Potential conflicts of interest: Nothing to report.
The intermediate filament cytoskeleton of hepatocytes is composed of keratin (K) 8 and K18 and has important mechanical and nonmechanical functions. However, the potential role of the K8/K18 network for proper membrane targeting of hepatocellular adenosine triphosphate–binding cassette transporters and bile formation is unknown. We therefore designed a comparative study in K8 and K18 knockout mice and respective wild-type controls to test the hypothesis that intermediate filaments of hepatocytes play a role in normal bile formation. In addition, we challenged mice either with a 1% cholic acid–supplemented diet or a diet containing the porphyrinogenic xenobiotic 3,5-diethoxycarbonyl-1,4-dihydrocollidine to determine the effect of K8/K18 loss on bile flow/composition and liver injury under different physiological and toxic stress stimuli. Protein expression levels and membrane localization of various transporters and anion exchangers were compared using western blotting and immunofluorescence microscopy, respectively, and bile flow and composition were determined under various experimental conditions. Our results demonstrate that loss of the intermediate filament network had no significant effect on bile formation and composition, as well as expression levels and membrane targeting of key hepatobiliary transporters under baseline and stress conditions. However, loss of K8 significantly increased liver injury in response to toxic stress. Conclusion: The intermediate filament network of hepatocytes is not specifically required for proper bile formation in mice. (HEPATOLOGY 2009.)
In differentiated hepatocytes, keratin (K) 8 and K18 assemble in equimolar ratios to form the intermediate filament (IF) cytoskeleton.1 It is now well established that the IF cytoskeleton plays a major role in protection from mechanical and nonmechanical forms of stress.2, 3 In addition, keratin IFs were identified as major cellular structures that are affected in different liver diseases, including alcoholic steatohepatitis, nonalcoholic steatohepatitis, copper toxicosis, and cholestatic liver diseases (for review, see Zatloukal et al.3). However, studies on the role of K variants/mutations possibly predisposing individuals to liver disease (such as cryptogenic cirrhosis) have revealed conflicting results.4–7
Based on findings in knockout mice, there is an increasing body of evidence that K8 and K18 provide resistance to various forms of liver injury.8–11 These findings are primarily thought to reflect a modulator function of Ks on the membrane targeting of different receptors engaged in apoptosis and interactions of Ks in the regulation of several signaling cascades.8, 12 Ks are also involved in the execution of apoptosis because they are targets of caspases,13 and resulting K network alterations can be used to detect this type of cell death in liver.14, 15
The secretion of bile depends on the proper function and localization of membrane transport systems of hepatocytes and cholangiocytes and on the structural and functional integrity of the bile secretory apparatus.16 Human cholestatic liver diseases and corresponding animal models are frequently associated with or even caused by altered expression and localization of hepatocyte transport systems, but also profound changes in the cytoskeleton of the hepatocytes, including increases in IFs.16–18 We have shown that in mice bile duct ligation and feeding of potentially toxic cholic acid (CA) results in overexpression and hyperphosphorylation of Ks, demonstrating that Ks are targets for bile acid toxicity.19 Comparable findings where obtained in primary biliary cirrhosis representing a prototypic cholestatic liver disease.20 In addition, the keratin network of hepatocytes may be essential for regular cellular trafficking processes of different proteins, including transporters such as the cystic fibrosis transmembrane reglator and the cell junction protein desmoplakin.21, 22 This led to the speculation that increases in hepatic K8/K18 expression may represent a protective response of cholestatic hepatocytes against accumulating potentially toxic biliary constituents (such as bile acids), increases in biliary pressure, or facilitating alternative transporter expression. However, the significance of K8/K18 overexpression in cholestasis remains unknown. Previous studies have reported a 20% reduction of bile flow in K8 mutant mice,23 indicating a potential role of the hepatocyte keratin network for bile formation, such as proper cell polarity or membrane targeting of adenosine triphosphate–binding cassette (ABC) transporters. We therefore compared bile secretion and composition of K8 and K18 knockout mice with corresponding wild-type (WT) controls to determine the impact of hepatocyte K IF loss on ABC transporter targeting and bile formation. In addition, we challenged these mice either with a CA-supplemented diet or a diet containing the porphyrinogenic xenobiotic 3,5-diethoxycarbonyl-1,4-dihydrocollidine (DDC) to determine the impact of hepatocytic IF loss on bile formation under toxic stress.
ABC, adenosine triphosphate–binding cassette; ANOVA, analysis of variance; Bsep, bile salt export pump; CA, cholic acid; DDC, 3,5-diethoxycarbonyl-1,4-dihydrocollidine; IF, intermediate filament; K, keratin, Mrp, multidrug-related protein; WT, wild-type.
Materials and Methods
Homozygous K8 (K8−/−) and K18 (K18−/−) knockout mice were bred in the 129 SV Ola strain, and genotypes of offsprings were determined by way of polymerase chain reaction as described.24 Two-month-old wild-type K8−/− and K18−/− mice (25-30 g) were fed a standard diet (Sniff, Soest, Germany) and water ad libitum. In addition, mice were either fed a 1% CA-supplemented diet for 1 week to determine the impact of IF network loss on bile formation in response to bile acid challenge or a 0.1% DDC-supplemented diet to mimic toxic stress. The experimental protocols were approved by the local Animal Care and Use Committee according to criteria outlined in the Guide for the Care and Use of Laboratory Animals prepared by the National Academy of Sciences and published by the National Institutes of Health (NIH publication 86-23, revised 1985).
Serum biochemical analysis and determination of bile flow and composition were performed as described.25, 26
Preparation of liver membranes and analysis of transporter protein levels by way of western blotting and immunofluorescence microscopy for transport proteins were performed as described.25, 27 Antibodies used for the determination of several protein levels are listed in Supporting Table 1. Western blotting for transport proteins was performed in triplicate for each experimental situation.
Western blotting for K8 and K18 was perfomed as described19 using specific anti-K8 (Clone Ks 8.7, dilution 1:1,000; Progen, Heidelberg, Germany) and anti-K18 (Clone Ks 18.04, dilution 1:1,000; Progen) antibodies.
Immunofluorescence microscopy for K8 and K18 was performed as described19 using the described specific antibodies (dilution 1:100).
Morphometric analysis of canalicular bile salt export pump (Bsep), multidrug-related protein 2 (Mrp2), and ZO-1 localization was performed using the profile tool of the LSM 510 (Zeiss, Oberkochen, Germany) to measure the intensity of the signal distribution reflecting the protein concentration across the canalicular region and in submembranous regions. To compare the canalicular localization of Bsep and Mrp2 between different genotypes and experimental situations, only longitudinally cut canaliculi (having one ZO-1 band on each side) were analyzed to avoid misinterpretation in regard to proper canalicuar targeting.
Data are reported as the arithmetic means ± standard deviation of three to 15 animals in each group. For comparison between controls versus DDC-fed animals versus CA-fed animals, statistical analysis was performed using analysis of variance (ANOVA) with Bonferroni posttesting or ANOVA on ranks with Dunn's posttesting if a normality test or equal variance test failed (using the Sigmastat statistics; Jandel Scientific, San Rafael, CA). For comparison between WT versus K8−/− and WT versus K18−/− mice, an unpaired t test or Mann-Whitney rank sum test was performed, the latter if an equal variance test failed.
Loss of the Hepatocellular IF Network Has No Impact on Bile Secretion and Composition Under Physiological Conditions.
For characterization of the models used, we performed western blotting and immunofluorescence microscopy using specific antibodies against K8 and K18 in both knockout strains showing the loss of the hepatocyte IF network in both strains (Supporting Fig. 1). In addition, deletion of K8 resulted in disappearance of the specific K18 protein band (probably due to degradation of the K18 protein24), whereas K8 protein expression was preserved in K18−/− mice. Because previous data have pointed to an at least indirect role of K8/K18 in bile formation,23, 28 we determined bile flow and composition in K8−/− and K18−/− mice and compared these with respective WT controls. As outlined in Table 1, we observed no differences in bile flow as well as biliary output of bile acids, phospholipids, cholesterol, and glutathione. These findings clearly demonstrate that loss of K8/K18 has no major effect on normal bile secretion and composition in mice under baseline conditions.
Table 1. Bile Flow and Composition in K8 and K18 Knockout Mice and WT Controls Under Various Experimental Conditions
WT (n = 6)
K8−/− (n = 4)
WT (n = 7)
K8−/− (n = 6)
WT (n = 5)
K8−/− (n = 3)
P < 0.05, control versus DDC versus CA using ANOVA with Bonferroni posttesting or ANOVA on ranks with Dunn's posttesting if normality test or equal variance test failed.
P < 0.05, WT versus K8−/−/18−/− using unpaired t test or Mann-Whitney rank sum test if equal variance test failed.
Loss of the Hepatocellular IF Network Has No Effect on Protein Expression Levels and Proper Localization of Canalicular ABC Transporters.
We compared protein levels and localization of several transport proteins, especially of the two major canalicular export systems for bile acids and organic anions Bsep and Mrp2, respectively, to determine whether the loss of the K IF network has an impact on their expression and proper membrane targeting. However, all tested transport proteins showed no significant differences in their expression levels between different genotypes under baseline conditions (Figs. 1 and 2; Supporting Figs. 2 and 3). Because bile acid excretion by Bsep and glutathione excretion by Mrp2 represent the major driving forces for canalicular bile formation, we compared Bsep and Mrp2 localization as revealed by their immunofluorescence pattern. Again, we observed no differences in the canalicular staining pattern for Bsep (Fig. 3) and Mrp2 (Fig. 4) between knockout mice and WT controls. Regular canalicular Bsep and Mrp2 targeting was also confirmed by morphometric analysis showing proper canalicular localization of these transporters in all chow-fed genotypes (Supporting Figs. 4 and 5). These findings are in line with the observed unchanged bile acid–dependent and bile acid–independent bile formation in these animals under baseline conditions (Table 1). In addition, we observed a comparable staining pattern for basolateral Mrp3 in all genotypes tested (Supporting Fig. 6). From those experiments, it can be concluded that loss of the K IF network does not result in impaired bile formation through defects in the expression or targeting of major hepatocellular transport systems.
CA Feeding Induces Bile Secretion Independent of K8/K18 Loss.
Because the liver is able to adapt to physiological and pathological bile acid challenge through up-regulation of canalicular ABC transporter expression resulting in increased biliary bile acid output in response to CA feeding,25, 29 we next tested the hypothesis that the K8/K18 hepatocyte network is essential for this compensatory mechanism. Choleresis was stimulated by CA feeding; however, we observed no significant differences in bile flow and composition (Table 1) between K8−/− and K18−/− mice and their respective WT controls. In line with these results, no differences in regard to protein expression levels (Figs. 1 and 2) and canalicular localization patterns of Bsep and Mrp2 were observed (Figs. 5 and 6). This was again confirmed by morphometric analysis showing proper canalicular targeting of these transporters in response to CA feeding in all genotypes (Supporting Figs. 4 and 5). However, compensatory overexpression of Mrp3 was significantly less pronounced in both knockout strains tested (Figs. 1 and 2). These quantitative differences were not detectable on immunofluorescence microscopy showing a similar pattern of increased basolateral Mrp3 staining in all CA-fed genotypes, suggesting no qualitative differences in basolateral targeting of this transport protein (Supporting Fig. 6). In addition, we observed no significant differences in serum alanine aminotransferase, alkaline phosphatase, and bilirubin levels between genotypes in response to CA feeding (Table 2), indicating equal susceptibility to potentially toxic bile acids. These data therefore demonstrate that the keratin IF network is not necessary to target increased amounts of the ABC transporters to the canalicular and basolateral membrane under conditions of increased bile acid load. The fact that K8 and K18 knockout mice showed no increased sensitivity to CA feeding precludes a pivotal direct role of Ks in defense against toxic bile acids.
Table 2. Serum Liver Enzyme and Serum Bile Acid Levels in K8 and K18 Knockout Mice and WT Controls Under Various Experimental Conditions
DDC Feeding Leads to Significantly Reduced Bile Flow in K8−/− and K18−/− Mice Independent of Transport Protein Alterations.
We next determined bile flow and composition as well as transporter expression and localization in DDC-fed knockout mice to explore whether these mice adapt differently in response to this xenobiotic. In line with previous data obtained in a different mouse strain,30 DDC did not affect bile flow after 7 days of feeding despite reduction in biliary glutathione levels in WT mice (Table 1). However, in K8 and K18 knockout mice, DDC feeding significantly reduced bile flow (Table 1). Further exploration of transporter expression revealed similar findings in knockout and WT animals with significant down-regulation of basolateral Ntcp and Oatp-1 as well as canalicular Bsep and Mrp2 expression (Supporting Figs. 7 and 8). In addition, double label immunofluorescence microscopy demonstrated severely alterated ZO-1 staining pattern indicating dilatation and partial disruption of canliculi and showed comparable results regarding membrane targeting of canalicular Bsep and Mrp2 (Supporting Figs. 9 and 10). Again, we found no differences in regard to the canalicular targeting of Mrp2 and Bsep comparing morphometric analysis in DDC-fed genotypes (Supporting Figs 4 and 5). K8−/− mice showed elevated serum alanine aminotransferase levels and both knockout strains significantly elevated serum bilirubin levels indicating an, at least in part, increased susceptibility of the knockout mice in response to DDC poisoning (Table 2). Taken together, these findings suggest that significantly reduced bile flow in K8−/− and K18−/− mice in response to DDC feeding may reflect an increased susceptibility of these animals to DDC rather than transporter-mediated alterations in canalicular bile formation. However, we cannot rule out differences in the contribution of ductules to bile formation in response to DDC feeding.
Previous findings of reduced bile flow in K8−/− mice pointed toward a potential role of the IF cytoskeleton in bile formation.23 In addition, the recent finding of intestinal anion exchanger 1/2 mistargeting in K8−/− mice with consecutive decreased Na+ net absorption and resulting diarrhea31 led us to the hypothesis that the K8/K18 network of hepatocytes may be engaged in physiological bile formation via a direct or indirect role in proper ABC transporter targeting. The current studies allow the conclusions that the hepatocyte K8/K18 network is (1) dispensable for proper bile secretion and composition under baseline conditions, (2) has no impact on compensatory overexpression and/or targeting of the tested canalicular and basolateral ABC transporters in response to CA feeding, (3) is not involved in defense against bile acid–induced toxicity, and (4) is not directly related to reduced bile flow in DDC-induced reduction of bile flow in respective knockout mice. Therefore, the findings of the current study make it unlikely that the IF cytoskeleton of hepatocytes participates in membrane targeting processes of the transporter proteins tested herein. However, loss of K8 increased liver injury in response to toxic stress.
We compared the effects of the porpyhrinogenic substance DDC in different genotypes, because this represents a well-characterized mouse model for a xenobiotic-induced cholestatic liver disease with features of steatohepatitis associated with alterations of transporter expression and consequently bile formation.30 After 7 days of DDC feeding, basolateral Ntcp and Oatp1 as well as canalicular Bsep and Mrp2 were significantly down-regulated to a comparable degree in both types of knockout animals and their corresponding WT controls. Tight junction morphology determined by ZO-1 staining was severely altered in all genotypes with expansion, elongation, and partial disruption, suggesting dilatation and distortion of bile canaliculi comparable to findings in common bile duct–ligated mice.14 This may also reflect an obstructive component of DDC-induced cholestasis via porphyrin plugs.30 However, costaining with antibodies against Bsep as well as Mrp2 revealed no structural differences between DDC-fed K8 and K18 knockout animals and WT controls. These results were contrasted by significantly reduced bile flow in DDC-fed keratin knockout mice, which cannot be explained by differences in hepatocellular transporter expression, as well as localization as shown herein. However, these findings cannot exclude differences in ductular bile formation in response to DDC feeding. It is more likely that K8 and K18 knockout mice may have increased susceptibility to DDC feeding through an undetermined mechanism as indicated by significantly serum bilirubin levels.
The liver may neutralize an increased bile acid load via induction of canalicular and alternative basolateral transporter expression (for example, through Mrp3, 4, Ostα/β).32 We and others have demonstrated this compensatory mechanism in CA-fed mice33, 34 and additionally in human cholestatic liver diseases underlining the general importance of this concept.35 Because it was not clear whether the intermediate filament network of hepatocytes is necessary for this compensatory process (for example, through a role of K8/K18 in transporter targeting to the canalicular membrane), we compared K8 and K18 knockout mice and controls in regard to their response to CA feeding. The data presented above clearly demonstrate that the K IF cytosekeleton is not necessary to target increased amounts of the tested ABC transporters to the canalicular and basolateral membrane under the situation of increased bile acid load in mice. In addition, we observed no significant differences in serum liver enzymes between the different genotypes, indicating equal susceptibility to bile acid feeding and arguing against a central role of K8 and K18 in defense against bile acids. The lack of Mrp2 protein induction in response to CA feeding in the 129 SV Ola strain contrasts with previous findings in Swiss albino and C57/BL6 mice33, 34 and may primarily be attributed to strain differences in bile acid handling. Previous in vitro findings of decreased horseradish peroxidase secretion in NiCl2-treated primary hepatocyte cultures, together with disturbances in the IF cytoskeleton structure,28 may therefore be at least partly related to cytotoxic effects of NiCl2 rather than representing alterations in a hypothetical IF cytoskeleton–membrane transport protein complex.
In conclusion, the findings of this study reject a pivotal role for the hepatocyte IF cytoskeleton in bile formation in mice under physiological and toxic stress situations, at least on a hepatocellular level. Consequently, previously observed alterations of this cell structure in cholestatic liver diseases may represent consequences rather than primary causes.
We gratefully acknowledge W. Erwa and colleagues for performing biochemical analysis of serum liver tests. The excellent technical assistance of Judith Gumhold is also gratefully acknowledged.