to the manufacturer's instructions [20].
Genes of interest, all over two-fold up/down-regulated genes (p,0.0005), were analyzed using the Database for Annotation, Visualization and Integrated Discovery functional annotation tool [21] with KEGG pathway analysis.
Functional annotations were also carried out using the Ingenuity Pathway Analysis (IPA; Ingenuity Systems, Redwood City, CA, http://www.ingenuity.com/), in which gene symbols and fold changes of the up-and down-regulated genes were imported.
All microarray data comply with MIAME guidelines and are deposited in ArrayExpress.
Identification of enriched cytobands. 2292 induced genes and 2076 repressed genes in the AD groups (Table 1) were analyzed for their enrichment in human cytoband regions and gene ontology (GO) terms as defined using the DAVID bioinformatics resources [21], with an individual cutoff for each gene of p,0.0005.

Statistical analysis
To identify differentially expressed genes between the different experimental groups in the microarray analysis, a two-way analysis of variance (ANOVA) was performed for each patient group compared to the healthy control group using Partek Genomics Suite 6.4. For each comparison between two experimental groups the fold change of every annotated gene, together with their corresponding p-value, was exported to Microsoft Office Excel. For quantitative Real-Time PCR, the relative mRNA expression and statistical significance were calculated using the REST 2009 software (available at www.qiagen.com) using Fisher's exact test. For genes chosen for pathway analysis, significance was corrected with Bonferroni multiple testing. Statistical significance for SCORAD was calculated using student's t -test. P-values,0.05 were considered as significant.

Genotyping and clinical presentation
Among the included patients (n = 43) fourteen carried none of the prevalent FLG mutations tested and were included in the AD FLG+/+ group. Fourteen AD patients carried one prevalent heterozygous FLG mutation (one R501X; thirteen 2282del4) and were included in the AD FLG+/2 group. Fifteen AD/IV patients carried either a homozygous or a compound heterozygous FLG mutation (two R501X; nine 2282del4; one S3247X; two 2282del4/S3247X and one 2282del4/R501X) and were included in the AD/IV FLG2/2 group (of these fifteen patients, three had IV phenotype without AD at the time of examination). One patient was excluded from the AD/IV FLG2/2group after no FLG mutations were detected, and subsequent STS genotyping revealed a point mutation, recently published elsewhere [16]. In the AD FLG+/+ group 78.6% were females, the average age was 56 (range 28-78) and the mean SCORAD was 7.6 (range 0-14.7). In the AD FLG+/2 group, 64.3% were female, the average age was 54 (range 28-71) and the mean SCORAD was 15.4 (range 6.2-25.8). For the AD/IV FLG2/2 group 46.7% were female,

TEWL and pH
Significantly higher TEWL was observed in the AD/IV FLG2/2 and AD FLG+/2 than in the healthy control group. The mean TEWL was higher also in the AD FLG+/+ group, although this did not reach statistical significance. pH was significantly higher in the AD/IV FLG2/2 group than in the healthy control group. Mean pH was higher also for the AD FLG+/2 and FLG+/+ groups, although this did not reach statistical significance (Fig. 1a).

FLG mRNA expression depending on genotype
All patient -groups showed lower mRNA expression of FLG than the control group, both with microarray analysis and with quantitative real-time PCR (qPCR). The mRNA expression levels were lowest in the FLG2/2 group (array p = 0.000008; qPCR p = 0.001), but significantly reduced also in the FLG+/2 (array p = 0.04; qPCR p = 0.001) and with qPCR also in the FLG+/+ group (array p = 0.59; qPCR p = 0.04) than in the healthy control group (Fig. 1b).

Altered expression profiles in AD
The microarrays representing 28869 annotated genes with 764885 distinct probes were used to identify and compare the gene expression of AD skin compared to healthy skin, and the difference in expression pattern depending on FLG genotype. The design of the Human Gene 1.0 ST array was based on the March 2006 (UCSC hg18, NCBI Build 36) human sequence assembly, containing over 99 percent coverage of sequences present in the RefSeq database. A full list of significantly altered genes is provided as Table S1. Among these, the Partek Genomics suite 6.4 was used to detect 4368 differentially expressed genes (minimum 2-fold change and p-value,0.0005) ( Table 1). These differentially expressed genes were distributed according to the chromosomal enrichment illustrated in Table S2.

Distribution of differentially expressed genes
The distribution of differentially expressed genes in all patient groups depends on their FLG genotype (Fig. 2a). Hierarchical clustering was used to group these differentially expressed genes, based on similarity in expression across the samples and to group individuals on the basis of similarities in gene-expression patterns (Fig. 2b). Each column represents a single array experiment and clusters from Fig. 2a are marked I-VII, respectively.
Altered pathways for cellular development and differentiation, inflammatory response and cell-to-cell signaling in AD/IV skin compared to healthy controls, regardless of FLG status. Ingenuity Pathway Analysis of differentially expressed genes for all patient groups reveals a pathway mapped to inflammatory response that was significantly induced compared to the control group (Fig. 3). In addition, there were several altered pathways mapped to cellular development and differentiation compared to healthy controls (Fig. 3).
Significantly altered pathways depending on FLG genotype status. In AD/IV skin with FLG2/2 genotype, several pathways were significantly altered compared to the healthy control group. Focal adhesion, extracellular matrix receptor interaction, regulation of actin cytoskeleton and calcium signaling pathways showed significantly altered expression ( Table 2).
In AD skin with FLG+/2 genotype, focal adhesion and extracellular matrix receptor interaction pathways displayed, similarly to the FLG2/2 group, a significant deviation compared to the healthy controls, together with ABC transporting pathway and actin cytoskeleton regulation ( Table 2).
For AD skin without FLG mutations focal adhesion, ECM receptor interaction and regulation of actin cytoskeleton show a deviating trend (Table 2), but this was not statistically significant. For a list of candidate genes mapped to altered pathways depending on FLG genotype see Table S3. Quantitative Real-Time PCR Identification of candidate genes. For complimentary testing of selected significantly differentially expressed genes mapped in the pathway analysis qPCR was performed from genes in cytobands previously associated to AD (Table S2 and  Table S3) and with fold changes close to two-fold up-or downregulation. Of significantly altered pathways, ITGA3 and CTNNA1 were mapped to focal adhesion, LAMB3 to extracellular matrix receptor interaction, ITGAE to actin cytoskeleton, JAM2 to tight junction pathway, VAV1 and CD28 to T cell receptor signaling pathway, and, PTK2B4 was mapped to calcium signaling pathway; all mapped using DAVID bioinformatics resources [21] with the KEGG pathway analysis option. TLR2 and STAT2 represent genes associated with immunological response mapped using Ingenuity Pathways Analysis (Fig. 3).

Discussion
FLG was shown to be down-regulated both by microarray analysis and qPCR in all AD/IV groups compared to the healthy controls. Although there was a gradient of down-regulation depending on FLG genotype with the lowest FLG expression in patients with FLG2/2 genotype followed by the FLG+/2 . Genes with altered expression in patients with FLG wild type genotype (FLG+/+), heterozygote FLG mutation genotype (FLG+/2), and homozygote FLG mutation genotype (FLG2/2). Clusters containing differentially expressed genes in marked regions I-VII are corresponding regions in a) and b). A list of p-values and fold changes for all differentially expressed genes are described in Table S1. Hierarchal clustering analysis was performed in both the gene (row) and experiment (column) dimension. Contrast value for each gene is shown, e.g. the standardized mean difference between the gene's expression in the group versus overall expression. doi:10.1371/journal.pone.0028254.g002 genotype, the FLG+/+ group also displayed down-regulation compared to healthy control skin. Recent studies have shown that pro-inflammatory cytokines may modulate the expression of FLG, even in patients without FLG mutations, which might be one of the underlying explanations of our finding [22,23].
Many of the potential AD candidate genes significantly altered in our study were located in chromosomal regions previously linked to AD [24] (Table S2), further highlighting these regions as interesting loci for potential candidate genes involved in AD susceptibility. The distributions of these differentially expressed genes in our study depended on FLG genotype, where several clusters were unique for each group, and others overlapped (Fig. 2). Genes from these groups are mapped to significantly altered pathways in each patient group. The functional alterations evident from the significantly higher TEWL and pH (Fig. 1) in the FLGdeficient groups may influence the number of induced or repressed genes involved in tightly regulated processes such as inflammatory response following a more permeable barrier, as well as enzymatic activity where the pH level is important [11].
The importance of changes in TEWL and pH has recently been highlighted in FLG deficient skin; where reduced levels of FLG degradation products are proposed to increase TEWL and pH; decreasing stratum corneum hydration and altering enzymatic activity [25,26]. This may account for alterations in corneocyte and lipid organization within the SC [26]. Given the frequent phenotypic overlap between dry skin, IV and AD (evident in our FLG2/2 group as well); it is proposed that these functional alterations are important for the pathogenesis in both IV and AD skin with FLG deficiency. In support of this hypothesis, our AD patients without FLG mutations displayed lower functional barrier impairment measured by TEWL, lower pH and significantly lower mean SCORAD than AD patients with FLG mutations (FLGrepeat variation may also influence the phenotype [27]. However, we did not investigate this). In addition, the lowest number of significantly altered genes was detected in our AD FLG+/+ group. This suggests a correlation between number of affected genes, barrier impairment and disease severity among included AD patients.
Of the many genes previously associated to AD [24] several were also dysregulated in our array data, such as serine protease inhibitor kazal-type 5 (SPINK-5), mast cell chymase (CMA1 and interleukin 4 (IL-4) ( Table S1). Any discrepancies regarding expression of inflammatory mediators commonly found in AD may at least in part be due to lower expression of these genes in non-lesional skin. CD28 and STAT2 are two inflammatory markers that were confirmed to be altered also by qPCR. CD28 has been suggested to be involved in the inflammatory response in AD [28] and STAT2 has been described as a candidate gene involved in mediating pro-inflammatory cytokines [29]. In addition, genes mapped to adhesion such as CTNNA1, JAM2 and LAMB3 were also confirmed to be significantly altered. Defects in cell adhesion have recently been highlighted as important in AD pathogenesis, with the finding of impairment in tight junctions contributing to the barrier dysfunction and immune dysregulation [30]. Downregulation of tight-junction proteins such as occludin and ZO-1 has also been demonstrated in IV skin recently [26] and tallies with our gene-expression pattern in FLG-deficient skin (Table S1). The LAMB3 gene encodes laminin-5, a glycoprotein that anchors basal cells to the underlying basal membrane [31], whereas CTNNA1 and JAM2, in addition to their cell adhesion function, have been suggested to be involved in cell differentiation [32] and lymphocyte homing [33], respectively. Interestingly CTNNA1 and LAMB3 were significantly altered in the AD group without FLG mutations. The underlying explanation could in part be the effects of putative down-regulation of FLG also in this group, but as the expression of LAMB3 did not reach significance in the AD FLG2/2 group and neither did CTNNA1 or LAMB3 in the AD FLG+/2 group, other explanations are plausible, including that these genes are candidates for the primary pathogenesis in AD in addition to FLG deficiency. The molecular mechanisms involved in the phenotype of AD following the functional barrier impairment in our material involve altered pathways such as cytoskeleton structure, calciumand phospatidylinositol signaling and ATP binding cassette (ABC) transport system ( Table 2). It has been suggested that FLG is of importance for cytoskeleton organization by aggregating keratin intermediate filaments (KIFs); and that FLG deficiency may cause cytoskeleton abnormalities such as perinuclear keratin retraction in granular cells [26]. KIF polymerization is actin-dependent [34] and subsequently actin-cytoskeleton aberrations may contribute to the peripheral KIF retraction previously demonstrated in FLGdeficient skin [26]. However, the role of FLG in impaired intermediate filament aggregation has been challenged [35] and other factors than FLG deficiency may explain the alterations in the pathway for the regulation of the actin cytoskeleton. Our findings support this, as pathways for actin-cytoskeleton regulation were altered in all our patient groups including the group without FLG mutations. In addition, several keratins (including KRT1 and KRT10) were significantly down-regulated in AD patients both with and without FLG mutations (Table S1). As the actin filament system has been suggested to be involved in KIF transport [34], it is possible that increased actin cytoskeleton regulation is a compensatory mechanism following a lower keratin expression. Altogether, our data suggests that both keratin expression and KIF regulation are subject to modulation in AD skin independently of FLG mutations.
FLG may be involved in calcium metabolism in the skin [36], and the calcium gradient is important for epidermal differentiation -a loss of this gradient increases keratinocyte proliferation and decreases differentiation [37]. Impaired calcium metabolism has been demonstrated in other conditions where the skin barrier is disrupted, such as Hailey-Hailey disease [38] and in psoriatic skin [39]. Further, defective lipid transportation and defects in lamellar body extrusion have previously been reported in AD [40,41] and mutations in this pathway may cause severe ichthyotic conditions such as Harlequin Ichthyosis [42]. Our FLG-deficient groups show alterations both in the pathway for calcium signaling and for ABC transport system, indicating that alterations in these pathways are involved in the pathogenesis of IV and AD with FLG deficiency.
In conclusion, we have demonstrated that several functional and molecular mechanisms in vivo in patients suffering from AD and IV depend on FLG genotype. Disease severity of AD, the gradient of TEWL and pH follow loss of FLG expression in the skin; and the number of altered genes and pathways may be correlated to FLG mRNA expression. We here emphasize further the role of FLG for the functional integrity of the skin barrier and the complex subsequent signaling systems involving inflammation, epidermal differentiation, lipid metabolism, cell signalling and adhesion that are affected in response to FLG deficiency.