Keratins are cytoskeletal filament-forming proteins found in skin and other epithelial (sheet) tissues (Table 1). Keratins (type I and II), and other highly related (types III–VI) intermediate filament or nanofilament proteins, used to be thought of as very inert because they form filament networks that are not easily disrupted, in contrast to actin and microtubule cytoskeleton systems, which can be rapidly and completely dismantled by many drugs. The identification of causative pathogenic mutations in the keratin genes KRT5 and KRT14 in the human blistering skin disease epidermolysis bullosa simplex (EBS) changed this view and demonstrated the importance of these proteins for maintaining tissue resilience (Bonifas et al., 1991; Coulombe et al., 1991; Lane et al., 1992). As the first-reported disease caused by keratin mutations, EBS set the pattern for all the other keratin disorders, as well as several non-keratin genetic skin diseases. Many ‘keratinopathies’ have now been identified with a variety of disease phenotypes, often predicted by tissue-specific keratin expression, with pathology usually involving some form of tissue fragility (Szeverenyi et al., 2008; see poster).
By reviewing the essential elements of keratin protein structure and filament assembly, we consider how keratin filaments provide tissues with mechanical resilience, and how mutations might impact upon these processes and compromise tissue function. The rare human keratinopathies have provided important clues to keratin function, revealing the potential roles for keratins in many types of stress response, and even effects on membrane trafficking and cell and tissue differentiation. This knowledge has begun to inform development of therapeutic options for keratin diseases, and successful first-in-man siRNA trials have been performed. We conclude by considering the issues and challenges now facing the field, and how they are or can be addressed.
Keratin protein and filament structure
The keratin family consists of 54 proteins (Schweizer et al., 2006) in two families (types I, acidic, and II, neutral or basic), each being expressed in a defined sequence during differentiation. The greatest expression diversity lies in skin (see poster). Keratin proteins can be divided into three functional groups: ‘simple’ keratins, expressed in embryonic and one-layered epithelia, including liver, intestine and glandular secretory cells; ‘barrier’ keratins, expressed in multilayered stratified squamous epithelia, such as the epidermis of the skin; and the harder ‘structural’ keratins that form hair and nail (Table 1).
Like all intermediate filaments, a keratin protein has three clear domains: a central α-helical rod domain, flanked by non-helical head and tail regions that contain most phosphorylation sites (Geisler et al., 1982; Omary et al., 2006). Keratins begin assembly as heterodimers of one type I and one type II keratin monomer (Hatzfeld and Weber, 1990; reviewed by Herrmann and Aebi, 2000), with rod domains aligned in parallel and in register. These dimers form antiparallel tetramers by overlapping the N-terminal half of their rod domains. Tetramers then assemble into ‘unit length filaments' that are 60 nm in length (see poster). These rapidly assemble end-to-end, within 10 seconds in vitro (Herrmann et al., 2002), forming rope-like nanofilaments that are ∼10 nm thick. The antiparallel nature of the tetramers results in filaments that do not have polarity, which is again in contrast to actin filaments or microtubules, and implies that keratin filaments cannot function as unidirectional tracks for molecular motors. At either end of the rod domain are the highly conserved helix initiation and termination motifs, which are hotspots for the most severe disease mutations (Lane and McLean, 2004; Uitto et al., 2007).
Dynamic behaviour of keratins
The cytoplasmic ropes of keratin form branching bundles in the cell and anchor into junctions around the cell perimeter. Keratins link through desmoplakin into desmosomes (Kouklis et al., 1994), which connect neighbouring cells, and through plectin (Andrä et al., 2003) into hemidesmosomes, which connect cells to their attachment substrates. This filament–junction network anchors cells in three dimensions through the epithelium. However, the keratin filament cytoskeleton must be dynamic, plastic and flexible to allow cells to proliferate during growth and to migrate during wound healing. Dynamic assembly and disassembly of filaments is also needed to allow epithelial cells to maintain an intact network while they alter their keratin expression profile during differentiation or in response to stress. For example, in the epidermis, K5 and K14 are the major keratins synthesised in basal cells (which can proliferate), whereas K1 and K10 are expressed in suprabasal cells (which are ‘locked’ into differentiation). However, keratin proteins have a long half-life, around 4 days for simple keratins (Denk et al., 1987), and K5 and K14 can persist in suprabasal cells (see poster) alongside K1 and K10 (Kartasova et al., 1993; Eriksson et al., 2009; Windoffer et al., 2011). A gradual transition, rather than complete disassembly and re-assembly of the network, preserves the structural integrity of the filament network during differentiation.
Where do keratins assemble in the cell? Intermediate filaments do not have defined organizing centres as microtubules do; they can be synthesised throughout the cell as tiny subfilamentous particles that occur throughout the cytoplasm (Liovic et al., 2003). The unit length filament (see poster) might be the nucleation precursor, but this has not been confirmed as its size (60 nm) is below the resolution limit of standard light microscopy. New super-resolution microscopy techniques (reviewed by Schermelleh et al., 2010) could soon shed light on this. Turnover might be accelerated at the cell periphery, and a model of filament assembly has been recently described that proposes that filaments are nucleated in the cell periphery at focal adhesions, before elongating and then integrating into the pre-existing peripheral network (Windoffer et al., 2011).
A probable cofactor for localising the assembling keratins is the linker protein plectin because plectin isoforms bind actin or intermediate filaments at different cell locations. Plectin is known to influence the formation and dynamics of vimentin filaments (the intermediate filament protein in fibroblasts) (Spurny et al., 2008; Kostan et al., 2009), and plectin has also been implicated in the actin-dependent inward movement of keratins in the cell periphery (Kölsch et al., 2009). Keratin assembly at focal adhesions (Windoffer et al., 2006) might also depend on phosphorylation by the p38 mitogen-activated protein kinases, as inhibition of these kinases prevents the formation of filament precursors in the cell periphery (Wöll et al., 2007).
A wide range of post-translational modifications have been described on keratins, including phosphorylation to ubiquitylation, sumoylation and acetylation (Omary et al., 2006; Ku et al., 2010; Srikanth et al., 2010; Snider et al., 2011), which are likely to modulate the solubility of keratins in specific situations. The most well-documented modification is phosphorylation, with many sites known on simple keratins (reviewed by Omary et al., 2006) but fewer defined on barrier keratins. Phosphorylation of keratin filaments by protein kinase Cζ is required for K8 and K18 filament remodelling in response to shear stress (Flitney et al., 2009; Sivaramakrishnan et al., 2009).
Considering all the above, even a single amino acid change in keratins can interrupt cell function in many ways, by altering their post-translational modifications, their integration into junctions or their filament assembly kinetics (Herrmann et al., 2002; Owens et al., 2004) to yield a less-stable filament network and cause disease. For example, some of the severe K14 mutations that cause Dowling–Meara EBS, such as the hotspot mutation p.Arg125Pro, result in the formation of cytoplasmic keratin aggregates (Ishida-Yamamoto et al., 1991). In vitro, this mutation reduces the ability of reconstituted filaments to bundle under cross-linking conditions (Ma et al., 2001). The impact of mutations on the mechanical resilience of epithelial tissues is discussed below.
Keratins provide epithelia with mechanical resilience
The keratin filament network can withstand significant mechanical forces, particularly in stratified epithelia such as the epidermis (Beriault et al., 2012). By conferring mechanical continuity across an epithelial sheet, the keratin–desmosome and keratin–hemidesmosome network generates mechanical resilience across the tissue, as the network will dissipate mechanical stress away from a source into the surrounding epithelium. Investigations of single-filament mechanical properties show that keratin filaments are flexible and tough, being less rigid than actin filaments at low strains, and less brittle than microtubules at high strains, where they show strain hardening (Janmey et al., 1991; Kreplak and Fudge, 2007). Bundling of keratin filaments appears to increase when cells are subjected to shear stress (Flitney et al., 2009).
The tissue fragility observed in many keratin diseases reflects the crucial role of keratins in providing mechanical stability to cells and tissues (McLean and Moore, 2011). However, studies suggest that mutations in keratins do not reduce mechanical resilience simply by preventing filament formation. A keratinocyte cell line expressing a GFP-tagged K14 carrying the p.Arg125Pro mutation can withstand uni-directional stretch (greater than 100%) without significant damage or cell death as much as cells expressing wild-type filaments can (Fudge et al., 2008; Beriault et al., 2012). However, the cells expressing mutant keratins are much less able to survive a repetitive stretch than those expressing wild-type keratins (Russell et al., 2004). It appears likely that mutations alter the dynamics of filament formation, giving rise to a keratin network that is less stable. Furthermore, cell lines generated from EBS patients with mutations in K5 or K14 migrate faster in tissue culture scratch wound assays than their wild-type counterparts, perhaps because the network is more dynamic, allowing the cell to re-organise its keratin filaments more quickly for migration (Morley et al., 2003). The presence of keratin mutations, or a reduction in keratin expression, also reduces expression of desmosome components and cytoskeletal linker proteins (Long et al., 2006; Liovic et al., 2009; Wagner et al., 2012). This suggests an alternative disease mechanism in which keratin mutations might result in tissue fragility due to a reduction in junction proteins, leading to decreased stability of the tissue.
Keratin disease phenotypes suggest that there are multiple functions for keratins
Evidence from human keratin diseases that are not associated with tissue fragility, as well as data from experimental models such as knockout and transgenic mice, has suggested that keratins are essential for many different cellular processes, including non-mechanical stress responses, organelle positioning and tissue differentiation (Gu and Coulombe, 2007; Omary et al., 2009; Toivola et al., 2010). These are discussed in the following sections.
Mutant keratins alter stress response
Expression of many keratins is upregulated in response to stress, suggesting an important role for these proteins in the stress response (reviewed by Toivola et al., 2010). In particular, the expression of keratins K6a, K6b, K16 and K17 is induced by inflammatory cytokines, or in response to wound healing, and oxidative or UV stress (Freedberg et al., 2001; DePianto and Coulombe, 2004).
Mutations in the simple epithelial keratins K8 and K18 have been identified as risk factors for some patients with inflammatory bowel disease (Owens et al., 2004) or liver disease (Ku et al., 2003). Loss of mechanical resilience in the intestine (required to withstand peristaltic movement during digestion) caused by keratin mutations might initiate tissue damage (Owens and Lane, 2004). By contrast, liver is less mechanically stressed but is vulnerable to damage by toxins such as alcohol, and, in liver, keratin mutations reduce the tolerance to such assaults (Ku et al., 2007; Omary et al., 2009). This might help explain why mutations in these keratins are silent in most individuals, but still predispose carriers to liver injury mediated by toxins or viruses, which constitute situations of cell stress (Omary et al., 2009). It has been suggested K8 acts as a ‘phosphate sponge’, undergoing hyperphosphorylation, which absorbs phosphorylated stress-activated protein kinase (SAPK) activity, thus preventing apoptosis (see poster; Ku and Omary, 2006).
Mutations in K5 and K14 that cause EBS also alter the response of a cell to stress. Cells expressing mutated K5 or K14 show amplified and accelerated SAPK signalling in response to external stresses (D'Alessandro et al., 2002), and the constitutive upregulation of extracellular-signal-regulated kinase (ERK) signalling in these cells contributes to their increased resistance to apoptosis (Russell et al., 2010). The targeting of keratin filaments for degradation is increased during stress and by the presence of mutated or misfolded keratins that cannot integrate efficiently into the keratin network (Jaitovich et al., 2008; Löffek et al., 2010; Na et al., 2010; Rogel et al., 2010).
Keratins and organelle transport
Mutations in some keratins have revealed a role for these proteins in certain membrane trafficking events (see poster) (Kumemura et al., 2004; Toivola et al., 2005; Kumemura et al., 2008). A group of rare keratin diseases, including Dowling–Degos disease, that are caused by mutations in K5 or K14 have a skin pigment phenotype (patches of hyper- and hypo-pigmented skin; see poster) that is not directly linked to skin blistering (when present) (Uttam et al., 1996; Betz et al., 2006; Lugassy et al., 2006). Melanosomes (pigment-containing granules) are produced by melanocytes and transferred into basal keratinocytes where they are arranged in a distal cap over the nucleus. A defect in the transfer of melanosomes, or their arrangement in keratinocytes, can lead to changes in skin pigmentation. Although the mechanism by which mutations in epidermal keratins alter melanosome arrangement is not fully understood, it might involve an interaction of K5 with the chaperone HSC70, which is involved in vesicle uncoating (Planko et al., 2007).
Tissue differentiation is affected by keratin expression
The tissue-specific expression pattern of keratins is set early in development and differentiation and is therefore tightly controlled. Histopathologists have historically used keratin expression in epithelial tumours to identify the tissue of origin (Lane and Alexander, 1990), and more recently as prognostic markers (reviewed by Karantza, 2011). Changing the keratins that are expressed in a cell can have wide-ranging effects; these might be related to a change in cell fate that is reflected in the altered keratin profile and might lead to a subsequent alteration in physical properties (discussed in Owens and Lane, 2003). For example, mice expressing K1 in pancreatic β cells develop diabetes characterised by a reduction of insulin-secretory vesicles (Blessing et al., 1993). Deletion of K17 in a mouse model leads to defects in hair follicle cycling, wound repair (through a decrease in cell size and protein synthesis), and an altered inflammatory cytokine profile (Kim et al., 2006; Tong and Coulombe, 2006; DePianto et al., 2010). Although interactions of K17 with the adaptor protein 14-3-3σ and TRADD (tumour necrosis factor receptor type 1-associated death domain) suggest potential mechanisms by which loss of K17 could give rise to some of these phenotypes, it seems likely that deletion of K17 has a more general effect on tissue differentiation. Studies of ‘unnatural’ keratin pairs, such as K5 paired with K16 or K18 rather than its normal partner, K14, suggest that the assembly and mechanical properties of the filaments can be drastically altered by changing the type I or type II keratin present (Yamada et al., 2002; Lee and Coulombe, 2009). This is reflected in the mouse knockout of K14, in which expression of K16 or K18 only leads to partial rescue of the phenotype, to differing degrees (Hutton et al., 1998; Paladini and Coulombe, 1999), indicating that expression of the correct keratins is crucial for normal tissue function. The correct expression of keratins is also important in the development and maintenance of epidermal appendages (hair follicles and sebaceous glands). Mutations in the keratins expressed in these appendages lead to various defects including the hair fragility syndrome monilethrix, or pachyonychia congenita, a disorder characterised by nail dystrophy, painful palmoplantar keratoderma, and either hair-follicle-associated cysts or cell fragility in mucous surfaces (McLean et al., 2005; Schweizer et al., 2007) (see poster).
Understanding the precise role of keratins in these processes, and how mutations give rise to the defects described will be crucial to the development of therapies for the keratin diseases.
Opportunities for therapy
So far, therapeutic approaches to keratin diseases have focussed on two different areas, (i) genetic ablation of the mutant protein, and (ii) small-molecule therapies to stabilise the keratin network in some way. A successful trial of small interfering RNA (siRNA) against mutant K6a in a pachyonychia congenita patient recently demonstrated the feasibility of the first therapeutic route (Leachman et al., 2010), and without doubt further siRNA-based approaches will follow (Atkinson et al., 2011; Liao et al., 2011). However, significant concerns have been expressed regarding potential off-target effects of siRNA (Jackson and Linsley, 2010). As changes in keratin expression can have dramatic effects on tissue differentiation (see above), attempts to alter keratin expression for therapeutic purposes must be approached with caution. Nevertheless, small-molecule therapies are also being investigated to manipulate keratin levels in the cell. Several small molecules have now been shown to modulate keratin expression, and they either alter the expression of several keratins or target a specific keratin. They include statins, which moderately downregulate the activity of the promoter for K6a (Kerns et al., 2007; Törmä, 2011; Zhao et al., 2011). In severe keratin disorders, where aggregates are observed, it might be possible to ameliorate the symptoms by reducing the amount of aggregation with the use of chemical chaperones, which could ‘clear the way’ for the keratin filament cytoskeleton to reform correctly (Löffek et al., 2010; Chamcheu et al., 2011).
Keratin filaments have a crucial function in providing mechanical resilience to epithelial tissues. Associated with this function, keratins are involved in the cell stress response, tissue differentiation and organelle transport. Despite significant effort, the details of the substructure of keratin filaments, and in particular the role of the head and tail domains, are poorly understood (Strelkov et al., 2003), as is the control of filament assembly and dynamics within the cell. One model of filament dynamics, in which filaments are nucleated in the cell periphery and elongate and mature as they move towards the centre of the cell, was presented above. However, filament dynamics are likely to be different in a migrating or dividing cell, where the keratin cytoskeleton is highly dynamic, compared with cells in an intact tissue, where filaments appear to be stable and are anchored at cell junctions. A combination of localised highly dynamic filament re-organisation [as has been observed for vimentin (Helfand et al., 2011)] and some form of subunit exchange within the network seems more likely, but the mechanisms by which keratin dynamics are controlled are not yet understood. Disease-causing keratin mutations might provide essential information in this area, particularly in explaining how changing filament formation and dynamics give rise to fragile epithelial tissues. In order to address these issues, it will be necessary to improve laboratory models of keratin diseases to better reflect the physiological condition. Understanding the connection between the role of keratin filaments in providing mechanical resilience and the apparently unrelated phenotypes that are observed in certain keratin diseases is a priority, and further investigations in this area will hopefully yield new therapeutic interventions for keratin diseases.
The authors would like to thank John Common for critical reading of the manuscript, and John Common, Declan Lunny and Graham Wright (IMB Microscopy Unit) for assistance in producing images.
Work in the authors' laboratory is supported by the Biomedical Research Council, Agency for Science, Technology and Research (A*STAR), Singapore.
A high-resolution version of the poster is available for downloading in the online version of this article at jcs.biologists.org. Individual poster panels are available as JPEG files at http://jcs.biologists.org/lookup/suppl/ doi:10.1242/jcs.099655/-/DC1
- © 2012. Published by The Company of Biologists Ltd