Initial Characterization of the Glutamate-Cysteine Ligase Modifier Subunit Gclm(−/−) Knockout Mouse

Glutamate-cysteine ligase (GCL) is the rate-limiting enzyme in the GSH biosynthesis pathway. In higher eukaryotes, this enzyme is a heterodimer comprising a catalytic subunit (GCLC) and a modifier subunit (GCLM), which change the catalytic characteristics of the holoenzyme. To define the cellular function of GCLM, we disrupted the mouse Gclm gene to create a null allele. Gclm(−/−) mice are viable and fertile and have no overt phenotype. In liver, lung, pancreas, erythrocytes, and plasma, however, GSH levels in Gclm(−/−) mice were 9–16% of that in Gclm(+/+) littermates. Cysteine levels inGclm(−/−) mice were 9, 35, and 40% of that inGclm(+/+) mice in kidney, pancreas, and plasma, respectively, but remained unchanged in the liver and erythrocytes. Comparing the hepatic GCL holoenzyme with GCLC in the genetic absence of GCLM, we found the latter had an ∼2-fold increase inK m for glutamate and a dramatically enhanced sensitivity to GSH inhibition. The major decrease in GSH, combined with diminished GCL activity, rendered Gclm(−/−) fetal fibroblasts strikingly more sensitive to chemical oxidants such as H2O2. We conclude that theGclm(−/−) mouse represents a model of chronic GSH depletion that will be very useful in evaluating the role of the GCLM subunit and GSH in numerous pathophysiological conditions as well as in environmental toxicity associated with oxidant insult.

GSH is synthesized from its precursor amino acids in two sequential enzymatic reactions. Glutamate-cysteine ligase (GCL) 1 catalyzes the formation of ␥-glutamylcysteine (␥-GC) from glutamate and cysteine. Glutathione synthetase then couples glycine to ␥-GC to form GSH. The reaction catalyzed by GCL is rate-limiting in GSH biosynthesis, and the product GSH is a feedback inhibitor of GCL activity. Higher eukaryotes contain GCL as a heterodimer comprised of a 72.8-kDa catalytic subunit (GCLC) and a 30.8-kDa modifier subunit (GCLM), which are encoded by genes on different chromosomes (8 -10). Studies with both native and recombinant GCL protein demonstrate that the catalytic subunit is necessary and sufficient for ␥-GC biosynthesis (11,12). Targeted disruption of the Gclc gene in mice demonstrated that GCLC and most likely GSH are essential for embryonic development but not survival of cells in culture (13,14).
Much of our understanding about the function of the modifier subunit has been the result of studies examining the catalytic activity of the purified GCLC protein versus the GCL holoenzyme. Experiments with both rat and human preparations indicated that GCLM modifies GCLC catalytic properties by decreasing the K m of glutamate and diminishing the inhibition by GSH (15,16). Because intracellular levels of glutamate are typically lower and those of GSH are typically higher than the K m and K i for GCLC, respectively, GCLC alone has been predicted to function poorly in maintaining cellular GSH levels (15). The experimental support for this hypothesis, however, is mixed. Overexpression of GCLM increases the cellular GSH by 2-fold, rendering cells resistant to oxidative stress (17). In agreement with this result, down-regulation of GCLM mRNA by ribozyme expression leads to a decrease of GSH levels in cultured pancreatic islet cells (18). On the other hand, up-regulation of GCLC alone has also been reported to support high levels of intracellular GSH (19,20). Furthermore, despite a decrease in GCLM following antisense RNA inhibition of GCLM translation, no changes in GSH were noted in human hepatoblastoma HepG2 cell cultures (21,22).
To define better the cellular function of GCLM, we have generated a mouse line with a targeted disruption of the Gclm gene. Mice homozygous for the null allele are viable and fertile; we show here that this mouse appears to be an excellent model system for studying animals that are severely compromised in their response to oxidative stress.
Preparation of Targeting Construct, Targeting, and Generation of Gclm(Ϫ/Ϫ) Mice-A -Fix II (Stratagene, La Jolla, CA) 129/SvJ mouse genomic library was screened with a GCLM cDNA probe (24), and several overlapping clones were mapped. A targeting construct was then prepared using portions of these clones (Fig. 1). 2 The Gclm gene was targeted in the same J1 and D embryonic stem cell lines as described previously (14). C57BL/6J blastocysts were used for embryonic stem cell injection (14). Resultant male chimeric mice were mated with C57BL/6J females, and the offspring were of mixed (C57BL/6J and 129/SvJ) genetic background. All mouse experiments were conducted on littermates having the three possible genotypes following the Gclm(ϩ/Ϫ) ϫ Gclm(ϩ/Ϫ) intercross, and all studies were approved by the University of Cincinnati Medical Center (UCMC) Institutional Animal Care and Use Committee (IACUC).
Genotyping Mice by Southern Blot Analysis and by PCR-Genomic DNA isolated from embryonic stem cells or mouse spleen was digested overnight with BglII and processed for Southern blot analysis using a 32 P-labeled 3Ј probe outside the region encompassed by the targeting construct ( Fig. 1). Following targeting, the mice were routinely genotyped using PCR. The targeted allele was detected as a 250-bp fragment using primers OL183a (5Ј-AACGTTGCAAGCTACTGC-3Ј) external to the targeted region of intron 1 and OL184a (5Ј-ATATGCGAAGTGGAC-CTG-3Ј) within the NEO gene. The wild-type allele was detected as a 200-bp fragment using primers OL183a and OL185s (5Ј-AGTT-GAGAGCCTTCACTG-3Ј) within the deleted region of intron 1.
Northern Blot and Western Immunoblot Analysis-Mice were killed by CO 2 asphyxiation, and tissues were frozen immediately on dry ice and stored at Ϫ70°C until use. For Northern blot analysis, total RNA was isolated from various tissues and analyzed for the presence of GCLC and GCLM mRNA as described previously (25). Equal loading was determined by rehybridizing blots with a probe for ␤-actin mRNA. For detection of GCLC protein, a rabbit polyclonal antibody was raised against the synthetic peptide QEKGERTNPNHPC (amino acids 79 -91 of the mouse GCLC) coupled to keyhole limpet hemocyanin. The cytosolic fractions were resolved by 12% SDS-PAGE and blotted with the above antibody (1:5000). The amount of GCLM protein was determined as described previously (14) using anti-GCLM antibody (a gift of T. Kavanagh).
Gel Filtration Chromatography-Liver samples were homogenized immediately in 10 volumes of homogenization buffer (154 mM KCl, 5 mM diethylenetriaminepentaacetic acid, 0.1 M potassium phosphate buffer, 10 mM MgCl 2 , and 5 mM ␤-mercaptoethanol, pH 6.8) using a Teflon homogenizer. The cytosolic fractions were prepared as described previously (14) and then separated by fast protein liquid chromatography (FPLC) (Amersham Biosciences). Samples were applied to a Superose-6 gel filtration column pre-equilibrated with the homogenization buffer, and the protein was separated with homogenization buffer.
Analysis of GCL Activity-The formation of ␥-GC from L-glutamate and L-cysteine was performed at 37°C under the conditions described previously (28). The reaction was stopped by adding a 50-l aliquot of the reaction mixture to 50 l of 10 mM diethylenetriaminepentaacetic acid in 40 mM HCl and 10% trichloroacetic acid. A fluorescent derivative of ␥-GC was generated using o-phthalaldehyde (26), and the derivative was separated from GSH and quantified using HPLC. The reversephase HPLC utilized isocratic elution from a Nova-Pak (4 m, 60 Å, 3.9 ϫ 150 mm) C18 column (Waters Corp., Milford, MA) with a mobile phase consisting of 7.5% methanol, 92.5% 150 mM ammonium acetate, pH 7.0. Fluorescence was determined at 365 nm excitation and 430 nm emission against standards of known concentration. Under the conditions utilized, the assay was linear with protein concentration and with time for at least 60 min.
Preparation of Mouse Fetal Fibroblasts (MFFs)-MFFs were prepared from gestational day 14.5 fetuses (29) derived from the Gclm(ϩ/Ϫ) x Gclm(ϩ/Ϫ) intercross. The cells from individual fetuses were cultured until the Gclm genotype was determined by PCR (as above) following which the cells from three or four fetuses were pooled for subsequent analyses.
Cell Treatments-For toxicity measurements, MFF cells were seeded in 24-well plates (4 ϫ 10 4 cells/well) 14 h before treatment. Cells were then administered by direct addition into the culture medium, 100ϫ concentrated stocks of the indicated compound in phosphate-buffered saline. The cultures were then incubated for an additional 8 h before the toxicity measurements; longer incubations did not lead to greater toxicity.
Statistical Analysis-All data are expressed as the means Ϯ S.E. Group means were compared by one-way analysis of variance using the SAS program (Windows version 8.0). p values Ͻ 0.05 were considered statistically significant. When the overall test of significance led to rejection of the null hypothesis, a multiple comparison was performed to determine the source of the effect.

Generation of Gclm(Ϫ/Ϫ) Mice-
We constructed a targeting vector in which the NEO minigene cassette disrupts Gclm exon 1 and removes the 3Ј splice donor site of intron 1 (Fig. 1A). Transcripts initiated at either of the two previously identified transcriptional start clusters of the Gclm gene (24) would be expected to encode only the N-terminal 61 amino acids of GCLM and terminate in the out-of-phase NEO gene, producing, if any, a nonfunctional protein. Of the 270 (agouti, 129/ SvJ-derived) embryonic stem clones resistant to both G418 and ganciclovir, four homologous recombinant clones were identified (data not shown); these clones were used for injecting into the (non-agouti C57BL/6J) blastocyst to generate chimeric mice. Such mice were mated with C57BL/6J female mice, and agouti animals were tested for germ line transmission of the targeted Gclm(Ϫ) allele, which would result in Gclm(ϩ/Ϫ) heterozygous mice. Gclm(ϩ/Ϫ) mice were then intercrossed, and the resultant offspring were genotyped by Southern blot (Fig.  1B) and PCR analysis (Fig. 1C). Because the probe used was outside the targeting construct, Southern blots confirmed Gclm targeting as opposed to random integration of the targeting construct.
GSH and Cysteine Levels-GCLM is known to bind to GCLC and change the catalytic characteristics of GCLC in vitro. In the context of cellular glutamate and GSH concentrations, GCLC alone (in a Gclm(Ϫ/Ϫ) mouse) is predicted to function poorly in synthesizing ␥-GC. Would GCLC alone lead to lowered levels of GSH? We therefore measured the intracellular levels of GSH in several tissues (Table I). In liver, kidney, pancreas, red cells, and plasma, GSH levels were dramatically decreased in Gclm(Ϫ/Ϫ) mice (9 -16% of that in wild-type animals) and significantly diminished in Gclm(ϩ/Ϫ) heterozygotes (43-82% of that in wild-type animals). Alterations in GSSG levels in Gclm(Ϫ/Ϫ) mice followed a trend similar to that of GSH (data not shown); consequently, there were no significant differences in the ratio of GSSG/GSH ϩ GSSG among the three genotypes. Lower GSH synthesis, presumably by liver, resulted in a dramatic decrease of GSH in plasma, and compared with that in Gclm(ϩ/ϩ) mice, plasma GSH levels in Gclm(Ϫ/Ϫ) and Gclm(ϩ/Ϫ) mice dropped to 16 and 48%, respectively, of that in Gclm(ϩ/ϩ) wild-type mice.
Circulating GSH is believed to replenish the intracellular cysteine pools via the membrane-bound enzymes ␥-glutamyltranspeptidase and dipeptidase, which are especially abundant in kidney and pancreas (30). Correspondingly, we found that cysteine levels in the kidney, pancreas, and plasma of Gclm(Ϫ/Ϫ) mice were markedly decreased, whereas no significant changes in cysteine levels were seen in liver or erythrocytes (Table I).
Physical and Kinetic Properties of Hepatic GCL-GCLM (30.8 kDa) interacts with GCLC (72.8 kDa) to produce the GCL holoenzyme (100 kDa). Using Gclm(ϩ/ϩ) hepatic cytosol and anti-GCLC and anti-GCLM (Fig. 4A), we found two overlapping chromatographic peaks, one ϳ100 kDa and the other ϳ72 kDa. The occurrence of the 100-kDa GCLC peak corresponded with the elution of GCLM (Fig. 4A), and no additional GCLM peak was detected in the remaining eluent or void-volume peak (not shown). These data indicate that the 100-kDa peak represents the GCL holoenzyme. Chromatography of Gclm(Ϫ/Ϫ) hepatic cytosol (Fig. 4B) supports this notion because the 100-kDa GCLC peak is absent in this sample as is the associated GCLM. These analyses support previous data implicating GCLM as Measurement of GCL specific activity from the hepatic cytosol of Gclm(ϩ/ϩ) and Gclm(Ϫ/Ϫ) mice (1.4 Ϯ 0.2 and 2.0 Ϯ 0.3 nmol/h/mg of protein, respectively) revealed a modest increase in the knockout mouse. This result is not surprising because we used saturating glutamate concentrations (15 mM) and no GSH in our assay system. Thus, the measured activity actually reflects the amount of GCLC present in the hepatic cytosol, which is somewhat elevated in Gclm(Ϫ/Ϫ) mice (Fig. 2B).
Following gel filtration, partially purified fractions containing only the GCL holoenzyme or GCLC alone were isolated (Fig. 4), and the catalytic characteristics of these fractions were tested to investigate the role of GCLM in mouse GCL holoenzyme function. These analyses were performed on GCL holoenzyme partially purified as shown in Fig. 4 or on GCLC prepared by three different methods: 1) GCLC from a Gclm(Ϫ/Ϫ) hepatic cytosol without purification, 2) GCLC from a Gclm(ϩ/ϩ) hepatic cytosol chromatographically separated from GCLM (Fig.  4), and 3) partially purified GCLC from a Gclm(Ϫ/Ϫ) hepatic cytosol (Fig. 4). All preparations provided similar results. Shown are data from GCLC prepared by method 3. Under all conditions for kinetic analysis the production of ␥-GC was linear for at least 1 h, and data were collected for these analyses 30 min following the start of enzymatic reactions. Shown for comparison are results from two previous studies that examined the K m for glutamate and cysteine for the recombinant rat and human GCL holoenzyme and GCLC alone (Table II). The apparent K m for L-glutamate is increased for GCLC com-pared with that for the GCL holoenzyme, whereas the apparent K m for L-cysteine does not differ significantly (Table II). Except for the large difference between the rat and both mouse and human with regard to a change in the K m for glutamate, the kinetic constants are similar between species. A striking difference between the GCL holoenzyme and GCLC is the dramatic inhibition of the latter by GSH (15,16,31). Using partially purified mouse GCL holoenzyme versus GCLC alone, we obtained a similar result. GCLC was more sensitive than the GCL holoenzyme to GSH inhibition (Fig. 5). Taken together these data suggest that GSH deficiency in the Gclm(Ϫ/Ϫ) mice is at least in part due to differences in catalytic characteristics between the GCL holoenzyme and the GCLC subunit. To test the possibility that cells without GCLM might be sensitive to oxidant stress, we measured MFF viability following exposure to the cellular oxidant H 2 O 2 (Fig. 6A). This doseresponse analysis demonstrated that Gclm(Ϫ/Ϫ) cells are strikingly more sensitive (ϳ10-fold) than Gclm(ϩ/ϩ) cells to cell killing by H 2 O 2 . In an effort to normalize toxicity between Gclm(ϩ/ϩ) and Gclm(Ϫ/Ϫ) MFFs, GSH was depleted in the former using phorone (32,33) and replenished in the latter using the cell-permeable GSH-EE (1,23). Preliminary doseresponse experiments first defined the concentrations of phorone and GSH-EE needed to normalize within 5% the GSH concentrations between Gclm(Ϫ/Ϫ) and Gclm(ϩ/ϩ) cells (data not shown). Thus, at the time MFFs were exposed to H 2 O 2 , the Gclm(ϩ/ϩ) MFFs treated with phorone contained GSH levels similar to that in Gclm(Ϫ/Ϫ) MFFs, and the Gclm(Ϫ/Ϫ) MFFs treated with GSH-EE contained GSH levels similar to that in Gclm(ϩ/ϩ) MFFs. Depletion of GSH by phorone enhanced the sensitivity of Gclm(ϩ/ϩ) cells to H 2 O 2 (Fig. 6B); however, these cells were still significantly more resistant than Gclm(Ϫ/Ϫ) cells. On the other hand, increases of intracellular GSH level by GSH-EE offered only moderate protection against H 2 O 2 for the Gclm(Ϫ/Ϫ) cells. DISCUSSION In this study, we have described the generation and initial characterization of the Gclm(Ϫ/Ϫ) mouse line. Gclm(Ϫ/Ϫ) mice are viable and fertile and show no overt phenotype compared with their Gclm(ϩ/ϩ) or Gclm(ϩ/Ϫ) littermates. GSH levels in several tissues and cells from Gclm(Ϫ/Ϫ) animals, however, are severely depressed. Based on the enzymatic characteristics of GCLC, previous reports (15) had suggested that loss of the GCLM protein might lead to extreme phenotypic abnormalities or even early mortality, similar to the phenotype found for the Gclc(Ϫ/Ϫ) homozygote (13,14). Gclc(Ϫ/Ϫ) mice die between gestational days 7.5 and 8.5 apparently as the result of apoptosis of cells in the distal portion of the developing embryo (13). Since Gclm(Ϫ/Ϫ) homozygotes do not suffer such consequences, it is possible that Gclm(Ϫ/Ϫ) mice maintain GSH above some critical level below which cell death signaling pathways would be activated. It should be noted that indeed most, if not all, cellular roles previously ascribed to GSH are redundant, and cells can survive in culture without GSH provided they are supplied with an alternate sulfur source (13). On the other hand, another obvious difference between Gclc(Ϫ/Ϫ) and Gclm(Ϫ/Ϫ) cells is the capacity of the latter to synthesize ␥-GC, and although the concentration of this intermediate is ex- tremely low in cells, one cannot dismiss the possibility that ␥-GC serves some critical cellular function(s) aside from GSH biosynthesis.

Sensitivity of Gclm(Ϫ/Ϫ) Cells to Oxidant Stress-To
Various studies suggest that intracellular GCLC and GCLM do not exist in equimolar amounts (21,34). The lowered GSH levels in tissues from Gclm(ϩ/Ϫ) mice further support this premise. One must be cautious with the interpretation of these data, however, because diminished GSH concentrations may also be attributed to lower levels of the GCL substrate, cysteine. This is very likely to be the case for kidney and pancreas where cysteine levels are well below normal; yet, this is unlikely for liver and erythrocytes. Further, the chromatographic separation of GCL in Gclm(ϩ/ϩ) hepatic cytosol shows an excess of GCLC (Fig. 4) (Fig. 6A).
Support for the idea that GCLM may be limiting can also be gleaned from the literature. For example, in HIV-1 Tat transgenic mice, a 59% decrease of GCLM leads to lowered hepatic GSH levels (35). In addition, a mutation in the 5Ј-flanking region of the human GCLM gene results in a diminished reporter gene response to oxidant insult and is associated with susceptibility to ischemic heart disease (36). Since GCLC and GCLM are regulated at the transcriptional level by oxidative stress and regulation of the two proteins might not be coordi-    5. GSH inhibition of activity derived from the GCL holoenzyme or from GCLC alone. Partially purified GCL holoenzyme or GCLC was assayed for ␥-GC synthesis in the presence of saturating glutamate and cysteine and the indicated concentrations of GSH. Assays were conducted in triplicate, and data are expressed as the percentage of the control activity in the absence of GSH (means Ϯ S.E.). In some instances, the circle denoting the mean is larger than the brackets denoting the S.E. nated, it is intriguing to speculate that some cell types may respond by up-regulating GCLM and others by up-regulating GCLC in contexts where their respective dimerization partner is in excess. Such regulation remains to be rigorously demonstrated.
The present study clearly shows that GCLM is not essential for viability but that loss of GCLM severely lowers GSH and renders cells susceptible to oxidant stress. Is lower GSH, per se, a principle reason for sensitivity to oxidants? Several studies have shown that the resistance to drug and/or radiation treatment of tumor cells does not necessarily correlate with cellular GSH levels but rather with GCL activity (33,37). Consistent with this notion, decreasing GSH in Gclm(ϩ/ϩ) MFFs or elevating GSH in Gclm(Ϫ/Ϫ) MFFs is only modestly effective in normalizing the phenotypes of these cells (Fig. 6B). This finding may suggest that recovery from oxidative insult requires a robust GSH synthetic potential and that the GCL holoenzyme is necessary for this capacity. Thus, a major function of GCLM may be to improve GSH synthetic capacity as a defense against oxidative insult. Such a role may be predictable based on kinetic properties of the GCL holoenzyme versus GCLC provided by this and previous reports (15,16).
The long term physiological consequence of chronic GSH deficiency has not yet been investigated. As mentioned above, a functional polymorphism in the 5Ј-flanking region of human GCLM gene has been identified and is associated with increased risk of myocardial infarction (36). Chronic GSH deficiency has also been suspected as a causative factor in a number of pathologies including neurodegenerative diseases, aging, diabetes mellitus, cataracts, and AIDS (38 -41). Most evidence supporting a pivotal role for decreased GSH in these conditions is correlative, however, rather than direct. Gclm(Ϫ/Ϫ) mice have incredibly lowered GSH levels but seem to be normal. Observation of this mouse line for another 12-24 months of its life span may prove interesting. We thus believe that these mice will provide an exciting model system for the study of chronic GSH deficiency and enhanced sensitivity to oxidative stress, especially that caused by environmental chemicals.