Introduction

Skin is the largest organ in the body and serves a multitude of physiological roles including barrier function and thermal regulation. Epidermis, the skin’s outer layer, is derived from the embryonic ectoderm through a series of steps that involve commitment, proliferation and differentiation (Blanpain and Fuchs 2009; Fuchs 1990; Niemann and Watt 2002). After achieving complete maturation, the epidermis is continuously regenerated throughout life to replenish skin cells that are sloughed at a rate of 2 million per hour (Milstone 2004). The mechanisms that underpin the development and maintenance of the epidermis are the subject of intense investigation, the ultimate goal of which is to better understand the disease mechanisms and to promote skin health.

Ectodermal dysplasia is a collective term that encompasses developmental disorders of the skin and its appendages as well as other ectoderm-derived tissues e.g., teeth, hair, nails and craniofacial structures. Although ectodermal dysplasia is broadly classified as hidrotic (when sweat glands function properly) and hypohidrotic, the phenotype is usually highly variable even within classes (Wright et al. 2014). On the other hand, recent molecular classifications provide a more meaningful categorization that allows for comparison based on molecular pathoetiology (Pagnan and Visinoni 2014).

Many genes are known to be mutated in ectodermal dysplasia ranging from the more common e.g., EDA to the very rare e.g., COG6 (Döffinger et al. 2001; Shaheen et al. 2013). Despite the marked progress made in recent years in deciphering the genetics of ectodermal dysplasia, most patients remain without a molecular diagnosis. TP63 is a particularly interesting ectodermal dysplasia gene, because it encodes a master regulator that serves as a molecular hub for virtually all known regulatory networks that control skin development and maintenance (Koster et al. 2004; Koster and Roop 2004; Truong et al. 2006). Not surprisingly, many of the genes that cause ectodermal dysplasia and other disorders of the integument encode proteins that regulate and/or are regulated by p63 (Koster 2010; Mitchell et al. 2012). Therefore, genes that are necessary for p63 regulation are attractive candidates for human skin diseases, especially when compelling animal models are available. In this study, we describe unbiased positional mapping in a large human family with a novel, autosomal dominant form of ectodermal dysplasia that led us to identify a novel, likely pathogenic variant in KDF1, encoding a recently described keratinocyte differentiation factor that acts upstream of p63 (Lee et al. 2013). The recapitulation of the human phenotype by the two available knockout mouse models supports a potentially causal link between the variant we identified and this novel form of ectodermal dysplasia.

Materials and methods

Human subjects

All patients were evaluated clinically by board-certified clinical geneticists and dermatologists. Clinical evaluation involved history taking, including family history, physical examination, electronic microscopic examination of hair shafts, histopathological examination of the skin and clinical exome sequencing. When clinical exome sequencing failed to reveal a likely causal mutation (for details about potential causes, please refer to (Shamseldin et al. 2016)), the entire family was recruited under an IRB-approved research protocol (KFSHRC RAC# 2121053) with informed consent. Blood was obtained from all affected and unaffected family members for subsequent genetic studies. Clinical photographs were obtained with a separate consent.

Positional mapping, exome sequencing and variant filtering

Axiom SNP chip genotyping was performed on all affected and unaffected members of the pedigree. AutoSNPa was used to exclude the remote possibility of pseudodominant inheritance, and DominantMapper was implemented to identify haplotypes that are exclusively shared by the affected members assuming a fully penetrant model (Carr et al. 2011). Exome capture was performed using TruSeq Exome Enrichment kit (Illumina) following the manufacturer’s protocol. Samples were prepared as an Illumina sequencing library, and in the second step, the sequencing libraries were enriched for the desired target using the Illumina Exome Enrichment protocol. The captured libraries were sequenced using Illumina HiSeq 2000 Sequencer. The reads were mapped against UCSC hg19 (http://genome.ucsc.edu/) by BWA (http://bio-bwa.sourceforge.net/). The SNPs and Indels were detected by SAMtools (http://samtools.sourceforge.net/). Variants from WES were filtered such that only novel (or very low frequency <0.1%), coding/splicing, homozygous variants that are within the candidate haplotypes of the affected individual and are predicted to be pathogenic were considered as likely causal variants. Frequency of variants was determined using publicly available variant databases (1000 Genomes, Exome Variant Server and ExAC) as well as a database of 817 in-house ethnically-matched exomes. Pathogenicity was likely if the mutation is loss-of-function (splicing/truncating) or, in the case of missense/in-frame indels, removes a highly conserved amino acid and is predicted to be pathogenic by the three in silico prediction modules PolyPhen, SIFT and CADD.

Computational structural analysis

Sequences were retrieved from the UniProt database. BLAST and SwissModel (Arnold et al. 2006) were used to search for suitable structural templates in the protein data bank (PDB). SwissModel and RaptorX (Kallberg et al. 2014) were used to produce homology models. Models were manually inspected, and mutations evaluated, using the PyMOL program (pymol.org). Disorder and secondary structure elements were predicted using RaptorX. Quark (Xu and Zhang 2012) was used for ab inito structure predictions. The presence of transmembrane helices and signal peptides was assessed using Phobius (Kall et al. 2004). Functional information was compiled from various resources, including UniProt and literature searches.

Results

Clinical report

A 30-year-old female, product of a non-consanguineous marriage was referred for evaluation because of suspected ectodermal dysplasia. She complained of thin eyebrows, absent teeth (status post dental implantation at age 23 years), nail changes, recurrent abscesses in the axillae and groin and decreased sweating. She was healthy otherwise. Her family history was significant for an affected father and multiple affected siblings who shared the same phenotype except for a sister who additionally had cleft palate status post surgical repair (Fig. 1; Table 1). Furthermore, she had one affected son who, in addition to his mother’s features, had natal teeth status post extraction at 2 weeks of age. Physical examination revealed lusterless scalp hair without scarring alopecia, thinned eyebrows and lack of natural teeth. She had keratosis pilaris, marked accentuation of palmar creases, and multiple scars of old healed lesions of hidradenitis suppurativa. Fingernails were normal, while toenails showed pincer nail deformity in the big toes and mild subungual hyperkeratosis in the others (Fig. 1). Complete blood count, liver profile and bone profile were normal. Lumbar spine coccyx X-ray showed sclerotic changes in the right femoral neck, and vertebral endplates with multiple stress fractures of long bones. Bone densitometry was normal. Her abdominal and renal ultrasounds were unremarkable. Table 1 summarizes the clinical features of all affected members. A skin biopsy from her 16-year-old affected brother showed hypoplastic sweat glands with epidermal orthokeratosis and papillomatosis, and some hair follicles showed follicular plug with distorted infundibulum. These findings were consistent with hypohidrotic ectodermal dysplasia and keratosis pilaris, respectively. Clinical exome sequencing did not reveal a likely causal variant in any known ectodermal dysplasia gene and did not suggest any likely candidate gene.

Fig. 1
figure 1

a Pedigree of the study family. b Facial dysmorphism (II:8) in the form of thinning of lateral eyebrows, saddle nose, and short philtrum. c Keratosis pilaris and scars of hidradenitis suppurativa (II:8). d Prominent hair follicles with resolving hidradenitis suppurativa lesion (II:7). e Natal teeth (III:2). f Absence of teeth (I:1). g Dystrophic toenails (II:8)

Full size image
Table 1 Clinical findings of the patients with ectodermal dysplasia
Full size table

Identification of a novel variant in KDF1

Lack of consanguinity, vertical transmission and involvement of both genders in the study family were highly consistent with an autosomal dominant inheritance. This was further supported by autozygosity mapping that was not consistent with pseudodominant inheritance. Assuming a fully penetrant model, we searched for all haplotypes that are exclusively shared by all seven affected members (Fig. 2). Within these candidate haplotypes, exome sequencing revealed 682 heterozygous coding/splicing variants including 77 that are absent in dbSNP (Fig. 2). Of these, there were four variants that are absent in 817 Saudi exomes and in ExAC (Table S1). Fortunately, there are available knockout mouse for three of these genes. The mouse phenotype for ASTN1 and ARF4 knockout did not involve any ectodermal findings, making the variants identified in these two genes unlikely candidates. No mouse models are available for IFIT5; so, the variant we identified therein (NM_012420:exon2:c.823G>T:p.A275S) remained a positional candidate with no further biological support from the literature. On the other hand, the phenotype of Kdf1−/− was highly reminiscent of the phenotype we report in the study family (see below). The variant therein (NM_152365:exon2:c.753C>A:p.F251L) was predicted to be pathogenic by PolyPhen (0.99), SIFT (0.01) and CADD (26). Sanger sequencing confirmed that all affected members were heterozygous for the change but none of the unaffected members. As expected from the relatively small family size, LOD score under the fully penetrant dominant model for this variant (1.8) did not reach significance.

Fig. 2
figure 2

a Chromosome ideograms with the candidate haplotypes as indicated by DominantMapper denoted by red lines. b Filtering scheme of the exome variants. c Cartoon of KDF1 and sequence chromatogram of the candidate variant. d Multisequence alignment showing strong conservation of the affected amino acid residue

Full size image

We used structural bioinformatics to investigate the mechanistic basis for the F251L mutation. There is no homologous experimental structure available that would serve as a good-confidence support for modeling the 3D structure of KDF1, or larger fragments thereof. Prediction of secondary structure and disorder shows that KDF1 is a largely flexible protein with only a few isolated secondary structures, mostly helices (Figure S1). No transmembrane helices are predicted. This indicates that KDF1 functions as a protein-binding adaptor, scaffold and/or cofactor, rather than an enzyme or transmembrane transporter or receptor. F251 is located in a region that is strongly predicted to form a 25-residue-long helical segment. Ab initio 3D structure prediction suggested that this helical segment is located in a 90-residue-long domain that may adopt a small folded region (Figure S1). Within this region, F251 is predicted to be buried (solvent inaccessible), and be part of a hydrophobic patch on an amphipathic helical region. S218, located at the N-terminal end of the predicted helical fragment, has been reported to be phosphorylated. Therefore, S218 interacts with kinases, and might modulate binding functions (to membrane or ligands) of the helical fragment containing F251. F251 might, therefore, be involved in stabilising a hydrophobic core of a small domain and/or be part of a helical ligand-binding motif. In either case, its substitution by the much smaller leucine might perturb its (intra- or intermolecular) interactions, and, hence, compromise the function of KDF1.

Discussion

Skin development is highly conserved between humans and mouse, which makes the latter an attractive model for studying the pathoetiology of skin developmental disorders using forward and reverse genetics approaches. Recently, Lee and colleagues reported the identification of a mouse mutant they called shorthand (shd) in an ENU mutagenesis screen (Lee et al. 2013). Grossly, shd had taught thick skin. However, that skin had severely impaired barrier function, and histological examination revealed marked disorganization of the layers of the epidermis. Specifically, the outermost layer of the epidermis (cornified layer) is lacking and the rest of the suprabasal layers have mixed identities. The basal layer itself was intact but with markedly increased mitotic index compared to controls. The phenotype is highly reminiscent of three other mutants: Irf6, Ikka and Stratifin, and to a lesser extent Ripk4. They mapped shd to a splicing mutation in a poorly characterized gene, which they termed Kdf1 for keratinocyte differentiation factor 1. Reassuringly, a reverse genetics approach whereby Kdf1 was knocked out using a genetrap vector recapitulated the phenotype of shd. In view of the resemblance between shd and the knockout mice for Stratifin, genetic interaction between Kdf1 and Staratifin was tested, and double heterozygotes were found to, indeed, recapitulate many phenotypic aspects of single homozygotes. Of note, Kdf1 knockout mice showed enhanced expression of p63, and deficiency of p63 was found to rescue shd, which is otherwise uniformly lethal due to esophageal adhesions and cleft palate (Lee et al. 2013).

The study by Lee and colleagues clearly show a non-redundant role of KDF1 in skin development that is mediated at least in part by its regulatory influence on p63. KDF1 has no clear homologues despite its very strong conservation in mammals and, to a lesser extent, in other animals. Therefore, the genetic interaction with the much more extensively studied Stratifin offers an opportunity to place KDF1 in a molecular developmental context. Stratifin (also known as 14-3-3σ) is a tumor suppressor that was originally identified as an exclusive epithelial marker (Prasad et al. 1992). Stratifin is a target of p63 but it also exerts a feedback effect and targets p63 to proteasome degradation (Fomenkov et al. 2004; Westfall et al. 2003). Mice deficient in stratifin display a remarkably similar histological phenotype to shd, i.e., thickened skin with enhanced proliferation and impaired proliferation that is similarly rescued by p63 deficiency (Li et al. 2011).

Although milder, the phenotype we describe in this study bears remarkable resemblance to that of shd, which suggests that the candidate variant we identified in KDF1 may be causally linked to the phenotype. Unfortunately, Kdf1−/− die shortly after birth; so, it is not possible to comment on the dental and hair phenotype that are prominent in the human patients we describe here. Also, Lee and colleagues did not comment on the morphology of sweat glands in their histological examination of the skin although they did demonstrate strong expression of Kdf1 in developing hair follicles (Lee et al. 2013). However, the highly similar phenotype between kdf1−/− and stratifin−/−, and the marked hair loss (sweat gland morphology was not reported) in stratifin−/+ prompting the name repeated epilation (stratifin mutant is semidominant) (Guenet et al. 1979; Herron et al. 2005; Holbrook et al. 1982; Li et al. 2005) suggest that KDF1 deficiency is similarly likely to cause hair loss. The dominant nature of the variant we identified in humans can be reconciled with the apparently haplosufficient kdf1−/+ mice by invoking interspecies differences.

In conclusion, we describe a family with a novel, autosomal dominant form of ectodermal dysplasia that we suggest is the mild human equivalent of the shd mutant phenotype in mouse based on the finding of a novel KDF1 variant by positional mapping and exome sequencing. Future patients with KDF1 mutations, including those with copy number losses involving KDF1, will help delineate the phenotype of KDF1 deficiency in humans and its mutational mechanism.