Keratins let liver live: Mutations predispose to liver disease and crosslinking generates Mallory-Denk bodies


  • Nam-On Ku,

    Corresponding author
    1. Department of Medicine, Palo Alto VA Medical Center and Stanford University Digestive Disease Center, Palo Alto, CA
    • Nam-On Ku, Palo Alto VA Medical Center, 3801 Miranda Avenue, Mail code 154J, Palo Alto, CA 94304===

      M. Bishr Omary, Palo Alto VA Medical Center, 3801 Miranda Avenue, Mail code 154J, Palo Alto, CA 94304===

    Search for more papers by this author
  • Pavel Strnad,

    1. Department of Medicine, Palo Alto VA Medical Center and Stanford University Digestive Disease Center, Palo Alto, CA
    Current affiliation:
    1. Department of Internal Medicine I, University of Ulm, Ulm, Germany
    Search for more papers by this author
  • Bi-Hui Zhong,

    1. The First Affiliated Hospital of Sun Yat-sen University, Guangzhou, PR China
    Search for more papers by this author
  • Guo-Zhong Tao,

    1. Department of Medicine, Palo Alto VA Medical Center and Stanford University Digestive Disease Center, Palo Alto, CA
    Search for more papers by this author
  • M. Bishr Omary

    Corresponding author
    1. Department of Medicine, Palo Alto VA Medical Center and Stanford University Digestive Disease Center, Palo Alto, CA
    • Nam-On Ku, Palo Alto VA Medical Center, 3801 Miranda Avenue, Mail code 154J, Palo Alto, CA 94304===

      M. Bishr Omary, Palo Alto VA Medical Center, 3801 Miranda Avenue, Mail code 154J, Palo Alto, CA 94304===

    Search for more papers by this author

  • Potential conflict of interest: Nothing to report.


Keratin polypeptides 8 and 18 (K8/K18) are the cytoskeletal intermediate filament proteins of hepatocytes while K8/K18/K19 are the keratins of hepatobiliary ductal cells. Hepatocyte K8/K18 are highly abundant and behave as stress proteins with injury-inducible expression. Human association studies show that K8/K18 germline heterozygous mutations predispose to end-stage liver disease of multiple etiologies (≈3 fold increased risk), and to liver disease progression in patients with chronic hepatitis C infection. These findings are supported by extensive transgenic mouse and ex vivo primary hepatocyte culture studies showing that K8 or K18 mutations predispose the liver to acute or subacute injury and promote apoptosis and fibrosis. Mutation-associated predisposition to liver injury is likely related to mechanical and nonmechanical keratin functions including maintenance of cell integrity, protection from apoptosis and oxidative injury, serving as a phosphate sponge, regulation of mitochondrial organization/function and protein targeting. These functions are altered by mutation-induced changes in keratin phosphorylation, solubility and filament organization/reorganization. Keratins are also the major constituents of Mallory-Denk bodies (MDBs). A toxin-induced K8>K18 ratio, and keratin crosslinking by transglutaminase-2 play essential roles in MDB formation. Furthermore, intracellular or cell-released K18 fragments, generated by caspase-mediated proteolysis during apoptosis serve as markers of liver injury. Therefore, K8 and K18 are cytoprotective stress proteins that play a central role in guarding hepatocytes from apoptosis. Keratin involvement in liver disease is multi-faceted and includes modulating disease progression upon mutation, formation of MDBs in response to unique forms of injury, and serving as markers of epithelial cell death. (HEPATOLOGY 2007;46:1639–1649.)

Intermediate filaments (IFs), microfilaments and microtubules are the three major cytoskeletal protein families in digestive and non-digestive organs.1 IFs are cell type-specific as exemplified by keratins in epithelial cells, vimentin in endothelial and other mesenchymal cells and desmin in myogenic cells.2 Keratins are the largest subfamily of cytoplasmic IFs and include type-I keratin polypeptides (K9-K20) and type-II keratins 1-8 (K1-K8). Epithelial cells express at least one type-I and one type-II unique keratins that associate as obligate noncovalent heteropolymers. For example, keratinocytes express K1/K5/K10/K14, while glandular single-layered epithelia co-express K8 and K18 with or without K7/K19/K20 depending on the cell type.3 Adult hepatocytes are unique in that they express K8 and K18 proteins (by KRT8/KRT18 genes) exclusively while hepatobiliary ductal cells express K7/K8/K18/K19/K20.3 Keratins (and all other IFs) share a common structure consisting of a conserved α-helical “rod” domain that is flanked by non-α-helical N-terminal “head” and C-terminal “tail” domains (Fig. 1A).

Figure 1.

Schematic of K8 and K18 protein domains, and immune staining of keratin apoptotic fragments and Mallory-Denk bodies in liver disease. (A) All IFs have characteristic tripartite structural motifs that consist of a central α-helical “rod” domain which is flanked by non-α-helical N-terminal “head” and C-terminal “tail” domains. The rod subdomains 1A, 1B and 2 are separated by non-α-helical linker L1 and L12 short regions. In some cases, domain assignments also include a linker (L2) that subdivides domain 2 into 2A and 2B.52 K8 and K18 associate as obligate non-covalent heterodimers after which dimers form antiparallel tetramers that extend laterally to yield the ≈10 nm diameter filaments. The black bars at the beginning and end of the rod domain represent highly conserved regions of IFs, and are unique in that they include mutation hot spots for epidermal keratins and other IF proteins. The head and tail domains, as compared with the rod, are the most variable among IFs and contain most of the posttranslational modifications [phosphorylation (phos) and glycosylation (gly)]. The L12 region of type-I (but not type-II) keratins and other IFs contains a conserved caspase-cleavage site (VEVD in K18, K19 and other IFs), while the tail domain of K18 contains a unique DALD caspase-cleavage site that is not found in other IFs. (B) Double immunofluorescence staining depicts localization of mouse K8 and apoptotic K18 fragment that is digested at Asp237. Livers were isolated from control FVB/n mice (a), or from Fas antibody (Jo-2)-injected mice (0.15 μg of antibody per gram mouse weight) (b). Livers were harvested 4.5 h after Fas antibody intraperitoneal injection. Arrows point to apoptotic cells, which appear yellow-orange due to co-localization of K8 with the K18 Asp237 fragment. In these cells, keratins exhibit either a filamentous pattern or a disrupted network with keratin-positive aggregates. Note that the control liver shows only keratin (green) but not the keratin apoptotic fragment (red) staining. (C) Triple immunofluorescence staining (using antibodies to the indicated antigens) of control human liver (a) and a liver from a patient with alcoholic cirrhosis (b). Note the presence of MDBs in image “b” (arrows) that co-stain with antibodies to K8 (green), ubiquitin (blue) and p62 (red). Only keratin (green) staining is visible in the control since staining to ubiquitin and p62 are difficult to appreciate because of their diffuse cytoplasmic distribution. Scale bar = 25 microns (B,C).

K8/K18 are highly abundant proteins in post-natal liver (e.g., they make up ≈0.3% of total mouse liver protein), while K19 is stoichiometrically relatively minor.4 Keratins are typically organized in filamentous arrays that extend from the nucleus to the plasma membrane (Fig. 1B,C). Despite their abundance, K8/K18 mRNA and protein levels increase ≈3-fold in response to liver injury as noted in mice exposed to agents that induce Mallory-Denk body (MDB) formation [previously called Mallory bodies]5 (Fig. 1C). In addition, liver K8/K18 protein but not mRNA increases 2-4 fold in patients with primary biliary cirrhosis.6 The regulation of keratins occurs via posttranslational modifications such as phosphorylation that involve the head/tail domains.7 Another type of keratin posttranslational modification is caspase-mediated proteolysis,8 whereby type-I (eg, K18/K19) but not type-II keratins are cleaved during apoptosis (Fig. 1A,B).

The importance of keratins in liver disease was last highlighted by HEPATOLOGY in 20023 and by others.9 More recent work provided support that keratin variants pose a risk for human liver disease progression, and our understanding of how this occurs is growing. We summarize herein the experimental evidence that implicates keratins in human liver disease and highlight recent mechanistic and functional insights on how keratins predispose to injury. We also cover other aspects of liver disease that involve keratins, including their essential role in MDB formation and how keratins and their phosphorylation and apoptotic fragments serve as markers of liver injury.

Animal Studies Implicate K8/K18 in Liver Disease Predisposition.

Animal models provided powerful tools for studying keratin function and linking keratin mutations with human disease. The initial landmark experiment involved overexpression of mutant K14 in mice which led to the identification of heterozygous K14 mutations as the cause of human epidermolysis bullosa simplex (EBS).10 To begin addressing K8/K18 function and potential disease relevance, transgenic mice were generated that expressed a K18 mutation at a highly conserved residue (Arg89)11 that is mutated (Arg-to-Cys/His) in ≈40% of epidermal-keratin-disease patients.2 The K18 Arg89Cys mice had remarkably fragile hepatocytes when isolated by liver perfusion and a mild chronic hepatitis but exquisite predisposition to apoptosis and various forms of liver injury (Table 1). The K18 Arg89Cys mutation results in keratin filament disruption that leads to impaired hepatocyte mechanical integrity, which is reminiscent to the homologous K14 Arg125Cys mutation that causes blister formation in patients with EBS. The importance of a hepatocyte intact keratin cytoskeleton was supported by ectopic hepatocyte expression of K14, which disrupted keratin filaments in association with inflammation and ballooning (Table 1). These studies were supported by findings in K8-null mice, which are highly susceptible to various forms of liver injury, and generated the hypothesis that K8/K18 variants cause or predispose to human liver disease.

Table 1. Summary of Keratin-related Animal Models with a Liver Phenotype
Genotype (Target)Keratin FilamentsLiver Phenotype (Basal Conditions)Hepatocyte FragilityPredisposition to Liver InjuryTested Injury ModelsStudy
  1. Several K8/K18-related mouse models have been generated that involve K8 or K18 ablation or expression of mutated K8 or K18. The only mouse model of a natural K8/K18 variant is the K8 G61C transgenic mouse. Hepatocytes of the mouse models manifest normal, absent or disrupted cytoplasmic keratin filaments depending on the genotype. Hepatocyte fragility is based on viability after liver perfusion. MLR, microcystin-LR; PH, partial hepatectomy; DDC, 3,5-diethoxycarbonyl-1,4-dihydrocollidine; TAA, thioacetamide.

K8-/- (C57B1/6)absentliver hemorrhage with embryolethalitynot applicablenot applicablenoneBaribault et al.53
K8-/- (FVB/N)absentmild hepatitis↑, resistant to MDB formationPentobarbital, MLR, high fat dietOmary et al.,3 Zatloukal et al.,5 Baribault et al.,54 Loranger et al.,55 Tao et al.56
K8 overexpressionnormalspontaneous MDBnormalpredisposition to MDB formationDDCZatloukal et al.5
K8 G61C (natural variant)normalnormalnormalFas, MLRKu et al.20
K8 S73A (phospho-site)normalnormalnormalFasKu et al.20
K18 -/-absentmild hepatitis, MDBsnot testednoneKu et al.,20 Magin et al.57
K18 overexpressionnormalnormalnormalresistant to MDB formationDDCHarada et al.33
K18 R89C (comparable site mutated in epidermal keratins)disruptedmild hepatitis↑, also to fibrosisGriseofulvin, acetaminophen, MLR, Fas, TAAOmary et al.,3, 7 Ku et al.,11 Strnad et al. (submitted)
K18 S33A (phospho-site for 14-3-3 binding)normalnormalnormalabnormal mitotic bodies after PHPHOmary et al.7
K18 S52A (phospho-site)normalnormalnormalMLROmary et al.7
Ectopic hepatocyte K14disruptedmild hepatitis, ↓ bile acid secretionnot testednot testednoneAlbers et al.58

CI, confidence interval; CYFRA 21-1, CYtokeratin FRAgment 21-1; EBS, epidermolysis bullosa simplex; hsp, heat shock proteins; HCV, hepatitis C virus; IF, intermediate filament; IHBs, intracytoplasmic hyaline bodies; K, keratin; MDBs, Mallory-Denk bodies; SAPK, stress-activated protein kinase; TPS, tissue-polypeptide-specific antigen.

KRT8 and KRT18 Are Susceptibility Genes for Human Liver Disease and Its Progression.

The liver-selective phenotype of K18 R89C mice and other genetic models, coupled with the identification of skin-specific keratin mutations in patients with EBS, led to the search and identification of K8/K18 mutations in patients with a broad range of liver diseases. Specifically, K8/K18 germline heterozygous variants were found to strongly associate with end-stage liver disease [P < 0.0001, odds ratio = 3.7 (95% confidence interval [CI], 2.0–6.8)]12, 13 and liver disease progression14 (Table 2). Notably, the risk of liver disease association with K8/K18 variants is higher than the risk of acute myocardial infarction in smokers (10-19 cigarettes/day) versus nonsmokers [P < 0.0001, odds ratio = 2.59 (95% CI, 2.4–2.9)].15

Table 2. Keratin Variants in Liver Disease
 Patient Cohort# of Mutation Carriers/Total Screened (%)P ValueOR* (95% CI)Study
Liver DiseaseControls
  • All the listed patient cohorts included patients with a broad range of liver diseases.

  • The only exceptions were:

  • 1

    included only patients with hepatitis C virus infection,

  • 2

    included patients with hemochromatosis (homozygous C282Y);

  • *

    When a cell contained 0 patients, 0.5 was added to each cell to calculate an odds ratio (OR);

  • **

    Combined German and Austrian disease cohorts;

  • For fibrosis progression per stage of fibrosis (e.g., stage 1 to 2; stage 2 to 3).

  • $

    Since there is no reported German control group studied for this variant, we used a control group from England.64

Screened domains/variants      
 All K8 and K18 exonsUS58/467 (12.4%)13/349 (3.7%)<0.00013.7 (2.0–6.8)Ku et al.12, 13
 Germany119/329 (5.8%)-0.0012.05 (1.33–3.15)Strnad et al.14
 K8 Y53H and G61CGermany29/1668 (1.7%)13/679 (1.9%)>0.70.9 (0.5–1.8)Halangk et al.59
 Austria3/107 (2.8%)1/45 (2.2%)>0.91.3 (0.1–12.5)Schoniger-Hekele et al.60
 K8 R340HUS29/119 (7.6%)0/116 (0%)<0.00418.9 (1.1–329.9)Lee et al.61
 K8 exon 1 and 6Germany11/151 (7.3%)---Strnad et al.62
Specific variants      
 K8 R340HUS39/586 (6.7%)25/808 (3.1%)<0.032.2 (1.3–3.7)Ku et al.,13 Lee et al.,61 Strnad et al.63
 Germany17/480 (3.5%)0/97 (0%)$>0.097.1 (0.4–119.0)Strnad et al.,14, 62 Owens et al.64
 K8 G61CUS6/467 (1.3%)9/963 (0.9%)>0.51.4 (0.5–3.9)Ku et al.,12 Tao et al.,63 Schneider et al.65
 Europe**37/2255 (1.6%)88/6061 (1.5%)>0.51.1 (0.8–1.7)Strnad et al.,14 Treiber et al.,16 Halangk et al.,59, Schoniger-Hekele et al.,60 Strnad et al.,62 Owens et al.,64 Buning et al.,66 Cavestro et al.67

The frequency of K8/K18 variants in liver disease does not have gender differences, but does for some variants associate with unique ethnic or population groups. In American and European populations, the most common K8/K18 variant is K8 R340H followed by K8 G61C (Fig. 2, Table 2). For example, K8 R340H is found in 6.7% of U.S. patients with liver disease versus 3.1% of controls (P < 0.03) and in 3.5% of German patients while absent in another small control group (P < 0.09) (Table 2). The physiologic consequences of K8 R340H are unknown but protein structure analysis predicts that the mutation has a potential destabilization effect13 and animal models of K8 R340 mutation should be forthcoming (unpublished observations). The prevalence of the second most common keratin variant, K8 G61C, is not significantly different in patients with liver disease and controls (Table 2). Interestingly, the frequency of K8 G61C in a European control group was variable and ranges from 0.8% (4/486) in the Czech Republic to 1.5% (23/1532) in Germany to 3.5% (7/200) in England.16 K8 G61C is completely silent in mice (and humans) unless challenged with hepatotoxic stress (Table 1). Hence, K8/K18 mutations can be considered “sleeper” biologically-silent variants unless they meet an appropriate stress stimulus.

Figure 2.

Distribution of keratin variants within the K8 and K18 protein backbone. The amino acid positions of the K8 and K18 subdomains (1A, 1B, 2) and linkers (L1, L12) are shown. Single-letter standard abbreviations are used to represent amino acids. The numbering scheme excludes the first methionine since this review covers not only genetic mutations but also phosphorylation and protease digestion sites. The amino acid residue assignment of the individual subdomains was derived from This assignment can vary depending on the source of the database. The indicated K8/K18 variants in the liver disease and control cohorts are derived from the studies summarized in Table 2, and are displayed above or below the schematic protein backbone, respectively. Keratin variants are categorized as those: (i) likely to be pathogenic in terms of risk for liver disease (red), (ii) unlikely to be pathogenic since they are found in similar or high frequency in the control cohort (black), (iii) “silent” without changing the amino acid (gray), and (iv) those found exclusively in the control group which may play a pathogenic role because of the nature of the amino acid alteration (green). The most common K8/K18 variant in Caucasians (and likely several other ethnic groups) is K8 R340H followed by K8 G61C as marked by the large arrowheads. Both K8 R340 and G61 are highly conserved among type-II keratins, while K8 Y53H and G433S are less conserved but are found almost exclusively in individuals of African descent (highlighted by encircled asterisk). Black bars at the end and beginning of the rod domain represent mutation hot spots for epidermal keratins and other IFs, and do not involve any known K8/K18 human variants. We define variants that are likely to be pathogenic (i.e., those displayed in red or green color) based on their preferential but not necessarily exclusive distribution in patients with liver disease rather than controls, and/or based on a mutation-induced change in the biochemical properties of K8/K18 such as crosslinking due to introduction of Cys (normal K8 and K18 do not have any Cys), or a change in solubility or phosphorylation.

A high frequency of the K8 Y53H and G433S variants in the U.S. liver disease cohort is noted among African Americans (6/26, 23.1%) compared with Caucasians (1/274, 0.4%) (P < 0.0001) and other ethnic groups.13 Further studies will need to validate the G433S variant frequency but Y53H is likely specific to those of African decent since it was seen in 0.1% (1/1532) of Germans but in 2.4% (17/722) of Africans.16 Both variants are likely to be pathogenic since transfection of K8 Y53H into cultured cells results in unstable keratin filaments12 while the G433S variant decreases K8 S431 phosphorylation in response to extracellular-regulated-kinase-1/2.13

K8 variants also associate with fibrosis progression during chronic HCV infection in a well-characterized German cohort.14 The overall K8/K18 mutation frequency of 5.8% in this cohort, which includes a broad range of liver disease severity, is lower than the 12.4% frequency (Table 2) in the end-stage liver disease U.S. cohort. However and albeit the numbers are relatively small, 10.3% (8/78) of patients with stage-4 fibrosis had K8 variants (i.e., similar to the 12.4% of patients with end-stage disease). The association of keratin mutation with liver fibrosis progression is supported by the significant predisposition of K18 R89C mice to thioacetamide-induced fibrosis (Table 1). However, the role of keratin mutation in promoting fibrosis is selective since K18 R89C mice have similar susceptibility to CCl4-mediated fibrosis as wildtype mice (Strnad et al., submitted).

Several other points regarding K8/K18 variants and liver disease deserve mention. First and for unclear reasons, K8 variants are far more common than K18 variants. Second, several nonexonic variants in KRT8/KRT18 have been identified but their association with liver disease is unclear. Notably, there is a unique and uniform association of an intronic IVS7+10delC deletion in all patients with the K8 R340H variant.14 Third, the animal data suggest that keratin variants are likely to predispose to acute liver injury and potential failure, which is currently being investigated in human studies. Fourth, epidermal/ocular keratin heterozygous mutations cause numerous tissue-specific diseases with near complete penetrance. In contrast, K8/K18 mutations pose a disease risk rather than promote disease per se since they exclude domains that are generally considered epidermal keratin and other IF mutation hotspots (Fig. 2).2 Although K8/K18 are expressed in several digestive organs, their exclusive expression in hepatocytes (without co-expression of additional keratins) explains why the liver is preferentially involved upon K8/K18 mutation.

Hepatocyte Keratin Functions and Their Modulation by Keratin Mutation

The induction, posttranslational and organizational modifications of keratins in response to injury in human (and mouse) liver disease, and their importance as cytoprotectors of the liver, implicates them as unique stress proteins. Liver keratins serve as stress markers by increasing their expression up to 3-fold, akin to heat shock protein (hsp)-70 which they bind to,3 and by undergoing posttranslational modifications such as phosphorylation, caspase-mediated fragmentation, ubiquitination and cross-linking.7, 17 Dynamic increased/decreased keratin site-specific phosphorylation serve as reliable markers for human liver disease progression/regression,18 and for epithelial cell injury in general.7 The keratin-related genetic animal models3, 7, 19, 20 and ex vivo cell culture studies20, 21 demonstrated the importance of keratins in protecting hepatocytes from oxidative and other hepatotoxic injuries that promote apoptosis, and also highlighted several important keratin functions (Fig. 3). Many of these functions ultimately feed into an over-arching function of K8/K18, namely to protect hepatocytes from apoptosis via mechanisms that include endowing mechanical integrity, serving as a phosphate sponge to divert undesirable stress-activated protein kinase (SAPK) substrate phosphorylation, modulation of cell signaling, appropriate targeting of cellular proteins and maintenance of mitochondrial function (Fig. 3).

Figure 3.

Summary of keratin functions and their alteration in the context of K8/K18 mutation. The diagram summarizes the effects of keratin mutation on normal keratin functions based on findings obtained using the keratin-targeted genetic animal models (Table 1) and related studies.

Keratin functions are fueled by phosphorylation and binding to associated proteins such as the 14-3-3 adaptor proteins, and are facilitated to a significant extent by keratin filament organization dynamics. Extreme cases of keratin filament disorganization are their absence in the context of K18- or K8-null hepatocytes, or their disruption in the context of K18 R89C. In these three contexts, hepatocytes become markedly fragile (Table 1) which provides direct evidence for the importance of K8/K18 in the mechanical cytoprotection of hepatocytes. This mechanical function is likely related to the unique biophysical flexible properties of the keratin cytoskeleton, as compared with microfilaments and microtubules, that allows it to withstand large deformations.22 To date, natural patient-associated mutations that cause K8/K18 absence or filament disruption have not been identified, in contrast to other keratins, which suggests that such mutations are likely to be lethal. However, more subtle keratin filament organizational dynamics are tightly linked to in vivo keratin phosphorylation, such as the role of K8 Ser73 phosphorylation by SAPKs p38 and JNK in keratin filament re-organization.7 In addition, the natural K8 variants Y53H and G61C interfere with keratin filament reorganization in response to oxidative and other stresses as noted in transfected cell culture models.12

Changes in keratin phosphorylation are central to hepatocyte injury responses as supported by transgenic mice that express phospho-mutant K18/K8. For example, mutation of the major K18 phospho-site (Ser52) predisposed mice to hepatotoxicity, while mutation of a nearby K18 phospho-site (Ser33) that regulates K18/14-3-3 binding blocked nuclear-to-cytoplasmic movement of 14-3-3 proteins during mitosis and resulted in the accumulation of abnormal mitotic figures without altering overall liver regeneration (Table 1). The importance of keratin/14-3-3 binding is further highlighted in studies of keratinocytes showing that such binding contributes to regulating protein synthesis and cell size,23 with similar findings in K8-null hepatocytes which are smaller and have decreased rates of protein synthesis than their normal counterparts.24 Furthermore, Raf-1 kinase, which also binds to 14-3-3 proteins, phosphorylates K18 and binds to K8 in a phosphorylation-dependent manner with release of the bound kinase in response to oxidative stress.25 In addition, expression of the K8 phospho-mutant Ser73Ala leads to marked predisposition to Fas-mediated injury.20 Therefore, keratins play an important role in regulating hepatocyte cell signaling via several modalities.

K8/K18 mutations can affect keratin phosphorylation directly or indirectly. Two patient-related examples include the K8 G61C/G433S mutations that induce conformational changes at nearby or somewhat distant serines that are physiologic SAPK substrates. For example, G433S mutation blocks phosphorylation at nearby K8 S431 when tested in vitro using extracellular-regulated-kinase-1/2.13 Similarly, K8 G61C expression in transgenic mice renders K8 S73 inaccessible to phosphorylation and shunts SAPKs to other substrates such as CREB, p90RSK and JNK thereby driving hepatocytes to undergo premature apoptosis. Therefore, one keratin non-mechanical function that was unmasked by keratin mutation is to serve as a phosphate sponge during stress conditions.20

Other important and emerging keratin functions that may be impacted by mutation are roles in protein-targeting and organelle functions.26 For example, K8-null liver shows increased basolateral distribution of the bile canaliculus Ecto-ATPase,27 an increased cell surface distribution of Fas receptor,21 and altered desmoplakin distribution.28 Furthermore, K18 R89C expression in cultured cells alters Golgi distribution29 and redistributes peripheral membrane (desmoplakin/zonula occludens-1/β-catenin) and cytosolic (14-3-3 ζ/glucose-6-phosphate dehydrogenase) proteins into keratin aggregates.30 In addition, proteomic analysis of K8-null livers shows significant changes in several mitochondrial proteins, and indeed isolated or in-situ K8-null mitochondria are smaller, show altered cytoplasmic distribution and are leaky when compared with control livers (unpublished observation).

Keratins Are Essential for Mallory-Denk Body Formation

MDBs were first described by Frank Mallory in 1911 as cytoplasmic hyaline inclusions characteristic of alcoholic liver disease and were recently re-named Mallory-Denk bodies to honor Helmut Denk's contributions.5 In addition to alcohol-related liver disease, MDBs are found in nonalcoholic steatohepatitis, Indian childhood cirrhosis, Wilson disease, primary biliary cirrhosis and hepatocellular carcinoma and share important structural and morphological features with aggregates of several nonepithelial disorders including neurodegenerative and muscular diseases.5 In mice, MDBs are induced by feeding with griseofulvin or diethoxycarbonyl-1,4-dihydrocollidine (DDC).5 These drugs result in distinct pathologic features to what is seen in human MDBs, but mouse models have contributed significantly to our understanding of the mechanism of MDB formation (Table 1, Fig. 4). Patient and experimental MDB formation is reversible after discontinuation of alcohol or the inciting drug but can rapidly reform upon re-exposure to toxic insult.5

Figure 4.

Properties and pathogenesis of MDBs. (A) MDBs are found in several human liver diseases and can be induced in mice by long-term feeding with toxins. Despite the varied histological features of different MDB-inducing conditions in humans and mouse models, MDBs are found primarily in ballooned cells (but not all ballooned cells contain MDBs) and share the same molecular constituents and morphologic appearance. Ultrastructurally, MDBs consist of randomly oriented short filaments (type II MDB) though much less common MDBs contain amorphous, electron-dense material (Type III) or parallel filaments (Type I).5 The assessment of MDBs in clinical contexts varies from the methods used in research settings. In routine clinical evaluation, MDBs are identified by their typical morphology and eosinophilic appearance using H&E staining, whereas indirect immunofluorescence keratin co-staining with ubiquitin and p62 is typically used to visualize MDBs in experimental models.5 H&E detects mainly large perinuclear inclusions whereas smaller aggregates can be visualized by indirect immunofluorescence. In addition, immunoblotting provides a useful biochemical approach to identify large molecular-sized complexes.17 ASH, alcoholic steatohepatitis; NASH, nonalcoholic steatohepatitis; PBC, primary biliary cirrhosis; HCC, hepatocellular carcinoma; DDC, 3,5-diethoxycarbonyl-1,4-dihydrocollidine; UBB+1, a frameshift mutation of ubiquitin B; H&E, hematoxylin and eosin staining; IHC, immunohistochemistry. (B) MDBs are induced in genetically predisposed individuals during specific chronic hepatocyte modes of injury. The critical determinants predisposing to MDB formation include the induction of an unbalanced K8>K18 ratio, K8 crosslinking by transamidation, keratin hyperphosphorylation and p62 overexpression. MDB formation is reversible and factors contributing to MDB resolution are toxin removal and proteasomal/autophagic degradation. It is not known if MDBs found in different clinical settings are identical in composition or precise mechanism of formation.

MDBs are defined by their morphological appearance and molecular composition. The bulk of MDBs consists of ubiquitinated and transamidated phospho-K8/K18, with limited amounts of chaperones (hsp-25/60/70) and the polyubiquitin-binding protein p625 (Fig. 4). MDBs typically form irregularly shaped large perinuclear aggregates that are routinely identified in clinical settings as eosinophilic or trichrome aniline blue-positive structures. They can be occasionally mistaken for intracytoplasmic hyaline bodies, which have a globular shape and a surrounding halo but are keratin-negative.5, 31 MDBs also have to be distinguished from apoptotic bodies and megamitochondria.5 Immunohistochemistry or immunofluorescence staining offer the most sensitive MDB detection method but are used infrequently in routine clinical diagnosis. Such methods allow easy detection of small MDBs, located throughout the cytoplasm or in the cell periphery in proximity to desmosomes, but the clinical significance of such small pre-MDBs or possibly resorbing MDBs remains to be defined.5

The mechanism of MDB formation is becoming increasingly understood. The most important contributors to MDB formation are achieving a K8>K18 protein state, facilitated by keratin induction and possible turnover, in association with K8/K18 hyperphosphorylation and crosslinking by transamidation. The importance of a K8>K18 ratio has been validated using several mouse genetic models.5 Keratin enzymatic crosslinking to itself (and likely to other proteins) is also required for MDB formation and occurs preferentially by transglutaminase-2, which accounts for ≈75% of mouse liver transglutaminase activity.17, 32 Site-specific keratin phosphorylation also plays an important role in MDB formation as exemplified by the inhibition of MDB formation in mice that express the phospho-mutant K8 S73A (unpublished observations) but not the phospho-mutants K18 S33A/S52A.33 Other factors likely to be important for MDB formation include the genetic background, extent of protein misfolding and p62 induction.5 Protein misfolding may occur in response to oxidative injury which renders proteins aggregation prone.5 Damaged proteins can be repaired through chaperones or degraded by the proteasomal/autophagic machinery. However, chaperones can be modified and their function compromised during alcoholic/hepatotoxic liver injury and the proteasome may be inhibited by UBB+1, a frameshift mutation of ubiquitin B.5, 34 Alternatively, the production of misfolded proteins can exceed the degradation capacity of the cell. MDB formation is also dependent on genetic factors as seen in different mouse strains, which manifest variable MDB amounts after DDC feeding (unpublished observations). Understanding these genetic differences in experimental animals is likely to contribute to our understanding of why MDBs are seen in only a subset of at-risk patients.

It is currently unknown, whether MDB formation is a bystander event, is protective or helps perpetuate liver injury. MDBs are often found in ballooned, empty hepatocytes and ballooning together with necroinflammation is an indicator of increased disease activity.35 In addition, MDB-containing hepatocytes are often surrounded and/or infiltrated by neutrophils and may represent a neoantigen immune-reaction activator. However, MDBs do not appear to diminish hepatocyte viability, and there is restoration of a proper keratin network after cessation of injury. It is possible that one role of MDB formation is to protect hepatocytes by sequestering potentially toxic and misfolded proteins into an inert inclusion body, but this possibility has not been validated directly.5

Serologic Testing of Keratins and Their Fragments in Liver Disease

Keratins are currently used in serology diagnostics pertaining to the liver in some settings and Centers.36 K8/K18/K19 or their nonapoptotic fragments, which are released from carcinoma cells into blood, constitute established tumor markers and include tissue-polypeptide-antigen (TPA, represents total K8/K18/K19), tissue-polypeptide-specific antigen (TPS, derived from K18), and CYtokeratin FRAgment 21-1 (CYFRA 21-1, derived from K19). Their main clinical use is to monitor treatment responses and tumor recurrence and to provide potential prognostic information for some cancers.37 However, serum keratin levels are also elevated in nonmalignant diseases thereby limiting their diagnostic specificity. For example, TPS and CYFRA 21-1 levels increase in alcoholic liver disease38 and high TPS levels occur in patients with nonalcoholic steatohepatitis.39

In contrast to assays that non-selectively measure released keratin fragments and intact keratins, methods that use antibodies specifically recognizing apoptotic K18 or K19 fragments but not intact keratins may hold promise. This strategy is based on the finding that type-I keratins are cleaved at the highly conserved VEMD/VEVD sequence in the L12 linker (Fig. 1A) during apoptosis.8, 40, 41 K18 has an additional caspase site at Asp396 (Fig. 1A) which is cut early during apoptosis (prior to K18 Asp237) before annexin-V reactivity and DNA fragmentation.40–42 Digestion of human K18 Asp396 can be monitored using the M30-antibody which recognizes newly exposed Asp396 after proteolysis.42 The M30-antibody ELISA may be a discriminatory serologic marker for determining liver disease severity, such as distinguishing simple steatosis from non-alcoholic steatohepatitis43 and predicting disease severity, treatment response and fibrosis progression in patients with chronic HCV infection.44, 45 Similarly, an antibody specific to exposed K18 Asp237 can be used to detect caspase-digested K18 and may offer unique features including applicability to mouse and human K18 and K19, fragment stability and high sensitivity (Tao et al., submitted). An important caveat is that apoptosis- and necrosis-generated keratin fragments may have overlapping epitopes.36 Therefore, the use of selective antibodies to measure released keratins or their fragments in sera of patients with unique liver diseases may offer clinically-useful advantages. Additional studies should help clarify the diagnostic and prognostic benefits of single or combined use of such reagents.

Another potential serologic marker is the formation of anti-keratin antibodies. Cytoskeleton components, particularly actin but also IF proteins, represent some of the molecular targets of anti-smooth muscle autoantibodies in patients with autoimmune liver disease.46, 47 For example, antibodies to K8/K18/K19 and their immune complexes have been described in some patients with autoimmune hepatitis, and titers of these antibodies decrease after steroid therapy.48 Also, proteomic analysis found an increased frequency of K8 antibodies in patients with hepatocellular carcinoma when compared to chronic hepatitis B/C carriers.49 However, the clinical utility of keratin auto-antibodies remains unclear although their presence in the context of some liver diseases does occur and warrants further investigation.

Summary and Future Directions

The importance of keratins in human liver disease is now clearly established based on strong human association studies and highly supportive genetic animal models. Keratins can be considered hepatocyte stress proteins due to their induction upon liver injury and their cytoprotective roles in preventing hepatocyte apoptosis and other forms of injury. The IF field, as applied to liver biology and disease, is an exciting area that offers many venues for study. Open areas of investigation with regard to human keratin mutations include: (i) defining whether a subset of liver diseases are more likely to be involved than others, (ii) clarifying the role of K8/K18 in acute liver failure, (iii) determining whether K19 variants play a role in biliary liver disease, (iv) defining the importance of unique keratin variants in specific ethnic backgrounds and (v) addressing therapeutic and prevention options for keratin-related liver diseases. Further understanding of keratin functions is likely to provide insights into how their mutations predispose to liver disease.

At the protein level, keratins are not only the most abundant components of MDBs but are also essential for MDB formation. Refining our understanding of the precise mechanism and genetics of MDB formation should allow determining whether MDBs serve a protective cell response, and whether they offer prognostic/diagnostic/therapeutic values beyond their current utility as histological markers of some liver diseases. Keratin abundance and the selectivity of K18 and K19 as caspase substrates render them attractive candidate diagnostic markers of liver epithelial cell death. Furthermore, future studies will likely unfold the identity of novel keratin-binding proteins and posttranslational modifications that may be involved in regulating unique keratin functions. Also, assessment of the context (acute versus chronic; necrosis versus apoptosis) and type of liver disease with regard to the predictive value and utility of intact or fragment keratin accumulation will need to be determined.

Other relevant potential areas of future investigation include the role of keratins and their mutations in modulating cross-talk between liver epithelial and non-epithelial cells. For example, interleukin-6 up-regulates K8 expression in cultured colonic cell lines50 and similar keratin regulation by interleukin-6 or other cytokines may occur in the liver. In addition, the role of other IF proteins in modulating the function of stellate, endothelial and Kupffer cells (e.g., vimentin/desmin/synemin51) and their cross-talk with surrounding epithelial cells is a wide open area for investigation.


We are grateful to Helene Baribault, Thomas Magin and Robert Oshima for their generous sharing of the keratin genetic animal models that have helped advance the field; Evelyn Resurrection for assistance with immune staining; and to Kris Morrow for assistance with figure preparation. We also thank all the patients, Blood Bank donors, families and other volunteers who have participated in the genetic studies described herein. Because of space constraints, we apologize for not being able to include all potentially relevant references.