Keratin mutation primes mouse liver to oxidative injury


  • Qin Zhou,

    1. Department of Medicine, Palo Alto Veterans Affairs Medical Center and Stanford University Digestive Disease Center, Palo Alto, CA
    Search for more papers by this author
  • Xuhuai Ji,

    1. Department of Medicine, Palo Alto Veterans Affairs Medical Center and Stanford University Digestive Disease Center, Palo Alto, CA
    Search for more papers by this author
  • Lixin Chen,

    1. Division of Gastroenterology and Liver Diseases, University of Southern California Research Center for Liver Diseases, USC-UCLA Research Center for Alcoholic Liver and Pancreatic Diseases, Keck School of Medicine USC, Los Angeles, CA
    Search for more papers by this author
  • Harry B. Greenberg,

    1. Department of Medicine, Palo Alto Veterans Affairs Medical Center and Stanford University Digestive Disease Center, Palo Alto, CA
    Search for more papers by this author
  • Shelly C. Lu,

    1. Division of Gastroenterology and Liver Diseases, University of Southern California Research Center for Liver Diseases, USC-UCLA Research Center for Alcoholic Liver and Pancreatic Diseases, Keck School of Medicine USC, Los Angeles, CA
    Search for more papers by this author
  • M. Bishr Omary

    Corresponding author
    1. Department of Medicine, Palo Alto Veterans Affairs Medical Center and Stanford University Digestive Disease Center, Palo Alto, CA
    • Palo Alto VA Medical Center, 3801 Miranda Avenue, Mail Code 154J, Palo Alto, CA 94304
    Search for more papers by this author
    • Fax: 650-852-3259

  • Conflict of interest: Nothing to report.


Mutation of the cytoskeletal intermediate filament proteins keratin 8 and keratin 18 (K8/K18) is associated with cirrhosis in humans, whereas transgenic mice that overexpress K18 Arg89→Cys (R89C) have significant predisposition to liver injury. To study the mechanism of keratin-associated predisposition to liver injury, we used mouse microarrays to examine genetic changes associated with hepatocyte keratin mutation and assessed the consequences of such changes. Liver gene expression was compared in R89C versus nontransgenic or wild-type K18-overexpressing mice. Microarray-defined genetic changes were confirmed by quantitative polymerase chain reaction. Nineteen genes had a more than two-fold altered expression (nine downregulated, 10 upregulated). Upregulated genes in keratin-mutant hepatocytes included the oxidative metabolism genes cytochrome P450, S-adenosylhomocysteine (SAH) hydrolase, cysteine sulfinic acid decarboxylase, and oxidation-reduction pathway genes. Downregulated genes included fatty acid binding protein 5, cyclin D1, and some signaling molecules. Several methionine metabolism-related and glutathione synthetic pathway intermediates, including S-adenosylmethionine (SAMe) and SAH, were modulated in R89C versus control mice. R89C livers had higher lipid and protein oxidation by-products as reflected by increased malondialdehyde and oxidized albumin. In conclusion, K18 point mutation in transgenic mice modulates several hepatocyte oxidative stress-related genes and leads to lipid and protein oxidative by-products. Mutation-associated decreases in SAH and SAMe could compromise needed cysteine availability to generate glutathione during oxidative stress. Hence keratin mutations may prime hepatocytes to oxidative injury, which provides a new potential mechanism for how keratin mutations may predispose patients to cirrhosis. (HEPATOLOGY 2005;41:517–525.)

Keratins (Ks) make up the intermediate filament cytoskeleton of epithelial cells and include type I (K9-K20) and type II (K1-K8) keratins.1–3 Most if not all epithelial cells express at least one type I and one type II keratin as obligate noncovalent heteropolymers in an epithelial cell– and differentiation-specific manner. For example: intestinal epithelia express K8/K19 with variable levels of K7/K18/K20, depending on the location within the crypt–villus axis; keratinocytes express K5/K14 basally and K1/K10 suprabasally; and hepatocytes express K8/K18 exclusively. Keratin mutations cause a variety of skin (K1/K5/K10/K14), oral/esophageal (K4/K13), and ocular (K3/K12) diseases and are associated with cryptogenic and noncryptogenic forms of end-stage liver disease (K8/K18).4–6

Involvement of keratin mutations with human disease, as a cause or predisposition, was initially suggested by the results of animal studies in which mutant keratins expressed in transgenic mice resulted in phenotypes that mirrored several human diseases.3, 7, 8 For example, transgenic mice that overexpressed a K14 deletion mutant developed a blistering skin disease9 that led to identifying mutations in K14 or its partner K5 as the cause of epidermolysis bullosa simplex.10, 11 Similarly, transgenic mice (R89C mice) that overexpress human K18 that is mutated at a highly conserved arginine (Arg89→Cys) develop mild chronic hepatitis but a marked predisposition to drug-induced liver injury in association with hepatocyte keratin filament disruption.12–14 Arg89 of K18 was chosen for mutation targeting because it is highly conserved in intermediate filament proteins, and its homolog is often mutated in several epidermal keratins and accounts for approximately 40% of the mutations involving epidermal keratins.15 Findings from animal models also provided the insight that one important function of K8/K18 is to protect hepatocytes from a variety of mechanical and nonmechanical forms of stress and led to the search and subsequent identification of K8/K18 mutations in patients with liver disease.16–18

The mechanism of how K8/K18 mutations predispose to liver disease is partially understood. A heightened susceptibility to injury after apoptotic stimulation was clearly demonstrated in livers of mice that express K18 R89C or in mice/hepatocytes that lack keratins.19–22 Keratin filament instability or “hyperstability” has also been observed upon transfection of cells with K8 mutations that are found in patients with liver disease, but how keratin filament alterations translate to the ability of a cell to better cope with stress is unknown.5, 17, 23 The liver appears to be selectively susceptible to injury as a consequence of K8/K18 mutation or absence as compared with the pancreas24, 25 and gallbladder,26 which appear to be spared when challenged with tissue-specific insults; this is likely due to compensation by other cytoskeletal and possibly noncytoskeletal proteins.27 To further address how keratin mutations predispose to liver injury, we used gene profiling and several biochemical tools to assess genetic changes that occur as a consequence of the K18 R89C mutation, which in turn causes keratin filament collapse and increased susceptibility to liver injury.


K, keratin; R89C, transgenic mice that overexpress an Arg89-to-Cys K18 mutant; SAH, S-adenosylhomocysteine; SAMe, S-adenosylmethionine; WT, wild-type; cDNA, complementary DNA; SAM, significance analysis of microarray; FDR, false discovery rate; PCR, polymerase chain reaction; GSH, glutathione; MDA, malondialdehyde; DNP, dinitrophenyl; mRNA, messenger RNA.

Materials and Methods

Mouse Strains.

Three homozygous transgenic mouse lines in an FVB/n genetic background (all male, 8-10 weeks old) were used: TG2 mice, which express wild-type (WT) human K18 (as control mice), and the F22 and F50 lines that express human K18 R89C.12, 13, 21 The transgene copy number is 34 for TG2, 20 for F22, and 36 for F50.13 Nontransgenic FVB/n mice were used as an additional control. All animals received care according to standard accepted criteria and guidelines.

Fluorescence-Labeled Targets and Complementary DNA Microarray Hybridization.

Total RNA was extracted from livers using an RNase mid kit (Qiagen, Valencia, CA) as recommended by the supplier. Total RNA (80-100 μg) was reverse transcribed into complementary DNA (cDNA) using Superscript II (Invitrogen, Carlsbad, CA) with random hexamer primers while incorporating Cy3-deoxyuridine triphosphate (Amersham Biosciences, Piscataway, NJ) for F22-, F50-, and FVB/n-derived cDNA or Cy5-deoxyuridine triphosphate for TG2-derived cDNA. Each experiment was independent in that it compared hybridization of labeled cDNA derived from separate mice. A total of 14 microarray experiment datasets (8 for F22/TG2, 6 for F50/TG2) were generated and analyzed. Fluorescence signals were acquired using a GenePix 4000b microarray scanner (Axon Instruments, Foster City, CA), and the images were processed using the GenePix Pro 3.0 software (Axon Instruments).28 The data files were entered into the Stanford Microarray Database and normalized. A filter was set to select array elements with a regression correlation of r > 0.6.

Significance Analysis of Microarray and Gene Function Annotation.

To identify the upregulated or downregulated genes in F22/F50 versus TG2 and FVB/n versus TG2 mouse livers, 1-class significance analysis of microarray (SAM) was performed.29 This analysis generated a list of genes with an average Cy5/Cy3 ratio that is significantly different from 1.0, along with an estimate (at 90% confidence) of the percentage of such genes identified by chance, or the false discovery rate (FDR), which is based on permutations of repeated measurements. The FDR can be adjusted by setting a threshold Δ at different stringencies, resulting in different numbers of genes whose change in expression are classified as significant. A list of genes identified by 1-class SAM analysis was compiled. The criteria for a gene to be included were an FDR of less than 1% and an average fold change of 2 or more. Also, 2-class SAM analysis was performed to identify genes regulated differently between the F22/TG2 and F50/TG2 comparisons to assess the difference in the expression pattern of F22 and F50 compared with TG2. Genes were assigned manually into functional pathway based on information retrieved from the Stanford Online Universal Resource for Clones and Expressed sequence tags (

Quantitative Reverse-Transcriptase Polymerase Chain Reaction.

Several genes were chosen for quantitative expression analysis. For this, 2 μg total RNA was reverse transcribed into single-strand cDNA using random primers as previously described.30 cDNA products were amplified using gene-specific primers (Table 1), and real-time quantitative polymerase chain reaction (PCR) was performed with an ABI Prism 7700 Sequence Detection System and SYBR green PCR Master mix (PE Biosystems, Foster City, CA). Cycling parameters were 95°C for 10 minutes, 95°C for 15 seconds, then 60°C for 1 minute (40 cycles). Each sample was analyzed three times, each time in triplicate, and each experiment was performed three times using independent mouse total RNA. Ribosomal protein L7 was used as an internal control.

Table 1. Quantitative Reverse-Transcriptase PCR Primers
GeneSequenceProduct Size (bp)
  1. NOTE. The primers shown were used to amplify several genes, including those that were modulated two-fold or more as determined by microarray analysis. Also shown is the predicted size of the amplified products.

  2. Abbreviations: Csad, cysteine sulfinic acid decarboxylase; Fabp5, fatty acid binding protein 5; Mat1a, methionine adenosyltransferase 1a; ccnd1, cyclin D1; GSTpi, glutathione S-transferase pi.

Riken 4632419J125′-CAAAGGGGCTTGTACGACAT-3′97

Glutathione, S-adenosylmethionine, S-adenosyl- homocysteine, and Malondialdehyde Measurement.

S-adenosylmethionine (SAMe), S-adenosyl-homocysteine (SAH), and glutathione (GSH) were purchased from Sigma (St. Louis, MO). GSH levels were determined using the recycling method of Tietze.31 SAMe and SAH levels were measured as described with slight modification. Liver specimens were homogenized (1:3 vol/vol) in 10 mmol/L phosphate (pH 7.4), 0.25 mol/L sucrose and an aliquot was used for protein determination.32 Liver homogenates were treated with 0.5 mol/L perchloric acid (1:1 vol/vol) and pelleted (1,000g for 15 minutes); the aqueous layer was then quantitatively removed and neutralized with 3 mol/L potassium hydroxide. SAMe and SAH were then determined by high-performance liquid chromatography (Series 410 LC pump) with an LC-90 UV detector and an LC-100 integrator (Perkin Elmer, Boston, MA) using a Partisil SCX 10-μm column (Whatman Chemical Separations, Cleveland, OH). SAMe was eluted isocratically at 1 mL/min with 0.19 M NH4H2PO4 (adjusted to pH 2.6 with 2 M H3PO4). SAH was eluted isocratically at 1 mL/min with 0.03 mol/L NH4H2PO4 containing 2% vol/vol acetonitrile (pH 2.6). SAMe and SAH were identified by measuring the absorbance at 254 nm, then quantified with standard curves of pure SAMe and SAH. Identity of the SAMe and SAH peaks was confirmed by spiking the sample with known standards, and their levels are reported as nmol/mg protein. Cysteine levels were measured as previously described.33

For malondialdehyde (MDA) estimation, livers were homogenized in phosphate-buffered saline, and MDA levels were determined using a Bioxytech MDA-586 kit (Northwest Life Science Specialties, Vancouver, WA) as recommended by the supplier. P values were determined using the Student t test.

Immunofluorescence Staining.

Livers were snap-frozen in optimum cutting temperature compound, followed by sectioning (6 μm/section) and fixing with acetone (10 minutes at −20°C). Tissues were double stained using mouse monoclonal antibody L2A1, which recognizes human K18, and rabbit antibody 8592, which recognizes human and mouse K8/K18.21 Antibody staining was visualized using Texas red–conjugated goat antirabbit and fluorescein isothiocyanate–conjugated goat antimouse antibodies and confocal microscopy.

Dinitrophenyl Derivatization, Two-Dimensional Gel Electrophoresis and Immunoblotting.

Livers were homogenized in a buffer containing 7 mol/L urea, 2 mol/L thiourea, 4% CHAPS, 0.5% IPG, and a protease inhibitor cocktail (Sigma), followed by centrifugation (13,000g for 10 minutes at 22°C). Samples were sonicated (three times for 20 seconds) to break up DNA, then repelleted. The supernatants were separated using an IPGphor isoelectric focus power supply and pre-cast Immobilin DryStrip pI 3-10 gel strips as recommended by the manufacturer (Amersham Biosciences, Piscataway, NJ). Following isoelectric focusing, the strips were subjected to in-strip dinitrophenyl (DNP) derivatization by incubating (15 minutes at 22°C) in 2 N HCl/10 mmol/L 2,4-dinitrophenyl-hydrazine (Sigma).34, 35 Strips were washed with 2 mol/L Tris-base/30% glycerol (15 minutes); equilibrated in 50 mmol/L Tris-HCl (pH 8.8) containing 6 mol/L urea, 2% sodium dodecyl sulfate, 30% glycerol, and 1% dithiothreitol (15 minutes); then re-equilibrated in the same buffer containing 2.5% iodoacetamide instead of dithiothreitol (15 minutes). The strips were then subjected to a second dimension gradient gel SDS-PAGE, transferred to polyvinylidene difluoride membranes, then immunoblotted with goat anti-DNP antibody (Bethyl Laboratories, Inc., Montgomery, TX). Spots of interest from Coomassie-stained duplicate gels were excised from the gels, then processed for protein identification using mass spectrometry.


We compared the gene profile of livers from transgenic mice that overexpress WT K18 (TG2 line) versus mice that overexpress human K18 Arg89→Cys (F22 and F50 lines), all of which were in an FVB/n mouse strain background. As shown in Fig. 1, the R89C mutation results in cytoplasmic keratin filament disruption (panels C and F vs. panels A, B, and E) as described previously.13 Expression of the human K18 R89C transgene behaves in a dominant-negative fashion with tissue-specific disruption of the endogenous keratin network (Fig. 1). The gene profile data were analyzed to identify genes that are significantly different between mutant versus WT keratin-overexpressing livers. The numbers of total affected genes (up- and downregulated) are shown in Fig. 2A, which compares mutant (F22 and F50) with WT (TG2) (1-class SAM analysis). When the FDR was set at less than 5%, 847 genes were upregulated and 150 genes were downregulated. At the more stringent FDR of less than 1%, the numbers decreased to 201 and 68, respectively (Fig. 2A). We confirmed that the F22 and F50 mutant K18–overexpressing lines behave similarly by using 2-class SAM to compare differential gene expression changes between F22/TG2 versus F50/TG2, and no such genes were identified with an FDR cutoff of less than 30% (Fig. 2A). Furthermore, no differences were noted when comparing nontransgenic FVB/n and WT K18–overexpressing (TG2) mice using 1-class SAM analysis (with an FDR cutoff of <10%; see Fig. 2A). Hence, in terms of liver expression profiling, the two mutant transgenic lines behave similarly when each is compared with the WT line, and the WT transgenic line behaves similarly to the nontransgenic line.

Figure 1.

Keratin filament staining in livers of mice that overexpress human WT or mutant K18. Livers were double stained using antibodies that recognize human and mouse (A-C) or human (D-F) K18. Panels A and D, B and E, and C and F represent sections that were double-stained as described in Materials and Methods. Bar = 30 microns. K, keratin.

Figure 2.

(A) The number of microarray elements that emerge as significant as related to the indicated mouse genotypes and the selected parameters. The total number of upregulated and downregulated genes is shown. Note the difference in the affected number of genes depending on FDR stringency. (B) Distribution of fold change values in mRNA levels of genes that are upregulated (positive values) or downregulated (negative values) when comparing F22 with TG2, or F50 with TG2 mice. The histograms represent the genes that were identified by 1-class SAM analysis at a FDR of less than 1% and a fold change of 1.5× to 2× (black bars) or 2× or more (gray bars). FDR, false discovery rate; SAM, significance analysis of microarray; Up-, upregulated; Down-, downregulated.

The average fold change in messenger RNA (mRNA) level was determined for all the up- and downregulated genes identified at a FDR of less than 1% (Fig. 2B). Using strict cutoff criteria, overexpression of human R89C K18 in transgenic mice when compared with mice that overexpress WT K18 results in the downregulation of 9 to 40 genes and the upregulation of 8 to 64 genes, depending on the strictness of the analysis criteria (Fig. 2B). We focused on genes that had a more than two-fold change in their expression (Table 2), even though this cutoff value excludes approximately 90% of the genes that were otherwise significantly up- or downregulated by SAM criteria. Interestingly, many of the affected genes are involved in oxidative metabolism and stress pathways, with the most significant pathways involving hepatic methionine metabolism and GSH synthesis. We then used quantitative reverse-transcriptase PCR to confirm the microarray finding for some of the altered genes. As shown in Fig. 3 and summarized in Table 2 for the transcripts we tested, the two upregulated (cysteine sulfinic acid decarboxylase and SAH hydroxylase) and four downregulated (cyclin D1, 4632419J12, GSH S-transferase pi, and fatty acid binding protein 5) genes had comparable mRNA alterations when using microarray analysis or quantitative reverse-transcriptase PCR. Not all gene alterations noted by array analysis were confirmed. For example, methionine adenosyltransferase 1A mRNA appeared to be increased 2.2-fold in F22 compared with TG2 (Table 2), but the mRNA levels were similar when tested with quantitative PCR (Fig. 3) or Northern blotting (not shown).

Table 2. Gene Alterations When Comparing Transgenic Mice That Overexpress WT or Mutant K18
Genbank Accession No.ProductFold Change (Array)FunctionFold Change (PCR)
  1. NOTE. This table lists the genes that are up- or downregulated when comparing F22 or F50 with TG2 mouse livers, and the fold changes as determined by array analysis or real-time quantitative PCR. A representative number of microarray-identified genes from each genotype comparison were analyzed via PCR, and the fold change is shown in the right column for those genes that were tested.

  2. Abbreviations: Csad, cysteine sulfinic acid decarboxylase; NADPH, reduced nicotinamide-adenine dinucleotide phosphate; Mat1a, methionine adenosyltransferase 1a; Ccnd1, cyclin D1; Fabp5, fatty acid binding protein 5.

F22 vs. TG2 upregulated    
 H3067A08Csad2.9L-cysteine catabolism to taurine2.4
 H3045H01Riken cDNA 0610039N193.7Unknown 
 1810030B1Cytochrome P450, 2b203.9NADPH-dependent electron transport 
 H3051A05Riken cDNA 1300002P222.5Fatty acid beta-oxidation cycle 
 H3014E01SAH hydroxylase2.0Key role in transsulfuration pathway2.0
 MPC:668G0/G1 switch gene 22.5Cell cycle 
 H3128D01MAT1a2.2Formation of SAMe1.1
 H3131A07Ech12.0Auxiliary fatty acid beta-oxidation 
F22 vs. TG2 downregulated    
 2810401B1Ccnd12.1Cell cycle control1.7
 2310034D1Anxa52.3Blood coagulation cascade 
 Image:33237Riken cDNA 4632419J122.9Inhibitor of serine proteases1.6
 2410004108Copz12.1Protein transport through Golgi network 
 2310007M2Riken cDNA 1190006C122.0Unknown 
 Image:60427GSH S-transferase2.0GSH transferase activity1.6
 2610014N07Riken cDNA 2610014N072.4Ubiquitin carboxyl-terminal hydrolase 
 2610025H03Fabp52.5Lipid-binding activity and transport2.9
 2210404C19nudt72.3Acetyl-CoA catabolism 
F50 vs. TG2 upregulated    
 H3067A08Csad2.2L-cysteine catabolism to taurine2.9
 H3045H01Riken cDNA 0610039N192.1Unknown 
 1810074H01Riken cDNA 1810074H014.6Homophilic adhesion 
F50 vs. TG2 downregulated    
 2810401B1Ccnd12.6Cell cycle control1.4
 2310034D1Anxa52.3Blood coagulation cascade 
 H3136A07Procollagen, type1, alpha 22.7Extracellular matrix, cell adhesion 
Figure 3.

Analysis of quantitative reverse-transcriptase PCR products. The primers listed in Table 1 were used to amplify RNA isolated from livers of the indicated mouse genotypes. Amplified products were separated using 2% agarose gels and were then visualized. All products had the predicted base pair size (Table 1). SAH, S-adenosylhomocysteine; Csad, cysteine sulfinic acid decarboxylase; Mat1a, methionine adenosyltransferase 1a; ccnd1, cyclin D1; GSTpi, glutathione S-transferase pi; Fabp5, fatty acid binding protein 5.

Given the enzymatic properties of the products of the two altered expressed genes, SAH hydroxylase and cysteine sulfinic acid decarboxylase (Table 2 and Fig. 3), we measured the liver tissue concentration of several precursors and products that are involved in cysteine and GSH formation and catabolism (see Fig. 4 for the biosynthetic pathway). As shown in Table 3, livers that express mutant K18 (F22 and F50) have similar levels of cysteine and GSH as livers that express WT K18. In contrast, SAMe and SAH liver tissue levels were reproducibly lower by approximately 15% to 22% in the F22 and F50 mice compared with TG2 mice (Table 3). This suggests that precursors to cysteine and GSH may be depleted more rapidly under conditions of stress and that the threshold to generate oxidation products may be lower in the keratin-mutant livers.

Figure 4.

Hepatic methionine metabolism and GSH biosynthesis pathways and effect of the K18 R89C mutation. The transsulfuration pathway converts methionine to cysteine, which is then converted to GSH via the GSH synthetic pathway. Increases and decreases in gene expression are represented by upward and downward arrows, respectively. Although basal GSH levels are similar when comparing WT- versus mutant-expressing K18 livers (Table 3), our proposed model is that under stress conditions the decreased SAMe and increased cysteine sulfinic acid decarboxylase in the setting of keratin mutation shunt or deplete cysteine from GSH production. In turn, the decrease in available GSH during stress would render hepatocytes more prone to oxidative injury. SAMe, S-adenosylmethionine; SAH, S-adenosylhomocysteine; Csad, cysteine sulfinic acid decarboxylase; GSH, glutathione.

Table 3. SAMe/SAH/GSH/Cysteine Concentrations in TG2, F22, and F50 Mouse Livers
  1. NOTE. For each transgenic mouse line, livers were removed from 4 individual mice (1–4), then processed to measure the concentration of the indicated products as described in Materials and Methods. The mean concentration is also listed for each product. The P values compare the measured concentration in (F22 + F50) versus TG2. All values are represented as nmol/L per mg protein.

Mean0.97 ± 0.080.35 ± 0.0224.89 ± 0.750.55 ± 0.07
Mean0.82 ± 0.030.30 ± 0.0124.30 ± 2.100.57 ± 0.06
Mean0.75 ± 0.050.29 ± 0.0127.95 ± 0.510.59 ± 0.03
P value.

The profile of the altered gene expression in the keratin-mutant livers led us to hypothesize that altered protein and lipid modifications related to oxidative stress may be generated because of the keratin mutation. Lipid peroxides, derived from polyunsaturated fatty acids, are unstable and decompose to form complex by-products including reactive carbonyl compounds, the most abundant of which is MDA. As shown in Table 4, MDA levels are significantly higher at basal levels in livers harboring the keratin mutation. Reactive oxygen species can also generate protein carbonyls, which can be estimated via coupling with 2,4-dinitrophenylhydrazine. To investigate the potential formation of protein reactive oxygen species by-products, we performed comparative high-resolution, two-dimensional gel electrophoresis on liver homogenates isolated from TG2, F22, and F50 mice. As shown in Fig. 5, although the Coomassie-stained protein profiles are similar when comparing TG2 with F22 and F50 livers, DNP staining using anti-DNP antibody showed a prominent increase in a 70-kd protein (highlighted by arrows; compare panels E and F with panel D) with a pI of 5.5 to 5.6. Excision of this protein followed by trypsin digestion and mass spectrometry of the tryptic fragments confirmed its identity as albumin, which is also consistent with its reported pI of 5.7 and molecular weight. Taken together, R89C mouse livers do indeed contain by-products of protein and lipid modification, as suggested by the gene expression profiling results.

Table 4. MDA Levels in Livers of TG2, F22, and F50 Mice
  1. NOTE. Livers were separately harvested from 4 mice of the indicated genotypes followed by estimation of MDA levels as described in Materials and Methods. The P values for the various MDA-level comparisons are .05 for F22 versus TG2, .007 for F50 versus TG2, and .0006 for (F22 + F50) versus TG2.

Mouse no.   
Mean94 ± 15152 ± 46153 ± 25
Figure 5.

Effect of keratin mutation on formation of protein carbonyls. Livers from the indicated mouse genotypes were homogenized followed by separation of proteins using two-dimensional gels (isoelectric focusing in the first [horizontal] dimension, then SDS-PAGE in the second [vertical] dimension). Gels were run in duplicate: one was stained by Coommassie blue and the second was transferred to a membrane followed by immunoblotting using anti-DNP antibody as described in Materials and Methods. The spots highlighted by arrows in panels A-C were then processed for mass spectrometry analysis and subsequent identification as albumin. Arrowheads in panels E and F represent a DNP-stained product that was not identified because it was not clearly visualized with Coomassie staining. DNP, dinitrophenyl; IEF, isoelectric focusing.


We used gene expression profiling to identify a modest number of genes that become altered in transgenic mouse liver as a consequence of the K18 Arg89→Cys mutation. Several K18 mutations (including deletions) have been identified at a much higher frequency in patients with end-stage liver disease compared with controls (Ku et al.18 and Ku et al., unpublished observations, November 2004). Although the K18 R89C mutation has not yet been observed in humans because it is likely to be embryolethal,5, 15 its pathological phenotype in mice is subtle unless the animal is exposed to hepatotoxic injury. As such, exposure to the hepatotoxins griseofulvin, acetaminophen, phosphatases inhibitors, or FAS antibody renders the R89C mice markedly more susceptible to liver injury compared with nontransgenic mice or the TG2 mice that overexpress WT K18.5, 21 In this study, we used the R89C mice as a model to study potential mechanisms that may be involved in keratin mutation–induced predisposition to liver injury, including the biological setting of keratin filament disruption.

The only defined mechanism that appears to play a clear role in the predisposition to mouse liver injury in the contexts of keratin mutation (K18 R89C) or absence (K8-null mice) is the increased predisposition to undergo apoptosis.19–22, 36 This increased predisposition to undergo apoptosis may be related to alteration in the c-Flip and Erk1/2 signaling pathway,22 or to changes in polarity and targeting of proteins to their normal compartments.37, 38 Given the established predisposition of R89C mice to various forms of oxidative and nonoxidative liver injury, the findings herein add a new likely mechanism as to how keratin mutations may predispose a liver to injury. Hence, we posit that keratin mutations may lower the threshold to undergo oxidative injury. Our results show that SAMe and SAH levels are consistently depressed (15%-22%) in basal “nonchallenging” conditions in livers harboring mutant K18, coupled with an increase in cysteine sulfinic acid decarboxylase. In the liver, cysteine is derived from the diet, protein breakdown, cysteine and cystine transport, and methionine via transsulfuration (Fig. 4). The ability of hepatocytes to convert methionine to cysteine is relevant because the liver serves as the major storage site for GSH and as a reservoir for cysteine, and the intracellular cysteine level critically influences the rate of GSH biosynthesis.39–41 Therefore, oxidative stress conditions, coupled with possible shunting of cysteine and decreased cysteine precursors in the setting of K8/K18 mutation, may place the liver at a higher risk for oxidative injury. The proposed priming mechanism toward oxidative injury, via keratin mutation, may also be potentiated by the observed increase in cytochrome P450 (Table 2), and is also supported by the accumulation of lipid (malondialdehyde) and protein (albumin) oxidation products. Accumulation of oxidation products is likely related to spikes of oxidative stress, and such accumulation is noted in patients with hepatitis C who consume alcohol42 and in brains of patients with Alzheimer's disease or Parkinson's disease.35 The normal GSH and cysteine levels (Table 3) in R89C mice is likely related to adequate compensation via diet and other contributions (Fig. 4) under nonstress conditions.

In conclusion, the mechanism through which keratin mutations alter the transcription of a variety of genes that increase susceptibility to oxidative injury may involve modulation of the keratin cytoskeleton interaction with, and regulation of, several signaling cascades.3, 43 For example, K18 binds to 14-3-3 proteins in a keratin phosphorylation-dependent fashion,44 and K8 binds to Raf in a Raf phosphorylation-dependent manner.45 In addition, absence of K8/K18 proteins as a consequence of K8 gene ablation in mice results in alterations in cell cycle regulation.46 Also, cell lines expressing epidermal keratin mutants that cause epidermolysis bullosa simplex manifest an accentuated stress-activated protein kinase response compared with cells that express their WT counterparts.47 Although the details of how keratin mutations influence gene transcription has yet to be investigated, apoptosis is generally a downstream consequence of oxidative injury.48, 49 The evidence of oxidative injury in the presence of a keratin mutation, even at baseline conditions, may be a trigger to apoptosis and as such may also contribute to the observed increased susceptibility to undergo apoptosis.


We are very grateful to Dr. Che-hong Chen for his assistance with the two-dimensional gel analysis, Drs. Sailaja Elchuri and Ting Ting Huang for advice regarding markers of oxidative stress, and Kris Morrow for expert figure preparation.