1. Keratins

1.1 Structure, Function and Expression

The cytoskeleton of all eukaryotic cells consists of three main networks: microfilaments (5–7nm diameter); microtubules (approximately 25nm diameter); and intermediate filaments (approximately 10nm diameter). Microfilaments and microtubules are involved in various cellular events including cell division, contraction, orientation and polarization, and anchorage. However, the function of intermediate filaments was largely undefined until pathogenic keratin mutations were identified in various skin fragility disorders. This provided increasing evidence that intermediate filaments represent a truly structural cytoskeleton within cells, making them resilient to everyday stress and physical trauma. There are more than 50 intermediate filament proteins, which are classified into at least six types on the basis of sequence homology, tissue specificity and immunologic characteristics. In the cytoplasm of epithelial cells, keratin intermediate filaments form a complex array of filaments from the nucleus to the plasma membrane. At this point they interact with specific protein junctions, desmosomes at intercellular membrane sites and hemidesmosomes at the basal cell surface (figure 1).

Fig. 1
figure 1

(a) Antibody staining of keratin filaments (green) in epithelial cells in culture, nuclei stained with 4c6-diamidino-2-phenylindole⋅2HCI (DAPI) (blue). (b) Abnormal clumping of keratin filaments (green) caused by a mutation in K14 (R125C).

Full size image

Keratins, of which there are at least 30 cytokeratins and trichocyte keratins (hair/nail keratins), form the type I and II groups of intermediate filament proteins. Analysis of recent sequence from the human genome identified some novel keratins bringing the total number to at least 49.[1] The type I, acidic keratins (K9–K20) are clustered on chromosome 17q12-q21[2,3] and type II, basic keratins (K1–K8) are located in a gene cluster on chromosome 12q11q-14.[2,4] K18 is the exception and maps to the type II gene cluster on chromosome 12.[5] Structurally keratins consist of a central α-helical rod domain of 310 amino acids flanked by non-helical head (V1) and tail (V2) domains (figure 2).[6] The rod domain is divided into the 1A, 1B, 2A and 2B domains each consisting of heptad repeats. Connecting these are three non-helical linkers, L1, L12 and L2 which are thought to provide flexibility to the rod domain. The type II keratins in addition have two subdomains, H1 and H2, which lie between the rod domain and the V1 and V2 domains, respectively. Within the 2B domain there is a ‘stutter’ sequence where the helix polarity is reversed. At the start and end of the rod domain are two short regions: the helix initiation motif at the amino terminus; and the helix termination motif at the carboxy terminus. The sequences of these regions are very highly conserved throughout the intermediate filament protein family. These helix boundary motifs are thought to be critical in overlap interactions during filament assembly.[7] All keratins of a given type have the same genomic organization and position of introns. Type I keratins have 8 exons and 7 introns and type II keratins have 9 exons and 8 introns. The standard nomenclature that is used throughout this review is, for example for keratin 1, K1 for the keratin protein and KRT1 for the corresponding gene.

Fig. 2
figure 2

Schematic diagram showing the basic protein structure of a keratin filament. The α-helical rod domain is divided into four domains, the 1A, 1B, 2A and 2B. Connecting these are the linkers, L1, L12 and L2. Flanking the rod domain are the variable V1 and V2 domains. At either end of the rod domain are the highly conserved helix initiation motif (amino terminus) and helix termination motif (carboxy terminus), shown in red. H1 and H2 subdomains are present in type II keratins. The ‘stutter’ sequence is marked by S.

Full size image

1.2 Keratin Filament Assembly

The first stage in assembly of a 10nm intermediate filament is the formation of a dimer. Keratins form obligatory heterodimers consisting of a type I and type II keratin. These align in parallel and in axial register to form a coiled coil α-helical dimer. The next stage is the formation of a tetramer by the alignment of two dimers. The precise spatial arrangement of these dimers is still uncertain and has been controversial. Cross-linking experiments indicate that there are three possible ways that keratin dimers could align.[8,9] Tetramers polymerize, both laterally and longitudinally to form 2–3nm protofilaments and when two of these align a 4.5nm protofibril is formed. The inter-twining of protofibrils results in a complete intermediate filament consisting, in the case of keratins, of 32 monomer chains in cross section.[10] The precise arrangement of the keratin filaments to form a three dimensional structure is not yet clear. Recently, some progress has been made in solving the crystal structure of intermediate filament proteins[11] although it is likely to take some years to solve the complete atomic structure of the highly insoluble proteins.

1.3 Keratin Expression Patterns

Keratins are generally expressed in specific pairs, consisting of a type I and a type II keratin in a tissue-specific and differentiation-specific manner[12] as summarized in table I.

Table I
figure Tab1

Expression patterns of keratins

Full size image

2. Keratin Diseases

2.1 Epidermolysis Bullosa Simplex (EBS): The First Identified Human Keratin Disease

Epidermolysis bullosa simplex (EBS) was the first inherited disorder to be associated with keratin mutations. To date, mutations in 18 keratin genes are associated with human disorders.

EBS is one of the three main forms of epidermolysis bullosa (EB). The other forms are junctional EB and dystrophic EB. EB classification is on the basis of the ultrastructural level of separation within the skin, an assessment of the clinical features, and the mode of inheritance (autosomal dominant or recessive). The characteristic intra-epidermal blistering of EBS is due to cytolysis in the sub-nuclear region of the basal keratinocytes. EBS is normally inherited as an autosomal dominant trait, although there are rare cases of recessive EBS (see below). In general EBS is considered to be the mildest form of EB with an incidence of approximately 1 : 50 000.[13–15]

The three most common types of autosomal dominant EBS are the Dowling-Meara (section 2.1.1), Weber-Cockayne (section 2.1.2) and Köbner (section 2.1.3) variants, as follows.

2.1.1 EBS Dowling-Meara

The Dowling-Meara variant (EBS-DM, On-line Mendelian Inheritance in Man [OMIM] 131760) is generally considered to be the most severe form of EBS.[16] Blistering may be extensive at birth and the disorder can be fatal in the first few months, therefore early diagnosis is important to enable appropriate care during this critical period. EBS-DM is inherited in an autosomal dominant manner, although many cases are sporadic. Blisters occur in herpetiform clusters on the trunk and proximal extremities and heal with minor scarring (figure 3) and milia may occur. Oral blistering is common, and the teeth and esophagus may be affected. There is often progressive hyperkeratosis of the palms and soles. Nails may be dystrophic or shed.[17] Histologically, EBS-DM differs from other forms of EBS. Abnormal clumping of keratin filaments within basal keratinocytes is observed. These precede basal cell cytolysis and intraepidermal blister formation.[18–20]

Fig. 3
figure 3

Clinical phenotypes caused by mutations in epidermal keratins. (a) Herpetiform pattern of blisters characteristic of the Dowling-Meara form of epidermolysis bullosa simplex (EBS) [EBS-DM]. Mutations in K5 and K14 underlie EBS. (b) Thickened hyperkeratotic skin typical of bullous congenital ichthyosiform erythroderma (BCIE), is caused by mutations in K1 and K10. (c) Epidermolytic palmoplantar keratoderma (EPPK), hyperkeratosis of the palms showing marked erythematous border is caused by mutations in K9.

Full size image

2.1.2 EBS Weber-Cockayne

In the Weber-Cockayne form of EBS (EBS-WC, OMIM 131800), blistering predominantly affects the hands and feet.[21,22] Data from the establishment of national EB registries in the UK and in the US suggests that this is the most common form of EBS.[14,15] Blisters develop in early or late childhood, often when a child is learning to walk. The blisters are induced by trauma from routine minor physical activities but can also develop spontaneously. The blisters sometimes bleed but heal without scarring or milia formation. Calluses may develop over areas of recurrent blistering. Hyperhidrosis of the feet is common and there may be mild hyperkeratosis of palms and soles.[23] There is a marked seasonal variation, with more severe blistering occurring in the warmer summer months.

2.1.3 EBS Köbner

EBS Köbner (EBS-K, OMIM 131900) is a subtype similar to EBS-WC but with generalized blistering on all parts of the body and in the oral cavity.[24] However, there is often no clear clinical distinction between EBS-WC and EBS-K as blistering in many patients with EBS-WC is not restricted to the feet and hands. Blisters develop at birth or in early infancy and heal without scarring. Secondary bacterial infections are a common problem. There may be mild to moderate hyperkeratosis of soles and palms and often hyperhidrosis of the feet. The nails may show mild involvement. There is a seasonal variation, like in EBS-WC, with more severe blistering in warm weather.

2.1.4 Identification of Keratin Mutations in EBS

Ultrastructural studies on EBS-DM provided the first indication that an abnormality in a basal cell structural protein could underlie at least some forms of EBS. Abnormal clumping of tonofilaments (keratin filaments) was observed in the basal keratinocytes in association with blistering.[18] Later studies showed tonofilament aggregation in some non-lesional, as well as in peri-lesional skin and in adnexal epithelia including sweat ducts, outer hair root sheaths and sebaceous glands.[19,20] Antibody staining demonstrated that the clumped tonofilaments consisted of the basal cell keratins, K5 and K14. Abnormal round bodies that stained for K5 and K14 were also present in cultured keratinocytes of EBS-DM patients.[19] It was proposed that the filament clumping was a primary event due to specific keratin abnormalities rather than as a consequence of blistering. However, the cause of keratin aggregation and the resulting cell fragility was unknown until studies in several independent laboratories identified the genetic defect in EBS. Using transgenic animals[25,26] and analysis of patient material[27–29] it was predicted and confirmed that mutations in the genes KRT5 and KRT14 that encode keratins K5 and K14 respectively, were the underlying cause of EBS.

Fuchs and colleagues had established the effects of keratin mutants in an in vitro system.[30–32] Expression of keratin mutants in epithelial cells caused aggregation of keratin filaments, (figure 1). To see the effect in vivo, K14 mutations were expressed in the skin of transgenic mice.[25,26] The resultant mice developed intraepidermal blistering at birth or after mild mechanical trauma. A range of EBS-like phenotypes was produced, dependent on the mutation. The most severe phenotype showed aggregation of tonofilaments in skin and in cultured keratinocytes, similar to that observed in EBS-DM.

These findings predicted that different EBS variants could be caused by mutations in K14. DNA sequence analysis was performed on two EBS-DM families. Heterozygous missense mutations were identified in K14, at the same codon arginine 125 (R125) but different bases changes.[28] This arginine is highly conserved throughout evolution. It is located in the helix initiation motif, a region already shown by in vitro deletion experiments to be critical for intermediate filament assembly.

Concurrently, Epstein and colleagues confirmed by genetic linkage analysis and mutation detection that the milder variants, EBS-K and EBS-WC, were also caused by keratin mutations. Linkage of EBS-K in one family to the type I keratin gene cluster on chromosome 17 implicated an abnormality in K14.[33] Consequently, a heterozygous missense mutation in the 2B domain of K14 was identified in this family.[27] In a family affected by EBS-WC, the disorder mapped to the type II keratin gene cluster on chromosome 12. This implied a defect in K5, the expression partner of K14.[33] Additional EBS-K and EBS-WC families were mapped to the keratin gene clusters on chromosomes 12 and 17.[34–36]

A third group simultaneously detected a K5 abnormality in EBS-DM.[29] Keratins expressed by cultured keratinocytes were analyzed by Western blot and stained with monoclonal antibodies. Loss of a K5 epitope on Western blots identified the region of the protein that was mutated. By DNA sequencing, a heterozygous missense mutation was found in the highly conserved helix termination motif of K5.[29]

These first pathogenic keratin mutations in K5 and K14 not only identified the cause of EBS but also confirmed the hypothesis that keratins have a primarily structural role within epithelial cells. They demonstrated that mutations in different members of the heterodimer (K5 or K14) could produce the same phenotype (phenocopies). Also, as predicted by transgenic studies,[25] different EBS phenotypes could be caused by mutations within the same gene.

There are now many reports of mutations in K5 and K14 in EBS-DM and in the milder forms, EBS-K and EBS-WC (keratin mutation database, table II).[37] The majority of these are heterozygous missense mutations. Most mutations in EBS-DM are in the highly conserved helix initiation motif of K14 (table III). Codon R125 has been identified as a ‘hot spot’ for mutations. Mutations at this position occur in a CpG dinucleotide leading to chemical instability, and this may account for the number of mutations found here.[38] Mutations in the conserved helix termination motif of K14 also cause EBS-DM.[39,40] There are fewer mutations in K5 underlying EBS-DM but those reported are also in the helix boundary motifs (keratin mutation database, table II). Some of the more unusual mutations in EBS-DM include a premature stop codon in the helix termination motif of K5,[41] and a donor splice site mutation leading to the deletion of 22 amino acids in the H1 and 1A domains of K5.[42] Another report of a premature stop codon in the helix termination motif of K5 leads to an EBS phenotype with severe blistering in childhood and unusually severe palmoplantar hyperkeratosis.[43] Mutations producing the milder phenotypes EBS-K and EBS-WC have been found throughout the rod domains of K5 and K14 (table III). Surprisingly, some of these are in the L12 linker domain, regions thought to provide some flexibility to the molecule. Mutations in these regions had not been predicted to produce a clinical phenotype. A number of mutations causing EBS-WC have also been found in the H1 head domain of K5. Here, codon 161 is particularly susceptible to mutation (I161S).[44] There is also one report of a deletion mutation in K14 in a case of EBS-WC.[45]

Table II
figure Tab2

Useful websites for research and diagnosis of keratin disorders

Full size image
Table III
figure Tab3

Human keratin gene mutations as at January 2002

Full size image

Correlations drawn between the positions of mutations versus phenotype suggest the more severe phenotype EBS-DM is caused by mutations in the helix boundary motifs. These domains are predicted to be critical in overlap interactions during filament assembly.[7] Mutations in other regions of the rod domain and just outside in the H1 domain can be better tolerated leading to the milder phenotypes, EBS-K and EBS-WC. However, there are a few exceptions where mutations in the helix termination motif of K14 produce the milder EBS-K and EBS-WC phenotypes.[39] There are also instances where the actual base change together with the position within the rod domain may be important in determining the phenotype. In K14, mutations at codon 119 (M119) can produce all 3 EBS phenotypes. M119T causes EBS-DM, [46,47] M119V causes EBS-K[47] and M119I leads to EBS-WC.[48] Interestingly, one patient from the EBS-WC family was homozygous for the mutation due to a consanguineous marriage and developed a more severe EBS-K phenotype.[49] In this case the mutation is partially dominant unlike the fully recessive ablation mutations (see section 2.1.5) or fully dominant mutations where homozygotes have a similar phenotype to heterozygotes.[50]

2.1.5 Other Forms of EBS with Mutations in K5 and K14

EBS with Mottled Pigmentation

EBS with mottled pigmentation (EBS-MP, OMIM 131960) has been described as a variant of EBS-K.[51] Generalized non-scarring blisters develop at birth or during early infancy. The skin also has a mottled appearance due to a pigmentary abnormality, which fades with age to become barely recognizable in adults. Other associated characteristics include skin atrophy on the trunk and extremities, hyperkeratosis of palms and soles and dystrophic nails. Patients were analyzed for mutations in K5 and K14. The same missense mutation in K5 (L25P) has been identified in all seven cases reported at this time. This mutation is unusual in that it is in the non-conserved variable V1 domain.[52,53] The function of the non-helical V1 and V2 domains at either end of the rod domain are largely unknown, but they are thought to determine the specific function of each keratin. How this mutation produces the pigmentary changes is unknown.

Recessive EBS

In rare instances, EBS can be expressed as a recessive trait (R-EBS, OMIM 601001), in some cases as a result of consanguineous marriages. K14 mutations have been identified in several patients. In the first, a very mild EBS-WC like phenotype was observed in homozygotes. The heterozygous parents were clinically unaffected. The causative mutation was a homozygous missense mutation just outside the helix initiation motif of K14.[54] Several other patients with more severe and generalized blistering have been reported. Complete loss of K14 expression was demonstrated in these patients by immunohistochemistry, immunoelectron microscopy and immunoblotting. K5 expression was normal. These individuals therefore represent human ‘knockouts’ of K14. Homozygous nonsense mutations were detected in two families[55,56] and a homozygous 2 nucleotide deletion mutation in a third family.[57] A homozygous splice site mutation was found in a fourth family.[58] Each mutation resulted in a premature termination codon predicting a truncated K14. Interestingly, a fifth case with a milder phenotype has a homozygous mutation that also results in a premature termination codon. In this case a more truncated protein than in the previous cases is predicted, a K14 protein of only 30 amino acids.[59] A more severe phenotype might have been predicted from this latter mutation but the number of such cases is still too small to draw firm conclusions. In one of the above patients, increased expression of K15 was proposed to compensate for the lack of K14.[58] It appears that the functional relationship between K14 and K15, both type I keratins of basal keratinocytes differs between mice and humans. In contrast to R-EBS patients lacking K14 who can live to old age,[58] transgenic mice in which K14 has been ablated die at about 3 months.[60] In mice, K15 expression decreases with age,[60] whereas in humans, its expression appears to persist and therefore compensates for loss of K14.[58]

As yet there are no cases of recessive EBS with mutations in K5. It is possible that such mutations could be lethal. Evidence for this comes from the recent generation of transgenic mice lacking K5.[61] These mice die at birth due to lack of keratins in the basal cells of the epidermis. Again, this result underlines the ability of K5/K15 filaments to partially rescue the phenotype in neonatal K14 knockout mice, and to a greater degree in R-EBS patients lacking K14. Interestingly, no phenotypes associated with K15 mutations have been reported in either humans or mice, to date.

The identification of keratin mutations in EBS led the way for the molecular basis of other keratin disorders to be elucidated. Disorders where the clinical phenotype affects tissues expressing specific keratins, and that ultrastructurally show aggregation of keratin filaments, were investigated. Mutations have now been identified in 18 keratin genes, either as missense mutations, deletion mutations, insertion mutations, splice site mutations or nonsense mutations. In general, keratin disorders are inherited in an autosomal dominant manner, therefore most keratin mutations act by dominant—negative interference at the protein level. Only a small amount of one copolymerizing keratin need be defective to disrupt the filament network and cause cell fragility. Described below are further keratin diseases for which the molecular pathogenesis has been solved.

2.2 Bullous Congenital Ichthyosiform Erythroderma

Bullous congenital ichthyosiform erythroderma (BCIE), also known as epidermolytic hyperkeratosis (EHK, OMIM 113800), is characterized by erythroderma and blistering at birth, leading to severe generalized hyperkeratosis in adulthood (figure 3). Cytolysis develops in the suprabasal layers of the epidermis, quite distinct from the basal cell blistering in EBS. Ultrastructurally, the basal cells appear normal. However, in the suprabasal cells there is abnormal clumping of tonofilaments leading to a collapsed cytoskeleton in these cells.[62,63] As cells migrate up through the epidermis, K5 and K14 are down regulated and K1 and K10 become the predominant keratins expressed in the suprabasal cells. These keratins became the focus of several studies to determine the molecular basis of BCIE. Again, data from transgenic mice studies supported the idea of a keratin defect. Transgenic mice expressing a truncated human K10 gene gave rise to a phenotype characteristically similar to BCIE.[64] The mutant K10 disrupted normal filament assembly/function in the suprabasal cells producing cytolysis in these upper layers of the epidermis. The first molecular data from patients demonstrated linkage of BCIE to the type II keratin gene cluster on chromosome 12.[65,66] Heterozygous mutations were subsequently identified in several BCIE families in K1 and K10.[67–69] These mutations were in the helix boundary motifs and in the H1 domain of K1. In K10, codon R156, the analogous codon to R125 in K14, was found to be a ‘hot spot’ for mutations.

There have since been numerous reports of mutations in K1 and K10 (table III and keratin mutation database, table II) the majority of them heterozygous missense mutations. In K10, R156 has continued to be the predominant ‘hot spot’ for mutations with 24 of the 38 published K10 mutations occurring here. Mutations in K10 have also been found elsewhere in the helix initiation motif and in the helix termination motif (keratin mutation database, table II). K1 mutations in BCIE are found predominantly in the helix boundary motifs.

From the apparent genotype/phenotype correlation seen in EBS it was predicted that mutations in other regions of the rod domain of K1 or K10 might produce milder BCIE phenotypes. So far there is only one report of such a mutation, a patient with a mild case of BCIE with a mutation in the L12 linker of K1.[70] It may be that some mutations in these regions produce such a mild phenotype that they go undiagnosed. From the mutations reported, BCIE can, however, be divided into two groups. Those with hyperkeratosis of the palms and soles usually have mutations in K1[71] and those without palmoplantar involvement in general are accounted for by mutations in K10.[72] It is postulated that the type I keratin, K9 may be able to partially compensate for disruption to the K10 network. K9 is specifically expressed in palmoplantar skin and this could explain why individuals with mutations in K10 have little or no palmoplantar involvement.[73,74]

Mutations in K1 and K10 have also been found in several patients that did not present with the classical BCIE phenotype; these represent variants of BCIE rather than distinct diseases. One of these is a variant where patients present with epidermal nevi. The areas of skin affected are characterized by a mosaic pattern that follows the lines of Blaschko. Mutations in K1 or K10 have been identified in keratinocytes from lesional skin but not in unaffected skin. These mutations have probably arisen from postzygotic mutations during embryogenesis.[75–77] It is possible that the offspring of these patients may present with generalized BCIE. Individuals with epidermal nevi and with known K1 or K10 mutations should therefore receive appropriate genetic counseling and prenatal diagnosis may be desirable.

Another variant of BCIE described as annular epidermolytic ichthyosis (AEI) is histologically and clinically similar to BCIE.[78] In addition to the blistering, erythema and hyperkeratosis typical of BCIE, numerous annular and polycyclic erythematous hyperkeratotic plaques develop on the trunk and proximal extremities. Mutations in K10 have been found in several patients with AEI.[79,80] More recently K1 mutations have been reported in patients with a similar phenotype, BCIE with cyclic ichthyosis.[81,82]

2.3 Ichthyosis Bullosa of Siemens

Ichthyosis bullosa of Siemens (IBS, OMIM 146800) is a type of epidermolytic hyperkeratosis with epidermal thickening and superficial blistering occurring predominantly on the flexures.[83] Characteristic ‘Mauserung’, or moulting of the outer layers of the epidermis in IBS distinguishes this condition from BCIE. Tonofilament aggregation and cytolysis limited to the upper spinous and granular cell layers of the epidermis was observed in patients with IBS.[84] This made K2e, which is expressed specifically in these upper layers of the epidermis[85] a good candidate gene for IBS. K2e has no obvious expression partner, but is thought to pair with K10. Linkage of the disease to the type II keratin cluster on chromosome 12 supported the idea of KRT2e as a candidate gene.[86] The first of many reports identifying mutations in K2e in IBS followed shortly thereafter.[84,87,88] The mutations were again clustered in the helix boundary motifs. The majority were in the helix termination motif with codon 493, a particular ‘hot spot’ for mutation. In more recent reports this pattern has continued. Of the 26 published cases, 15 patients had codon 493 affected with different substitutions. Of these, 13 were in a CpG dinucleotide sequence involving a G-A transition, E493K (keratin mutation database, table II).

One of the early reports describes K2e mutations in two families that were originally misdiagnosed as BCIE.[88] This is a reminder that there can be overlap in clinical phenotype. Patients with milder forms of BCIE can appear similar clinically to IBS. This should be remembered when determining candidate genes for mutation screening. Using molecular techniques to screen all candidate genes it is possible to confirm the diagnosis in patients that may be difficult to distinguish clinically. It is also well known that for many of these keratin disorders there can be interfamilial phenotypic variation that may make diagnosis more confusing.

2.4 Pachyonychia Congenita (PC)

Pachyonychia congenita (PC) encompasses a range of ectodermal dysplasias where hypertrophic nail dystrophy is a prominent clinical feature. PC is normally inherited in an autosomal dominant manner and is subdivided into two main types:[89] PC-1 (Jadassohn-Lewandowsky type, OMIM 167200)[90] and PC-2 (Jackson-Lawler type, OMIM 167210)[91] Both types are associated with mutations in keratin genes. Mutations in the type II keratin, K6a and its type I expression partner K16 produce PC-1. Mutations in K17 and its recently identified type II expression partner, K6b, produce PC-2.

2.4.1 PC-1

PC-1 is characterized by hypertrophic nail dystrophy, focal non-epidermolytic palmoplantar keratoderma, and oral leukokeratosis (figure 4). Oral leukokeratosis may to a lesser extent be present in PC-2 and is somewhat variable in both forms. By electron microscopy of palmoplantar skin, tonofilaments in suprabasal keratinocytes appear abnormal, as condensed clumps, although these are morphologically different from the aggregates seen in EBS-DM and BCIE.[92,93] These abnormal keratin clumps were proposed to consist of the K6a/K16 expression pair. These keratins are expressed in the nail bed, mucosa and palmoplantar epidermis,[12] the epithelial tissues affected in PC-1. Consequently, K6a and K16 mutation analysis was performed on several PC-1 families. This led to the identification of the first mutations in K6a[94] and K16.[92] Mutations found in additional families are predominantly clustered to the helix initiation motifs (table III). As well as heterozygous missense mutations, several deletion mutations have also been reported in this region. In K6a, codon 171 is a hotspot for either deletion mutation or missense mutation.[95,96]

Fig. 4
figure 4

Clinical phenotypes of pachyonychia congenita (PC) caused by mutations in K6/K16 (PC-1) and K6b/K17 (PC-2). (a) Hypertrophic nail dystrophy typical of PC-1 and PC-2. (b) Oral leukokeratosis, normally present in PC-1, and may to a lesser extent occur in PC-2. (c) Multiple pilosebaceous cysts characteristic of PC-2 patients, occurring post-puberty.

Full size image

Several reports of mutations in K16 have been found in more unusual PC-1 phenotypes. Firstly, in a case of late onset PC-1 where there were no skin or nail changes until 6 years of age. A missense mutation was found in the central 2B domain of K16.[97] Mutations in this region of K14 cause the milder EBS phenotypes. Secondly, a case described as unilateral palmoplantar verrucous nevus was caused by a mosaic mutation.[98] Although histologically similar to epidermolytic palmoplantar keratoderma (EPPK, see below), the lesions were only present on one side of the body and followed the lines of Blaschko. This is an example of an unusual phenotype where there was no obvious candidate gene. Several candidate genes were screened including K1, K9 and K16. No mutations were found in K1 or K9. However, an in-frame 12 base pair (bp) deletion mutation was identified in the variable V1 domain of K16.[98] This mutation was only detected in cDNA derived from the affected skin of the palm. The normal K16 sequence was found in unaffected skin. Due to the mosaicism observed it was concluded to be a postzygotic mutation.

2.4.2 PC-2

In PC-2, in addition to hypertrophic nail dystrophy, characteristic multiple pilosebaceous cysts develop during puberty (figure 4). These are the main distinguishing features between the two types of PC. Mild focal non-epidermolytic palmoplantar keratoderma, natal teeth, pili torti (twisted hair), and some oral lesions may also be present. In early childhood it can be difficult to distinguish between PC-1 and PC-2. The possibility of a keratin defect underlying PC-2 was explored after linkage analysis of a large Scottish pedigree showed co-segregation of the disease to the type I keratin cluster on chromosome 17.[99] Candidate genes included KRT17, expressed in the outer root sheath of hair follicles, nail bed and sebaceous glands.[100,101] A heterozygous missense mutation in K17 was subsequently found in a PC-2 family.[92] Heterozygous missense mutations in K17 have now been identified in many patients with PC-2 (keratin mutation database, table II). To date, all are in the helix initiation motif, with mutations in codons 92 or 94 accounting for the majority of them. Two deletion mutations, also in the helix initiation motif domain, produce the same PC-2 phenotype as the missense mutations.[96,102] There are no reports of mutations in the helix termination motif of K17 but they are predicted to produce PC-2.

In one family with PC-2, genetic linkage analysis excluded the type I keratin gene cluster, thereby excluding KRT17 as a candidate gene. However, the disease mapped to the type II keratin gene cluster on chromosome 12. There was no obvious type II expression partner for K17 but the K6b isoform of K6 was considered a candidate. It was demonstrated experimentally that K17 and K6b are co-expressed in many epidermal appendages. By mutation analysis, a missense mutation was identified in the helix termination motif of K6b in this family.[103] Thus a second gene for PC-2 was identified.

2.4.3 Steatocystoma Multiplex

Steatocystoma multiplex (OMIM 184500), a variant of PC-2, is characterized by multiple pilosebaceous cysts like those seen in PC-2. There are no additional features other than perhaps very mild nail involvement. Missense mutations in K17 have been found in three families.[104,105] Interestingly, the mutations reported occur at the same codons as those that can also produce PC-2. In one case the mutation was the same, R94C.[104]

As seen for other keratins phenotypic variation is also seen with mutations in K17. Different phenotypes (PC-2 or steatocystoma multiplex) can arise from the same mutation, indicating that the phenotype must depend on a combination of other genetic and environmental factors.

2.5 Palmoplantar Keratoderma

The inherited palmoplantar keratodermas encompass a range of disorders that can be difficult to distinguish clinically. Hyperkeratosis of the palms and soles occurs in a diffuse, focal or punctate pattern and may be associated with other ectodermal features.[106] Different forms of palmoplantar keratoderma are caused by mutations in different keratin genes, for example epidermolytic palmoplantar keratoderma is caused by mutations in K9 (see section 2.5.1). Mutations in other forms of palmoplantar keratoderma have been found in keratin genes that are principally involved in more common diseases. For example, focal non-epidermolytic palmoplantar keratoderma is caused by mutations in K16 (see below), mutations in K16 also cause PC-1, as described above. The majority of mutations in K1 result in BCIE but a few unusual cases of palmoplantar keratoderma have been reported that also have K1 mutations. These findings demonstrate the heterogeneous phenotype that can result from mutations in some keratin genes.

2.5.1 Epidermolytic Palmoplantar Keratoderma (EPPK)

Characterized by epidermolytic hyperkeratosis of palmoplantar epidermis, epidermolytic palmoplantar keratoderma (EPPK, OMIM 144200) is an autosomal dominant disorder that presents early in life.[107] Histologically, cytolysis occurs in the suprabasal layers of thick epidermis and aggregates of keratin filaments are seen by electron microscopy.[108] The distinct localization of the diffuse thick yellowish hyperkeratosis to the palms and soles with no other clinical phenotype (figure 3) pointed towards KRT9 as the candidate gene because of its specific expression pattern in the suprabasal cells of palmoplantar skin.[73,74] Like K2e, K9 appears to be an accessory keratin probably pairing with K1 but with no other obvious type II keratin expression partner. The first EPPK families studied mapped to the type I keratin cluster on chromosome 17.[109] Following the cloning of the K9 cDNA and corresponding gene, mutation analysis was performed in patients with EPPK.[110] Heterozygous missense mutations were identified in the helix initiation motif of K9 in several EPPK families.[110] To date, 28 of the 29 mutations reported for EPPK are in the helix initiation motif (table III), 16 of these are in codon 162 (R162) the analogous codon to R125 in K14 and R156 in K10. These are CpG ‘hot spot’ mutations in EBS and BCIE respectively. Similarly, mutations in K9 R162 occur in a CpG dinucleotide. The only K9 mutation reported that does not occur in the 1A domain is a 3 bp insertion mutation in the helix termination motif.[111] This was the first report of an insertion mutation in a human keratin and the clinical phenotype was shown to be similar to those caused by missense mutations.

2.5.2 Mild EPPK

In three families with a mild form of EPPK, linkage of the disease to the type II keratin gene cluster eliminated KRT9 as a candidate gene. Analysis of KRT1 identified a splice site mutation at the exon 6 splice donor site. This resulted in a new splice site 54 bases downstream and subsequent insertion of 18 bp.[112] Genotyping data suggested that all three families could have the same ancestral mutation. The disease phenotype in these families is quite distinct from BCIE.

2.5.3 Diffuse Non-Epidermolytic Palmoplantar Keratoderma

An unusual mutation has been reported in the Unna-Thost form of PPK. Linkage analysis excluded the type I keratin locus and therefore KRT9 was excluded as a candidate gene.[113] KRT1 was screened and a missense mutation was found. But unlike all other K1 mutations this is in the non-conserved variable V1 domain at the amino terminus. No further mutations have been reported in non-epidermolytic palmoplantar keratoderma (NEPPK).

2.5.4 Ichthyosis Hystrix Curth-Macklin

A case of severe mutilating palmoplantar keratoderma, present through several generations in an African-American family, was recently reported as the Curth-Macklin form of ichthyosis hystrix (IHCM, OMIM 146590). Histologically, the keratin filaments in the spinous and granular layer appeared abnormal. Linkage analysis of several candidate regions showed complete co-segregation of the disease with the type II keratin cluster on chromosome 12.[114] Several type II keratin genes were screened for mutations. A mutation was identified in K1 in the non-conserved variable V2 domain at the carboxy terminus. Substitution of two guanine bases for an adenine resulted in a frameshift and premature termination codon 229 bp downstream.[114] This is the first pathogenic mutation occurring in the V2 domain of a keratin. Interestingly, the keratin loci were eliminated in another family diagnosed as IHCM.[115]

2.5.5 Focal Non-Epidermolytic Palmoplantar Keratoderma

Focal non-epidermolytic palmoplantar keratoderma (FNEPPK) is characterized by the presence of focal hyperkeratosis on the pressure points of the palms and soles similar to that observed in PC-1. There is minimal or no nail involvement and no oral involvement. In two individuals presenting with FNEPPK, heterozygous missense mutations were found in the helix initiation motif of K16.[116] These mutations are similar to those causing PC-1 and might have been predicted to give a more severe phenotype. This again demonstrates the complexity of keratin mutations. A third case is an unusual heterozygous mutation, a 23 bp deletion and a separate 1 bp deletion in the helix termination motif.[117] This removes most of the helix termination motif, the most highly conserved motif sequence in keratins and all intermediate filaments. The unexpectedly mild phenotype suggests that loss of most of this critical domain has a milder dominant-negative effect on filament assembly than certain missense mutations. Data from transient transfections of mutant K16 protein and other K16 constructs into epithelial cells in culture supported this hypothesis.[117] Mutations in K6a are also predicted to give a FNEPPK phenotype.

2.6 Non-Epidermal Keratin Disorders

Certain keratins are expressed in tissues other than the epidermis and epidermal appendages. Mutations in these keratins were also proposed to result in a clinical phenotype. The first mutations to be identified in a non-epidermal keratin were in the disorder white sponge nevus.

2.6.1 White Sponge Nevus

White sponge nevus of Canon (WSN, OMIM 193900) occurs as a benign autosomal dominant disorder affecting non-cornifying stratified squamous epithelia. White ‘spongy’ plaques are observed, predominantly in the mouth (oral leukokeratosis) (figure 5) but may also be present in the esophagus and anogenital mucosa.[118] Onset is often in early childhood. Reports of tonofilament aggregation in patients with WSN indicated a defect in K4 or K13.[119,120] These keratins are specifically expressed in the mucosal non-cornifying stratified epithelia. Mutation analysis of two Scottish families revealed a 3 bp heterozygous deletion mutation in the helix initiation motif of K4.[121] Independently, another group identified a heterozygous missense mutation in K13 in a family with WSN.[122] Further heterozygous mutations have been identified in other WSN families including a 3 bp insertion in K4[123] and missense mutations in K13.[124,125] To date, the mutations in K4 and K13 have all been found in the helix initiation motif (table III) but one would predict that mutations in the helix termination motif would produce a similar phenotype.

Fig. 5
figure 5

(a) Oral white sponge nevus (WSN), oral leukokeratosis is due to mutations in K4 and K13 (reproduced from Munro,[126] with permission from Dermatology in Practice). (b) Meesmaan’s epithelial corneal dystrophy (MECD), multiple fine microcysts in the cornea are seen by slit lamp due to mutations in K3 and K12.

Full size image

2.6.2 Meesmann’s Epithelial Corneal Dystrophy

Meesmann’s epithelial corneal dystrophy (MECD, OMIM 122100) is an autosomal dominant disorder of the corneal epithelium originally described in a large German family.[127] There are characteristic intraepithelial cysts in the anterior cornea seen by slit lamp (figure 5). These cysts are filled with intracellular debris and probable keratin aggregates.[128] Onset is in early childhood although it may sometimes be delayed until late childhood/early adulthood. The disorder is often asymptomatic, and vision is not normally affected. However, because of the fragility of the corneal epithelium, patients can be intolerant of contact lenses. K3 and K12 are specifically expressed in the corneal epithelial cells.[12,129] This, together with the fact that K12 knockout mice have extremely fragile corneal keratinocytes,[130] led to the discovery that mutations in K3 and K12 can cause this disorder.[131] Descendants of Meesmann’s original German kindred were studied and by genetic linkage analysis MECD was mapped to the KRT12 locus. Co-segregation of this locus was also found in a small family from Northern Ireland. Mutation analysis identified heterozygous missense mutations in the helix initiation motif of K12 in both families. In a second family from Northern Ireland, the disorder co-segregated with the KRT3 locus and consequently a mutation was identified in helix termination motif of K3.[131] Mutations have been identified in other families, all occurring in the helix boundary motifs of K12.[132–135]

2.6.3 Cryptogenic Cirrhosis

The simple epithelial keratins K8 and K18 are expressed in the liver, pancreas and intestinal epithelium. For some time, keratin researchers wondered what clinical phenotype would be produced by mutations in these genes. Good evidence for association of these genes and liver disorders came from the discovery that transgenic mice expressing mutant K18 develop chronic hepatitis.[136] DNA was then analyzed from patients with liver disease of unknown cause. In one out of 28 patients with cryptogenic cirrhosis, a missense mutation was identified in K18 (H127L).[137] This mutation is unusual in that it is in the L1 linker domain, a site where no other pathogenic keratin mutations have been reported. A follow up study of a large number of patients from three groups: patients with cryptogenic liver disease, noncryptogenic liver disease, and randomly selected patients who provided blood to the hematology department, were analyzed for mutations in K8 or K18. Five mutations were identified in the H1 domain of K8 in patients with cryptogenic liver disease, three were G61C and two were Y53H.[138] Mutations in the H1 domain of other type II keratins have been shown to be pathogenic. In K1 and K5, mutations in the H1 domain underlie some cases of BCIE and EBS-WC respectively (keratin mutation database, table II). The effect of these K8 mutations on filament reorganization were analyzed in tissue culture experiments.[138] Together with data from transgenic studies, the results indicate that mutations in K8 or K18 may predispose patients to late onset liver disease.

2.6.4 Trichocyte (Hair) Keratins

The type I and type II trichocyte or hair keratin genes co-localize with the cytokeratins on chromosome 17 and 12 respectively. These keratins form the highly cross-linked tough polymer of which hair, nail and the hard tips of the lingual papillae are composed. These proteins share the same basic structure as epithelial keratins but are very rich in cysteine residues, allowing a high degree of disulfide cross-linking. The proposal that mutations in these genes could produce an abnormal hair phenotype has been investigated.

Monilethrix

One candidate disease for mutations in a trichocyte keratin was monilethrix (OMIM 158000). This is a rare autosomal dominant disease with variable phenotypic expression, from mild hair loss to near total alopecia (figure 6). The hair is fragile and by light microscopy, periodic thinning of the hair shaft gives the hair a beaded appearance. By electron microscopy, defects are found in the microfibrillar structure of the hair shaft. Cytolysis and tonofilament clumping in the cortical cells of the bulb of the hair follicle are also observed.[139] Follicular keratosis and nail abnormalities may be present. Evidence for a defect in a trichocyte keratin came from genetic linkage between monilethrix and the type II keratin gene cluster on chromosome 12.[140,141] Identification of heterozygous missense mutations in the type II hair keratin hHb6 followed.[142] In all patients, the mutation occurred in the helix termination motif in codon 413 but by different substitutions. A subsequent report identified further mutations in hHb6, again in codon 413 thereby identifying it as a ‘hot spot’ for mutations.[143] Additionally, a heterozygous missense mutation was identified in another family in the corresponding residue of the highly related type II keratin, hHb1.[143] Mutations have now been identified in many families (keratin mutation database, table II). To date, all are missense mutations occurring predominantly in the helix termination motif of hHb6 (table III). A few have been reported in the helix initiation motif of hHb6, and several in the helix termination motif of hHb1 (table III).

Fig. 6
figure 6

Monilethrix, variable alopecia, caused by mutations in hair keratins hHb6 or Hhb1.

Full size image

3. Molecular Genetics in Diagnosis of Keratin Disorders

Before identification of keratin mutations in the disorders described above, diagnosis was often confirmed by histological analysis of a skin biopsy. For example, to determine the level of splitting within the blister in EBS, or to look for the characteristic clumping of keratin filaments in EBS-DM. These histological methods are important and are sometimes still necessary. However, by using molecular techniques the exact gene defect can be identified, providing an unequivocal diagnosis. These can be performed on genomic DNA derived from blood samples, alleviating the need to take biopsy material. Analysis at the molecular level is by DNA sequencing of candidate genes. In patients not fitting the phenotype classification, genetic linkage analysis is performed to determine the keratin gene cluster to which the disease maps. Candidate genes are then screened by DNA sequencing. Mutations should be present in all affected family members but absent in unaffected members. Confirmation of a mutation also requires screening of at least 50 ethnically matched, unrelated, control DNA samples to exclude the mutation as a common polymorphism occurring within the general population. For all keratins associated with diseases at this time (except K8 and K18) there are two ‘hot spot’ regions for mutations, the highly conserved helix boundary motifs. Mutation screening should focus on these regions first. In addition, the H1 domain of type II keratins and the L12 linker domain of both types are frequent sites for mutations.

In rare cases where inheritance of the disease is recessive, staining of tissue from patients with appropriate antibodies may reveal loss of expression of a particular keratin. DNA sequencing of this candidate gene will identify the mutation. In consanguineous families, linkage analysis by homozygosity mapping may be useful.

3.1 Pseudogenes

The presence of multiple pseudogenes for some keratins, some of which are also expressed at low levels, hampered early mutation detection studies using genomic DNA samples. It is important to be aware of pseudogenes to ensure that mutation analysis is only performed on the functional gene. Pseudogenes have been reported for K6,[144] K8,[1] K14,[145] K16,[3] K17,[146] K18,[1] and K19.[147] The sequences of some of these pseudogenes are now known. It is therefore possible to design specific polymerase chain reaction (PCR) primers for the respective functional gene to avoid amplification of pseudogenes.

There is at least one KRT14 pseudogene.[145] A genomic mutation detection system has been developed for exons 1, 4 and 6 which encode the 1A, L12 and 2B domains of K14 containing the mutation ‘hot spots’.[39] In this case, before PCR amplification, genomic DNA is digested with restriction enzymes that cleave only the KRT14 pseudogene, leaving the genomic sequences of the functional KRT14 intact.

Several discrepancies in the published K16 coding sequence.[3,148,149] led to a further study.[150] This confirmed the K16 cDNA sequence reported by Paladini et al.[149] and described cloning of the corresponding KRT16 gene and 2 pseudogenes. PCR primers were designed to specifically amplify the functional KRT16.[150] Mutations identified in KRT16 in PC-1 families were also confirmed by reverse transcription (RT)-PCR in mRNA from the same individuals demonstrating specificity of genomic PCR for the functional gene

Likewise for KRT17, specific PCR conditions have been developed[92,96] to avoid amplification of the two reported KRT17 pseudogenes.[146]

For K6 there are at least six copies of the gene, of which at least two are expressed.[144] A mutation detection strategy has been developed for all of KRT6a,[151] and some exons of KRT6b.[103] Several mutations identified in genomic DNA from PC patients have been confirmed at the mRNA level, demonstrating specificity of these specific PCR primers and conditions.

4. Future Studies

As more patients are studied, additional mutations may be found outside the hotspot regions. This should provide further information regarding the function of these domains. In some cases there may be no mutation in the candidate keratin genes. In these situations it is worth considering the few known keratins that are not yet associated with a human disorder, for example, K15. It is also probable that some of the novel keratin genes identified from sequence data from the human genome project may be associated with human diseases.[1] Alternatively mutations in other genes such as those encoding intermediate filament associated proteins may produce similar phenotypes to those of some keratin genes. For example, a form of EBS with muscular dystrophy is due to mutations in plectin, a high molecular weight cytoskeleton associated protein.[152,153] Mutations in the desmosomal protein, plakophilin 1, lead to an ectodermal dysplasia phenotype with skin fragility.[154]

5. Prenatal Diagnosis

Now that many keratin disorders can be diagnosed at the molecular level, the focus of several research groups is on treatment and prevention of these diseases. Treatment in the form of gene therapy is difficult as most keratin mutations are dominant acting and the mutant gene needs to be inactivated. Ways of achieving this are currently being studied. A mouse model has recently been developed that provides a realistic model of EBS-DM. This has important implications for gene therapy of EBS and will provide a system to test different approaches in vivo.[155]

While many of the disorders caused by keratin mutations are not severe enough to warrant prenatal diagnosis, in some of the more severe phenotypes such as EBS-DM, BCIE or severe PC, it may be requested. Previously diagnosis would have been from a fetal skin biopsy, for example in the case of EBS, to determine the presence of blistering. However, reliable mutation detection methods from genomic DNA are now available. Therefore, it is possible when the familial mutation is known, to perform prenatal diagnosis from chorionic villus samples (CVS) at an early stage of the pregnancy. CVS DNA is obtained at 11–12 weeks gestation for mutation analysis. Multiple villi are dissected out from the CVS and DNA is isolated using standard DNA extraction protocols. Care is needed to prevent contamination of the CVS DNA with maternal cells as this can lead to problems of diagnosis. CVS DNA is screened for the known familial keratin mutation using standard PCR and DNA sequencing protocols. Prenatal diagnosis by mutation analysis has been performed for several patients with BCIE,[156,157] EBS,[40] and one patient with PC-1.[150] In the UK, genetic diagnosis is performed by National Health Service laboratories, and in the USA by local genetics laboratories or by private companies (see table II).

6. Intermediate Filament Mutation Database

With the increasing number of mutations identified within the keratins, together with the pathogenic mutations that are emerging in other members of the intermediate filament gene superfamily, there was a need for a readily accessible database of this information. The Epithelial Genetics group at Dundee University, at the suggestion of the HUGO Human Genome Variation Society, has established the Intermediate Filament Mutation Database to address this problem. This catalog of mutations will provide a valuable resource when analyzing mutations in terms of their frequency, position and clinical phenotype, and allow correlations to be made between genotype and phenotype. A summary of the number of keratin mutations as at January 2002 is shown in table III. This database is regarded as the property of all keratin and intermediate filament researchers and will be run as an international consortium. Comments and suggestions are welcome, as well as direct submission of mutations and polymorphisms, at the website (http://www.interfil.org), see table II.

7. Conclusion

During the last decade mutations in 18 keratin genes have been associated with epithelial diseases. This has led to improved classification of these disorders and diagnosis by molecular techniques combined with clinical examination and histological analysis. Current research in many groups is now focussed on developing treatments to alleviate the debilitating symptoms caused by these diseases.