Simple epithelia

Two types of simple epithelia are distinguished, i.e. single-layered and multi-layered epithelia. In single-layered epithelia, all cells are attached to the basement membrane and extend to the surface of the epithelium (e.g. endothelium, mesothelium, epithelium in the renal tubules and alveoli). In multi-layered epithelia (i.e. pseudostratified epithelia), all cells are in contact with the basement membrane but do not necessarily extend to the surface of the epithelium (e.g. respiratory epithelium).

Transitional epithelium

In transitional epithelia, at least some cells attach to the basement membrane and also extend to the surface of the epithelium [e.g. epithelium of the renal pelvis, ureter and urinary bladder (also called ‘urothelium’ by Banks, 1993)].

Microfilaments

Microfilaments are the smallest filaments of the cytoskeleton with a diameter of 7 nm. They are assembled from globular actin molecules, which join to form filamentous actin by using energy from adenosine triphosphate or guanosine triphosphate. Two filamentous actin molecules then combine into a helical microfilament. Microfilaments are polarized so that while one end is elongated by the addition of more filamentous actin molecules, the opposite end is disassembled. Therefore, actin microfilaments can act as conveyor belts for micro-motor molecules, such as myosin. Actin filaments are anchored via actin-binding proteins (e.g. plectin) to the cell membrane by focal adhesions, which are part of tight junctions.

Microtubules

Microtubules are the largest filamentous structures of the cytoskeleton with a diameter of about 20 nm. They are assembled from α- and β-tubulin molecules, which form heterodimers. These heterodimers are assembled using energy from guanosine triphosphate to form protofilaments, which in turn are combined to form a hollow microtubule. Microtubules are polarized and, thus, can act as conveyor belts for micro-motor molecules (e.g. dynein, kinesin) similar to microfilaments. Microtubule formation is initiated by so-called microtubule-organizing centers, such as the basal bodies in kinocilia or the centrosomes involved in the formation of the spindle apparatus of cells. The network of microtubules is stabilized within a cell by the surrounding microfilaments and intermediate filaments. The microtubule-organizing centers also interact with keratin filaments (see below) by binding to them through mediator proteins and thereby maintaining the polarity of certain epithelial cells.

Intermediate filaments in general

Intermediate filaments differ fundamentally from microfilaments and microtubules. Unlike microfilaments and microtubules, intermediate filaments can aggregate into bundles of varying diameter, ranging from 7 to 12 nm. Intermediate filaments are not polarized. Therefore, intermediate filaments are not involved in intracellular transport but serve as a scaffold for the cytoskeleton. Recently, however, keratin filaments in mammals have been suggested to be involved in the transport of melanosomes from their site of endocytosis in the cell periphery to the cell center (Planko et al. 2007). Keratins that form intermediate filaments are expressed exclusively in epithelial cells sensu lato (Moll et al. 1982; Steinert, 2001), regardless of the germ layer origin of these cells. In non-epithelial cells, there are various types of intracellular intermediate filaments, including desmin (in myogenic cells), vimentin (in fibroblasts and fibrocytes), lamin (in cell nuclei) and neurofilaments (in neuronal cells) (Moll et al. 1982). In addition, cells in connective tissues (e.g. fibroblasts, chondroblasts) produce extracellular intermediate filaments (i.e. the collagen filaments).

Previously, it was assumed that the process of filament aggregation could start anywhere in a cell. Recent observations suggest, however, that the formation of intermediate filaments requires some kind of organizer, such as the focal adhesions of microfilaments to the cell membrane (Gu & Coulombe, 2007). The assembly and turnover of intermediate filaments require the assistance of enzymes, such as kinases and phosphatases, as well as chaperone molecules, such as heat shock proteins (Herrmann et al. 2007).

Intermediate filaments are a crucial component of the so-called cell signalosome, the multifunctional protein complex that is essential for cell development and possibly for the regulation of protein degradation, which translates environmental cues into modifications of gene expression (Herrmann et al. 2007). In the cytoskeleton, intermediate filaments are attached to the nuclear membrane next to nuclear pores and to cell membrane modifications, such as desmosomes and hemidesmosomes (Steinert, 2001). Electron microscopic images suggest that intermediate filaments loop through the attachment plaques of desmosomes (Pekny & Lane, 2007).

Primary structure of keratins

Each keratin is characterized by a chain of amino acids as the primary structure of the keratin protein, which may vary in the number and sequence of amino acids, as well as in polarity, charge and size (Brown, 1950; Makar et al. 2007). However, the amino acid sequence of a particular keratin is remarkably similar in different species (Fuchs, 1983). For example, bovine, rat and human skin keratins are all rich in the amino acids glycine, serine, leucine and glutamic acid (Fuchs, 1983). The amino acid sequences of keratins in soft- or hard-cornified epithelia of mammals are also similar (Fuchs, 1983). This similarity is the basis for the concept of orthologous keratins (i.e. similarity in immunological epitope, MW, pI, amino acid sequence and location within tissues).

The amino acid sequence of a keratin influences the properties and functions of the keratin filament (Roop et al. 1984). Furthermore, the position of a particular amino acid within the chain of amino acids can influence the entire three-dimensional architecture of the keratin molecule (Wu et al. 2000). For example, lysine in position 23 and glutamate in position 106 are required for the proper assembly of K5/K14 keratin filaments in vitro and in vivo (Wu et al. 2000).

The amino acid sequence of a specific keratin determines the molecular structure and properties of the secondary, tertiary and quaternary structures of keratins, as well as the nature of the bonds (e.g. covalent or ionic) to other components of the cytoskeleton (Steinert et al. 1984; Coulombe & Omary, 2002). In particular, the sulfur-containing amino acids methionine, histidine and cysteine are instrumental in establishing disulfide bonds within an amino acid chain, between two keratins or between keratins and KFAPs (e.g. filaggrin, involucrin, see section ‘Keratins and keratin filament-associated proteins’) in keratinized cornifying cells of stratified epithelia.

The amino acid chain of β-keratins [see section ‘Other keratins (e.g. β-keratin and thread keratins)’], which are characteristic of hard-keratinized and hard-cornified modified epidermis in reptiles and birds, is shorter than that of α-keratins (which are characteristic of the soft-keratinized and soft-cornified epidermis of reptiles and birds, as well as of all keratins in mammals). For example, in the β-keratin of the emu feather, only 32 amino acids form the central rod domain, 23 amino acids form the head domain and 47 amino acids form the tail domain (Fraser & Parry, 2008). Feather keratins also characteristically contain large amounts of the amino acids serine, proline, valine, leucine, glutamate and aspartate (Woodin, 1956).

Post-translational changes of the primary structure of keratins

The amino acid sequence (i.e. the primary structure of a keratin encoded by a gene) is slightly longer than the amino acid sequence of the mature keratin, which indicates a post-translational modification of the keratin prior to the formation of the keratin filaments (Hesse et al. 2001).

Comparative analyses of keratins extracted from the living cells of the Stratum spinosum et granulosum and from the dead cells of the Stratum corneum in humans and mice demonstrate a precursor/product relationship, also indicating post-translational modifications as well as other modifications during cell differentiation (i.e. keratinization and cornification) (Bowden et al. 1984). The MW of a keratin extracted from the cells of the Stratum corneum is about 2–5 kDa smaller than that of its precursor in the living epidermal cells, this possibly being the result of a loss of amino acids from the tail domain (Bowden et al. 1984). The MW of keratins can also increase as a result of their post-translational modification, sometimes altering the pI (Schweizer & Winter, 1982). Post-translational modifications influence the physicochemical properties of keratins and their assembly to form heterodimers and tetramers (Yamada et al. 2002). Such alterations primarily affect the head and tail domains of keratins, whereas the rod domain may be modified by caspase-mediated cleavage (Coulombe & Omary, 2002).

Post-translational modifications of keratins, such as the formation of disulfide bonds, phosphorylation, glycosylation, deimination or inter- and intrachain peptide bonds, can influence the conformation of the molecule and the formation of keratin filaments.

Cross-linking of keratins via inter- and intrachain peptide bonds occurs during epidermal differentiation (Kubilus & Baden, 1983) by connecting an ɛ-lysine to a γ-glutamyl amino acid residue mediated by transglutaminases (Buxman & Wuepper, 1975; Baden et al. 1976; Ahvazi et al. 2003). Keratins may be cross-linked by transglutaminases to one another or to constituents of the cornified envelope, such as the Stratum corneum basic protein or involucrin to form insoluble aggregates (Kubilus & Baden, 1983). For example, transglutaminases cross-link a lysine residue of the V1 subdomain in the head domain (see section ‘Primary structure of keratins’) of the type II keratins K1, K2e, K5 and K6 to involucrin and loricrin in the cornified envelope in stratified squamous epithelia (Candi et al. 1998; Eckert et al. 2005). Three different transglutaminases have been identified so far in the epidermis of mammals (Eckert et al. 2005), all of which depend on calcium as a co-factor (Ogawa & Goldsmith, 1976). Transglutaminases are also present in the avian epidermis (Alibardi & Toni, 2004).

Intra- and intermolecular disulfide bonds are formed in keratins by connecting two sulfhydryl residues of two amino acids (such as two cysteines) enzymatically via the enzyme sulfhydryl oxidase (Hashimoto et al. 2000). This enzyme (MW 65 kDa) is expressed in keratinocytes of the Stratum granulosum of the rat and mouse epidermis (Hashimoto et al. 2000; Matsuba et al. 2002). Disulfide bonds are also formed in the rod domain of K4 and K13, which are produced by suprabasal cells of the esophageal epithelium (Pang et al. 1993). In addition to intrachain disulfide bonds, interchain disulfide bonds are formed in heterodimers of keratins (Pang et al. 1993). All keratins in the cornified cells of the human epidermis are cross-linked by interchain disulfide bonds, whereas the keratins in the keratinizing epidermal cell are not (Sun & Green, 1978).

Phosphorylation and dephosphorylation are perhaps the most important post-translational modifications of keratins because they affect the equilibrium of soluble keratins deposited in granular aggregates (Strnad et al. 2002). The degree of phosphorylation is variable (Sun & Green, 1978) and may enable or prevent the interaction of keratins with other molecules, such as signaling molecules, receptor molecules, etc. (Kirfel et al. 2003). Potential consequences of keratin phosphorylation include changes in the solubility of keratins, in the organization of keratin filaments and in the interactions with other proteins (Liao et al. 1995; Strnad et al. 2002). Phosphorylation modifies the conformation of keratin filaments as a prerequisite for remodeling the cytoskeleton (Owens & Lane, 2003; Pekny & Lane, 2007). Specifically, the phosphorylation of amino acids in the head domain changes the overall net charge of this region and, thus, prevents any interactions with the rod domains of adjacent keratins, thereby preventing the assembly of keratin filaments (Wöll et al. 2007). The head domains of type II keratins are more easily phosphorylated enzymatically than those of type I keratins (Liao et al. 1995). For example, the amino acids serine or threonine of the head domain of type II keratins K4, K5 and K6 are phosphorylated enzymatically in epithelial cells of the esophagus and epidermis (Ku & Omary, 2006). The enzyme p38 mitogen-activated phosphokinase catalyses the phosphorylation of serine in the head domain of the type II keratin K8 (Wöll et al. 2007). Similarly, the serine or threonine residues of both the type II keratin K8 and type I keratin K18 are phosphorylated in hepatocytes (Ku et al. 2002). Phosphorylation of K8 and K18 in the liver prevents the attachment of ubiquitin molecules and therefore protects them from proteolysis (Ku & Omary, 2000). Phosphorylation also contributes to the depolymerization (i.e. disassembly) of keratin filaments in hepatocytes. As a result of this depolymerization, the pool of soluble tetramers increases (Ku et al. 1996; Steinert, 2001). Enzymes, such as mitogen-activated protein kinase and protein kinase C-related kinase, are involved in the phosphorylation of keratins, which are then able to bind to heat-shock proteins and to signaling molecules in order to intercept signals (Ku et al. 1996). In addition, keratin phosphorylation sequesters phosphates, thereby influencing signaling cascades (Ku et al. 1996, 2002). In contrast to phosphorylation, dephosphorylation promotes keratin filament reassembly (Steinert, 2001).

Glycosylation is another post-translational modification of proteins. The glycosylation of keratins has been documented in human hepatocytes (Chou et al. 1992) and in the epidermis of pigs (King, 1986). For example, a sugar moiety, such as N-acetylglucosamine, is enzymatically linked to the amino acid residue of serine or threonine of K8 and K18 in hepatocytes (King & Hounsell, 1989; Chou et al. 1992). Glycosylation has also been observed in K13 in the esophageal epithelium (King & Hounsell, 1989). Glycosylated amino acid residues are not available for other chemical processes, such as phosphorylation, which results in an alteration of the binding or signaling functions of keratin (Chou et al. 1992).

Deimination, also known as citrullination, is another form of post-translational modification of keratins and keratin filaments (Kubilus et al. 1980). During this process, the enzyme peptidylarginine deiminase converts arginine to citrulline. Because citrulline is neutral, the loss of a positive charge may lead to conformational changes of the keratin and keratin filaments. For example, deimination of K1 takes place during terminal differentiation of human epidermal cells (Senshu et al. 1996). K10 and the KFAP filaggrin are also deiminated (Senshu et al. 1996).

Proteolytic cleavage of proteins may also occur post-translationally. Keratins and keratin filaments are quite resistant to cleavage by common proteolytic enzymes (e.g. trypsin and pepsin) due to their stabilization by numerous cross-linking disulfide bonds (Gupta & Ramnani, 2006). Proteolytic cleavage of keratins and keratin filaments is achieved by specific enzymes, such as caspase or cathepsin, which are released from lysosomes (Presland & Jurevic, 2002). Proteolysis can occur in the proteasome as part of the disassembly of keratin filaments and proteins in simple epithelia (e.g. hepatocytes) or as part of the processes of keratinization and cornification in stratified epithelia (e.g. the epidermis).

Keratin proteolysis (i.e. keratinolysis) is also achieved by specific proteases (i.e. keratinases) that are produced by keratinolytic fungi or bacteria known to affect fingernails, hairs, beaks and feathers (Gupta & Ramnani, 2006; Thys & Brandelli, 2006). Such keratinases are related to serine proteases and have important biotechnological applications, such as for the degradation of feathers as a waste byproduct in the poultry industry (Gupta & Ramnani, 2006).

Secondary structure of keratins

All proteins that form intermediate filaments have a tripartite secondary structure consisting of an N-terminal head domain, a central α-helical rod domain and C-terminal tail domain, and all proteins are able to self-assemble into filaments (Steinert et al. 1984; Coulombe et al. 2004). The secondary structure of keratins is also divided into three parts (Fig. 2a), i.e. the head domain (towards the N-terminal of the molecule), the rod domain in the center and the tail domain (Steinert et al. 1984; Presland & Dale, 2000; Parry et al. 2007). Each of these three domains is divided into subdomains (Steinert et al. 1985; Hatzfeld & Burba, 1994). Domains and subdomains are determined by the amino acid sequence of the keratin and serve various functions in the assembly of keratin filaments and in the binding of keratins and keratin filaments to cell adhesion complexes or to signaling molecules (Steinert et al. 1984; Hatzfeld & Burba, 1994). The distribution of ionic charges may vary in the different domains and subdomains, with the head and tail being positively charged (Steinert et al. 1984).

Changes in the secondary structure of keratins

The secondary molecular structure (e.g. α-helix, β-sheet or β-turn) of keratins and filaments can be changed under the influence of physicochemical forces, such as mechanical forces or chemical processes.

Mechanical forces, such as tension, compression and probably also shearing, can alter the secondary molecular structure of keratins. For example, the secondary molecular structure of keratin filaments in sheep hair is changed from an α- to a β-conformation when a hair bundle is stretched (Bendit, 1957). As revealed by X-ray diffraction pattern analysis, low humidity accelerates this transformation, possibly through changes of the internal hydraulic pressure within the cells. X-ray diffraction patterns prior to and after release from stretching revealed no significant differences, suggesting that the original secondary molecular structure of the keratins is recovered (Bendit, 1957). If a horse hair is stretched for more than 60% of its length, the α-helix of the keratin molecule is transformed into a β-sheet, possibly due to an unraveling of the coiled-coil of the heterodimer (Kreplak et al. 2001).

Certain direct modifications (e.g. phosphorylation, intra- and interchain peptide bonds) of keratins or indirect changes in the chemical milieu (e.g. pH, ionic strength, osmolarity, etc.) of the cell compartments affect the molecular structure and functions of keratins (Wilson et al. 1992).

Functions of the domains and subdomains of keratins

The domains and subdomains of one keratin molecule (Fig. 2a) will interact with those of adjacent keratin molecules in the formation of subunits, such as heterodimers, tetramers and finally of keratin filaments. Although the α-helical rod domain is sufficient for the formation of heterodimers (Fig. 2b) or tetramers (Fig. 2c), the assembly of keratin filaments requires the non-helical head and tail domains (Wilson et al. 1992).

Heterodimers are formed by a pair of keratins that are oriented parallel so that the head of one adjoins the head of the other. The rod domain assists in the formation of the c. 45 nm long coiled heterodimer. The amino acid sequences in the rod domain are largely homologous or, in other words, the rods are only minimally variable. This means that their structure is highly deterministic of its function and that variants may not be able to function with partner keratins (Steinert, 2001).

The head and tail domains establish the correct parallel and staggered orientation of the rod domains in a heterodimer and possibly also in a tetramer. In addition, the head domain of type II keratins is covalently bound to proteins in the cornified envelope of keratinizing epithelial cells, whereas the tail domain anchors the intermediate filaments to the desmosome via so-called linker proteins (Candi et al. 1998). The tail also interacts with KFAPs, such as filaggrin or trichohyalin, forming a filament–matrix complex that is similar to reinforced concrete to impart stability to the cytoskeleton (Hatzfeld & Burba, 1994).

Tertiary structure of keratins: heterodimer

The tertiary structure of keratins is a heterodimer (Fig. 2b) that is formed by the rod domains of one acidic and one basic keratin in parallel orientation (Er Rafik et al. 2004). This heterodimer is the first building block of a keratin filament (Eichner et al. 1986). Keratin filaments are obligatory heteropolymers containing equimolar amounts of type I and type II keratins (Moll et al. 1982; Hatzfeld & Franke, 1985). The heteropolymeric nature of the keratins is established at the level of the double-stranded coiled-coil (i.e. a heterodimer) (Steinert, 1990), as predicted by Crick (1953) and Pauling & Corey (1953). Therefore, synthesis of keratins in the cells is tightly controlled in order to obtain stoichiometry of the acidic and basic keratin pairs and to produce the keratin filaments that are specific to particular stages of the differentiating epithelial cells (Navarro et al. 1995). All epithelial cells at all stages of differentiation express mRNA sequences of both acidic and basic types of keratins in a coordinated fashion (Fuchs & Marchuk, 1983; Fuchs et al. 1983).

Formation of a heterodimer by the parallel alignment of a single pair of type I and type II keratin is the first step in the assembly of a keratin intermediate filament (Coulombe & Fuchs, 1990; Steinert, 2001). In a heterodimer, only the α-helical rod domains of the keratins align with each other (Steinert et al. 1984; Hatzfeld & Burba, 1994; Kammerer et al. 1998) and the alignment and structure are stabilized by the hydrophobic interactions of certain amino acid residues (Coulombe & Fuchs, 1990; Meng et al. 1998). Jones et al. (1997) observed that the keratin molecules in heterodimers of soft-keratinizing-cornifying cells are aligned in parallel but slightly out of phase. This shift of about seven to eight amino acids in the alignment allows an overlap of the head and tail of the keratins where two heterodimers come together. In contrast, the keratins in heterodimers of hard-keratinizing-cornifying cells are aligned in register and do not overlap the tail and head domains where two heterodimers come together. In vitro, Coulombe & Fuchs (1990) found that heterodimers (and tetramers) are the most stable building blocks of keratin filaments. The heterodimers of keratins are still soluble in the cytoplasm but this solubility depends on the type of keratin and on the physicochemical characteristics of the cytoplasm (Herrmann & Aebi, 2004). For example, the K8/K18 heterodimers are soluble in vitro in a medium with a pH of 9, whereas, in the same conditions, the heterodimers K5/K14 extensively form filaments (Herrmann & Aebi, 2004).

Mechanical functions of keratins

Keratins fundamentally influence the architecture (e.g. cell polarity and cell shape) and mitotic activity of epithelial cells (Magin et al. 2007). The best-known function of keratins and keratin filaments is to provide a scaffold (through self-bundling and by forming thicker strands) for epithelial cells and tissues to sustain mechanical stress, maintain their structural integrity, ensure mechanical resilience, protect against variations in hydrostatic pressure and establish cell polarity (Coulombe & Omary, 2002; Gu & Coulombe, 2007). Keratin filaments can be rapidly disassembled and reassembled, thereby providing flexibility to the cytoskeleton. Keratins as building blocks of keratin filaments are concentrated in the cell periphery near the focal adhesions of the cell membrane, and the polymerization of these keratins forming filaments is regulated by signaling molecules (e.g. heat shock proteins, 14-3-3 regulatory proteins, and various kinases and phosphatases) (Magin et al. 2007).

Keratins are functionally redundant, hence they can assemble in various combinations to form keratin filaments (e.g. in the epidermis K1/K10, K1/K9, K2/K9 or K2/K10), and some keratins, at least, can be replaced by others without a loss of functionality of the keratin filaments (Coulombe & Omary, 2002). The regulation of keratin function is achieved by the post-translational modification of the keratins (Coulombe & Omary, 2002). Keratins and keratin filaments interact with cell-adhesive complexes, i.e. desmosomes, hemidesmosomes and focal adhesions (Maniotis et al. 1997; Coulombe & Omary, 2002). Keratin filaments are also connected to microtubules, which indicates that the constituents of the cytoskeleton interact with one another (Kim & Coulombe, 2007; Oriolo et al. 2007). This interaction of keratin filaments with constituents of the microtubule organizing center is involved in the apicobasal orientation of microtubules, which maintains the cell polarity of enterocytes, i.e. epithelial cells of the intestinal mucosa (Oriolo et al. 2007).

To adapt to changing conditions within cells, keratins interact with various other proteins, such as heat shock proteins (i.e. chaperones), desmoplakins of desmosomes, tubulin complex proteins, etc. (Izawa et al. 2000; Garmyn et al. 2001; Oriolo et al. 2007).

Non-mechanical functions of keratins

Keratins and keratin filaments are involved in other than mechanical functions, such as cell signaling, cell transport, cell compartmentalization and cell differentiation (Oshima, 2007; Vaidya & Kanojia, 2007). For example, K10 inhibits cell proliferation in suprabasal cells but induces cell differentiation (Paramio et al. 1999). Keratin filaments also influence the cell responses to intrinsic and extrinsic signals, such as pro-apoptotic signals and the proper distribution of membrane proteins in polarized epithelial cells (Coulombe et al. 2004; Kim & Coulombe, 2007). For example, in hepatocytes, K8 and K18 bind to signaling molecules, thereby disrupting the apoptotic signaling cascade that would initiate apoptosis (Gu & Coulombe, 2007). The rod domain of K18 can bind to the C-terminal of a cell membrane-bound signal receptor, thereby blocking the activation of the second messenger, i.e. an enzyme involved in the apoptosis in hepatocytes (Inada et al. 2001). Similarly, K8 and K18 bind to the tumor necrosis factor receptor 2 and, thereby, modify the apoptotic signaling cascade (Paramio & Jorcano, 2002). Keratins bind various signaling molecules, such as protein kinases, phosphatases and proteins of the 14-3-3 family (Paramio & Jorcano, 2002). In epidermal wounds, K1 and K10, which are normally expressed in suprabasal cells, are downregulated, whereas the expression of K6, K16 and K17 is initiated (Coulombe et al. 2004). Keratins in modified epithelial sensory cells (e.g. Merkel cells, hair cells) could be involved in the signal transduction from the apical part to the basolateral part, where transmembrane channels or transmitter-containing vesicles are opened to release signaling molecules transducing the signal to associated neurons (Lumpkin & Caterina, 2007). K8 is a marker for cells developing taste buds (Mbiene & Roberts, 2003), and denervation of the gustatory fibers eliminates the keratins that are specific for taste buds (Oakley, 1993). In certain supporting epithelial cells in the organ of Corti, keratin filaments may be involved in micromechanical functions of transmitting vibrations to the sensory hair cells (Mogensen et al. 1998).

Keratin filaments also influence cell metabolic processes by regulating protein synthesis and cell growth (Gu & Coulombe, 2007). K17 regulates the protein synthesis and cell growth in injured stratified epithelia by binding to a signaling molecule (Kim et al. 2006).

Keratins may also be involved in the transport of membrane-bound vesicles in the cytoplasm of epithelial cells. For example, the distribution of membrane-bound melanosomes in epidermal basal cells is disrupted in cells that produce a depleted K5 due to a ‘loss of function’ mutation in the gene affecting the amino acid sequence of the head domain of K5 (Planko et al. 2007).

Keratins in simple epithelia

Keratins of simple epithelia make up less than 5% of the total protein content in an epithelial cell (Nelson & Sun, 1983). Simple epithelia usually express the basic keratin K8 and its acidic partner K18, along with various other keratins (e.g. K7, K19, K20 and K23) during cell differentiation following cell division (Achtstätter et al. 1985; Bosch et al. 1988; Kirfel et al. 2003; Owens & Lane, 2003; Lu et al. 2005; Pekny & Lane, 2007). In venules, lymphatics and capillaries, endothelial cells express K7 and K18 but not K8 and K19, which are characteristic of many simple epithelia, or K14, which is characteristic of basal cells of all stratified epithelia (Miettinen & Fetsch, 2000). During the development of simple epithelia, the translation of mRNAs encoding the basic keratins K7 and K8 precedes that of the acidic partners K18 and K19, and consequently there are no keratin filaments in the cell periphery at first but instead many granulated aggregations of K7 and K8 (Lu et al. 2005). Different keratins are produced in various parts of the simple epithelia lining sweat glands (Schön et al. 1999). Cells in the secretory unit synthesize K7, K8 and K19, whereas cells forming the duct produce K5, K14 and K19 (Schön et al. 1999). Unlike the keratins of cornified stratified epithelia, those of simple epithelia are not cross-linked by disulfide bonds (Owens & Lane, 2003). In apoptotic cells of simple epithelia, keratin filaments and proteins are digested enzymatically. For example, K18 in apoptotic hepatocytes is cleaved by the enzyme caspase generating a neoepitope of K18, and this neoepitope can be recognized immunohistochemically to detect apoptotic hepatocytes for diagnostic purposes (Leers et al. 1999).

Keratins in stratified epithelia

Keratins in stratified epithelia (Fig. 3) make up 30–40% of the total protein content in an epithelial cell (Nelson & Sun, 1983), with different sets of keratins being synthesized at specific stages of cell differentiation. Keratins in stratified epithelia are classified according to their position within a stratum of keratin-producing cells. Basal keratins (i.e. ‘B-type keratins’) are produced in the basal stratum, whereas the differentiation-specific keratins (i.e. ‘d-type keratins’) are produced in suprabasal strata (Schermer et al. 1989). Suprabasal cells can also produce ‘hyperproliferative’ keratins (i.e. ‘H-type’ keratins), such as K6 and K16 (Schermer et al. 1989). Stratified epithelial cells produce keratins (e.g. K1, K5, K10, K14) in such a manner that each of the acidic keratins (e.g. K1, K5) is always matched by a particular basic keratin (e.g. K10, K14) to form a heterodimer (Steinert, 2001). Keratins characteristic of simple epithelia, such as K8 and K18, can be co-expressed with keratins characteristic of stratified epithelia, such as K4, K13, K14 and K15 (Bosch et al. 1988).

Keratins typical of basal cells of stratified epithelia include the basic keratin K5 and the acidic keratin K14, which combine to form heterodimers (Nelson & Sun, 1983; Dale et al. 1985). Stratified epithelia express additional keratins starting in the suprabasal cell layers (Bowden et al. 1984). Different types of stratified epithelia are characterized by different keratin pairs in the suprabasal cells (Eichner & Kahn, 1990). There are at least four differentiation types for the suprabasal cells of stratified epithelia (Sun, 2006), i.e. the skin type, cornea type, esophagus type and hyperproliferative type. For example, the suprabasal cells in the epidermis of the skin specifically produce K1 and K10, whereas those of the anterior corneal epithelium produce K3 and K12, and those of the esophageal epithelium produce K4 and K13 (Eichner & Kahn, 1990). The hyperproliferative type of suprabasal cells is characterized by K6 and K16 (Sun, 2006).

The synthesis of the keratins that are characteristic of a specific type of differentiation of suprabasal cell may be reversible. For example, the expression of K1 and K10 in the suprabasal cells of the epidermis can be modulated reversibly (Nelson & Sun, 1983), so that the suprabasal cells of the keratinizing and cornifying epidermis can produce K2 and K9 in addition to K1 and K10 (Moll et al. 1987). The suprabasal cells of esophageal epithelium produce K4 and K13 but are also able to co-express K6 and K16 (Pang et al. 1993).

Keratins in stem cells

Stem cells of epithelia reside in niches, such as the deep epidermal ridges or the bulge of the hair follicle, where they are insulated from physical and mechanical influences (Watt, 1998). These stem cells produce specific keratins, such as K15 and K19 (Narisawa et al. 1994; Watt, 1998; Waseem et al. 1999; Webb et al. 2004). Originally, all basal cells in stratified epithelia (Fig. 4) were thought to be able to produce new cells to maintain the epidermis because mitotic figures were observed in the entire basal layer of the fetal human epidermis, in which all basal cells produce the keratin K15, which is characteristic of epidermal stem cells (Webb et al. 2004). However, the concepts of epidermal stem cells, transient amplifying cells and epidermal proliferative units were developed by Alonso & Fuchs (2003) when it was recognized that the telomeres of chromosomes limit the number of possible cell divisions to about 30 within the lifetime of a cell lineage and that this number of epidermal cell divisions would not suffice to compensate for the loss of cells through normal desquamation and shedding (Harley, 1991; Ueda, 2000).

Under the new concept, stem cells have a special capacity to renew themselves over the lifetime of an organism, and they divide into two different daughter cells, namely one stem cell and one transient amplifying cell (Watt, 1998; Alonso & Fuchs, 2003; Webb et al. 2004). Primary cell cultures of stem cells from hair follicle and the epidermis generate a variety of cell types, namely stem cells (10%), transient amplifying cells (40%) and keratinocytes (50%), with each cell type having different potentials to proliferate and generate new cell clones (Papini et al. 2003). In the interfollicular epidermis, 2–5% of the basal cells are stem cells and produce K15 (Webb et al. 2004). Some basal cells in the interfollicular epidermis cells are mitotically active keratinoblasts and produce the keratin K14. Some other basal cells are differentiating cells or keratinocytes that downregulate the production of K14 and upregulate the production of K10 (Webb et al. 2004). Drawing upon the observations by Alonso & Fuchs (2003), the new model of epidermal renewal proposes that the basal layer of stratified epithelia is comprised not only of stem cells and transient amplifying cells but also of cells that have already started their differentiation into keratinocytes. The transient amplifying cells replicate themselves four to eight times before they finally divide into keratinocytes. The descendants of one stem cell (i.e. the transient amplifying cells and keratinocytes) are collectively called an ‘epidermal proliferative unit’ (Strachan & Ghadially, 2008). Hence, the presence of transient amplifying cells would generate 480–1920 keratinocytes from a single stem cell over the lifetime of an organism (i.e. 16 × 30 to 64 × 30), which is about 16–64 times more than if stem cells produced only keratinocytes without transient amplifying cells. The epidermal proliferative unit is not simply a theoretical concept but also explains the occurrence of distinctly localized skin modifications in skin diseases.

In contrast, Clayton et al. (2007), using a different approach, proposed a model of epidermal cell amplification and maturation that includes only stem cells and keratinocytes. In this model, stem cells divide symmetrically into stem cells (8%) or keratinocytes (8%) and asymmetrically into a stem cell and a keratinocyte (84%). Thus, one cell produces maximally 1 billion keratinocytes over the lifetime of an organism [i.e. 2³⁰= (1024)³ ∼10⁹].

Epidermal stem cells, which are also called keratinoblasts, produce the keratins K5, K14, K15 and K19 (Lyle et al. 1998). Stores of these keratinoblasts are found in the bulge of hair follicles, in epidermal glands (i.e. sebaceous, sweat and mammary glands) and possibly in the interfollicular epidermis (Jones, 1996). In the hair bulge, the stem cells in the basal layer synthesize the keratins K5, K15, K17 and K19 but not K14 (Larouche et al. 2008). The keratinoblasts in the first suprabasal layer of the hair bulge express K5 and K17 but not K14, K15, K16 or K19. The keratinoblasts in the second suprabasal layer of the hair bulge express K5, K14, K15 and K17 but not K19 (Larouche et al. 2008). These variations in keratin expression may support the concept of different cell types (i.e. stem cells, transient amplifying cells or undifferentiated keratinocytes).

The keratinocytes are pushed towards the surface of the epidermis by the newly produced cells in the basal layer and go through various stages of maturation (i.e. keratinization only or keratinization and cornification). However, there are several issues that remain poorly understood. One issue is the lifespan of an individual cell lineage and the duration of the various maturation stages of the tissue-specific cells. Another issue is whether keratinocytes are still mitotically active or not, although the models by Alonso & Fuchs (2003) and Clayton et al. (2007) assume that they are not as indicated by the production of K1/K10 instead of K5/K14.

Keratins in epithelial cells with special functions

Some epithelial cells do not fit within the traditional concept of epithelial tissues because they perform specialized tasks but nevertheless express keratins that are also expressed in epidermal cells. This fact provides a possible explanation for the connection between certain epidermal disorders and disorders of other organs, such as infertility. For example, spermatids express the keratins SaK57 (i.e. a homologue to the epidermal K5) and K9, which are expressed in the course of spermiogenesis (Kierszenbaum, 2002). Myoepithelial cells of the secretory units of exocrine serous glands, such as sweat, mammary, lacrimal and salivary glands, produce K17 (Freedberg et al. 2001). Merkel cells express K8, K18 and K19 (Moll et al. 1984; Narisawa et al. 1994; Larouche et al. 2008), and smooth muscle cells in the human placenta cross-react with an anti-K8 antibody, thus indicating that they contain K8 (Miettinen & Fetsch, 2000) or at least a molecule with an epitope that binds to this antibody. Similarly, in the transition from a non-sensory to a specialized sensory cell, the synthesis of keratins is downregulated (Cyr et al. 2000).

Signaling molecules influencing and regulating the keratinization

The matrix metalloproteinase 19 is expressed in the basal keratinocytes of the epidermis. This metalloproteinase affects numerous processes in the epithelial cells of the epidermis, such as adhesion, proliferation, migration and cell differentiation (Beck et al. 2007). A decrease in the synthesis of matrix metalloproteinase 19 results in a decrease of cell proliferation and in an enhancement of cell differentiation, which is accompanied by an altered synthesis of keratins, of KFAPs, such as involucrin and loricrin, and of intercellular cementing substances such as sphingolipids, glucosylceramides and cholesterylsulfate.

The membrane-bound enzyme phosphatidylinositol 3-kinase is another regulator of differentiation of keratinocytes (Sayama et al. 2002). This enzyme mediates an adhesion signal in keratinocytes attached to the basement membrane suppressing the beginning of keratinization processes.

Keratins in different cell layers

Keratinizing keratinocytes of stratified non-cornifying or cornifying epithelia produce specific keratins that differ from those produced in the basal cells. The synthesis of K3 and K12 is characteristic of the keratinization in the stratified epithelium of the cornea, whereas the synthesis of K4 and K13 is characteristic of the keratinization of the stratified oral epithelium. The process of keratinization in the suprabasal cells of the stratified epidermis is defined by the production of K1 and K10. In parallel to the synthesis of those specific keratins, the keratinizing epithelial cells produce proteins of the keratin filament–matrix complex, such as profilaggrin, and of the cornified envelope, such as involucrin. Besides synthesizing proteins, the keratins produce special lipids, such as glucosylceramides, which are stored in membrane-coated vesicles and are later released into the intercellular space of the stratified epithelium via exocytosis. All of these processes are coordinated in parallel to the keratin synthesis.

Keratins and keratin filament-associated proteins

The KFAPs are non-filamentous structural proteins that are produced in keratinocytes of the Stratum granulosum of stratified keratinizing and cornifying epithelia and stored in keratohyalin granules (Jessen, 1970). In general, two types of KFAPs can be distinguished based on the structures to which the KFAPs contribute. The KFAPs (e.g. involucrin, loricrin) that are stored in a subset of keratohyalin granules (i.e. the l-granules) form the subcytolemmal cornified cell envelope. The KFAPs that are stored in another subset of keratohyalin granules (i.e. the F-granules) combine with keratin filaments to form the filament–matrix complex stabilizing the cytoskeleton of cornifying keratinocytes (Eckert et al. 2005). Different matrix-forming KFAPs are produced in the soft-keratinizing and cornifying keratinocytes of the interfollicular epidermis (e.g. profilaggrin/filaggrin) compared with those produced in the modified soft-keratinizing and cornifying cells of the hair follicle (e.g. trichohyalin, trichoplein) (Mack et al. 1993; Nishizawa et al. 2005). This corroborates that the soft keratinization and cornification processes in the interfollicular vs. those in the follicular epidermis are different.

The onset of the synthesis of KFAPs and keratins is coordinated. For example, the synthesis of the KFAP ‘profilaggrin’ starts shortly after the expression of K1 and K10 (Dale et al. 1985). This profilaggrin has a calcium-binding region in its N-terminal domain and is cleaved by the proteinase caspase-14 into several filaggrin molecules, which then bind to keratins and keratin filaments (Mack et al. 1993; Denecker et al. 2007). Caspase-14 is activated only in cornifying epithelia of terrestrial mammals but is not expressed in the nail corneocytes of humans (Denecker et al. 2008). The latter is not surprising, as Bragulla (1996) and Bragulla & Homberger (2007) have shown for the equine hoof that a Stratum granulosum with profilaggrin containing keratohyalingranula is absent in hard-cornifying stratified epithelia. The expression of caspase-14 may be activated by a transcription factor of the family of activator protein 1 (AP-1). During the embryonic development of the epidermis, the expression of caspase-14 coincides with the appearance of a Stratum corneum (Denecker et al. 2008). Hence, the observations so far indicate that profilaggrin/filaggrin (and caspase-14) are produced only in soft-keratinizing and -cornifying stratified epithelia (Bragulla & Budras, 1991). Filaggrin itself is a cationic protein (Mack et al. 1993). It is sandwiched between the rod domains of two keratin filaments with which it establishes ionic bonds (see ‘ionic zipper model’ of Mack et al. 1993).

Keratins and cell adhesive complexes

Keratin filaments are anchored to cell–cell adhesion complexes, such as desmosomes and hemidesmosomes, on the internal sides of cell membranes, thereby ensuring that keratinocytes remain connected to one another, transfer mechanical forces to neighboring cells and maintain the integrity of their cytoskeleton (Hanakawa et al. 2002). At the desmosomes, keratin filaments bind to the tail domain of desmosomal cadherin molecules, such as plakoglobin (Dusek et al. 2007), as well as plectin, periplakin, envoplakin and desmoplakin (Kazerounian et al. 2002), thereby anchoring the cytoskeleton to the cell membrane. Plakoglobin also acts as a signaling molecule within the ‘apoptotic’ signaling cascade during cornification, as cells that lacked plakoglobin are not able to undergo cornification (Dusek et al. 2007). The head domains of keratins bind to the C-terminal (i.e. the tail) of desmoplakins (Meng et al. 1998). For example, the head domains of the epidermal basic keratins K1, K2, K5 and K6 bind to the tail domain of desmoplakin 1, whereas the other basic keratins and all acidic keratins are not able to do so (Kouklis et al. 1994). The bonds of desmoplakin to keratins in the stratified-cornified epidermis are stronger than those in simple or non-cornified epithelia (Meng et al. 1998).

Keratins and the cornified envelope

During cornification, the keratinocyte cell membrane between the desmosomes is reinforced with special proteins (e.g. involucrin, loricrin, etc.) to form the subcytolemmal cornified cell envelope (see section ‘Keratins and keratin filament-associated proteins’). As a consequence, the keratin filaments can attach now not only to the desmosomes but also anywhere else on the cell wall of corneocytes. The keratin filaments do so via their head domains of basic keratins (Kirfel et al. 2003).

Programmed cell death of keratinocytes

The process of cornification involves the programmed death of the keratinocytes (Houben et al. 2007). This epidermal programmed cell death is different from the programmed cell death in apoptosis because the extra- and intracellular signaling cascade that is characteristic of apoptosis is not activated (Lippens et al. 2005; Denecker et al. 2007). The cornification of keratinocytes also differs from apoptosis because the terminally differentiated cells are dead but intact and not cell fragments as in apoptotic bodies (Weil et al. 1999). The only genuine apoptosis in the epidermis occurs in so-called sunburn cells, which are basal cells that are damaged by UVB rays (Denecker et al. 2007). Apoptosis can take place at any stage of cell differentiation but the process of cornification can start only after a cell has already gone through a certain differentiation. Hence, cornification is not a type of apoptosis (Denecker et al. 2008). However, apoptotic enzymes, such as caspase-3, are activated in cornifying cells in the transition zone between the granular layer and the layer of cornified cells (Weil et al. 1999). The enzyme caspase-3 may be involved in the dismantling of the cell nucleus and of the organelles (Weil et al. 1999). The apoptotic protease caspase-14, which is also activated in the cornifying keratinocytes of the epidermis (Demerjian et al. 2008), is not expressed in the keratinizing but not cornifying keratinocytes of the oral epithelium (Lippens et al. 2000). Caspase-14 is not activated in the cornifying cells of the nail matrix (Jäger et al. 2007), indicating differences in the processes of soft vs. hard keratinization and cornification.

Structure and location of keratin genes

In mammals, each of the keratins is encoded by a single gene (Krieg et al. 1985). Hence, the various keratins are the genuine products of gene transcription and translation and not of post-translational proteolysis of precursor molecules (Moll et al. 1982). However, as an exception to this general rule, in humans, bovines and murids, the isoforms of K6 are encoded by three genes (Rogers et al. 2005). The orthologue K6 isoforms have amino acid sequences that are more than 95% identical in these species (Wang et al. 2003). The mRNAs of KRT6a and KRT6h are translated into different isoforms (Rogers et al. 2005) as a result of either alternative splicing or of different post-translational modifications. K6c and K6d are the products of KRT6a, and the keratins K6e and K6f are the products of KRT6h (Rogers et al. 2005). The subfamily of genes encoding K6 isoforms has developed from genetic modifications that were independently acquired in various species (Navarro et al. 1995).

In tetrapods, the number of genes encoding acidic keratins is almost equal to that of genes encoding basic keratins, whereas in teleost fishes, the number of genes encoding the acidic keratins is three times the number of genes encoding the basic keratins (Schaffeld et al. 2007). The number of genes encoding keratins varies in mammals, e.g. it is higher in opossums than in humans (Zimek & Weber, 2006).

In humans, keratins are encoded in 54 genes and these genes are clustered on two chromosomes (Bowden, 2005; Schweizer et al. 2006). The acidic type I keratin genes are clustered on chromosome 17 and the basic type II keratin genes are clustered on chromosome 12 (Bowden et al. 1998; Hesse et al. 2001; Coulombe & Omary, 2002). The keratins K25–K28, which are specific for the cornifying cells of the inner root sheath of a hair follicle, are encoded by a cluster of genes on chromosome 17 (Bawden et al. 2001; Langbein et al. 2006). Similarly, the keratins K71–K74, specific for the inner root sheath of the human hair follicle, are encoded by genes arranged in a cluster on chromosome 12 (Langbein et al. 2003).

The gene encoding the acidic type I keratin K18 is an exception, as it is located on chromosome 12 next to the type II keratin genes, including the gene for K8, which is the partner of K18 (Waseem et al. 1990). This fact led to the hypothesis that KRT8 and KRT18 are close to ancestral keratin genes (Waseem et al. 1990). Additional support for this hypothesis comes from the fact that the sequences of these two genes are similar (Owens & Lane, 2003). K8 and K18 are found in all gnathostome vertebrates (Schaffeld et al. 2007). In humans, a phylogenetic analysis of the genes of the type I keratins (i.e. acidic soft, acidic hard, acidic inner root sheath and related keratins) suggested that the acidic keratin genes related to the inner root sheath originated earlier during evolution than the genes for epidermal keratin or hair keratin (Bawden et al. 2001). The criteria for establishing relationships among keratin genes are (i) a high sequence similarity (> 70%) of the region encoding the rod domain, (ii) the same orientation of the genes on the chromosome (Bowden, 2005) and (iii) the position of the keratin genes in, for example, the same cluster (Zimek & Weber, 2006).

The genes of the two types of keratins (i.e. the acidic and basic keratins) are also clustered on two chromosomes in other mammals, such as the rat, mouse (on chromosomes 11 and 15), chimpanzee, dog, cow and sheep (Hesse et al. 2004; Lu et al. 2006; Makar et al. 2007). The fact that keratin genes are clustered on only two chromosomes supports the hypothesis that tandem duplications have occurred during the evolution of keratin genes (Molloy et al. 1982; Coulombe & Omary, 2002). The genes encoding β-keratins of feathers are also arranged in clusters (Molloy et al. 1982). The nucleotide sequences of keratin-like genes have been identified in lower eukaryotes (e.g. yeast, species Saccharomyces) and these genes are encoding intermediate filament proteins (Fuchs & Marchuk, 1983; Fuchs et al. 1983).

Genes encoding proteins, which form intermediate filaments such as keratin filaments, consist of seven or eight exons (Bader et al. 1986). The basic type II keratin genes consist of seven exons, whereas the acidic type I keratin genes consist of eight exons (Makar et al. 2007), with the exception of the gene encoding the acidic keratin K19 with only five introns (Bader et al. 1986). Coulombe et al. (2004) explained the significance of the large number of keratin genes by the need for adaptable viscoelastic properties of epithelial cells through the production of different keratins forming keratin filaments of varying mechanical properties.

Mutations of keratin genes

Mutations of keratin genes may be responsible for the development of new keratins. For example, in the mouse, a segment deletion in the acidic hair keratin gene affects the amino acid sequence in the H2 subdomain in the tail domain of the keratins (Bertolino et al. 1990). Changes of the amino acid sequence of keratins due to mutations may have severe consequences for the assembly of keratin filaments (Steinert, 2001). Szeverenyi et al. (2008) compiled the information concerning the mutations of keratin genes in humans.

Duplications of keratin genes in combination with frame-shift mutations result in the formation of pseudogenes (Hesse et al. 2004). In humans, 87% of all keratin pseudogenes have similar sequences as KRT8 and KRT18 but they are scattered on various chromosomes (Hesse et al. 2001; Coulombe & Omary, 2002), possibly through crossing-overs or translocations. A total of 138 human pseudogenes are related to KRT8 or KRT18 (Hesse et al. 2004). According to Hesse et al. (2001), this observation supports the inference that KRT8 and KRT18 have a long history, i.e. may have been among the first keratin genes in ancestral vertebrates. In contrast, if these pseudogenes have similar sequences as the functional genes in humans, they may have a recent history, i.e. they may have been duplicated recently in the primate lineage. Pseudogenes related to other keratin genes, such as KRT14, KRT16 and KRT17, exist (Hesse et al. 2001).

Mutations of keratin genes are responsible for a number of disorders in the human and murine skin (Pekny & Lane, 2007). In the skin disorder ‘epidermis bullosa simplex’, which involves a breakdown of basal cells in stratified epithelia, either the KRT5 or KRT14 is defect. In case of a mutation of KRT14, the type I keratin K15 can compensate for K14 to combine with the type II keratin K5, the usual partner of K14; hence, the symptoms are less severe than when KRT5 is mutated, for which there is no compensating keratin (Steinert, 2001; Pekny & Lane, 2007). The same principle applies to K6 and its partners K16/17 (Coulombe et al. 2004). KRT1 and KRT10, characteristic of keratinizing suprabasal cells in the stratified epidermis, are mutated in the skin disorder epidermolytic hyperkeratosis (Pekny & Lane, 2007). The disease ‘white sponge naevus’, which affects the oral and genital stratified epithelia, is caused by mutations of KRT4 and KRT13 (Pekny & Lane, 2007).

We thank our editors Matthew Vickaryous and Jean-Yves Sire for inviting us to write this review and for their patience, suggestions and support during the preparation of our manuscript. Brooke A. Hopkins, Michelle L. Osborn and Margaret Jensen-Shoats proof-read the manuscript and provided suggestions and comments for its improvement. Sigrid N. Hamilton provided logistical support.