Non-canonical glutamate-cysteine ligase activity protects against ferroptosis

Cysteine is required for maintaining cellular redox homeostasis in both normal and transformed cells. Deprivation of cysteine induces the iron-dependent form of cell death known as ferroptosis; however, the metabolic consequences of cysteine starvation beyond impairment of glutathione synthesis are uncharacterized. Here, we find that cystine starvation promotes ferroptosis not only through the inhibition of glutathione (GSH) synthesis, but also through the accumulation of glutamate. Surprisingly, we find that glutamate-cysteine ligase catalytic subunit (GCLC) prevents glutamate accumulation through the generation of alternative γ-glutamyl peptides. Further, inhibition of GCLC accelerates ferroptosis under cystine starvation in a GSH-independent manner. These results indicate that GCLC has an additional, non-canonical role in the protection against ferroptosis to maintain glutamate homeostasis under cystine starvation.


Introduction
Amino acids can play critical biosynthetic functions beyond their use for protein synthesis. A notable example is the thiol-containing amino acid cysteine. Cysteine-derived molecules are crucial for multiple cellular processes as a consequence of their sulfur moiety, which facilitates diverse functions, including enzyme catalysis, energy transfer, and redox metabolism ( (Stipanuk et al., 2006), the most abundant intracellular antioxidant (Winterbourn and Hampton, 2008). GSH is a tripeptide consisting of the amino acids cysteine, glutamate and glycine. The synthesis of GSH occurs in two steps (Anderson, 1998). First, glutamate and cysteine are ligated by GCLC, producing the dipeptide γ-glutamyl cysteine (γ-Glu-Cys). Next, glycine is added to γ-Glu-Cys, producing the tripeptide GSH (γ-Glu-Cys-Gly). The antioxidant activity of GSH is a consequence of its function as a cofactor to multiple antioxidant proteins, including glutaredoxins (GRXs), GSH peroxidases (GPXs), and GSH S-transferases, thereby removing reactive oxygen species (ROS) (Harris and DeNicola, 2020).
Because of both its reactive thiol moiety and its essential function in redox homeostasis, cysteine levels are tightly regulated. While cysteine excess is prevented by overflow into the taurine pathway (Stipanuk et al., 2009), cysteine demand is met by inducible regulation of cystine import.
Following oxidative stress, the expression of the cystine/glutamate exchange transporter xCT is induced, (Habib et al., 2015) thereby facilitating the uptake of cystine and its reduction to cysteine.
In some tissues, most notably the liver, cysteine is also synthesized from homocysteine and serine via the transsulfuration pathway (Beatty and Reed, 1980;Rao et al., 1990; Reed and Orrenius, 1977). Given the important roles of cysteine, many cancers overexpress xCT (Ji et  Cysteine inadequacy can induce an iron-dependent form of cell death known as ferroptosis (Dixon et al., 2012). Ferroptosis is triggered by the reaction of polyunsaturated fatty acids (PUFA) in membrane lipids with peroxyl radicals produced from iron (Fe 2+ ) and ROS (Cao and Dixon, 2016;Yang et al., 2014), thereby inducing lipid peroxidation. Consistently, processes that promote ferroptosis include increased ferritin uptake (Gao et al., 2015), ferritin degradation (Mancias et al., 2014), synthesis of PUFA containing lipids (Dixon et al., 2015;Doll et al., 2017), and mitochondrial ROS production (Gao et al., 2015;Gao et al., 2019). However, while cysteine is directly linked to GSH synthesis, which can influence the levels of ROS and lipid peroxides via GPX4 (Yang et al., 2014), cysteine availability can also influence the levels of cofactors and metabolites within cells beyond its use for GSH synthesis. Importantly, the metabolic consequences of cystine starvation are poorly understood.
To understand the metabolic consequences of the cystine starvation, we performed quantitative stable isotope labeled metabolite tracing in non-small cell lung cancer (NSCLC) cells. We found that glutamate accumulated due to impaired GSH synthesis, and promoted ferroptosis. Further, we identified that under cysteine-deprived conditions, GCLC used other small, non-charged amino acids in place of cysteine to generate γ-glutamyl peptides. This promiscuous activity prevented glutamate accumulation to protect against ferroptosis. γ-glutamyl peptide synthesis by GCLC was also evident in mouse tissues.

Cystine starvation impairs GSH synthesis prior to the onset of ferroptosis
To evaluate the consequence of cystine starvation in NSCLC cells, we starved a panel of cell lines of extracellular cystine and first monitored viability over time. Cystine starvation induced the death of most cell lines between 24-48 hrs, with the exception of H460 and H1944 cells, which were more resistant ( Figure 1A). Cell death was confirmed to be ferroptosis due to both the ability of the ferroptosis inhibitor Ferrostatin-1 (Fer-1) and iron chelator DFO (Dixon et al., 2012) to rescue cell death ( Figure 1A) and the morphological changes characteristic of ferroptosis ( Figure S1A). Next, we examined the metabolic consequences of cystine starvation.
To understand the immediate consequences of cystine starvation, we starved cells for 4 hrs, which caused the rapid depletion of intracellular cysteine to almost undetectable levels in all cell lines ( Figure 1B). To determine whether cystine starvation influenced cellular processes, we first examined glutathione (GSH) synthesis. Cysteine is a rate limiting metabolite of GSH synthesis (Stipanuk et al., 2006) and GSH also plays an important role in ferroptosis via ROS metabolism (Dixon et al., 2012) and substrate of GPX4 (Conrad and Friedmann Angeli, 2015). Therefore, to evaluate the effect of extracellular cystine starvation on GSH synthesis, we conducted quantitative 13 C3-serine tracing. 13 C3-serine can be metabolized to 13 C2-glycine (M+2) and 13 C3cysteine (M+3), which are subsequently incorporated into GSH (M+2 and M+3, respectively, Figure 1C). After 4 hours labeling, most of the serine fraction and half of the glycine fraction were labeled ( Figure S1B and S1C). Importantly, while the amount of newly synthesized glycine was equivalent or increased in the cell lines following cystine starvation, the amount of M+2 glycine incorporated into GSH was dramatically depleted. Minimal M+3 labeling was detected. In addition, total GSH levels were lower, consistent with an inhibition of GSH synthesis. These results indicate that the extracellular cystine starvation rapidly depletes intracellular cysteine availability for GSH synthesis, which precedes the induction of ferroptosis.

Inhibition of GSH synthesis promotes glutamate accumulation
GSH synthesis consumes glutamate and glycine in addition to cysteine. We observed that inhibition of GSH synthesis was associated with an accumulation of intracellular glycine and glutamate in multiple NSCLC cell lines following cystine starvation ( Figure 2A). In addition, glutamate export is obligatory for cystine import and glutamate accumulation may also be influenced by reduced cystine/glutamate exchange. Consistently, we observed that cystine starvation decreased glutamate exportation (Figure 2A and D). Because glutamate was previously shown to contribute ferroptosis (Gao et al., 2015), we examined whether glutamate plays a causal role in cystine-starvation induced ferroptosis in NSCLC cells. We treated cells with 5 mM glutamate diethyl ester (GlutEE), a concentration we confirmed increases intracellular glutamate to similar levels as cystine starvation in A549 cells ( Figure S2A and 2A). We found that GlutEE treatment accelerated ferroptosis in multiple NSCLC cells ( Figure 2B), while glutamine starvation, which depleted intracellular glutamate ( Figure S2D), or treatment with the transaminase inhibitor AOA could rescue ferroptosis ( Figure 2C). Interestingly, the effects of AOA could be overridden by treatment with dimethyl-alpha-ketoglurate (DMαKG), suggesting αKG or its downstream metabolite mediates the effects of glutamate. Glutamate was previously shown to promote ferroptosis via the TCA cycle and ROS generated from the oxidative phosphorylation (Gao et al., 2015;Gao et al., 2019). Consistently, we found that GlutEE promoted ROS accumulation under cysteine starved conditions ( Figure S2B). Therefore, these results demonstrate that inhibition of GSH synthesis by cystine starvation not only depletes GSH, but also induces the accumulation of glutamate to promote ferroptosis.

GCLC prevents ferroptosis independent of GSH production
Multiple studies have demonstrated a potent, synergistic effect of GSH synthesis inhibition with BSO with limitation of cystine uptake or availability (Cramer et al., 2017;Harris et al., 2015).
Consistently, we observe that BSO treatment promotes ferroptosis under cystine starvation in most NSCLC cell lines ( Figure 3A). However, we find that cystine starvation rapidly inhibits intracellular GSH synthesis ( Figure 1B). Importantly, while BSO treatment depleted GSH in cystine replete cells as expected, it did not change GSH levels in cystine starved cells ( Figure   3B). Thus, the accelerated cell death induced by BSO under cystine starvation may not be explained by the depletion of GSH. Therefore, we examined whether GCLC could play a role in cystine-starvation induced ferroptosis independent of GSH synthesis. To evaluate this question, we generated GCLC and GSS KO A549 clones ( Figure 3C). Importantly, both clones were defective in GSH synthesis as evidenced by significantly reduced intracellular GSH levels effectively compared to parental cells ( Figure 3D). Importantly, the GCLC KO clones demonstrated accelerated ferroptosis induction under cystine starvation compared to parental cells, which could be rescued by GCLC cDNA, while the GSS KO clones did not ( Figure 3E).
Further, BSO treatment induced ferroptosis in the GSSKO clone ( Figure 3F), despite the absence of GSH ( Figure 3D), further confirming the GSH-independent role of GCLC in ferroptosis protection. Finally, GCLC KO, but not GSS KO, accelerated ferroptosis induction under cystine starvation following acute deletion in H1299 cells ( Figure S3A-C), suggesting this is not a consequence of adaptation in single cell clones. Together, these data indicate that GCLC can prevent cystine starvation-induced ferroptosis of NSCLC cells independent of GSH production ( Figure 3G).

Cystine starvation induces GCLC-dependent γ-glutamyl peptide accumulation
To determine the mechanism of GSH-independent protection against ferroptosis by GCLC, we conducted non-targeted metabolomics. Interestingly, we discovered a cluster of LC-MS peaks which were highly depleted by BSO treatment following cystine starvation in A549 cells ( Figure   4A). Further, these peaks were the same ones that were the most highly accumulated by extracellular cystine starvation ( Figure 4A). These unknown LC-MS peaks were identified as γglutamyl-di or tri-peptides, which all contain a glutamate-derived moiety ( Figure 4A). Authentic standards for γ-glutamyl threonine (γ-Glu-Thr) and γ-glutamyl-alanyl-glycine (γ-Glu-Ala-Gly) were not available, thus we further validated their identity via stable isotope labeled metabolite tracing. The 13 C5, 15 N2-glutamine tracing result indicated that both γ-Glu-Thr and γ-Glu-Ala-Gly were derived from glutamate ( Figure 4D). Further, 2, 3, 3-2 H3-serine tracing showed that γ-Glu-Ala-Gly was derived from the glycine ( Figure S4A). We extended these observations to other NSCLC cell lines and found that cystine starvation consistently promoted the accumulation of γglutamyl peptides, which was inhibited by treatment with BSO ( Figure 4B). These results suggest that cystine starvation promotes the accumulation of γ-glutamyl peptides by the GSH synthesis pathway.

γ-glutamyl peptide synthesis by GCLC scavenges glutamate to protect against ferroptosis
The tripeptide γ-glutamyl-2-aminobutyryl-glycine (γ-Glu-2AB-Gly) is known to be generated by GCLC and GSS (Huang et al., 1988;Oppenheimer et al., 1979) in a similar manner to GSH by substituting 2-aminobutyrate for cysteine ( Figure S4B). Consequently, the accumulation of γ-Glu-2AB-Gly under cysteine starvation can be explained by cysteine unavailability for GCLC ( Figure 4B and S4B). In contrast, γ-glutamyl-dipeptides are reported to be derived from GSH by γ-glutamyl transferase (GGT) extracellularly ( Figure S4B) (Hanigan and Pitot, 1985). However, 13 C5, 15 N2-glutamine tracing demonstrated that while the newly labeled GSH fraction was very small, as expected, glutamate and γ-glutamyl-dipeptides were newly labeled to ~ 50% in cystine starved A549 cells ( Figure 4C and D), suggesting that the γ-glutamyl dipeptides were synthesized from glutamate but not from GSH ( Figure S4B). Because γ-glutamyl-valine is synthesized by the Saccharomyces cerevisiae glutamate-cysteine ligase (Sofyanovich et al., 2019), and γ-glutamyl dipeptide synthesis by mouse liver extracts was recently shown to be GCLC-dependent (Kobayashi et al., 2020), we hypothesized that the γ-glutamyl dipeptides were directly generated by GCLC rather than GSH metabolism by GGT. To evaluate this, we evaluated the γ-glutamyl dipeptide levels in the GCLC and GSS KO clones. Importantly, the γglutamyl dipeptides that were accumulated following cystine starvation in parental cells were dramatically depleted only in the GCLC KO clones ( Figure 5A). Interestingly, the γ-glutamyl dipeptide levels were generally higher in GSS KO clones than parental lines, which can be explained by the feedback inhibition of GCLC by GSH, and more weakly by γ-Glu-2AB-Gly (Richman and Meister, 1975) ( Figure 5A). In addition, both γ-Glu-2AB-Gly and γ-Glu-Ala-Gly tripeptides were depleted by both GCLC and GSS KO compared to parental cells, as GSS activity is required for the ligation of glycine. Consistent alterations of γ-glutamyl peptides were observed in GCLC and GSS KO H1299 cells, which were rescued by GCLC or GSS restoration ( Figure S5A). These results indicate that γ-glutamyl dipeptides are directly generated by GCLC under cystine starved condition ( Figure 5B).
As we found that GCLC inhibition with BSO treatment or genetic KO could accelerate ferroptosis under cystine starvation ( Figure 3A and E), we examined whether GCLC-mediated γ-glutamyl dipeptide synthesis plays a causal role in this process. Because glutamate accumulation promoted ferroptosis (Figure 2), we evaluated the ability of γ-glutamyl dipeptides to serve as a glutamate sink. Both inhibition of GCLC with BSO treatment and GCLC KO increased intracellular glutamate levels under cystine starvation, while GSS KO was actually protective ( Figure 5C and D). We also observed an accumulation of glutamate in GCLC KO, but not GSS KO H1299 cells under cystine starvation ( Figure S5B). Importantly, the γ-glutamyl dipeptides themselves did not play a protective role against ferroptosis as their supplementation did not rescue cystine starvation-induced ferroptosis of GCLC KO clones ( Figure S5C). Finally, cystine starvation-induced ferroptosis of GCLC KO cells was rescued by both glutamine starvation and AOA treatment ( Figure S5D). Together, these results demonstrate that GCLC has a noncanonical role in ferroptosis to balance the glutamate pool to protect against ferroptosis under cystine starvation ( Figure S5E).

GCLC regulates glutamate homeostasis in vivo
Finally, we examined whether GCLC mediates the synthesis of γ-glutamyl peptides in vivo under normal physiological conditions. To this end, systemic Gclc deletion was induced in an adult mouse ( Figure 6A) and we examined the effect in the liver, kidney and serum. Efficacy of Gclc deletion was evident by the depletion of glutathione by 75-90% in these tissues ( Figures 6B-D).
While glutathione was present in the reduced (GSH) form in tissues, the serum had predominantly the oxidized form (GSSG), which may either be due to the oxidizing extracellular conditions or oxidation during sample preparation. Further, we found that Gclc KO liver, kidney, and serum were also depleted of γ-glutamyl-peptides, including both the dipeptides and tripeptides ( Figures 6B-D). In addition, deletion of Gclc increased glutamate levels in the liver and serum, but not the kidney ( Figures 6B-D). Overall, these results indicate that GCLC plays a causal role in the homeostatic control of glutamate and γ-glutamyl peptide metabolism in vivo ( Figure 6E).

Discussion
The findings reported herein demonstrate that g-glutamyl peptide synthesis by GCLC provides GSH-independent protection from ferroptosis following cystine starvation. While cystine starvation-induced ferroptosis has commonly been attributed to the depletion of cellular GSH, we show that cystine starvation induces complex metabolic changes within cells. Our work does not exclude the importance of GSH in the protection against ferroptosis. GSH is a major for the interpretation of studies using BSO to inhibit GCLC. While many of those results may be attributed to GSH depletion, the contribution of GCLC to g-glutamyl peptide synthesis and glutamate scavenging may also play a very important role, particularly in the context of xCT inhibition, where cells cannot export glutamate. Our findings warrant the development of potent GSS inhibitors for the study of ferroptosis to distinguish these mechanisms. These inhibitors would also be valuable for therapeutic combinations with ferroptosis inducers, although they may increase glutamate scavenging, which may affect cellular responses.
Our in vivo results provide direct genetic evidence to support the GCLC-mediated g-glutamyl peptide production that was recently been reported in mouse liver extracts (Kobayashi et al., 2020), where glutamate could be ligated with other amino acids in a reaction inhibited by BSO.
The promiscuity of GCLC toward amino acids other than cysteine is not a unique feature of this enzyme, and has been shown for many other metabolic enzymes. For example, serine palmitoyl transferase will also metabolize alanine or glycine when serine is limiting (Penno et al., 2010) and glutamate-aspartate aminotransferase will also metabolize cysteine sulfinic acid (Weinstein et al., 1988). In the case of GCLC, this feature may have been selected for during evolution, as the S. cerevisiae homolog (Gsh1p) also has the ability to at least use valine (Sofyanovich et al., 2019). Additional work is needed to determine which other amino acids are accepted by S.
cerevisiae Gsh1p. For the human enzyme, small, non-charged amino acids that are structurally similar to cysteine can be used based on their appearance in g-glutamyl peptides, although the full spectrum of amino acids has not been tested in a direct enzymatic assay.
Our findings may extend beyond conditions of cysteine deficiency. Systemic deletion of mouse Gclc revealed that Gclc plays a role in the regulation of glutamate and g-glutamyl peptides levels in normal tissue. Notably, glutamate accumulation was only observed in the liver but not the kidney. Liver plays a critical role in GSH synthesis to supply the rest of the organism, which may consume significantly more glutamate in liver than kidney (Ookhtens and Kaplowitz, 1998). It is important to note that, in contrast to cell culture, depletion of g-glutamyl peptides in Gclc KO tissue may be a consequence of both canonical, extracellular GGT mediated g-glutamyl dipeptide production and the intracellular GCLC-mediated pathway. However, the accumulation of glutamate and the depletion of g-glutamyl tripeptides, which require the activity of GSS, strongly suggests that these peptides are being produced intracellularly. Supportively, the activity of GGT is negligible in the mouse liver compared to kidney (Kobayashi et al., 2020). It is not known whether g-glutamyl peptides have additional functions in tissues beyond serving as a reservoir for glutamate, and potentially other amino acids. g-glutamyl peptides levels have been shown to be increased under conditions of liver injury, including drug-induced injury, hepatitis infection, liver cirrhosis, and hepatocellular carcinoma (Soga et al., 2011). Additional work is needed to understand the role of g-glutamyl peptide synthesis in these diseases.
We also find that GSS may regulate glycine homeostasis by producing g-glutamyl tripeptides, including g-Glu-2AB-Gly and g-Glu-Ala-Gly. Future work is needed to both understand whether GSS can use other amino acids besides glycine and determine the full spectrum of g-glutamyl tripeptides produced by GSS. Further, we find that GSS deficiency actually enhances g-glutamyl dipeptide synthesis, which can be explained by loss of feedback inhibition of GCLC by GSH.
These findings raise interesting implications for the metabolic phenotypes of patients with inborn errors of glutathione metabolism. Although extremely rare, mutations in GCLC and GSS result in hemolytic anemia. Interestingly, GSS mutant patients also present with 5-oxo-prolinuria, which is not observed in GCLC deficiency (Ristoff and Larsson, 2007). While this 5-oxo-prolinuria has been attributed to the accumulation of g-glutamylcysteine and its metabolism to 5-oxo-proline by g-glutamylcyclotransferase (GGCT) (Ristoff and Larsson, 2007), our work suggests that other gglutamyl-amino acids are likely produced in this situation to contribute to 5-oxo-prolinuria. This is likely to depend on the availability of cysteine, which would likely become limiting if the feedback inhibition of GCLC is lost due to an inability to synthesize GSH. via daily intraperitoneal injection for 5 consecutive days at 160 mg/kg (tamoxifen/mouse body mass). Mice were humanely euthanized (isoflurane inhalation followed by cervical dislocation)

Key
14-16 days later and necropsies were performed. Tissues (liver, kidney) were removed and serum was isolated; both were snap frozen on dry ice and stored at -80℃ prior to analysis.

Lentivirus generation
Lentiviruses were generated by overnight PEI transfection of 90% confluent Lenti-X 293T cells (Clonetech) with target lentiviral plasmid, and packaging plasmids pCMV-dR8.2 dvpr and pCMV-VSV-G in DMEM (10% FBS). The next day, the medium was changed to fresh DMEM (10% FBS). After 24 hrs, the first batch of virus contained medium was collected and filtered by 0.45 μm PES filter. The second batch of virus contained medium was collected as following above, further combined to the first batch and stored at -80°C until virus infection.

Lentiviral infection of NSCLC cells.
To increase of gene deletion efficiency in a polyclonal population, H1299 cells with stable

Crystal Violet based cell viability assay.
Cells were plated in 96-well plates at a density of 10,000-20,000 cells/well in 100 μL final volume.
The next day, the medium was changed to 100 μL medium with different experimental conditions as indicated. At the indicated time points, cells were fixed with 4% Paraformaldehyde, stained with crystal violet, washed and dried. The crystal violet was dissolved in 10% acetic acid and the absorbance was measured by 600 nm wavelength. The relative cell number was normalized to control cells of each experimental set.

Sample preparation for the liver tissue targeted metabolomics.
The frozen liver tissue samples were pulverized using a pre-chilled BioPulverizer (

Sample preparation for the intracellular non-targeted metabolomics, and Glycine and
Glutamate quantification.
NSCLC cells were plated in 6 well dishes so they were 70% confluent at extraction and preconditioned in RPMI medium containing dFBS (10%) overnight. The following day, the medium was aspirated and the cells were quickly washed with 1 mL of RPMI (10% dFBS, 1% P/S),

Quantitation of extracellular glutamate exportation
The extracellular medium collected above was transferred to the 96 well plates and the medium glutamate levels was measured by YSI 2900 (Yellow springs, OH, USA) using 2755 glutamate standard (5 mM) The extracellular glutamate secretion rate (nmol/µL of cell volume/hrs) determined from cell volume measurements above. The mobile phases were eluted as following gradient condition. 0-13min: 80% to 20% of mobile phase B, 13-15min: 20% of mobile phase B. The ESI ionization mode was positive or negative.

LC-MS analysis.
The MS scan range (m/z) was set to 60-900. The mass resolution was 120,000 and the AGC target was 3 × 10 6 . The capillary voltage and capillary temperature were set to 3.5 KV and 320°C,  Figure S4A and B), as described in result. dilution with 5% non-fat milk in TBST. After wash membrane 3 times using TBST for 10 min of each, the 10,000 times diluted secondary antibody in 5% non-fat milk in TBST (goat anti-rabbit or goat anti-mouse, Jackson ImmunoResearch) were attached to membrane for 1 hr. After wash the membrane in TBST for 3 times for 10 min of each, the enhanced chemo-luminescence signal was measured by exposing to the x-ray film followed by fixing and developing.

Statistical analysis
Statistical analyses were conducted with Graph Pad Prism 8. For the comparison of two groups, two-tailed Student's t-test was used. For the comparison of more than 3 experimental groups, one-way ANOVA was used with Bonferroni's multiple comparison test.

Lead Contact
Further information and requests for resources and reagents should be directed to and will be