Keratins modulate the shape and function of hepatocyte mitochondria: a mechanism for protection from apoptosis

Intermediate filaments (IFs), actin-containing microfilaments and microtubules comprise the three major cytoskeleton protein families in most mammalian cells (Fuchs and Cleveland, 1998). Compared with the limited number of actins and tubulins, IFs are encoded by nearly 70 unique genes and are divided into five major groups (I-V) on the basis of structural characteristics (Herrmann et al., 2007). IF gene mutations are associated with a wide spectrum of diseases that reflect their tissue-selective expression (Omary et al., 2004). The largest subgroup of IFs are keratins, which are selectively expressed in epithelial cells (Coulombe and Omary, 2002), and include keratins (K) type I (K9-K28) and type II (K1-K8 and K71-80) (Schweizer et al., 2006). Keratins carry out a variety of structural and non-structural functions, with the most prominent being protection of epithelial cells from mechanical and non-mechanical stresses (Coulombe and Omary, 2002; Magin et al., 2007; Omary et al., 2009).

In simple-type epithelia, as found in the liver, pancreas and intestine, K8 and K18 are the major keratins with additional variable levels of K7, K19, K20 and K23 depending on the cell type (Omary et al., 2009). Hepatocytes express only K8 and K18, which play an important cytoprotective role in the liver (Omary et al., 2009). For example, human mutations in K18 or K8 pose a threefold risk for the development of end-stage liver disease (Ku et al., 2001; Omary et al., 2009). Similarly, K8-null mice have a normal lifespan (Baribault et al., 1994) but harbor a marked susceptibility to liver injury (Caulin et al., 2000; Gilbert et al., 2001; Loranger et al., 1997; Zatloukal et al., 2000). The increased susceptibility of livers from K8-null mice, K18-null mice or mice with mutant K8 or K18 to injury is reflected by their enhanced susceptibility to undergo hepatocyte apoptosis. This might be related to redistribution of Fas cell surface receptors (Gilbert et al., 2001); decreased threshold to oxidative injury (Zhou et al., 2005); and the importance of keratins as a cytoprotective buffer by serving as a `phosphate sponge' towards stress-activated kinases (Ku and Omary, 2006).

Mitochondria play a central role in apoptotic or necrotic cell death (Green and Kroemer, 2005) and might be modulated by IFs (Capetanaki et al., 2007; Toivola et al., 2005). For example, there appears to be physical association between mitochondria and desmin in striated myocytes (Reipert et al., 1999), neurofilaments purified from bovine spinal cords (Wagner et al., 2003) or vimentin in cultured cell lines (Tang et al., 2008). The important functional link between IFs and mitochondria is supported by findings in desmin-null mice, which develop abnormally shaped mitochondria in heart and skeletal muscles, with functional changes that include decreased maximal respiration rate, lower ADP-stimulated oxygen consumption and adenine translocation coupling, and alterations in several mitochondrial proteins (Milner et al., 1996; Li et al., 1997; Fountoulakis et al., 2005; Capetanaki et al., 2007). Similarly, absence of K19, which is expressed at low levels in muscle, causes mitochondrial reorganization and myopathy (Stone et al., 2007). In addition, morphological changes of mitochondria are found in cultured cells transfected with K18 R89C (K18 in which arginine89 is exchanged for cysteine) (Kumemura et al., 2008). In order to understand the molecular mechanism for the predisposition of keratin-mutant mice to apoptosis, we carried out a proteomic approach to characterize the changes in livers isolated from K8-null and their littermate wild-type (K8-WT) control mice. Our overall findings reveal that keratin absence or mutation leads to changes in mitochondrial morphology and function. These findings indicate that keratin filaments might directly or indirectly regulate mitochondria in the liver, and provide a potential mechanism for the established predisposition to liver injury caused by K8 or K18 mutation.

We initially carried out an analysis of the mouse total liver proteome using two-dimensional differential-in-gel-electrophoresis (2D-DIGE) coupled with mass spectrometry and identified nearly 170 known major protein species (nearly 300 isoforms) from mouse liver (supplementary material Fig. S1 and Table S1). In order to probe the molecular mechanisms underlying the higher sensitivity of K8-null hepatocytes to apoptosis (Caulin et al., 2000; Gilbert et al., 2001), we compared the total liver protein expression profile of K8-null and K8-WT livers using 2D-DIGE and made protein identification assignments (supplementary material Table S2) by comparison with the proteome characterization shown in supplementary material Fig. S1. This comparison revealed changes in numerous proteins in K8-null versus wild-type livers (supplementary material Fig. S2). Several of the altered proteins were mitochondrial, which was further confirmed by 2D-DIGE analysis of mitochondria isolated from K8-WT or K8-null livers (Fig. 1A and supplementary material Table S3). Examples of the protein isoform alterations include spots 826a-d, which were identified as aldehyde dehydrogenase 4 family member A1 (ALDH4A1) (Fig. 1B). This enzyme is a mitochondrial matrix NAD-dependent dehydrogenase that is p53-inducible and has a protective role during cellular stress responses (Yoon et al., 2004). We validated the charged isoform changes in ALDH4A1 that were observed by 2D-DIGE independently using freshly isolated liver homogenates from K8-WT and K8-null livers (Fig. 1C).

The alterations in hepatic mitochondrial protein profiles led us to focus on this organelle. Ultrastructural analysis of mitochondria in situ using transmission electron microscopy demonstrated that absence of K8 results in significant reduction of hepatic mitochondrial size compared with that of wild-type mitochondria (Fig. 2A-D). Consistent with the in situ findings, isolated mitochondria (which provide a global but in vitro assessment) also showed a significantly smaller mitochondrial size in K8-null than in the wild-type livers (Fig. 2E,F). Notably, these changes in mitochondrial size were also observed in livers of mice that overexpress K18 R89C or that overexpress K8 G61C (supplementary material Fig. S3A). The K18 R89C mutation results in disruption of keratin filaments, hepatocyte fragility and predisposes the liver to injury and apoptosis (Ku et al., 1995; Ku et al., 2003). The K8 G61C is a natural K8 human variant that predisposes its carriers to liver disease progression (Ku et al., 2001; Omary et al., 2009), and when expressed in mice it predisposes hepatocytes to apoptosis (Ku and Omary, 2006).

In addition to the decrease in mitochondrial size in the livers of K8-null, K8-mutant or K18-mutant mice, immunofluorescence staining of mouse livers showed an irregular intracellular distribution pattern of mitochondria in the livers of K8-null versus K8-WT mice (Fig. 3). The major features of the mitochondrial alterations included clumping and cortical redistribution, with loss of the broad cytoplasmic distribution. Similar changes in the distribution of mitochondria were also noted in the K18 R89C livers but less so in the livers that overexpress K8 G61C (supplementary material Fig S3B), which is consistent with the less dramatic ultrastructural changes in mitochondria from these livers (supplementary material Fig. S3A). These findings suggest a direct or indirect regulation of mitochondria in hepatocytes by the keratin cytoskeleton.

We examined the function of K8-null mitochondria using several methods. First, we isolated mitochondria from mouse livers and measured their cytochrome c content under basal conditions, given that release of cytochrome c from mitochondria is a particularly important event in the induction of apoptosis. As shown in Fig. 4A, there was less cytochrome c in the mitochondria of K8-null mice (lanes 7 and 8) than in K8-WT mice (lanes 5 and 6), whereas total cytochrome c was similar in the two genotypes (lanes 1-4). In addition, cytochrome c mRNA levels in livers of K8-WT and K8-null mice were similar when measured using quantitative PCR (three livers from each mouse genotype were tested, not shown). Although no obvious differences were noted in mitochondrial fragility during the isolation process per se, these findings suggest that K8-null mitochondria might be more leaky and sensitive to the isolation process, which then leads to release of cytochrome c during preparation of mitochondria from the K8-null livers.

Release of cytochrome c from mitochondria occurs via Ca²⁺-independent and Ca²⁺-dependent mechanisms (Gogvadze et al., 2001). Because the buffer we used for isolation of the mitochondria contained EGTA, this suggests that a Ca²⁺-independent mechanism (e.g. via Bax) might be involved in the release of cytochrome c in K8-null mitochondria. However, it is also possible that cytochrome c simply leaks out from mitochondria due to mechanical injury during the preparation process.

In order to assess Ca²⁺-induced release of cytochrome c, we tested the effect of Ca²⁺ alone or in combination with tertiary-butyl hydroperoxide (t-BHP), an oxidative stress-inducer. As shown in Fig. 4B, Ca²⁺ alone induced more release of cytochrome c from mitochondria of K8-null than from wild-type (lanes 2 and 6) livers, whereas no release of cytochrome c was observed in K8-WT and K8-null controls (lanes 1 and 5). The Ca²⁺-induced release of cytochrome c was dramatically enhanced when combined with t-BHP (Fig 4B; lane 4 versus 2, or lane 8 versus 6), especially in K8-null mitochondria (lane 8). t-BHP alone had no effect on release of cytochrome c (Fig. 4B; lanes 3 and 7). These results indicate that mitochondria of K8-null livers are more susceptible to Ca²⁺-induced release of cytochrome c than those of K8-WT livers.

The mitochondrial inner membrane is normally impermeable to all but a few selected metabolites and ions. Mitochondrial permeability transition (MPT) represents a non-selective increase in the permeability of the inner mitochondrial membrane that allows free passage of molecules up to 1.5 kD in size (Crompton, 1999). To understand whether absence of keratins affects the MPT, we utilized Ca²⁺-induced mitochondrial swelling as a model because Ca²⁺ is a fundamental activator of the MPT pore opening and induction of MPT is readily detected by large-amplitude mitochondrial swelling. As shown in Fig. 4C, addition of Ca²⁺ decreased the 540 nm optical density for both K8-null and K8-WT mitochondria as compared to their controls. Importantly, the Ca²⁺-induced MPT was more rapid in K8-null versus K8-WT mitochondria, thereby indicating that keratins might play a role in prevention of MPT pore opening. Notably, a portion of keratins remained associated with the isolated mitochondrial fractions of mouse livers (Fig. 4A; lanes 5 and 6); but it is not known whether hepatic mitochondria associate with K8 and K18 directly or indirectly. At the ultrastructural level, the association of well-formed keratin filaments with mitochondria is not prominent (not shown).

Mitochondria produce the majority of cellular ATP. Given the altered morphology and function of K8-null mitochondria, we measured the ATP content of mouse liver tissues. As shown in Fig. 4D, ATP content in K8-null liver is ∼15% less than in K8-WT livers (P=0.003; Student's t-test), which correlates with the proteomic findings of up-regulated α/β-ATPase or the 75-kD subunit of complex I in K8-null mitochondria as a compensatory mechanism.

In conclusion, the decreased size of mitochondria that is observed in vivo as a consequence of the absence or disruption of keratin filaments suggests several potential underlying mechanisms to explain the findings. One possibility is that the mechanical function of keratins in maintaining cell integrity also extends to a role in maintaining the integrity of subcellular organelles (Toivola et al., 2005). Alternatively, mitochondrial fission or fusion (Berman et al., 2008) might involve keratins and possibly other IFs, although it is not known whether any of the proteins that are involved in mitochondrial morphology (Berman et al., 2008; Suen et al., 2008) associate with keratins. Although mitochondrial fission and fusion are continuous and dynamic processes in cells, mitochondrial fission is more prominent during apoptosis and therefore whether limited fusion contributes to the predisposition to apoptosis in hepatocytes that harbor keratin mutations is unclear. However, the findings herein demonstrate that the significant decrease in mitochondrial size correlates with a decrease in mitochondrial cytochrome c content, an increase in mitochondrial permeability and a heightened response to Ca²⁺-coupled oxidative injury. The observed changes in the mitochondrial proteins might also represent a metabolic stress risk that becomes exacerbated with exogenous stress. These findings provide an additional physiological explanation as to why keratin mutation predisposes hepatocytes to apoptotic injury.

It is important to highlight that the mechanism of hepatocyte susceptibility to apoptosis with keratin mutation is highly dependent on the nature of the mutation. For example, hepatocytes are remarkably fragile when they lack K8 (Loranger et al., 1997) or K18 (Ku and Omary, 2006), or have the K18 R89C mutation that causes disruption of keratin filaments (Ku et al., 1995). We tested two of these conditions (the absence of K8 and the K18 R89C mutation) and showed that they are accompanied by significant changes in mitochondrial size and organization. By contrast, K8 G61C mutation does not cause cell fragility and is associated with less dramatic changes in hepatocyte mitochondria. Yet, the K8 G61C mutation renders hepatocytes exquisitely sensitive to apoptotic stimuli via a mechanism that seems to be related, at least in part, to shunting of keratin phosphorylation to other cellular non-keratin substrates that in turn promote apoptosis (Ku and Omary, 2006).

Hepatic mitochondria were isolated using established methods (Ott et al., 2002). Mice (K8-null or their K8-WT littermates generated by interbreeding K8^+/– mice) were sacrificed by CO₂ inhalation and their livers removed. The livers were minced (4°C), followed by homogenization (teflon pestle) in ice-cold MSH buffer (210 mM mannitol, 70 mM sucrose, 5 mM HEPES, pH 7.4) with 1 mM EGTA. The homogenates were pelleted to remove the nuclear fraction (10 minutes; 680 g; 4°C) and the supernatant transferred to a pre-chilled tube and re-centrifuged (15 minutes; 6800 g; 4°C). The pellet was washed once in ice-cold MSH buffer (without EGTA) and used as the mitochondrial-enriched fraction. The resultant supernatants were further centrifuged at 100,000 g for 30 minutes to generate a cytosolic fraction.

Mouse liver tissues or freshly isolated mitochondria were fixed with 2% glutaraldehyde in 0.1 M sodium phosphate buffer (pH 7.2) (1 hour; 4°C). Samples were visualized and photographed using a Philips CM-12 transmission electron microscope. Mitochondrial size was analyzed with the NIH ImageJ program. Statistical comparisons were done using Student's t-test. Significance is defined as P<0.05.

Proteins were separated by SDS-PAGE then transferred to membranes followed by immunoblotting. The antibodies used were directed to K8 (M20) (LabVision) and cytochrome c (Santa Cruz Biotechnology). Blotted proteins were visualized by enhanced chemiluminescence. For tissue staining, livers were frozen and sectioned for subsequent immunofluorescence staining after acetone fixation (Ku et al., 2004).

Whole liver lysates or mitochondrial fractions of wild-type and K8-null livers were pre-labeled in vitro with Cy3 (green) and Cy5 (red), respectively, then mixed in equal fractions. Equal mixed samples were separated by isoelectric focusing (first dimension) then SDS-PAGE (second dimension). Gels were scanned using a Typhoon imager, and protein spots analyzed (DeCyder software) by comparison with the whole liver proteome characterization (supplementary material Fig. S1 and Table S1). 2D gel analysis was carried out as described (Ku et al., 2004).

Freshly isolated mitochondria (from K8-WT or K8-null livers) were washed and resuspended in MSH buffer. Equal amounts of mitochondria (2 mg protein/ml) were incubated (25°C) with control, CaCl₂ (Ca²⁺), t-BHP or Ca²⁺ + t-BHP for 10 minutes (0.2 mM final concentration for both Ca²⁺ and t-BHP). Small aliquots were removed from each tube to determine total cytochrome c, followed by pelleting the remaining mitochondria (4°C; 16,000 g; 5 minutes) and collecting the supernatants to determine the released cytochrome c. To assess MPT, swelling of mitochondria (1 mg protein/ml) was monitored in MSH buffer containing 2 mM Tris-phosphate, 5 mM succinate and 1 μM rotenone by continuous measurement of the decrease in absorbance (540 nm, 37°C) using a SpectraMax 340 96-well reader (Molecular Devices) (Choo et al., 2004; Zhao et al., 2004). A concentration of 4 nmol Ca²⁺/mg protein was used in the measurement. All presented data are representative of three independent experiments.

Samples were prepared as described previously (Yang et al., 2002) with slight modification. Equal weights of liver pieces from K8-WT and K8-null mice were rapidly processed for boiling in double-distilled water (10 minutes) followed by homogenization. ATP concentration was measured by the ATPlite luminescence ATP detection assay (PerkinElmer).