Introduction

Three filamentous systems make up the cytoskeleton of cells. These are microfilaments, microtubules and the intermediate filaments. Sixty-five homologous proteins with the ability to form 10 nm filaments define the intermediate filament family in humans.1,2 They likely arose by mutation of an ancestral nuclear lamin gene3 and are restricted in expression to metazoans.4 Individual IFs have diverged and specialized their N-terminal and C-terminal domains that flank the α-helical central region necessary for filamentous polymerization. The proteins are divided into at least five groups as shown in Table 1. The type I and type II keratins are the largest subgroups of the intermediate filament family. The keratins polymerize as obligate heteropolymers of individual type I and type II proteins. The keratins are further subdivided into epithelial and hair related keratins. Type III IF proteins are represented by vimentin, glial fibrillary acidic protein, desmin and peripherin. Type IV IF proteins include the three forms of neurofilament proteins, α-internexin, syncollin, nestin and synemin and the type V IF proteins represent the three nuclear lamins.

Table 1 Coding Intermediate Filament Genes

Keratins

Investigations of the functions of specific intermediate filament proteins and the identification of mutant IF proteins involved in human genetic diseases have demonstrated that many IF proteins function to provide mechanical strength to cells and tissues. This is clearly demonstrated in mammalian skin where mutations or the absence of specific epidermal keratins causes epidermal fragility resulting in blistering skin diseases5,6,7 and alterations in the mechanical properties of filaments.8 Similarly, muscle mechanical defects are evident in the absence of desmin.9 In addition, A2 and A3 IF proteins of Caenorhabditis elegans are necessary for locomotion.4 The keratins associated with hair are uniquely specialized for stability reinforced by extensive extracellular disulfide bridging. The specialized keratins of the outer epidermis (K1, K2, K9, K10) are distinguished by tail domains greatly enriched in glycine and by the lack of caspase cleavage sites within the type I extracellular keratins (K9, K10). These keratins contribute to the structural integrity essential for the barrier function of the skin. Intracellular epithelial keratins such as K5 and K14 crossbridge desmosomal junctions and thus provide structural continuity with the edges of epithelial cells. Individual keratins bind to the desmoplakin component of desmosomes with varying affinity and structural requirements.10,11 Additional adapter proteins also contribute to the association of IF with desmosomes.12 Keratins and E-cadherin are two defining characteristics of epithelial cells. Many simple epithelial cells are particularly sensitive to apoptosis when detached from extracellular matrix and neighboring cells in a process termed anoikis.13,14 A possible role of keratins in anoikis remains to be investigated. Multiple excellent reviews concerning the structure, dynamics and function of intermediate filaments are available.2,5,15,16,17,18,19,20 Further discussion here will focus on the possible roles of keratins in apoptosis.

K8 and K18

K8 and K18 are expressed in simple, or predominately single layered, internal epithelia. Several characteristics of K8 and K18 distinguish them from other members of the type I and II keratins. First, they are the first IF proteins to be expressed during mammalian development being associated with the differentiation to the trophectodermal layer of the blastocyst. However, the human epiblast, unlike the mouse inner cell mass, is an epithelium and expresses K8 and K18. In addition to the unique genes coding for human K8 and K18, each is represented by 35 and 62 processed pseudogenes respectively. These are distributed over many chromosomes.1,21 The large number of processed pseudogenes for K8 and K18 may reflect their expression in the human epiblast which provides a germ line target for retroviral reverse transcription.16 The smaller number of K8 and K18 pseudogenes found in mice21,22 is consistent with the difference in expression of these genes during early development. K18 is also the only type I keratin found within the type II keratin multigene locus found on human chromosome 12. The other type I keratins are found on chromosome 17. The genes for K8 and K18 are adjacent to each other and at the distal end of the Krt1 locus.

Another distinguishing characteristic of K8 and K18 is their common and persistent expression in carcinomas. Other keratins that are expressed in the tissue of origin of carcinomas tend to be more sporadically expressed or lost during the progression of the tumor. K8, K18 and K19 have proved to be useful carcinoma markers.

While keratins are promiscuous in their ability to heterpolymerize from different type I and II proteins in vitro,23,24 preferential assembly of certain keratin pairs in vivo has been revealed by gene knockout experiments. Thus the removal of K18 results in the loss of K7 as well as K8 protein in intestine.25 In the absence of a complementary partner K8 and K18 are normally degraded rapidly, although K8 is degraded less quickly than K18 in some cells.26 In K18 null animals, aggregates of precipitated K8 residue, known as Mallory bodies, are found in the livers of old animals. Thus the degradation of K8 may not necessarily be complete, at least under some pathological conditions.25

Gene targeting experiments have demonstrated that K8 is important for development because K8 deficient embryos die at E12.5.27 However, the viability of K8 deficient mice is dependent on the genetic background.28 In an FVB/N genetic background approximately 50% of the expected number of K8 deficient embryos are born. By contrast, the knockout of K18 or K19 does not alter viability.25,29 However, K18, K19 compound homozygote null embryos die at about 9 days of development.30 Thus K18 deficiency is compensated by K19 and vice versa (Table 2). K8 deficiency is also compensated to some degree by K19 because K8, K19 compound homozygote null embryos die earlier than K8 null embryos.29 The protective effect of K19 is likely due to low levels of K7, expressed in late trophoblast derivatives. This interpretation of K8, K18 and K19 knockouts is summarized in Table 2. However, the absence of all keratins (and very likely all IF) in K18, K19 compound homozygotes still permits keratin deficient blastocysts to implant, invade the uterine environment and develop for 9 days. If K8 and K18 primarily provide mechanical strength, it is perhaps surprising that development can proceed during trophectoderm expansion, implantation, and invasion. Furthermore, no abnormalities of parietal or visceral yolk sac development were detected during the differentiation of K8 deficient ES cells in vitro.31 Thus K8 is not necessary for the appearance or function of polarized epithelial yolk sac cells or the initial trophoblast derivatives. Other cytoskeletal elements may provide sufficient strength to implanting embryos until day 9 of development. However, the postimplantation failure of trophoblast derivatives may reflect additional functions for simple epithelial keratins.

Table 2 Keratin complementation in placenta

K8 and K18 associated disease

The incomplete penetrance of embryonic lethality of K8 null mice has permitted the evaluation of the importance of K8 in adult mice. Adult K8 null mice in the semipermissive FVB/N background develop a mild hepatitis and a more dramatic inflammatory bowel disease by about 6 months of age.28 This condition is similar to that seen in mice without the T cell receptor α32 or the IL-233 and IL-1034 cytokines. In addition, K8 null adults were found to be particularly sensitive to pentobarbital anesthesia, partial hepatectomy35 and to liver toxicity induced by griseofulvin, 3,5-diethoxycarbonyl-1, 4-dihydrocollidine36 acetaminophen37 or the phosphatase inhibitor microcystin-LR.38 In the case of griseofulvin treatment, which models the formation of Mallory bodies in alcoholic liver disease, the use of K8 null animals revealed that Mallory body formation was dependent on K8. However, K8 null animals developed liver disease even in the absence of Mallory bodies.36 Thus these dense aggregates of cross-linked and modified K8/K18 are not the causative agent of subsequent liver damage. An important role for K8/K18 in resistance to liver toxicity is reinforced by the increased sensitivity of transgenic mice which overexpress mutant forms of K18 to griseofulvin, acetaminophen and microcystin-LR.37,39,40 Recently, mutations of K8 in humans have been found to be associated with cryptogenic liver disease41 and inflammatory bowel disease (Birgit Lane, personal communication). In the case of crytogenic liver disease, two mutations of K8 have been identified. One changes a tyrosine at position 53 to a histidine (Y53H). The second changes a conserved glycine at position 61 to a cysteine (G61C). One consequence of the G61C is disulfide bridging of K8, apparently to itself. Both mutations appear to alter the stability of K8/K18 filaments when cells are stressed but not under normal growth conditions. Together, the mouse and human diseases associated with K8 mutations indicate an important role in resistance to liver toxicity and inflammatory bowel disease. The mechanism of the protective effects of K8/K18 remains to be determined.

K8 and TNF sensitivity

A recent observation has implicated K8 and K18 in resistance to apoptosis induced by TNF family receptors. Several epithelial cell lines deficient in K8/K18 IF are more sensitive to apoptosis induced by the combination of TNF and cycloheximide than their control counterparts.42 This resistance to TNF was correlated with the ability of K8 but not K18 to bind to the cytoplasmic domain of TNFr2 and the lymphotoxin-β receptor (LTβR) but not TNFr1 or Fas. Co-localization of TNFr2 and K8/K18 was detected at the periphery of cells. In addition, treatment of either K8 or K18 null animals with concanavalin A (ConA) showed that the hepatocytes of both types of keratin deficient animals were more sensitive to lymphocyte initiated apoptosis. K8 and K18 are the only IF proteins expressed in hepatocytes. Thus gene targeting of either gene results in the absence of hepatocyte IF. ConA induced hepatitis acts through circulating blood elements and involves both Fas and TNF.43,44,45 Increased levels of both TNF and Fas ligand are found in the livers of mice treated with ConA. The similar response of both K8 and K18 null animals to ConA was of particular interest because K18 null animals are phenotypically normal and do not develop the spontaneous mild hepatitis found in K8 null animals. The colonic hyperplasia of the K8 null animals may also reflect the involvement of the immune system. It is interesting that trophoblast derivatives are normally resistant to TNF induced apoptosis.46,47 It is possible that TNF sensitivity may be involved in the trophoblastic defects found in null K8 and K8, K19 and K18, K19 compound null animals.

K8 and Fas

Very recently, a similar role for K8 in protection from Fas mediated apoptosis of hepatocytes has been discovered.48 Explanted hepatocytes from K8 deficient mice are particularly sensitive to activating Fas antibody but not TNF or TRAIL. Furthermore, the increased sensitivity to Fas activating antibody was correlated with an increase in the fraction of cellular Fas on the cell surface. Total Fas levels in K8 null and wild-type hepatocytes appear similar. The increased sensitivity of K8 null animals to activating Fas antibody was found only at a moderate dose of antibody. Microtubule dissociating drugs helped protect hepatocytes from subsequent Fas antibody challenge. K8/K18 may be involved in the trafficking of Fas from the Golgi to the apical surface of polarized epithelia.

Increased Fas on the cell surface of K8 null hepatocytes may predispose them to apoptosis because over expression of death receptors can trigger the apoptotic cascade even in the absence of ligand. However, TNF is key for liver growth and regeneration as well as an apoptotic stimulus.49 Perhaps K8 hepatocytes are presensitized to Fas stimulation by previous exposure to TNF. It will be interesting to determine if K18 null mice also express elevated levels of hepatocyte surface Fas and to analyze the effects of TNF and Fas deficiencies on K8 null embryonic lethality.

The association of K8/K18 with death receptor family function may be related to earlier implications of K8/K18 with drug resistance in several cell lines.50,51 Some K8/K18 deficient cell lines are more sensitive to DNA damaging agents such as mitoxantrone, doxorubicin, melphalan, bleomycin, and mitomycin C than the controls that were engineered to express keratin. These drugs have been shown to induce surface Fas expression through a p53 dependent process in multiple cancer cells.52 Increased Fas expression sensitizes cells to FasL stimulated killing. Chemotherapeutic drug treatment sensitizes multiple cancer cell lines to either Fas or Trail killing.53 While several observations indicate that killing of cancer cells by DNA damaging agents is not exclusively dependent on Fas,54 Fas or other death receptors may contribute to the cellular damage. The role of keratin IFs in resistance to some drug induced apoptosis, may be linked to death receptors.

K8 and K18 have now been convincingly connected to resistance to multiple kinds of stress.15,19 This evidence includes the association of HSP70 with K18;55 the increased sensitivity of K8 null mice and K18 dominant negative transgenic mice to the toxicity of acetaminophen,37 microcystin-LR38 and griseofulvin36 and the increased sensitivity of K8 null mice to pentobarbital and liver perfusion.35 While the correlation of various forms of stress with K8/K18 filament phosphorylation is well established, the mechanism by which these keratins confer resistance or protection is less obvious. One potential common thread to many types of stress is TNF family involvement.

However, one potential contradiction to the function of K8/K18 in moderating TNF family ligand induced pathology is the surprising lack of protective effect of either K8 or K18 on experimentally induced pancreatitis.56 TNF is reported to contribute to the severity of the experimental disease.57 However, TNF may play a secondary role in the pancreatitis models and a more direct role in several liver toxicity models.

How might K8/K18 moderate TNF and Fas induced apoptosis?

Both TNF and FasL normally induce apoptotic signaling by the engagement of the cognate receptors. However, increased surface expression of receptor can also initiate signaling. The increased sensitivity of K8 null hepatocytes to activating Fas antibody has been suggested to be a reflection of the increased proportion of Fas expressed at the cell surface. This would imply that K8/K18 filaments might specifically moderate the transport of Fas to the cell surface. This suggestion is of particular interest in light of the recent identification of abnormalities in the distribution of apical surface markers in both the small intestine and liver of K8 null animals.58 In the small intestine of K8 null animals, regional specific differences in the expression of syntaxin-3, intestinal alkaline phosphatase and CFTR chloride channel proteins were found. Given the rapid turnover of intestinal cells, these regional differences on the villi may reflect alterations in the turnover of the proteins. Syntaxin-3 is a key element of the SNARE machinery involved in specific vesicular transport. Interestingly, sytaxin-3 is co-localized with vimentin filaments in fibroblasts and Hela cells59 but is not found associated with the K18 filaments of Hela cells. The mechanism by which keratin filaments might influence vesicular trafficking of specific surface molecules remains to be elucidated. However, the recently appreciated dynamic properties of IF including rapid intracellular movement60 facilitates consideration of many possibilities.

Interference with the transport of TNFr1 from the Golgi to the cell surface has been shown to be a mechanism of inhibition of TNF induced apoptosis in poliovirus infected cells.61 This function would place K8/K18 upstream of TNFr or Fas receptor signaling. Alternatively, K8 may play a role downstream of TNFr signaling by moderating the signaling from receptors. Indirect evidence of IF involvement in signaling includes the increased NF-κB and Jnk activation found in keratin deficient endodermal cells in response to TNF,42 the binding of 14-3-3 protein to K18 in a phosphorylation dependent manner,62 and MAP kinase phosphorylation of IF proteins.63 K8 is capable of binding the cytoplasmic domain of TNFr2. However, signaling through TNFr2 is significantly different from TNFr1, as TNFr2 has no death effector domain. One example of an indirect mechanism of influencing TNF signaling is the induction of TNF on the cell surface by TNFr2 signaling which then contributes to apoptosis.64

Very recently Inagaki and colleagues65 described an alternative and likely key mechanism by which keratin attenuates TNF induced death. Residues 77–128 of K18 and K14 both mediate an association with the C-terminal portion of TNFR1-associated death domain protein (TRADD). This association with a key adapter molecule was demonstrated biochemically by co-assembly, co-immunoprecipitation analysis, and by immunofluorescent localization of both endogenous and exogenous TRADD and K18. Importantly, this investigation provides evidence for the competition of ligand activated TNFR1 and keratin for TRADD that dissociates from keratin and associates with ligand bound TNFR1 to form a signaling complex that activates caspase 8. This report provides the best present explanation of the protective effect of K8/K18. The very interesting control of the association of TRADD with keratins will likely be an important focus of future investigation.

Intermediate filament proteins are cleaved by caspases during apoptosis

Nearly 30 years ago, Bjorklund developed an antibody test for tissue polypeptide antigen (TPA) that was found in the serum of cancer patients. The nature of this circulating antigen remained obscure in spite of significant use of the tumor marker, until 1984, when the antigen(s) was identified as a complex of keratin 8, 18 and 19 fragments.66 This discovery helped explain the correlation of TPA with carcinoma status because most carcinoma continued to express epithelial keratin intermediate filament proteins K8, K18 and K19.67 However, the origin of soluble fragments of insoluble keratin was still a puzzle. In 1997, Caulin et al. demonstrated that K18 but not K8 was a substrate for caspase digestion during the course of epithelial cell apoptosis.68 Caspases 3, 6 and 7 were all capable of cleaving after the aspartate 238 (D238) and mutation of the DEVD238 sequence rendered this site resistant to cleavage both in vitro and in vivo.68,69 However, the K18 internal cleavage site is utilized in apoptotic MCF-7 cells which are deficient in caspase 3.70 Thus caspase 3 is not essential for K18 cleavage at D238. A second cleavage site nearer the C-terminal end was deduced on the basis of a ten residue, monoclonal antibody epitope terminating with a potential caspase cleavage site, DALD,397 that rendered antibody binding dependent on apoptotic cleavage of K18.71 The identification of D397 as the second caspase cleavage site of K18 was confirmed by mutagenesis.72

The first intermediate filament protein to be identified as a caspase substrate was the nuclear lamin A.73,74 The internal cleavage site of lamin A is identical to that found in K18. This cleavage site of K18 and other IF proteins is found within the L1-2 linker region between two conserved, central, alpha helical domains (Figure 1).75 None of the type II keratins have similar potential cleavage sites. Simple alignment of other IF proteins lead to the prediction that other type I keratins and vimentin, desmin and NF-M would likely be caspase substrates. Caspase cleavage at the L1-2 linker region has been confirmed for K14, K17, K19, and vimentin.72,76 The L1-2 caspase cleavage site of multiple type I keratins is of particular interest because mutations of the caspase recognition site have been found in K14 of patients with epidermolysis bullosa simplex, a genetic blistering skin disease. However, a careful analysis of four spontaneous mutations within the context of K14 revealed that only the mutation of the required aspartate (VEVD or VEMD) prevented cleavage of K14 during apoptosis.72 Confirmation of the cleavage of the wt sequence VEMD/A of K14 suggests that other type I keratins, K12–17, desmin and neurofilament-L may also be caspase substrates.

Figure 1
figure 1

Conservation of caspase cleavage site of type I keratins and other IF protein. A schematic diagram of the domain structure of IF proteins is shown at the top. The L1-2 linker region connects the coil 1B and coil 2A subdomains. Note the conservation of the VEVD or VEMD caspase recognition sites except in K9 and K10 which function as extracellular barrier proteins in epidermis and the absence of similar sites in type II keratins (reprinted with permission from the Journal of Cell Biology)68

Consequences of caspase digestion of IF during apoptosis

The orderly packaging and disposal of cellular contents during cell death may be important to limit inflammatory reaction. The cleavage of IF during apoptosis may reflect the programmed disposal of relatively insoluble structural proteins. The engagement of Fas with FasL, or activating antibody results in the formation of a signaling complex involving proteins with death effector domains such as FADD and the subsequent activation of caspase 8. Caspase 8 translocates from either a membrane or mitochondrial location to plectin, an abundant cytoskeletal scaffolding protein which binds all three cytoskeletal fibers, actin, intermediate filaments and microtubules. The subsequent cleavage of plectin by caspase 8 results in a reorganization of the microfilament system.70 The early cleavage of plectin appears not to affect intermediate filament organization. Soon after an apoptotic challenge K8/K18 filaments are hyperphosphorylated and collapse to punctate inclusions of variable size (Figure 2). Similar K18 aggregations have been observed in some epithelial cells not undergoing apoptosis77 and may reflect the consequences of keratin phosphorylation.63 During TRAIL induced apoptosis, activated caspase 3 co-localizes with similar, though larger aggregates late in the process of death.78 This organization into dense aggregates without a limiting membrane appears to contain most of the activated caspase 3 and the K18 that has been cleaved at position D397. The status of the VEVD238 cleavage site in such aggregates is not known. If the association of active caspase 3 with these aggregates is strictly dependent on the presence of K18, it is possible that keratins may provide some protection against caspase 3 digestion by limiting the availability of active caspase 3.

Figure 2
figure 2

K18 reorganization during epithelial cell apoptosis. SNG-M human adenocarcinoma cells were treated with etoposide, then fixed and stained with K18 monoclonal antibody and propidium iodide for nuclear morphology. (A) K18. (B) DNA fluorescence. Arrows point to apoptotic cells. Note the reorganization of keratin filaments into granules and the apoptotic nuclear morphology. Similar granular structures also react preferentially with K18 phospho-serine 53 specific antibody68

The hyperphosphorylation of K18 in cells treated with okadaic acid, a phosphatase inhibitor, was found to render K18 resistant to caspase digestion at D238 but not D397.72 This is an interesting result in light of the rapid phosphorylation of K18 on serine 53 after apoptosis initiation.68,69,79 The reorganization of K18 filaments in apoptotic cells is associated with the cleavage of K18 at D397, K18 phosphorylation at S53, and the association of active caspase 3 with the K18 granules. However, it is not clear whether the reorganization of keratin filaments is caused by its phosphorylation alone, its cleavage or even an association with other proteins within the cells. Dissection of the roles of K18 phosphorylation, its association with caspase 3 and caspase cleavage of K18 may reveal new ways to regulate caspase activity.

While the initial observations linking intermediate filaments and apoptosis involved lamins and keratins, other IF proteins may also function during the process. For example, vimentin is a caspase substrate.76,80 Furthermore, the forced expression of the N-terminal caspase digestion product of vimentin as a GFP fusion protein has been reported to induce caspase dependent apoptosis.76 This effect is suggested to amplify the death signal. However, the effect has only been reported with the GFP fusion protein and the natural N-terminal caspase product of vimentin has not been detected in apoptotic cells suggesting that it may be further degraded rapidly. Additional investigation should reveal whether the subtle phenotypes of the vimentin knockout81,82 may be supplemented by a functional role in apoptosis.

In summary, simple epithelial keratins have been implicated in resistance to drug toxicity, hepatic stress and inflammatory bowel disease. Recent connections of death receptor involvement with epithelial apoptosis in both cell lines and mouse models should stimulate further advances in understanding the possible functions of programmed cell death.