Hepatocyte-specific Gclc deletion leads to rapid onset of steatosis with mitochondrial injury and liver failure


  • Potential conflict of interest: Nothing to report.


Oxidative stress is considered to be a critical mediator in liver injury of various etiologies. Depletion of glutathione (GSH), the major antioxidant in liver, has been associated with numerous liver diseases. To explore the specific role of hepatic GSH in vivo, we targeted Gclc, a gene essential for GSH synthesis, so that it was flanked by loxP sites and used the albumin-cyclization recombination (Alb-Cre) transgene to disrupt the Gclc gene specifically in hepatocytes. Deletion within the Gclc gene neared completion by postnatal day (PND)14, and loss of GCLC protein was complete by PND21. Cellular GSH was progressively depleted between PND14 and PND28—although loss of mitochondrial GSH was less severe. Nevertheless, ultrastructural examination of liver revealed dramatic changes in mitochondrial morphology; these alterations were accompanied by striking decreases in mitochondrial function in vitro, cellular ATP, and a marked increase in lipid peroxidation. Plasma liver biochemistry tests from these mice were consistent with progressive severe parenchymal damage. Starting at PND21, livers from hepatocyte-specific Gclc knockout [Gclc(h/h)] mice showed histological features of hepatic steatosis; this included inflammation and hepatocyte death, which progressed in severity such that mice died at approximately 1 month of age due to complications from liver failure. Conclusion: GSH is essential for hepatic function and loss of hepatocyte GSH synthesis leads to steatosis with mitochondrial injury and hepatic failure. (HEPATOLOGY 2007.)

Glutathione (GSH), the most abundant intracellular thiol, has several ascribed biological functions, most notably electrophile detoxication and oxidant elimination.1 GSH is synthesized within cells; cellular uptake of intact GSH does not occur or is a very inefficient process.2, 3 GSH is synthesized from its precursor amino acids by 2 consecutive ATP-consuming enzymatic reactions. The first reaction is catalyzed by glutamate cysteine ligase (GCL), the rate-limiting and regulatory enzyme in GSH synthesis and the second by GSH synthetase. In mammalian cells, both enzymes are believed to be cytosolic.4 Eukaryotic GCL is a heterodimer composed of a catalytic (GCLC) and a modifier (GCLM) subunit. GCLC possesses all the catalytic activity to form γ-glutamylcysteine; GCLM optimizes the catalytic properties of the holoenzyme.5–7 The overall rate of GSH synthesis is controlled by the amount of GCL, availability of L-cysteine, and the feedback inhibition by GSH on GCL.8, 9 Mice with a targeted disruption of the Gclc gene die between gestational day 7.5 and 8.5, indicating an essential role of GCLC, and likely GSH, during embryogenesis.10, 11 However, GSH is not essential for the growth of some cultured cells.11Gclm(−/−) knockout mice, on the other hand, reveal no overt phenotype, although the knockout mice possess only 10% to 40% of normal tissue GSH levels12

Liver has the highest GSH level among tissues and hepatocytes are the main source of GSH in liver. Due to its role in metabolism and detoxication, the hepatocyte is an important target of many ingested drugs and other xenobiotics. Biotransformation may result in liver injury by way of increased reactive oxygen species (ROS) formation. For example, acetaminophen-induced hepatotoxicity is mediated by increased generation of intrahepatocyte electrophiles, consequent to its metabolic activation by cytochrome P450 2E1 (CYP2E1). The toxicity of acetaminophen is intimately intertwined with the depletion of hepatocyte GSH; however, whether GSH depletion per se is a cause of toxicity is unknown.13, 14

The function of GSH in hepatocytes has been studied mostly by chemical depletion of GSH, using buthionine sulfoximine (BSO), which is an inhibitor of GCLC. BSO treatment of newborn rats leads to some hepatic abnormalities, including mitochondrial dysfunction,15 but these results are difficult to interpret because BSO depletes GSH in extrahepatic tissues, most being more sensitive to BSO depletion than liver. Moreover, BSO has nonspecific inhibitory activities.

To better understand the function(s) of hepatocyte GSH, we have developed mice harboring a conditional Gclc allele flanked by loxP sites (floxed Gclc), which we term Gclc(f). Our strategy in designing Gclc(f) is that cyclization recombination (Cre)-mediated recombination will result in loss of Gclc function and the consequent loss of GSH. By crossing an albumin-specific albumin (Alb)-Cre transgene, which drives Cre-recombinase expression exclusively in each hepatocyte16 having a Gclc(f/f) background, we herein study the consequences of GSH loss on hepatic function.


Alb, albumin; BSO, buthionine sulfoximine; Cre, cyclization recombination; GCL, glutamate cysteine ligase; GCLC, glutamate cysteine ligase catalytic subunit; GLCM, glutamate cysteine ligase modifier subunit; GSH, glutathione; PND, postnatal day; ROS, reactive oxygen species.

Materials and Methods

Generation of Gclc Floxed Mice

Details on targeting construct, targeting procedure and the characterization of conditional Gclc knockout [Gclc(f/f)] mice can be found in the Supplementary Materials (available at the Hepatology website: http://interscience.wiley.com/jpages/0270-9139/suppmat/index.html).

Generation of Hepatocyte-specific Gclc(h/h) Knockout Mice

We purchased Alb-Cre transgenic mice [C57BL/6-TgN(AlbCre)21Mgn] from The Jackson Laboratory (Bar Harbor, ME). We generated hepatocyte-specific Gclc knockout [Gclc(h/h)] mice by intercrossing Gclc(f/f) mice with Alb-Cre transgenic mice. The offspring were of mixed (C57BL/6J and 129S6/SvJ) genetic background. We conducted all studies on littermates and approved them by the University of Cincinnati Medical Center Institutional Animal Care and Use Committee.

Southern Blot and PCR Analysis

We performed Southern blotting and PCR analysis as described12 to detect the 3 Gclc alleles. Details on Southern blotting and primers used in PCR analysis can be found in the Supplementary Materials.

Northern Blot and Western Immunoblot Analysis

We collected mouse livers at the indicated ages of the mice, flash-frozen in liquid nitrogen, and stored them at −70°C until use. We isolated total RNA using Tri-reagent (Molecular Research Center, Cincinnati, OH), as described.17 We analyzed for the presence of GCLM, heme oxygenase-1, metallothionein-1, and β-Actin as described.18 To detect the mRNA of GCLC, we used the primers ACTAGGCTGTCCTGGATTCA (sense) and TTGCCCATCCCGAATCCCA (antisense) to generate a probe by amplifying the cDNA sequences of Gclc from exon 4 to exon 6, which we would delete after Cre/LoxP recombination. To detect GCLC and GCLM proteins, we raised a chicken polyclonal antibody against mouse recombinant GCLC or GCLM. We resolved liver cytosolic fractions by 12% SDS-PAGE and blotted them with the above antibodies (1:50,000 for anti-GCLM; 1:25,000 for anti-GCLC).

Histopathological Examination, Immunohistochemistry Study, and Oil Red O Staining

Tissue preparations for immunohistochemistry, light microscopy, and electron microscopy are described in the Supplementary Materials. We conducted immunostaining for GCLC and GCLM, using a TSA Biotin System Kit (NEN Life Science Products, Boston, MA). We detected GCLC and GCLM, using chicken polyclonal antisera at dilutions of 1:50 and 1:100, respectively. For oil red O staining, we stained frozen liver sections (7 μm thick), from 5 to 6 animals for each group, with oil red O and counterstained with hematoxylin.

Plasma Analysis

We collected mouse blood by cardiac puncture, using heparinized syringes. After centrifugation, we collected plasma and immediately assayed it for ALT, AST, and bilirubin levels in the Clinical Laboratories at Cincinnati Children's Hospital Medical Center.

Measurement of Cellular ATP, GSH, and Cysteine Levels

We measured liver ATP levels, using the ATP luminescence kit (Sigma).19 We determined GSH and cysteine levels as detailed.20, 21

Measurement of Mitochondria Respiration, Energy Coupling, and Membrane Potential

We isolated liver mitochondria and measured mitochondrial oxygen consumption as described.19 We estimated mitochondrial membrane potential using 5,5′,6,6′,-tetrachloro-1,1,3,3′-tetraethylbenzimidazolylcarbocyanine iodide (JC-1; Sigma) as described.22

Lipid Peroxidation

We estimated hepatic lipid peroxidation by thiobarbituric acid-reactive substances assay using 10% liver whole homogenate, as described.23

Protein Quantification

We determined protein concentration using the Bradford method (Biorad, Hercules, CA).

Statistical Analysis

We performed statistics using SigmaStat Statistical Analysis software (SPSS Inc., Chicago, IL). We compared group means by 1-way analysis of variance, followed by Student t test for pairwise comparison of means. All data were normally distributed and are reported as mean ± SEM. P values of <0.05 were regarded as statistically significant.


Generation of Gclc Floxed Mice

To overcome the embryonic lethality of global ablation of the Gclc gene, we generated a conditional Gclc knockout mouse line containing a genetically engineered Gclc(f) allele in which exons 4 to 6 are flanked by loxP sites. Removal of these essential coding sequences also eliminates in-frame splicing between exon 3 with the remaining exons of Gclc. The Gclc(f) allele, despite carrying foreign sequences in introns 3 and 6, was functionally indistinguishable from the Gclc(+) wild-type allele. Details on the generation and characterization of Gclc floxed [Gclc(f/f)] mice can be found in the Supplementary Materials.

Generation of Hepatocyte-specific Gclc(h/h) Knockout Mice

To generate mice harboring the hepatocyte-specific Gclc(h) allele, we intercrossed Alb-Cre mice with Gclc floxed mice (Fig. 1A). Mice of all 3 possible genotypes were born with Mendelian frequency, indicating no embryonic lethality due to hepatocyte-specific deletion of Gclc (data not shown). Gclc(f/f) and Gclc(+/h) mice were phenotypically normal throughout adulthood (data not shown). Gclc(h/h) mice, although phenotypically normal through postnatal day (PND)14, thereafter displayed a phenotype of growth retardation and early lethality (Fig. 1B). We believe this is the result of the hepatocyte-specific loss of Gclc gene expression.

Figure 1.

Generation of hepatocyte-specific Gclc(h/h) knockout mice. (A) Breeding scheme for generating the Gclc(h/h) mice. (B) Growth curves of Gclc(f/f) and Gclc(h/h) littermates. Data are reported as means ± SEM from 7 to 10 mice for each group.

Timing of Hepatocyte-specific Deletion of Gclc

In Alb-Cre-positive Gclc(f/f) mice, we detected liver-specific recombination of Gclc(f) with the Gclc(h) allele by PCR (Fig. 2A). As expected, we did not detect the Gclc(h) allele in hepatic DNA from Gclc(f/f) mice lacking Alb-Cre (Fig. 2A, lane 1) or in Alb-Cre-positive mice in any tissue, except liver (Fig. 2A, lanes 7-12). In Alb-Cre-positive mice, we detected the Gclc(h) allele in hepatic DNA as early as gestational day 16 and robustly by PND7.

Figure 2.

Hepatocyte-specific recombination of Gclc in Gclc(f/f) mice. (A) PCR analysis of genomic DNA for the Alb-Cre and the Gclc(h) allele. We prepared DNA (lanes 1-6) from liver harvested at the indicated time points from Gclc(f/f) mice (lane 1), Gclc(f/f) mice containing Alb-Cre (lanes 2-6), or from the indicated tissues from Gclc(f/f) mice containing Alb-Cre harvested at PND30 (lanes 7-12). GD refers to gestational day, and PND refers to postnatal day. K, kidney; P, pancreas; H, heart; L, lung; S, spleen; T, tail. (B) Southern blot analysis of liver DNA isolated from Gclc(f/f) or Gclc(h/h) mice at different time points. The 3.6-kb and 1.2-kb bands represent the Gclc(f) and Gclc(h) allele, respectively. (C) Northern blot analysis for mRNAs of GCLC, GCLM and β-Actin. (D) Western blot analysis for GCLC, GCLM β-actin protein from livers of Gclc(f/f) or Gclc(h/h) mice at PND14, PND21, and PND28.

To quantitate the conversion of Gclc(f) to Gclc(h), we used Southern blot hybridization analysis. Hepatic conversion of Gclc(f) to Gclc(h) was 65% complete by PND7 and reached 88% by PND30 (Fig. 2B). We presume that the presence of the remaining ∼10% of undeleted Gclc(f) allele reflects the Gclc(f) allele in nonparenchymal cells, consistent with previous estimates.24 Conversion of Gclc(f) to Gclc(h) resulted in loss of detectable GCLC mRNA (Fig. 2C) and protein (Fig. 2D). Although nonparenchymal cells still possess Gclc, their expression of this gene is modest relative to hepatocyte and below the level of detection under these conditions. As shown below, GCLM mRNA was increased as a consequence of Gclc ablation, whereas cytoplasmic GCLM protein accumulation was modestly increased at PND21 and PND28.

Timing of Hepatocyte Loss of GCLC

To examine hepatocyte loss of GCLC, we performed immunohistochemistry (Fig. 3). In control hepatic sections from Gclc(f/f) mice, both GCLC (Fig. 3B) and GCLM (Fig. 3C) were localized predominantly within hepatocytes, which is consistent with the role of this cell type in robust GSH synthesis. Staining for both GCLC and GCLM in Gclc(f/f) mice was indistinguishable from that in wild-type mice and was most intense in the periportal hepatic zone (data not shown). This is consistent with previous reports.25, 26 In Gclc(h/h) mice by PND14, a time when GCLC mRNA and protein are dramatically diminished in Gclc(h/h) mice (Fig. 2), we could detect GCLC in only 15 ± 3% of the hepatocyte population (Fig. 3E). In contrast, GCLM stained in all hepatocytes (Fig. 3F). By PND21, when the recombination of Gclc(f) to form Gclc(h) was complete according to Southern blot analysis (Fig. 2B), we detected GCLC immunoreactivity only in nonparenchymal cells (Fig. 3H, K). We estimate that, by PND21, <0.1% hepatocytes still contained detectable GCLC. On the contrary, we detected GCLM in hepatocytes from Gclc(h/h) mice (Fig. 3I,L). Taken together, these data demonstrate that the Gclc(f) allele is efficiently and specifically deleted in hepatocytes by Cre recombinase and that loss of immunoreactive GCLC was completed between PND14 and PND21.

Figure 3.

Immunohistochemistry for the GCLC and GCLM protein. Liver sections free of paraffin were stained with H&E (A, B, C & D) or immunostained (brown) for GCLC (B, E, H & K) or GCLM (C, F, I & J) and counterstained with hematoxylin. (Magnification = 1000×).

During immunohistochemical analysis, we also made observations regarding the localization of GCLC and GCLM within hepatocytes, which require explanation, so that we might better interpret Fig. 3. Based on cellular fractionation, both GCLC and GCLM are cytosolic proteins—suggesting that their staining might be diffusely localized to cytoplasm. As we showed in Fig. 3B,C, however, this is not the case. Both GCLC and GCLM display similar perinuclear focal localizations in hepatocytes. In Gclc(h/h) mice at PND14, a minority of hepatocytes still contain GCLC (Fig. 3E). In these hepatocytes, GCLC is focally localized. In contrast, at PND14 virtually all hepatocytes stain positive for GCLM and the majority stain diffusely. Furthermore, GCLM is no longer excluded from nuclei (Fig. 3F, right inset). Interestingly, the same percentage of hepatocytes in PND14 liver sections show focal staining for GCLM and GCLC. We have not determined by colocalization that these are the same hepatocyte populations. If they are, these data suggest that sequences within GCLC might dictate a discrete subcellular distribution for itself and GCLM. Alternatively, GCLM may display altered subcellular localization when GSH levels are low. Discrimination between these 2 possibilities will require experimentation beyond the scope of this report.

GSH and Cysteine Levels

GCLC possesses all the catalytic activity for the rate-limiting step in GSH de novo biosynthesis. Because hepatocytes account for ∼85% of the total cell volume in the liver plus the highest capacity for GSH biosynthesis, loss of GCLC in hepatocytes presumably will result in the loss of most of the hepatic GSH. Indeed, compared with Gclc(f/f) mice, hepatic GSH levels in Gclc(h/h) mice had fallen to 37 ± 0.5% at PND14 and 4.5 ± 0.4% at PND28 (Table 1).

Table 1. Tissue Thiol Levels
  • NOTE: GSH levels are expressed as μmol/g tissue in liver, kidney, pancreas, and lung and as μM in plasma. Cysteine levels are expressed as μmol/g tissue. Values are means ± SEM from 4 or 5 mice. Numbers in parentheses are % of that in Gclc(f/f) mice.

  • *

    P < 0.05, when comparing Gclc(f/f) and Gclc(h/h) mice.

PND14(f/f)4.9 ± 0.10.9 ± 0.10.7 ± 0.10.7 ± 0.026.0 ± 2.00.24 ± 0.010.53 ± 0.020.34 ± 0.020.17 ± 0.00
 (h/h)1.8 ± 0.05*0.9 ± 0.00.7 ± 0.00.6 ± 0.05.2 ± 0.5*0.26 ± 0.010.45 ± 0.070.28 ± 0.030.16 ± 0.01
  (37%)   (20%)
PND21(f/f)6.6 ± 0.21.5 ± 0.10.9 ± 0.20.7 ± 0.033.0 ± 2.60.28 ± 0.010.55 ± 0.050.87 ± 0.130.14 ± 0.00
 (h/h)0.5 ± 0.07*1.0 ± 0.1*0.4 ± 0.0*0.5 ± 0.0*5.4 ± 0.6*0.29 ± 0.040.51 ± 0.010.66 ± 0.050.14 ± 0.01
PND28(f/f)5.3 ± 0.21.9 ± 0.10.9 ± 0.00.8 ± 0.059.0 ± 2.20.28 ± 0.071.34 ± 0.380.40 ± 0.000.19 ± 0.01
 (h/h)0.2 ± 0.05*1.2 ± 0.1*0.5 ± 0.0*0.5 ± 0.0*1.0 ± 0.1*0.26 ± 0.031.09 ± 0.250.35 ± 0.030.15 ± 0.00

The hepatocyte has been hypothesized as the cell type most important for supplying plasma GSH, and consequently important in supplying the cells of peripheral tissues with GSH and cysteine through the γ-glutamyl cycle.11, 27 With hepatic GSH depletion, Gclc(h/h) mice showed extremely low plasma GSH levels, and a modest decrease of GSH in peripheral tissues; on the other hand, the level of cysteine in any peripheral tissue assayed was not affected (Table 1).

Gclc(h/h) Mice Develop Steatosis and Die of Complications Due to Liver Failure

In plasma from Gclc(h/h) mice, AST and ALT levels were not significantly different from control at PND14 but increased dramatically and progressively by PND21 and PND28 (Table 2). We observed a similar trend for plasma levels of bilirubin. The values for plasma enzymes and bilirubin were highly elevated by PND28—suggesting considerable hepatic damage. By approximately PND30, Gclc(h/h) mice became moribund, likely due to complications of hepatic failure, and we humanely killed them.

Table 2. Liver Biochemical Functions
AgeGclcConjugated Bilirubin (mg/dL)Unconjugated Bilirubin (mg/dL)AST (IU/L)ALT (IU/L)
  • NOTE: Values are means ± SEM from 3 or 4 mice.

  • *

    P < 0.05, when comparing Gclc(f/f) and Gclc(h/h) mice.

PND14(f/f)0.1 ± 0.050.0 ± 0.0109 ± 526 ± 12
 (h/h)0.2 ± 0.030.0 ± 0.0137 ± 1065 ± 24
PND21(f/f)0.0 ± 0.00.0 ± 0.0100 ± 530 ±3
 (h/h)0.4 ± 0.02*0.0 ± 0.0681 ± 25*165 ± 10*
PND28(f/f)0.0 ± 0.00.0 ± 0.0122 ± 837 ± 2
 (h/h)9.2 ± 0.6*0.2 ± 0.0*3400 ± 261*1720 ± 124*

Examination of liver sections by light microscopy revealed multiple morphologic changes in the mutant mice. Coincident with the depletion of hepatic GSH, hepatocytes of Gclc(h/h) mice displayed abnormalities by PND21 (Fig. 4A; vacuolated hepatocytes present in focal, periportal clusters), which became progressively more severe by PND28 (Fig. 4B; vacuolated hepatocytes diffusely present in widespread patches). Hepatocyte vacuolization was of both macrovesicular and microvesicular patterns (Fig. 4D, arrow and arrowhead, respectively), whereas the latter was predominant. Histochemical staining with oil Red O identified the vacuolated material as neutral lipid, which accumulated over time (Fig. 4E-H); steatosis was present in 26 ± 7% and 69 ± 7% of total hepatocytes by PND21 and PND28, respectively. In addition to steatosis, we identified both scattered single cell death (Fig. 4C, arrow) and areas of contiguous hepatocyte death, which accounted for 2 ± 0.4% of the overall hepatocyte population by PND28. In addition, there was an increase in binucleated hepatocytes (Fig. 4C, arrowhead). These features of parenchymal damage and repair spared the bile ducts (Fig. 4C,D; asterisks), consistent with a primary hepatocyte derangement. Focally present, and lessening with the onset of diffuse steatosis, was a mixed (acute and chronic) inflammatory infiltrate (Fig. 4C, periductal region at top left).

Figure 4.

Histopathology and fat staining of liver. We stained the liver sections from Gclc(f/f) mice (g and h) and Gclc(h/h) mice (A-F) with hematoxylin and eosin (H&E) (A-D) for histologic assessment, or oil red O (E-H) for fat staining. Hepatocytes became fat laden (A and B, high magnification insets at right) in a patchy distribution (A and B; see also E and F, low magnification insets at left). Macrovesicular and microvesicular steatosis (D, arrow and arrowhead, respectively), single hepatocyte death (C, arrow), scattered binucleated hepatocyte (C, arrow head), unaffected bile ducts (C and D, asterisks) and inflammatory infiltrate (C, periductal region at top left) are indicated. Original magnifications: ×100 (A and B, left insets of E-H) and ×600 (C, D, E-H, right insets of A and B).

We further examined the changes in hepatocytes by electron microscopy. By PND21, hepatocytes from Gclc(h/h) mice started to reveal abnormal ultrastructural changes, which deleteriously progressed afterwards. By PND28, compared with Gclc(f/f) mice (Fig. 5A,C), Gclc(h/h) mice showed loss of the lamellar rough endoplasmic reticulum, which was replaced by large quantities of small vesicles that were devoid of ribosomes (Fig. 5B); approximately 20% of hepatocyte mitochondria were morphologically abnormal—revealing ring-shaped (Fig. 5D, solid arrow), vesicular (Fig. 5D, open arrow), and ballooning (Fig. 5D, asterisks) morphologies. Despite unaffected bile ducts seen by light microscopy, we observed dilated bile canaliculi having inclusions inside ultrastructurally in Gclc(h/h) mice (compare Fig. 5E,F).

Figure 5.

Electron micrographs of hepatocytes at PND28. (A) The rough endoplasmic reticulum in Gclc(f/f) mice and (B) its vesicular transformation in Gclc(h/h) mice (bars = 1.0 μm). (C) Lamellar cristae in mitochondria in Gclc(f/f) mice (bar = 0.5 μm) and (D) gigantic (*), vesicular (open arrow), and ring shape (solid arrow) mitochondria found in Gclc(h/h) mice (bar = 1.0 μm). (E) Bilecanaliculus of normal size in Gclc(f/f) mice and (F) dilated bilecanaliculus with inclusion in Gclc(h/h) mice (bar = 1.0 μm). RER, rough endoplasmic reticulum; V, cytoplasmic vesicles; M, mitochondria; BC, bilecanaliculus.

Mitochondrial Dysfunction in Livers from Gclc(h/h) Mice

Mitochondria are a major source of endogenous ROS, which are byproducts of aerobic respiration. Mitochondrial GSH is considered critical in the elimination of mitochondria-derived H2O2 and in maintaining mitochondrial redox homeostasis.28, 29 Although GSH is synthesized in the cytoplasm, it is imported into the mitochondria against a concentration gradient and at the expense of cytoplasmic GSH.30 As described above, in Gclc(h/h) mice starting at PND21, electron microscopy revealed atypical mitochondria—suggesting the loss of structural integrity in these organelles. This led us to investigate further the functional integrity of liver mitochondria in Gclc(h/h) mice. In Gclc(h/h) mice (Fig. 6), liver mitochondrial GSH decreased from 42% at PND14 to 16% at PND28, compared with control levels.

Figure 6.

Mitochondrial GSH levels in liver from Gclc(f/f) and Gclc(h/h) mice. We used liver mitochondrial suspensions to measure GSH levels using o-phthalaldehyde. Results are reported as means ± SEM from 4 or 5 mice for each group. *P < 0.001, when comparing Gclc(f/f) and Gclc(h/h) mice. We indicate the percentage in Gclc(f/f) mice.

Accompanying this fall in mitochondrial GSH, both mitochondrial respiration and membrane potential were progressively impaired in these mice. Compared with Gclc(f/f) mice (Table 3), respiratory control ratios were decreased in liver mitochondria obtained from Gclc(h/h) mice—mostly due to decreased oxygen consumption in state 3 respiration. Furthermore, mitochondria from Gclc(h/h) mice at PND25 and PND28 were poorly coupled, as indicated by respiratory control ratios lower than 3.0. In a similar manner, we observed decreased mitochondrial membrane potentials in Gclc(h/h) mice, starting at PND21. In agreement with impaired mitochondrial respiration, liver ATP levels in Gclc(h/h) mice dropped from 76% at PND21 to 20% at PND28, of that seen in Gclc(f/f) mice (Fig. 7).

Table 3. Mitochondrial Respiration and Membrane Potential
AgeGclcState 3 (O2/min/mg protein)State 4 (O2/min/mg protein)RCR (state 3/state 4)Membrane Potential (FU ratio)
  • NOTE: Values are means ± SEM from 3 or 4 mice.

  • Abbreviation: FU, fluorescence units.

  • *

    P < 0.05, when comparing Gclc(f/f) and Gclc(h/h) mice.

PND14f/f343 ± 1783 ± 114.2 ± 0.35.3 ± 0.7
 h/h291 ± 272 ± 34.2 ± 0.34.9 ± 1.3
PND21f/f503 ± 89114 ± 184.4 ± 0.13.7 ± 0.8
 h/h301 ± 17*87 ± 63.5 ± 0.1*2.0 ± 0.1*
PND25f/f382 ± 5798 ± 173.9 ± 0.15.0 ± 0.4
 h/h289 ± 21*115 ± 92.5 ± 0.0*1.2 ± 0.2*
PND28f/f386 ± 7087 ± 154.4 ± 0.13.1 ± 0.3
 h/h242 ± 27*90 ± 112.7 ± 0.0*0.9 ± 0.2*
Figure 7.

Hepatic ATP levels. We measured ATP levels in liver homogenates from mice with the indicated genotype. Results are reported as means ± S.E. from 4 or 5 mice for each group. *P < 0.05, when comparing Gclc(f/f) and Gclc(h/h) mice. We indicate the percentage in Gclc(f/f) mice.

Oxidants and Oxidative Stress Response in the Liver of Gclc(h/h) Mice

GSH plays an important role in eliminating H2O2. The depletion of GSH observed in the hepatocytes from Gclc(h/h) mice may result in failure to effectively eliminate ROS, especially H2O2, which would result in oxidant accumulation and the concomitant oxidative stress response. We used the thiobarbituric acid-reactive substances assay to evaluate lipid peroxidation—commonly used as an indicator of oxidant generation in cells and tissues.31 Compared with Gclc(f/f) mice, Gclc(h/h) mice showed elevated levels of hepatic lipid peroxidation, starting at PND21, which increased further by PND28 (Fig. 8A). A hallmark of cellular oxidative stress response is the transcriptional upregulation of oxidant stress-responsive genes, such as heme oxygenase-1 (Hmox1),32Gclm,33 and metallothionein-1 (Mt1).34 As a measure of the hepatic oxidant stress response, we measured the steady-state mRNA levels from these genes in Gclc(h/h) livers (Fig. 8B). Compared with Gclc(f/f) mice, the mRNA levels for these genes were dramatically elevated in livers from Gclc(h/h) mice at all times studied.

Figure 8.

Oxidative stress response from liver of Gclc(h/h) mice. (A) Hepatic lipid peroxidation was estimated by the thiobarbituric acid-reactive substances (TBARS) assay. Results are expressed as nmol malondialdehyde (MDA) equivalents/g liver, according to a MDA-diacetal standard curve. *P < 0.05, **P < 0.001, when comparing Gclc(f/f) and Gclc(h/h) mice. (B) Northern hybridization analysis of liver mRNA levels for heme oxygenase-1 (HMOX1), metallothionein-1 (MT1), and β-Actin.


In the present study, using conditional-targeting strategies, we generated a hepatocyte-specific Gclc knockout mice line. We showed that the targeted Gclc(f) allele is expressed in several tissues, without showing any significant differences from wild-type Gclc (Supplementary Fig. 1). Moreover, we showed that recombination of Gclc(f) by Cre recombinase is efficient—with timing and specificity similar to previous studies using these Alb-Cre transgenic mice. Data in this manuscript are consistent with the Gclc(h) allele being a null allele. Taken together, our data suggest that Gclc(f) is a conditional null allele that should be useful for assessing the function of Gclc and, likely by extension, for assessing the function of GSH in a large number of cell types.

Herein, we demonstrate that hepatocyte Gclc and, most likely GSH, is essential for the normal functioning of liver. We also show a correlation between the loss of GSH, loss of mitochondrial function, onset of oxidant damage and progression of serious liver disease. In this model for conditional Gclc ablation, we were able to detect recombination of Gclc in hepatocytes as early as GD16, and it is very near complete by PND14—although the GCLC protein persists in some hepatocytes until PND21. Cre recombinase is driven by the Alb gene regulatory sequences. The timing of Gclc recombination is in agreement with the onset of expression of the Alb gene during hepatogenesis and liver maturation after birth, and with the Alb gene's hepatocyte-specific pattern of expression.35

Pharmacologic studies suggest that the mitochondrial GSH pool, although derived from cytoplasmic GSH, behaves as a metabolically separate pool36 and is, to a large degree, preserved—even though cytoplasmic GSH is depleted. Herein we show that, in Gclc(h/h) livers at PND14, total GSH and mitochondrial GSH are depleted to nearly equal percentages (Table 1; Fig. 6). We must exercise caution when interpreting these data, however, because at PND14 a considerable population of hepatocytes is still immunopositive for GCLC. Thus, at PND14 the hepatocytes represent a mixed population of cells at various stages of GCLC and GSH depletion. At later time points, our data are consistent with the following assertions, which have been previously hypothesized in GCLC inhibitor studies: (1) Mitochondrial GSH is an independent pool of GSH, whose depletion is delayed compared to the depletion of total cellular GSH. This is consistent with mitochondrial transport of GSH occurring against a concentration gradient; therefore, mitochondrial GSH represents a high-affinity pool of GSH3, 36 (2) A threshold of hepatic GSH (10% in previous studies) is needed to preserve mitochondrial GSH, which is above a level that preserves mitochondrial function37.

Studies with Gclm(−/−) mice support these assertions; loss of Gclm impairs GSH biosynthesis (12). In Gclm(−/−) mice, hepatic GSH is approximately 15% of wild-type GSH by PND21 and thereafter. In contrast, hepatic mitochondrial GSH is maintained at ∼60% of wild-type levels. Moreover, in liver, the mitochondria from Gclm(−/−) mice have normal structure and function (data not shown). Interestingly, in γ-glutamlytranspeptidase Ggt(−/−) knockout mice, who fail to metabolize and recycle the amino acids from extracellular GSH, hepatic GSH is maintained at approximately 20% of control levels—even though mitochondria from these livers display impaired function and abnormal morphology.38 We should note that high plasma GSH levels, in combination with low cellular cysteine levels, complicate this animal model. It is also noteworthy that cultured Gclc(−/−) cells possessing only 2% of wild-type GSH levels maintain mitochondrial function compatible with survival.39 This may suggest cell-specific differences in GSH requirement.

Hepatocytes are postulated to possess prolific GSH synthetic capacity, because they possess not only high levels of GCL but also the capacity to synthesize cysteine through trans-sulfuration.40 Studies in rodents suggest that the hepatocyte is of great importance in inter-organ homeostasis of GSH and cysteine.41, 42 GSH is exported from hepatocytes through sinusoidal and canalicular efflux, which accounts for the rapid turnover of hepatic GSH. The sinusoidal efflux of GSH is believed to be the principal source of circulating plasma GSH, which is subsequently metabolized in tissue sites that are rich in the ectoenzyme γ-glutamyltranspeptidase, resulting in liberation of cysteine for intracellular protein and GSH biosynthesis. In Gclc(h/h) mice, plasma GSH is dramatically lower—falling from 20% to 2% between PND14 and PND28. These data support the role for the hepatocyte in supplying plasma GSH.

Interestingly, in Gclc(h/h) mice, GSH levels in extrahepatic tissues are not dramatically compromised and cysteine levels are normal; we did not observe toxicity in extrahepatic tissues. Based on GSH levels we would anticipate this. In fact, Gclm(−/−) mice, which have impaired GSH synthesis, have dramatically lower GSH levels in all extrahepatic tissues assayed compared to Gclc(h/h) mice, but are also without phenotype.12 This result suggests that GSH in extrahepatic tissues does not rely in large extent on maintained plasma GSH and that toxicity due to GSH depletion in extrahepatic tissues does not confound this model.

Steatosis and necroinflammation in the liver may appear as the hepatic lesion in diseases of various etiologies; intrahepatocyte generation of oxidants is believed to be important in the pathogenesis of these hepatopathies, and mitochondria have been hypothesized to be the major affected organelle.43 Mitochondrial dysfunction not only disrupts fat homeostasis in the liver, but also causes overproduction of ROS, which trigger lipid peroxidation and cell death. In the liver from Gclc(h/h) mice, we observed progressive steatosis, inflammation and hepatocyte death. In Gclc(h/h) mice, we documented an oxidant stress response as early as PND14, with upregulation of a number of genes known to be oxidant-responsive. Thereafter, oxidized lipids are readily detected. Neutral lipid accumulation in hepatocytes accompanies oxidant generation. Moreover, dramatic morphological changes in hepatocyte mitochondria, and in vitro assessments of mitochondrial function, document mitochondrial damage in Gclc(h/h) mice. Similar changes in mitochondrial morphology and function have been documented, using chemical depletion of GSH by way of subchronic BSO administration. Importantly, such defects in mitochondria have been documented both in rodents and in patients.43

Gclc(h/h) mice failed to gain weight beginning on PND18. Proper nutrition is an important determinant of the hepatic GSH and energy status, and protein calorie malnutrition can result in similar histopathological changes in the liver.44 In the current study, we did not exclude malnutrition as a factor in the pathology. However, we do know that following weaning Gclc(h/h) mice consume similar levels of food compared to wild-type controls (Fig. 9A). Furthermore, we were able to prevent mortality of the Gclc(h/h) mice by supplementation with N-acetylcysteine in drinking water starting at PND18. Such mice also fail to gain weight (Fig. 9B), suggesting that impaired nutrient absorption, if it is occurring, is not the cause of hepatic failure (manuscript in preparation). Taken together, our observations in Gclc(h/h) mice are consistent with the notion that depletion of the hepatic mitochondrial GSH pool, due to impaired de novo GSH synthesis, is necessary and sufficient to cause steatosis and death.

Figure 9.

Food intake and mouse growth curve with N-acetylcysteine (NAC) supplementation in drinking water. (A) We recorded the amount of mouse chow consumed by Gclc(f/f) (•) or Gclc(h/h) mice (○) and we expressed it as gram/day/mouse. (B) We dissolved NAC (Sigma, St.Louis, MO) in regular tap water (10g/L) and adjusted it to pH 7.0 by 10N NaOH. Starting at PND18, we provided NAC supplemented drinking water to 1 group of Gclc(f/f) (•) and Gclc(h/h) mice (○), and we replaced it every 2 days. Gclc(h/h) mice with NAC supplementation (○) were able to survive to adulthood (at least 3 months). Data are reported as means ± SEM from 7 to 10 mice for each group. *P < 0.05, when comparing Gclc(h/h) mice with and without NAC supplementation.


We thank our colleagues for their careful reading of this manuscript. We thank Dr. Antonio Perez Atayde (Boston Children's Hospital) and Dr. Kevin Bove (Cincinnati Children's Hospital Medical Center) for their help in our analysis of liver histopathology. We thank Dr. Mario Medvedovic for valuable input into our statistical analysis. Much of this work was presented at the 44th Annual Meeting of the Society of Toxicology, New Orleans, LA (March, 2005).