AKT activation because of PTEN loss upregulates xCT via GSK3β/NRF2, leading to inhibition of ferroptosis in PTEN-mutant tumor cells

SUMMARY

Here, we show that the tumor suppressor phosphatase and tensin homolog deleted from chromosome 10 (PTEN) sensitizes cells to ferroptosis, an iron-dependent form of cell death, by restraining the expression and activity of the cystine/glutamate antiporter system X_c⁻ (xCT). Loss of PTEN activates AKT kinase to inhibit GSK3β, increasing NF-E2 p45-related factor 2 (NRF2) along with transcription of one of its known target genes encoding xCT. Elevated xCT in Pten-null mouse embryonic fibroblasts increases the flux of cystine transport and synthesis of glutathione, which enhances the steady-state levels of these metabolites. A pan-cancer analysis finds that loss of PTEN shows evidence of increased xCT, and PTEN-mutant cells are resistant to ferroptosis as a consequence of elevated xCT. These findings suggest that selection of PTEN mutation during tumor development may be due to its ability to confer resistance to ferroptosis in the setting of metabolic and oxidative stress that occurs during tumor initiation and progression.

Graphical abstract

In brief

Cahuzac et al. report a potential mode of PTEN tumor suppression through activation of ferroptosis in response to oxidative stress. PTEN loss elevates xCT via an AKT-GSK3β-NRF2 axis and consequently increases glutathione pools, which inhibits ferroptosis. These findings also identify a role of PTEN in cysteine and glutathione metabolism.

INTRODUCTION

Regulated cell death has been known for decades, and distinct molecular types of regulated cell death have been elucidated, including apoptosis, necrosis, and autophagy.^1–5 Ferroptosis, a relative newcomer to the field of programmed cell death pathways, has only recently become better appreciated.^6–10 While certain aspects have been observed since the 1950s, it was not until 2012 that ferroptosis was characterized and recognized as a morphologically, biochemically, and genetically distinct form of regulated cell death.^11–13 Ferroptosis is driven by iron-dependent peroxidation, which ultimately leads to loss of plasma membrane integrity and cell death. The physiological importance of ferroptosis in diseases such as cancer, stroke, and neurodegeneration has been established.^14–21

The selenoprotein glutathione peroxidase 4 (GPX4) and its cofactor glutathione (GSH) neutralize lipid peroxides and effectively protect cells from ferroptosis.^22,23 As an essential antioxidant and building block of GSH, cysteine also plays an integral role in ferroptosis regulation by modulating GSH levels. While cysteine can be imported into cells through neutral amino transporters, the bulk of extracellular cysteine exists in its oxidized form, cystine, which is solely imported by the cystine-glutamate transporter system X_c⁻ (xCT), encoded by the gene SLC7A11.^24–27 xCT is expressed as a heterodimer with CD98/4F2hc, which is required for localization to the plasma membrane and for amino acid transporter function.²⁸ When cystine is imported into the cell, it is rapidly reduced to cysteine and subsequently utilized for GSH synthesis as well as other metabolic processes.²⁹ Perturbing this cystine-xCT-GSH-GPX4 axis increases ferroptosis because of insufficient antioxidant defenses within the cell.¹⁰ One such example occurs through pharmacological inhibition of xCT with erastin or sulfasalazine (SAS), which reduces cystine import and diminishes GSH pools to effectively inhibit GPX4.¹⁰

As a key component of GSH synthesis and oxidative homeostasis, SLC7A11/xCT expression is modulated by multiple transcriptional and epigenetic regulators (NRF2, ATF3, ATF4, ARID1A, p53, BACH1, BRD4, and BAP1), protein stabilizers (OTUB1, SLC3A2, and CD44), and signaling pathways (EGFR, insulin growth factor [IGF], STAT3, transforming growth factor β [TGF-β], interferon γ [IFNγ], RAS, phosphatidylinositol 3-kinase [PI3K], and mTOR).^14,30–52 One of the transcriptional regulators often involved in regulation of SLC7A11 is NF-E2 p45-related factor 2 (NRF2), a transcription factor that acts as a master regulator of antioxidant response and promotes transcription of SLC7A11 in certain contexts, such as oxidative stress.^34,53 NRF2’s target genes create a network of coordinating enzymes that allow NRF2 to direct a multifaceted response to various types of cellular stress to maintain internal homeostasis and elevate intracellular GSH.⁵⁴ By reducing lipid peroxidation through its target genes, NRF2 acts as a negative regulator of ferroptosis.^55–57 NRF2 is regulated primarily at the protein level by E3 ubiquitin ligases, such as β-TrCP, which mark NRF2 for proteasomal degradation.^58–62 AKT activity also regulates NRF2 expression through a GSK3β/β-TrCP axis.^49,63 More specifically, GSK3β phosphorylates NRF2 in the absence of oxidative stress, creating a recognition binding site for β-TrCP.⁵⁸ β-TrCP then binds and mediates ubiquitination of NRF2, marking it for proteasomal degradation.⁵⁹ AKT is known to phosphorylate and deactivate GSK3β, allowing NRF2 to accumulate in the cell by avoiding recognition by β-TrCP.^49,63 This GSK3β-NRF2 axis has also been found to be important in ferroptosis regulation.^57,64

Phosphatase and tensin homolog deleted from chromosome 10 (PTEN) is one of the most frequently mutated tumor suppressor genes in cancer.^65–67 Aberrant overactivation of PI3K signaling, caused by loss of PTEN, promotes tumorigenesis by enhancing cell cycle progression and reducing apoptosis.^68–70 To carry out its tumor suppressor function, PTEN encodes a lipid phosphatase that dephosphorylates the second messenger phosphatidylinositol-3,4,5-trisphosphate (PIP3) to inhibit the PI3K pathway and its downstream effectors, such as AKT and mTOR.^71,72 PTEN has been found to elicit GSK3β-mediated phosphorylation of NRF2 via inactivation of AKT, leading to subsequent β-TrCP-mediated NRF2 degradation.^73,74 PTEN is also known to regulate the activity of the mTOR kinases mTORC1 and mTORC2, which have been linked to xCT and ferroptosis regulation.⁷⁵ For example, mTORC1 has been observed to indirectly upregulate SLC7A11 transcription, while mTORC2 has been observed to phosphorylate serine 26 on xCT, leading to lower transport activity. However, differing observations from various studies on the relationship between mTOR, xCT, and ferroptosis have been published.^{41,42,48,76,77} Whether PTEN’s negative regulation of NRF2 or mTOR affects SLC7A11 transcription, xCT transport function, and, subsequently, GSH synthesis or ferroptosis is not known.

Using isogenic Pten-null (knockout [KO]) and wild-type (WT) mouse embryonic fibroblasts (MEFs) as well as PTEN-mutant and WT cancer cell lines and patient tumor samples, we show that deletion of PTEN increases expression of SLC7A11 to lower susceptibility to ferroptosis. Absence of PTEN increased expression of xCT, import of cystine, and production of GSH. This was consistent with the ability of PTEN to regulate AKT, GSK3β, and NRF2. Mutation of PTEN leads to ferroptosis resistance that is rescued by AKT inhibition, which lowers the heightened expression of NRF2 and xCT in these cells. PTEN therefore creates a susceptibility to ferroptosis by keeping GSH pools low, which is needed for GPX4 to counteract lipid oxidation and prevent ferroptosis. These results demonstrate that PTEN regulates cysteine metabolism, redox balance, and ferroptosis and reveal a new tumor-suppressive property of PTEN.

RESULTS

PTEN loss desensitizes cells to ferroptosis induction by erastin and cystine deprivation

While research over the last decade has documented the role of ferroptosis in cancer, the contribution of PTEN to the modulation of ferroptosis has not been defined. Given PTEN’s known role as a regulator of the PI3K-AKT-GSK3β-NRF2 axis, we hypothesized that PTEN could play a role in modulating ferroptosis. To begin exploring the potential relationship between PTEN and ferroptosis, we first infected Pten flox/flox primary MEFs with an empty vector or Cre recombinase adenovirus and tested for sensitivity to erastin, a known ferroptosis inducer that works through inhibition of xCT.^8,10 Characterization of the effects of PTEN loss on proliferation and metabolism in this MEF model system has been conducted previously in our lab.⁷⁸ Pten KO MEFs were observed to be significantly more resistant to ferroptosis induction by erastin compared with Pten WT MEFs, with a 4-fold difference in lethal dose 50 (LD₅₀) (Figures 1A–1C). Importantly, the ferroptosis inhibitors N-acetylcysteine (NAC) and ferrostatin-1 were able to fully rescue the cell death observed in Pten WT MEFs while apoptosis, necrosis, and autophagy inhibitors were not, indicating ferroptosis as the mode of cell death (Figure 1D). We hypothesized that PTEN status may alter the redox balance of cells and monitored cell death in response to the oxidative stress triggered by hydrogen peroxide (H₂O₂) in MEFs. We observed that the absence of PTEN leads to a 2-fold decrease in cell death in MEFs treated with 750 μM H₂O₂ (Figure S1A). Interestingly, another ferroptosis inducer that targets xCT, SAS, elicited a similar difference in sensitivity, while RSL3, a ferroptosis inducer that directly inhibits GPX4, did not produce differential sensitivity between Pten WT and Pten KO MEF pairs (Figure 1E).⁶ These findings suggest that, during oxidative stress triggered by impaired xCT function, PTEN loss desensitizes cells to ferroptosis induction at the level of xCT.

Figure 1. PTEN loss desensitizes cells to ferroptosis inducers that target xCT in MEFs and cancer cell lines.

(A) Erastin LD₅₀ for Pten WT and Pten KO MEFs treated for 48 h with dose titrations of erastin and DRAQ7 to monitor cell death. Unpaired one-tailed Student’s t test.

(B) Percent cell death and corresponding images of Pten WT and Pten KO MEFs treated with DRAQ7 and DMSO or 300 nM erastin for 48 h. Fisher’s LSD statistical test.

(C) Crystal violet staining of Pten WT and Pten KO MEFs treated with DMSO or 300 nM erastin for 48 h.

(D) Crystal violet staining of Pten WT and Pten KO MEFs treated for 48 h with 300 nM erastin in combination with either 200 μM ferroptosis inhibitor N-acetylcysteine (NAC), 10 μM ferrostatin 1 (Fer1), 10 μM Z-VAD-FMK, 1 μM necrostatin-1 (Nec1), or 1 μM bafilomycin A1 (Baf-A1).

(E) Percent cell death in Pten WT/Pten KO MEFs stained with DRAQ7 and treated with 8 or 16 nM RSL3 or 65 μM SAS for 48 h. Fisher’s LSD statistical test.

(F) Crystal violet staining of MDA-MB-231 and SUM149 breast cancer cells treated with varying doses of erastin for 48 h.

(G) Crystal violet staining of Daoy and H4 brain cancer cells treated with varying doses of erastin for 48 h.

(H) Crystal violet staining of NCIH 226 and H520 lung cancer cells treated with varying doses of erastin for 48 h.

(I) Quantification of the crystal violet staining in (F)–(H) and Figures S1B and S1C for relative survival of cells when treated with 12.5 μM erastin. Mann-Whitney test.

ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Each experiment was conducted with 3 technical replicates and repeated twice unless otherwise noted.

To investigate whether PTEN mutation causes resistance to ferroptosis in human cells, we tested for erastin sensitivity in a panel of human cancer cell lines (Table S1). In agreement with our initial finding, PTEN-deficient cell lines were consistently more resistant to erastin-induced ferroptosis compared with their PTEN WT counterparts (Figures 1F–1I and S1B–S1D).

PTEN loss enhances SLC7A11 transcription and xCT protein expression

We next sought to determine the reason for PTEN’s ability to modulate sensitivity to ferroptosis. To probe whether PTEN could regulate the target of erastin, xCT, we performed qRT-PCR and analyzed previously published microarray data to examine transcription of SLC7A11 in our isogenic pairs of Pten KO and WT MEFs.⁷⁹ Pten KO MEFs had increased SLC7A11 transcription compared with Pten WT MEFS (Figures 2A and 2B), and importantly, when xCT protein was examined, we saw a nearly 7-fold increase in xCT protein expression in Pten KO MEFs relative to Pten WT MEFs (Figure 2C). On the other hand, no difference in GPX4 protein expression was observed (Figure 2D). xCT protein was also found to be upregulated in PTEN-deficient breast, brain, and lung cancer cell lines compared with their PTEN WT counterparts, while no difference in GPX4 protein was observed (Figures 2E and S2A). Interestingly, BT549, the PTEN-mutant breast cancer cell line that demonstrated sensitivity to erastin comparable with that of PTEN WT breast cancer lines (Figure S1D), expressed low levels of xCT, which could explain its atypical sensitivity to erastin compared with other PTEN-deficient breast cancer lines (Figure 2E).

Figure 2. PTEN loss upregulates SLC7A11 and xCT expression in MEFs, cancer cell lines, and patient tumor samples.

(A) Relative expression of SLC7A11 (microarray) in Pten WT/Pten KO MEFs. Unpaired two-tailed Student’s t test.

(B) Relative SLC7A11 expression (qRT-PCR) in Pten WT/Pten KO MEFs. Unpaired two-tailed Student’s t test.

(C) xCT Western blot images and quantification from Pten WT/Pten KO MEFs. Paired two-tailed Student’s t test, n = 5.

(D) GPX4 Western blot of Pten WT/Pten KO MEFs.

(E) xCT Western blot of breast, brain, and lung cancer cell lines.

(F–H) Expression of SLC7A11 by PTEN, PIK3CA, or AKT1/2 status in the ICGC/The Cancer Genome Atlas (TCGA) pan-cancer patient dataset (Nature 2020), analyzed by RNA sequencing (RNA-seq) and extracted from cBio Portal. Kruskal-Wallis test.

(I) Expression of SLC7A11 by PIK3CA, AKT1/2, and PTEN status in the ICGC/TCGA pan-cancer patient dataset (Nature 2020), analyzed by RNA-seq and extracted from cBio Portal. Kruskal-Wallis test.

(J) xCT Western blot of Dox-inducible xCT MDA-MB-231 cells treated with varying concentrations of Dox over 72 h.

(K) Percent cell death in Dox-inducible xCT MDA-MB-231 cells stained with DRAQ7 and treated with DMSO or 8 μM erastin for 72 h with or without 1 μg/mL Dox. Fisher’s LSD statistical test.

(L) Erastin LD₅₀ in Dox-inducible xCT MDA-MB-231 cells treated for 72 h with dose titrations of erastin alone or in combination with 1 μg/mL Dox. Unpaired two-tailed Student’s t test.

(M) xCT western blot of MDA-MB-468 cells treated with siRNA control or siRNA xCT.

(N) Percent cell death in MDA-MB-468 cells treated with DMSO or 5 μM erastin and control siRNA or xCT siRNA. Uncorrected Fisher’s LSD test.

Unaltered, WT (PIK3CA, AKT1, AKT2, and PTEN); PTEN Mut, PTEN mutation; PTEN HetLoss, loss of one allele; PTEN HomDel, loss of both alleles; PIK3CA AKT1 or AKT2 Altered, mutation, amplification, or gain of function; PTEN Altered, mutation, heterozygous loss, or homozygous deletion. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Each experiment was conducted with 3 technical replicates and repeated twice unless otherwise noted.

To determine how broadly relevant the finding of elevated SLC7A11 transcription in PTEN-deficient MEFs and cancer cell lines could be, we additionally explored SLC7A11 mRNA expression in patient tumor samples from the Cancer Genome Atlas (TCGA) PanCancer Atlas database using cBio Portal.^80,81 Patients, across multiple cancers and specifically in cohorts of brain, lung, and breast cancer, harboring tumors with different types of PTEN alterations had higher levels of SLC7A11 mRNA compared with patients with PTEN WT tumors (Figures 2F and S2B).⁸² This trend was not found for GPX4 mRNA (Figure S2C). While the increased expression of SLC7A11 observed in PTEN-deficient tumors samples in the brain glioblastoma cohort was not consistently significant, the difference was profound, and the lack of significance likely came from the limited number of WT PTEN tumor samples, which reduced the statistical power to detect a difference (Figure S2B). Because PTEN’s known canonical function negatively affects PI3K and AKT activity, we additionally explored how mutations, amplifications, and gain-of-function alterations of these genes were associated with SLC7A11 expression in patient tumor samples. In a pan-cancer cohort, we observed increased levels of SLC7A11 transcription in patients with PIK3CA and AKT alterations (Figures 2G–2I). When we examined an invasive breast cancer cohort of patients with and without alterations in these genes, we observed patient tumor samples with PIK3CA mutation to have no change in SLC7A11 expression compared with unaltered tumors, while tumors with amplification or gain-of-function alterations of PIK3CA were observed to have increased expression of SLC7A11 compared with unaltered tumors (Figure S2D). Additionally, in the invasive breast cancer cohort, we found tumor samples with AKT2 alterations, but not AKT1, to have increased SLC7A11 expression compared with unaltered tumors (Figure S2D). PTEN loss is known to activate AKT more strongly than PI3K mutation, which we also see in our cBio analysis (Figure S2E).⁸³

PTEN expression is enhanced by WT p53, which is also a negative regulator of SLC7A11.⁸⁴ Thus, p53 mutation, which lowers PTEN expression and raises SLC7A11, could be confounding. However, when we removed patients with p53 alteration from the pan-cancer cohort, we still observed significantly increased SLC7A11 transcription in tumors with PIK3CA, AKT1, or PTEN alterations compared with unaltered tumors (Figure S2). Furthermore, in a p53 dominant-negative MEF pair, Pten KO MEFs continued to exhibit diminished sensitivity to erastin-induced ferroptosis compared with Pten WT MEFs (Figure S2G).

To elucidate whether the sensitivity observed in PTEN WT tumor cell lines was caused by low levels of xCT, we next over-expressed xCT in the PTEN WT breast cancer cell line MDA-MB-231 with a lentiviral doxycycline-inducible xCT expression vector (Figure 2J) and tested for changes in sensitivity to erastin. MDA-MB-231 was the cell line most sensitive to erastin in our panel, with nearly 100% cell death when treated with 8 mM of erastin, and it had the lowest expression of xCT (Figures 2E, 2J, and S1D). Overexpression of xCT with doxycycline (Dox) treatment was observed to completely rescue this previously lethal dose of erastin and increased the erastin LD₅₀ over 10-fold, suggesting low xCT as a regulator of the ferroptosis sensitivity observed in these cells (Figures 2K and 2L). Furthermore, reducing xCT via small interfering RNA (siRNA) in the PTEN mutant MDA-MB-468 breast cancer cell line resensitized these previously resistant cells to erastin-induced ferroptosis (Figures 2M and 2N).

PTEN regulates cysteine metabolism and decreases GSH synthesis

Because xCT is an antiporter that exchanges intracellular glutamate for extracellular cystine and is the sole importer of cystine, we further examined whether PTEN KO cells are also differentially resistant to cystine deprivation. If erastin and SAS are triggering their phenotypic difference between PTEN-mutant and PTEN WT cells through their effect on xCT function, then, we hypothesized, cystine deprivation should also reveal this difference. Upon dilution of cystine in the medium, we found that Pten KO MEFs exhibit decreased cell death and required a lower concentration of cystine for growth compared with Pten WT MEFs (Figures 3A, 3B, and S3A), demonstrating a diminished sensitivity to cystine deprivation. These findings suggest that, during oxidative stress triggered by impairedxCT function and depleted cystine levels, PTEN loss desensitizes cells to ferroptosis induction.

Figure 3. PTEN restrains cysteine metabolism, and PTEN loss increases cystine uptake by xCT and promotes GSH synthesis.

(A) EC₅₀ values of Pten WT and Pten KO MEFs subjected to 72-h cystine starvation. Unpaired two-tailed Student’s t test.

(B) Percent cell death in Pten WT and Pten KO MEFs treated with DRAQ7 subjected to cystine starvation titration. Unpaired two-tailed Student’s t test.

(C–E) Relative metabolite abundance in Pten WT/Pten KO MEFs (liquid chromatography-tandem mass spectrometry [LC-MS/MS]). Unpaired two-tailed Student’s t test.

(F) Relative percent accumulation of ¹³C into cystine and GSH synthesis pathway metabolites in Pten WT/Pten KO MEFs (LC-MS/MS). Unpaired two-tailed Student’s t test.

(G) Cysteine metabolism overview.

(H) Percent cell death in Pten WT and Pten KO MEFs treated with DRAQ7 and 125 nM erastin alone or in combination with 3 mM BSO or 125 mM GEE for 48 h. Unpaired two-tailed Student’s t test.

(I) LD₅₀ values of Pten WT/Pten KO MEFs treated with DRAQ7 and dose titrations of erastin alone or in combination with 3 mM BSO or 125 mM GSH ethyl ester (GEE). Fisher’s LSD statistical test. (J) Buthionine sulfoximine (BSO) LD₅₀ in Pten WT and Pten KO MEFs. Unpaired two-tailed Student’s t test.

Met, methionine; SAM, S-adenosylmethionine; SAH, S-adenosyl homocysteine. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Each experiment was conducted with 3 technical replicates and repeated twice unless otherwise noted.

Furthermore, because PTEN loss increases the abundance of xCT, it would stand to reason that intracellular cystine would also be increased in PTEN KO cells and, as a result, that cysteine and GSH metabolism could be enhanced. Therefore, to test this hypothesis, we next explored whether loss of PTEN could upregulate cysteine metabolism and GSH synthesis by performing steady-state aqueous metabolite profiling. Steady-state metabolomics revealed that Pten KO MEFs have a 6-fold and 4-fold increase in intracellular cystine and cysteine abundance, respectively, as well as a higher abundance of GSH and the GSH synthesis intermediate γ-glutamylcysteine compared with Pten WT MEFs (Figure 3C). In addition to cystine import by xCT as a source of cysteine, cysteine can also be funneled into or recycled from the trans-sulfuration and choline oxidation pathways (Figure 3G). Pten KO MEFs were also found to have increased abundance of trans-sulfuration pathway metabolites as well as choline oxidation pathway metabolites (Figures 3D and 3E). Collectively, this suggests that PTEN regulates cysteine and GSH metabolism and that PTEN KO cells have more GSH compared with PTEN WT cells.

The heavy isotope ¹³C2-cystine was then utilized to investigate cystine flux to determine whether the increased GSH in the Pten KO MEFs was being synthesized from increased cystine being brought into the cell by xCT. Because cystine can only be imported through xCT, and the only source of ¹³C2-cystine is the extracellular cystine provided in the medium, then any metabolite found to contain the heavy isotope must have been converted from the cystine imported by xCT.^24–27 Pten KO MEFs were found to have a 4-fold and 3-fold higher accumulation of ¹³C in intracellular cystine and cysteine, respectively, than Pten WT MEFs, indicating that more extracellular cystine is being brought into the cell by xCT (Figure 3F). This result seems plausible, given that these cells were observed to have higher levels of xCT transporters compared with Pten WT MEFs. Furthermore, there was more cystine flux into GSH synthesis in Pten KO MEFs, as indicated by the 7-fold higher accumulation of heavy-isotope-labeled GSH and higher accumulation of its preceding intermediate γ-glutamylcysteine (Figure 3F). Together these findings suggest that PTEN loss heightened the cell’s ability to import cystine via xCT and, as a result, increased GSH pools. As a necessary cofactor for GPX4, lack of adequate GSH is known to inhibit GPX4 activity, creating a deficit in the anti-oxidant defenses of the cell to combat lipid peroxidation. If Pten KO MEFs have higher GSH levels and, as a result, higher GPX4 activity to neutralize reactive oxygen species (ROS), then one would predict that these cells would have lower accumulation of ROS compared with Pten WT MEFs. To investigate this, flow cytometry using DCFDA dye, which stains for ROS, was performed on two isogenic models: (1) p53 dominant-negative Pten KO/WT MEFs and (2) Pten KO/WT MCF10A cells. Pten KO MEFs and MCF10A cells were found to have nearly 3-fold and 2-fold decreases, respectively, in ROS compared with their Pten WT counterparts (Figure S3B).Therefore, this increase in GSH metabolism and subsequent decrease in ROS levels could explain the diminished sensitivity to ferroptosis we observe in PTEN KO cells.

If GSH levels are responsible for modulating the difference in ferroptosis sensitivity observed between PTEN-deficient and PTEN WT cells, then one would predict that either raising GSH in PTEN WT cells or lowering it in PTEN-deficient cells would remove this difference. To probe this hypothesis, we utilized GSH ethyl ester (GEE), a cell-permeable GSH that bolsters GSH levels in a cell, and buthionine sulfoximine (BSO), an inhibitor of γ-glutamylcysteine synthetase that blocks the first step of GSH synthesis and depletes GSH pools. We treated Pten KO and Pten WT MEFs with erastin combined with either GEE or BSO to see how these compounds would affect ferroptosis sensitivity. GEE was observed to increase erastin resistance in Pten WTMEFs, while BSO rescued erastin sensitivity in Pten KO MEFs, highlighting the importance of GSH pools in regulating this differential sensitivity observed between Pten WT and Pten KO MEFs (Figures 3H and 3I). Additionally, Pten KO MEFs were found to be significantly less sensitive to BSO treatment compared with Pten WT MEFs. This observation could be due to the low GSH pools PTEN WT cells have to start with (Figure 3J).

PTEN loss increases NRF2 expression and transcription of its target genes via AKT

We next sought to determine how PTEN negatively regulates SLC7A11 transcription. As noted previously, there are many transcriptional regulators that regulate SLC7A11/xCT. Therefore, to tackle this question, we analyzed the microarray data generated from our Pten WT/Pten KO MEFs to determine whether the transcription of any of the known transcriptional regulators of xCT were augmented in the context of PTEN status. Of these regulators, only 2 were significantly altered in a way that could explain the increase in SLC7A11 transcription observed in Pten KO cells compared with Pten WT MEFs: Bach1 and Nrf2 (Table 1). BACH1 is a transcriptional repressor of SLC7A11, whose expression was significantly downregulated in Pten KO MEFs, while NRF2 is a transcriptional activator of SLC7A11, whose expression was significantly increased in Pten KO MEFs. We further examined the protein abundance of NRF2 and BACH1 to see whether their levels changed in response to PTEN status. Pten KO MEFs were observed to have higher NRF2 protein expression compared with Pten WT MEFs (Figure 4A). This phenotype was not observed across breast, brain, or lung cancer cell lines (Figure S2A). The observed variability is likely because the cancer cell lines are not isogenic and have unique mutation signatures in which other mutations besides PTEN may be driving the cancer cell line’s NRF2 expression. In contrast, no difference in BACH1 protein expression was detected between Pten KO and WT MEFs (Figure S4A). Because a difference in NRF2 protein expression was observed, and PTEN is known to negatively regulate NRF2 expression, we decided to further explore this regulator of xCT to determine whether it was modulating the difference in SLC7A11 transcription observed between PTEN WT and PTEN KO cells.^49,58,63,73 Given the increased abundance of NRF2 expression in Pten KO MEFs, these cells would be predicted to have higher expression of NRF2’s target genes. While this was observed to be true for at least one of NRF2’s target genes, SLC7A11, to determine whether the activity of this transcription factor is upregulated by PTEN loss, we wanted to further explore whether other target genes of NRF2 were also upregulated. To pursue this, gene set enrichment analysis (GSEA) was conducted on microarray data generated from Pten KO and Pten WT MEFs. This differential gene expression analysis revealed that NRF2 target genes were upregulated in Pten KO MEFs compared with Pten WT MEFs (NES 1.54, q = 0.01), indicating that NRF2 transcriptional activity is increased in Pten KO MEF cells (Figure 4B).

Table 1. mRNA expression fold change of SLC7A11 transcriptional regulators in MEFs.

xCT regulator	Regulation type	Fold change in mRNA expression (KO/WT)	Significance (p value)
Atf3	transcriptional repressor46	1.1991	0.0538
Atf4	transcriptional promoter33,34	1.0599	0.0733
Bap1	Transcriptional repressor37	0.9993	0.9472
Trp53	transcriptional repressor36	1.0033	0.6464
Stat3	Transcriptional promoter44,45	1.0377	0.1450
Bach1	transcriptional repressor52	0.9131	0.0014*
Pou2f1 (Oct1)	transcriptional promoter43	0.7749	0.0007*
Nfe2l2 (Nrf2)	transcriptional promoter35	1.0825	0.0316*
Nrf1	Transcriptional repressor85	1.0010	0.9813
Arid1a	transcriptional promoter38	0.9579	0.1989
Brd4	transcriptional repressor47	0.9561	0.0820
Ifng	transcriptional repressor14	0.9847	0.7467
Igf1	transcriptional promoter51	0.9677	0.3629
Tgfb1	transcriptional repressor53	0.9797	0.7515
CD44	protein stabilizer39	1.0815	0.0138*
Slc3a2	protein stabilizer40	1.1016	0.0181*
Otub1	protein stabilizer48	1.0165	0.0900
Egfr	protein stabilizer41	0.8507	0.0046*

Shown is the mRNA expression fold change (KO/WT) of SLC7A11/xCT regulators. mRNA expression was extracted from microarray data on Pten WT/Pten KO MEFs; n = 4 per genotype. Students t test (unpaired, two tailed, unequal variance)

^*p < 0.05. Bach1 and Nfe2l2 are bolded because they are differentially expressed with p < 0.05 in the expected direction.

Figure 4. PTEN loss promotes SLC7A11 transcription via an AKT-GSK3β-NRF2 axis in MEFs and cancer cell lines.

(A) NRF2 Western blot in Pten WT/Pten KO MEFs.

(B) GSEA of NRF2 antioxidant response target genes in Pten WT/Pten KO MEFs.

(C) pGSK3β, total GSK3β, and β-actin western blots in Pten WT/Pten KO MEFs treated with DMSO or 1 μM AZD5363 for 48 h.

(D and E) NRF2 or xCT Western blot in Pten WT/Pten KO MEFs treated with PI3K pathway inhibitors for 48 h.

(F) pGSK3β^Ser−9, NRF2, and xCT western blots in MDA-MB-231 parental, empty vector (EV), or PTEN KO clones.

(G) Erastin LD₅₀ and percent cell death in MDA-MB-231 parental, EV, and PTEN KO clones treated with DRAQ7 and 8 μM erastin for 48 h. Fisher’s LSD statistical test.

(H) Percent cell death and cell death fold change in Dox-inducible PTEN U87 MG cells treated with 37.5 μM erastin for 48 h with or without 1 μg/mL of Dox. Mann-Whitney test.

(I) pGSK3β^Ser−9, NRF2, and xCT western blot in Dox-inducible PTEN U87 MG cells treated with or without 1 μg/mL dox.

(J) LD₉₀ values and fold change (KO/WT) in Pten WT/Pten KO MEFs treated with DRAQ7 and dose titrations of erastin in combination with PI3K pathway inhibitors (1 μM AZD5363, 125 pM rapamycin, or 5 nM torin1) for 48 h. Fisher’s LSD test.

(K) NRF2 and xCT western blots in Pten WT/Pten KO MEFs treated with 250 nM CHIR99021 and/or 1 μM AZD5363.

(L) Percent cell death in Pten WT/Pten KO MEFs treated for 48 h with DRAQ7 and 250 nM erastin in combination with 1 μM AZD5363 with or without 250 nM CHIR99021 and/or 10 μM Fer1. Unpaired two-tailed Student’s t test.

*p < 0.05, **p < 0.01, ***p < 0.0001, ****p < 0.00001. Each experiment was conducted with 3 technical replicates and repeated twice unless otherwise noted.

Next, we wanted to investigate how PTEN is regulating levels of NRF2. β-TrCP, an E3 ubiquitin ligase that marks NRF2 for proteasomal degradation, was of particular interest for the purpose of our study because it is regulated by AKT and GSK3β;^58,59 in the absence of oxidative stress, GSK3β phosphorylates NRF2, creating a recognition binding site for β-TrCP to bind and mark NRF2 for proteasome degradation. AKT is known to deactivate GSK3β through phosphorylation of its serine 9 residue, allowing NRF2 to avoid phosphorylation and recognition by β-TrCP and accumulate in the cell.⁶³ We hypothesized that PTEN’s known inhibitory effect on AKT was activating GSK3β and that this was resulting in the observed low NRF2 and xCT expression in PTEN WT cells, subsequently inducing ferroptosis in response to the oxidative stress triggered by erastin. We explored pGSK3β^Ser−9 levels in our MEFs and observed an increase in pGSK3β^Ser−9 in Pten KO MEFs compared with Pten WT MEFs, indicating that GSK3β is less active in these KO cells (Figure 4C). Notably, no differential expression in total GSK3β was observed. Next, we wanted to explore whether the decrease in active GSK3β in the Pten KO MEFs was a result of enhanced AKT activity because of the absence of PTEN. To probe this, we treated the MEF pair with AZD5363, an AKT inhibitor, at a dose that inhibits downstream signaling to PRAS40 and reduces growth (Figures S4B and S3C) to see how this would affect pGSK3β^Ser−9 abundance. After 48 h of treatment, the heightened abundance of pGSK3β^Ser—9 (inactive form) that was observed in Pten KO MEFs was markedly reduced. In agreement with these findings, Rojo et al.⁷³ observed that PTEN indirectly elicits GSK3β-mediated phosphorylation of NRF2 via inactivating AKT, leading to subsequent β-TrCP-mediated NRF2 degradation in a KEAP1-independent manner. Furthermore, we observed that AKT inhibition via 1 μM AZD5363 fully rescued the heightened NRF2 and xCT expression observed in Pten KO MEFs compared with Pten WT MEFs (Figures 4D–4E). Additionally, inhibiting PI3K, a positive regulator of AKT, with 250 nM dual PI3K inhibitor GDC0941 also lowered NRF2 and xCT expression in Pten KO MEFs (Figure S4C). To explore the potential role of mTOR, a downstream target upregulated by AKT, in regulating xCT protein levels, we tested in parallel the mTOR inhibitors rapamycin (125 pM), which inhibits TORC1, and torin1 (5 nM), which inhibits TORC1 and TORC2, for their ability to influence NRF2 and xCT expression. These inhibitors did not change the levels of these proteins despite their ability to reduce mTOR signaling to S6 at these doses and decrease growth (Figures 4D, 4E, S4B, and S3C). These results demonstrate that PTEN, through its control of AKT and GSK3β, regulates NRF2 abundance and, therefore, expression of xCT.

We next wanted to see whether the chronic activation of the PI3K signaling cascade because of PTEN inactivation would also lead to higher pGSK3β, NRF2, and xCT in a human cancer cell line and alter ferroptosis sensitivity via xCT inhibition. We therefore used a CRISPR-Cas9 lentivirus to knock out PTEN in MDA-MB-231 cells, the PTEN WT breast cancer cell line with the highest sensitivity to erastin. In agreement with our findings in MEFs, PTEN KO MDA-MB-231 cells exhibited heightened pGSK3β^Ser−9, NRF2, and xCT levels, as well as resistance to induction of ferroptosis by erastin compared with PTEN WT MDA-MB-231 cells, as shown by the 2-fold difference in LD50 (Figures 4F–4G). Additionally, we overexpressed PTEN in the PTEN mutant brain cancer cell line U87 MG using a lentiviral Dox-inducible PTEN expression vector and tested for changes in sensitivity to erastin as well as NRF2 and xCT protein expression. As expected, expressing PTEN in these U87 MG cells was observed to increase sensitivity to erastin as well as decrease pGSK3β^Ser−9, NRF2, and xCT expression (Figures 4H–4I).

AKT facilitates the differential erastin sensitivity between Pten KO and Pten WT MEFs

Next, we wanted to see whether this AKT/GSK3β/NRF2/xCT pathway was facilitating the differential ferroptosis sensitivity observed between Pten KO and Pten WT MEFs. Previously we had observed that PTEN WT cells could be made resistant to erastin by overexpressing xCT (Figures 2K and 2L), which demonstrated that low xCT facilitates the ferroptosis sensitivity observed in PTEN WT cells. Therefore, if AKT is mediating the excess xCT seen in PTEN deficient cells by inhibiting GSK3β function and allowing NRF2 to accumulate, then inhibiting AKT and subsequently activating GSK3β should re-sensitize these cells to erastin sensitivity. To test this line of reasoning, 1 mM AZD5363 was combined with erastin to determine whether AKT inhibition could rescue Pten KO MEFs from erastin resistance by mimicking the relatively low AKT activity observed in Pten WT MEFs. The expected differential sensitivity to erastin treatment alone was again observed between Pten KO MEFs and Pten WT MEFs. However, this difference was fully rescued by addition of 1 μM AZD5363, reducing the LD₉₀ of Pten KO and Pten WT MEFs nearly 6-fold and 2-fold, respectively (Figure 4J). This result highlights the importance of AKT in the differential sensitivity observed between Pten KO and WT MEFs. Significantly, PI3K inhibition via 250 nM GDC0941 was also found to rescue the differential erastin sensitivity between Pten KO and Pten WT MEFs (Figure S4D). Importantly, while these doses of AZD5363 and GDC0941 reduced growth and their respective downstream effector proteins, they did not affect cell death on their own (Figures S4B, S3C, and S4D).

Several studies have recently uncovered a potential link between mTOR, which is regulated by numerous upstream inputs, and ferroptosis.^{41,76,86–89} Because AKT is a known regulator of mTOR, it was prudent to further explore whether mTOR had a role in modulating the relationship between PTEN, AKT, and ferroptosis. We next tested the mTOR inhibitors rapamycin and torin1 for their ability to influence PTEN-dependent sensitivity to erastin. MEFs were treated with erastin alone or in combination with either mTOR inhibitor at doses that substantially reduced mTOR kinase activity within the cell: 125 pM rapamycin or 5 nM torin1. Although the doses of rapamycin and torin1 may seem low, their IC₅₀ values are only 100 pM and 2 nM, respectively. Like AZD5363, the chosen doses of rapamycin and torin1 reduced growth and downstream effectors but did not cause cell death on their own, allowing examination of their effect on erastin sensitivity (Figures S4B, S3C, and S4E). While an enhanced response was observed when combining either mTOR inhibitor with erastin in Pten KO and WT MEFs, as indicated by the reduced erastin LD₉₀ values (Figure 4J, top), the differential erastin sensitivity between the Pten KO and WT genotypes remained (Figure 4J, top and bottom panel). Specifically, rapamycin and torin1 reduced the erastin LD₉₀ in Pten WT MEFs by 17% and 56%, respectively, and 37% and 28% in Pten KO MEFs, demonstrating a synergy between mTOR inhibitors and erastin. However, when cotreating with rapamycin, the fold change (KO/WT) in erastin LD₉₀ between Pten KO and Pten WT MEFs was only reduced from 2.5 to 2, and when cotreating with torin1 it actually increased to 4-fold (Figure 4J, bottom panel). Additionally, as noted above, mTOR inhibition does not appear to significantly reduce NRF2 or xCT protein abundance (Figures 4D and 4E). These findings suggest that, while mTOR appears to have a role in ferroptosis, as indicated by the synergy of mTOR inhibitors with erastin, it does not explain the PTEN-dependent difference in ferroptosis sensitivity, nor does it appear to have a role in mediating the increased expression of NRF2 and xCT observed in PTEN-deficient cells.

Because drugs were added in combination with erastin for the above studies, we sought to confirm that ferroptosis was still the mode of cell death and that the AKT or mTOR inhibitors were not triggering apoptosis or some other form of cell death in the presence of erastin. To confirm this, MEFs were treated with the same combination treatments as above, but this time with or without ferrostatin-1. While all combination treatments were again observed to enhance erastin sensitivity in Pten WT MEFs, only the AKT inhibitor was able to re-sensitize Pten KO MEFs to erastin and rescue the difference in erastin sensitivity (Figure S4F). Importantly, ferrostatin-1 was able to rescue cells from death under all of the various conditions, indicating that ferroptosis was still the mode of cell death.

GSK3β inhibition blocks the synergistic effect of combining AZD5363 with erastin

Given that AKT regulates many proteins in addition to GSK3β, we decided to further examine whether GSK3β mediates AKT regulation of ferroptosis sensitivity in the context of PTEN. If inhibition of AKT was increasing ferroptosis sensitivity in PTEN-deficient cells through its effect on promoting GSK3β activity, then one would hypothesize that inhibiting GSK3β in combination with AKT inhibitors would counteract the AKT inhibition and allow the PTEN-deficient cells to remain resistant to ferroptosis. To assess this, Pten WT and Pten KO MEFs were treated with erastin alone or in combination with 1 μM AZD5363 or in combination with AZD5363 and 250 nM CHIR99021, a potent and selective inhibitor of GSK3β. Upon addition of CHIR99021, AZD5363 was no longer able to rescue ferroptosis induction in erastin-treated Pten KO MEFs nor decrease xCT or NRF2 protein levels in these cells (Figures 4K, 4L, and S4G). Moreover, adding CHIR99021 to erastin treatment significantly reduced ferroptosis in Pten WT MEFs. CHIR99021 solo treatment also increased basal levels of NRF2 and xCT in Pten WT MEFs but had no effect on these proteins in Pten KO cells. Therefore, PTEN’s ability to induce ferroptosis upon oxidative stress triggered by erastin hinges on reduced AKT activity and subsequent GSK3β activation.

DISCUSSION

PTEN acts as a tumor suppressor through its lipid phosphatase function on the PI3K pathway. Activation of AKT because of loss of PTEN inhibits cell death induced by a variety of apoptosis stimulators.⁹⁰ Here we show that inactivation of PTEN in fibroblasts, cancer cell lines, and patient tumor samples upregulates the cystine/glutamate antiporter xCT through the AKT-GSK3β-β-TrCP-NRF2 axis, which promotes cystine import and generation of GSH, protecting cells from ferroptotic death. The gene encoding xCT, SLC7A11, is upregulated in PTEN-mutant tumors and tumor cell lines. We further show that PTEN-mutant cancer cell lines exhibit increased resistance to ferroptosis relative to PTEN WT cell lines. These findings suggest that selection for mutations of PTEN in cancer could be in part due to resistance to ferroptosis, which is likely activated by the harsh metabolic environment that occurs during tumor development.

Furthermore, we document that PTEN sensitizes cells to ferroptosis through a previously unknown regulation of cysteine metabolism by PTEN. We observed that PTEN loss upregulates xCT transcription and protein expression, resulting in aberrant cysteine metabolism, increased cystine flux into GSH synthesis, and accumulation of GSH pools. Artificially bolstering GSH levels with GEE was observed to cause erastin resistance in Pten WT MEFs, while reducing GSH pools with BSO was able to rescue erastin sensitivity in Pten KO MEFs. GPX4, along with its necessary cofactor GSH, acts as a phospholipid hydroperoxidase to neutralize lipid peroxides, negatively regulating ferroptosis.^10,23 Therefore, lower levels of GSH that render GPX4 inactive create a void in the antioxidant defenses of a cell and effectively increase susceptibility to ferroptosis in response to oxidative stress.²² These low GSH pools in PTEN WT cells help explain the sensitivity to ferroptosis we observe in these cells. Because we did not observe differential sensitivity to the GPX4 inhibitor RSL3, we postulate that PTEN-deficient cells have increased GPX4 function because of larger GSH abundance. Therefore, directly inhibiting GPX4 renders the excess GSH useless and ablates the differential resistance observed between PTEN-deficient and PTEN WT cells. Our observation links PTEN not just to ferroptosis but also to cysteine metabolism as a whole through its regulation of cystine import via xCT. Antioxidant and redox balance need to be coordinated with nutrient uptake of cystine because nutrients are required to make antioxidants, but cystine import is also utilized for other pathways, such as the trans-sulfuration pathway and choline oxidation pathway. The PI3K/PTEN pathway is a major regulator of anabolic metabolism, and our findings add yet another way through which PTEN is affecting metabolism and its interaction with redox balance. Further studies exploring the additional consequences of PTEN loss on cysteine metabolism would be an interesting avenue of research and could provide insights into how cancer cells may by utilizing PTEN loss to modulate metabolism to meet their aberrant metabolic needs.

Our findings reveal a mechanism for PTEN’s ability to modulate sensitivity to ferroptosis that is dependent on PTEN’s well-known regulation of AKT and its substrate GSK3β and highlight PTEN’s influence on the GSK3β-NRF2 axis, known to be important in ferroptosis regulation. Previous work has rigorously established that the transcription factor NRF2, when phosphorylated by GSK3β, is degraded by β-TrCP. In 2014, Rojo et al.⁷³ observed in MEFs and HEK293T cells that loss of PTEN leads to elevation of NRF2. Specifically, they found that PTEN elicits GSK3β-mediated phosphorylation of NRF2 via inactivating AKT, leading to β-TrCP-mediated NRF2 degradation and reduced tumor growth.⁷³ More recently, Wu et al.⁶⁴ found that the GSK3β/NRF2 signaling pathway modulated erastin-induced ferroptosis in breast cancer.⁶⁴ Here we established, for the first time, that PTEN regulates xCT expression, cystine and GSH levels, and ferroptotic death and that this regulation occurs via a PTEN/AKT/GSK3β/NRF2/xCT signaling pathway. We found that loss of PTEN increased NRF2 protein levels and transcription of xCT and other NRF2 target genes. This heightened expression of NRF2, as well as expression of xCT, was found to be rescued with AKT inhibition. Furthermore, AKT inhibition was observed to rescue the resistance to ferroptosis observed in Pten KO MEFs treated with erastin, but co-treatment with a GSK3β inhibitor prevented this rescue. Moreover, inhibition of GSK3β was able to rescue Pten WT MEFs from sensitivity to erastin-induced ferroptosis. These observations demonstrate that PTEN’s ability to regulate NRF2 and xCT expression and, subsequently, promote ferroptosis hinges on inhibiting AKT activity and activating GSK3β. Our findings were corroborated by analysis of public tumor data, which showed that a wide variety of tumor types have elevated expression of SLC7A11, which is associated with alteration of PTEN, PIK3CA, or AKT2. Interestingly, studies using immortal MCF10A breast epithelial cells have reported that oncogenic PIK3CA lowers SLC7A11 expression and that oncogenic AKT2 does not change the level of SLC7A11 expression, which, based on our analysis of breast cancer patient tumor samples, is not typical.^48,49

mTOR signaling via the mTORC1 and mTORC2 complexes, which are known to be upregulated because of loss of PTEN, has also been implicated in xCT and ferroptosis regulation. Several studies have shown that mTOR signaling suppresses ferroptosis or promotes xCT.^{42,75,77,85,86,88,89,91} Conversely, several studies have determined that mTOR complexes promote ferroptosis or inhibit xCT.^{41,48,76,92–94} The discrepancies in the findings of these various studies suggest that mTOR’s role in ferroptosis regulation may be context dependent. For example, Conlon et al.⁷⁶ mainly used ferroptosis inducers targeting xCT and focused on sarcoma lines, while several other studies utilized ferroptosis inducers that target GPX4 and focused on cell lines from a wider range of cancer types. Yi et al.⁸⁸ found that inhibiting the PI3K-AKT-mTOR axis sensitized cells to ferroptosis induced by the GPX4 inhibitor RSL3 because of reducing SREBP1 activity and antagonizing SREBP1-mediated lipogenesis and monosaturated fatty acid production. While we saw an increase in ferroptosis when we combined mTOR inhibitors with erastin, aligning with the findings that mTOR represses ferroptosis, rescue of the differential sensitivity observed between Pten WT and Pten KO MEFs was not observed. mTOR inhibitors also did not appear to modify NRF2 or xCT expression in our MEF system. Therefore, the known difference in mTOR activity between PTEN WT and PTEN-mutant cells does not explain the difference in sensitivity to ferroptosis between these two cell genotypes. These findings suggest that, while mTOR certainly has an important role in ferroptosis, AKT regulation of GSK3β, NRF2, xCT, and GSH is a major contributing factor in PTEN regulation of ferroptosis as well. Although inhibiting mTOR partially rescued ferroptosis induction in Pten KO and WT MEFs treated with erastin, inhibiting AKT was able to completely rescue induction in PTEN KO MEFS, perhaps by targeting the NRF2 and mTOR regulatory pathways. We postulate that, under oxidative conditions, PTEN may be promoting ferroptosis through its inhibition of AKT (1) by activating GSK3β, reducing expression of NRF2 and its target gene xCT, and depleting GSH pools and (2) through reducing mTOR activity and its downstream ferroptosis inhibitory mechanisms, as reported in other studies.

Several other tumor suppressors, such as p53 and BAP1, have also been reported previously to promote ferroptosis, contributing to their tumor-suppressive properties. Growing evidence, like these findings, suggest an innate, physiological role of ferroptosis in tumor suppression.^{14,35,36,95–102} In this study, we identified that the tumor suppressor PTEN facilitates cellular ferroptosis, which is likely a novel mode of tumor suppression for PTEN. Tumors generate copious amounts of ROS that can trigger ferroptosis. Therefore, it is plausible that one way through which PTEN suppresses tumor formation is by keeping GSH levels in check and inducing ferroptosis when cells accumulate high levels of ROS. Loss of PTEN and subsequent accumulation of NRF2, xCT, and GSH would provide a selective advantage for cells in a harsh environment and may explain why PTEN loss is beneficial to tumorigenesis and is frequently observed in cancer.

Limitations of the study

One limitation of our study is a lack of in vivo validation in a mouse model of PTEN-deficient cancer. Doing so would further validate tumor-suppressive role of ferroptosis. Additionally, while we think ferroptosis is relevant in many types of cancer, not all cancer types show PTEN loss correlating with increased xCT expression in TCGA studies or cancer cell lines. For example, prostate cancer does not show elevated xCT expression in PTEN-mutant cells. Another limitation is that we did not measure lipid peroxides in our studies.

STAR★METHODS

RESOURCE AVAILABILITY

Lead contact

Requests for further information or reagents should be directed to lead contact Ramon Parsons (Ramon.Parsons@mssm.edu).

Materials availability

This study did not deposit materials to repositories. Please contact the lead contact to request plasmids or reagents from this study.

Data and code availability

• Steady state and cystine flux metabolomics data have been deposited with Metabolomics WorkBench, and are publicly available as of the date of publication. Accession IDs are listed in the key resources table as Metabolomic WorkBench: ST002572 and Metabolomic Workbench: ST002573, respectively. The microarray data that was reanalyzed in this study was previously generated and published by Steinbach et al., and deposited with Gene Expression Omnibus GSE120478.⁷⁹.

• This paper does not report original code

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Actin	Sigma	A5316;RRID:AB_476743
Vinculin	Sigma	V9131; RRID:AB_477629
xCT	Abcam	175186; RRID:AB_2722749
NRF2	Cell Signaling	D1Z9C; RRID:AB_2715528
PTEN	Millipore	04-035;RRID:AB_1163491
pGSK3β Serine 9	Cell Signaling	9336; RRID:AB_331405
total GSK3β	Cell Signaling	D5C5Z; RRID:AB_2636978
pAKT Serine 473	Cell Signaling	9271; RRID:AB_329825
total AKT	Cell Signaling	9272;RRID:AB_329827
pPRAS40	Cell Signaling	D4D2; RRID:AB_2798140
total PRAS40	Cell Signaling	D23C7; RRID:AB_2225033
pS6 Serine 240/244	Cell Signaling	D68F8; RRID:AB_2798089
total S6	Cell Signaling	5G10; RRID:AB_331355
BACH1	Santa Cruz	sc-271211; RRID:AB_10608972
GPX4	abcam	ab125066; RRID:AB_10973901
Rabbit Secondary Antibody	Fisher	31460; RRID:AB_228341
Mouse Secondary Antibody	Fisher	31431; RRID:AB_10960845
Bacterial and virus strains
NEB Stable Competent E. coli	New England Biolabs	C3040
Stbl3 Competent E. coli	ThermoFisher	C737303
Adenovirus Null	Vector BioLabs	1300
Adenovirus Cre-Recombinase	Vector BioLabs	1045
Chemicals, peptides, and recombinant proteins
DMEM Media	corning	10–013-CV
RPMI-1640 Media	corning	10–040-CV
DMEM/F-12 Media	corning	10–090-CMR
MEM Media	corning	10–009-CVR
Ham’s F12 Media	corning	10–080-CV
Trypsin	Cell Signaling	7406S
Penicillin-Streptomycin	Fisher	30002ci
Fetal Bovine Serum	Gibco	A52567-01
Rapamycin	Sigma	R8781
Torin1	Sigma	475,991
CHIR99021	Sigma	SML1046
AZD5363	Selleck Chem	S8019
Glutathione Ethyl Ester	Sigma	G1401
L-Buthionine-sulfoximine	Sigma	B2515
3,3’ - 13C2-Cystine	Cambridge isotope laboratories	CLM-520-PK
Cystine	Sigma	56–89-3
Doxycycline hyclate	Sigma	D9891-25G
Sulfasalazine	Sigma	S0883
RSL3	Sigma	SML2234
Erastin	Sigma	E7781
Bafilomycin	Sigma	SML1661
GDC0941	Sigma	957,054–30-7
L-glutamine	Fisher	MT25005CI
polybrene	Millipore Sigma	TR-1003-G
Blasticidin HCl	ThermoFisher	A1113903
BamHI-HF	New England BioLabs	R3136S
EcoRI-HF	New England BioLabs	R3101S
MluI-HF	New England BioLabs	R3198S
BsmBI	New England BioLabs	R0739S
NotI-HF	New England BioLabs	R3189S
SuperFi DNA polymerase	ThermoFisher	12,351,010
T4 DNA ligase	New England BioLabs	M0202S
Necrostatin-1	Sigma	N9037
Z-VAD-FMK	Sigma	V116
Ferrostatin-1	Sigma	SML0583
N-Acetylcysteine	Sigma	A7250
Lipofectamine 2000	ThermoFisher	11,668,019
RNAiMAX	ThermoFisher	13,778,150
Puromycin	Sigma	P7255
Critical commercial assays
Qiagen RNeasy Plus Micro Kit	Qiagen	74,034
iScript cDNA Synthesis Kit	BIO-RAD	1,708,890
Gibson Assembly Kit	New England BioLabs	E5510S
luminescence-based Lonza kit	BioScience	LT07-418
Deposited data
Metabolism data	Metabolomics Workbench	Accession IDs Metabolomics Workbench: ST002572 (Steady State) and Metabolomics Workbench: ST002573 (Cystine Flux)
Microarray data	Gene Expression Omnibus (GEO)	GEO: GSE120478
Experimental models: Cell lines
HCC1419	ATCC	CRL-2326
HCC1395	ATCC	CRL-2325
NCIH226	ATCC	CRL-5826
NCIH446	ATCC	HTB-171
ZR-75-1	ATCC	CRL-1500
HCC1937	ATCC	CRL-2337
HCC1187	ATCC	CRL-2322
MDA-MB-468	ATCC	HTB-132
SKBR3	ATCC	HTB-30
MDA-MB-231	ATCC	CRM-HTB-26
U87 MG	ATCC	HTB-14
Daoy	ATCC	HTB-186
T98G	ATCC	CRL-1690
HCC1806	ATCC	CRL-2335
BT549	ATCC	HTB-122
NCIH520	ATCC	HTB-182
MX1	ATCC	CRL-2258
H4	ATCC	HTB-148
NCIH2066	ATCC	CRL-5917
NCIH2126	ATCC	CCL-256
NCIH2085	ATCC	CRL-5921
LN229	ATCC	CRL-2611
LN18	ATCC	CRL-2610
HEK293T	ATCC	CRL-1573
SUM149	Parsons Lab	N/A
P53 Dominant-Negative, Pten fl/fl MEFs	Parsons Lab	N/A
Primary Pten fl/fl MEFs	Parsons Lab	N/A
Experimental models: Organisms/strains
Mouse (PTENfl/fl, B6.129S4)	Jackson Laboratory	006,440
Oligonucleotides
Mouse SLC7A11 forward primer 5’ - TGGGTGGAACTGCTCGTAAT-3′	This paper	N/A
Mouse SLC7A11 reverse primer 5’ - AGGATGTAGCGTCCAAATGC-3′	This paper	N/A
Mouse GAPDH forward primer 5’ - TCACCAGGGCTGCTTTTAAC-3′	This paper	N/A
Mouse GAPDH reverse primer 5’ - AATGAAGGGGTCATTGATGG-3′	This paper	N/A
Myc-DDK-xCT forward 5′ - TCGAGCTTGCGTTGGATTGCA CCGGTGAGGAGATCTGCCGCCGC-3′	This paper	N/A
Myc-DDK-xCT reverse 5′ - GAGGCCAGATCTGGAATTCATTA AACCTTATCGTCGTCATCCTTGTAATCC-3′	This paper	N/A
PTEN_Fwd 5′-GAATCTCAGGATCCCCACCATGACA GCGATCATCAAAGAGATCGTTAG-3′	This paper	N/A
PTEN_Rev 5′GTCCTGAATTCGACTTTTGTAATTTGTG TATGCTGATCTTCATCAAAAGGTTCATTC TCTGGATCAGAGTCAG-3′	This paper	N/A
sgRNA guide sequence for CRISPR/Cas9 PTEN KO 5’ - ATTCTTCATACCAGGACCAG-3′	Human Brunello Genome-Wide Library	(Transcript:NM_001304718.1)
pTRI-Bla_F_new 5′-CCGAGGTTCTAGACGAGTTTAC-3′	This paper	N/A
pTRIPZ_seq_R 5′-TCTGACGTGGCAGCGCTCGCC-3′	This paper	N/A
siRNA xCT	Millipore Sigma	SASI_Hs01_00158008
siControl	Millipore Sigma	SIC002
Recombinant DNA
pCMV VSV-G	Addgene	RRID:Addgene_8454 (Stewart, S. et al.)103
pMD2.G	AddGene	RRID:Addgene_12259 (Trono et al., unpublished)
psPAX2	Addgene	RRID:Addgene_12260 (Trono et al., unpublished)
doxycycline-inducible pTRI-blas vector	Poulikos Poulikakos Lab, Tisch Cancer Insitute	N/A
pcDNA3.1 PTEN plasmid	Parsons Lab, Tisch Cancer Institute	N/A
LentiCRISPRv2 blast	Addgene	RRID:Addgene_98293 (Stringer, B. et al.)104
TRE3G-MYC-Puro	David Dominguez-Sola Lab, Tisch Cancer Insitute	N/A
Myc-DDK- tagged xCT pLVX-Puro vector	Paul Mischel lab, University of California San Diego	N/A
Software and algorithms
Prism 9	GraphPad	https://www.graphpad.com/scientific-software/prism/; RRID: SCR_002798
Gene Set Enrichment Analysis (GSEA) 4.1.0 software	GSEA	https://www.pnas.org/content/102/43/15545RRID:SCR_003199
ImageJ	Schneider et al., 2,012,105	https://ImageJ.nih.gov/ij/download.html RRID:SCR_003070
cBio Portal	cBio Portal	RRID:SCR_014555
Incucyte Zoom	Sartorius	https://www.sartorius.com/en/products/live-cell-imaging-analysis/live-cell-analysis-software RRID:SCR_019874
Snapgene	Snapgene	https://www.snapgene.com/support/downloads RRID:SCR_015052
Other
ECL	Fisher	34,080
PVDF Membrane	Fisher	ipvh00010
Autoradiography Film	Denville	E3018
NuPAGE 4–12% Bis-Tris gels	ThermoFisher	NP0336BOX
SYBR Green Master Mix	BiMake	B21202
ROX reference dye	BiMake	B21202

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Mouse embryonic fibroblasts (MEFs)

All animal experiments were performed under the guidelines of the Institutional Animal Care and Use Committee (IACUC) at Mount Sinai.^103–105 A cross between 12 week old, B6.129S4 Pten fl/fl mice obtained from Jackson Laboratory was set up and 14 days later both male and female embryos were harvested from the pregnant females. Highly vascularized sections of the embryos were removed - head, extremities, and liver. The remainder was minced via a scalpel and resuspended in 0.25% trypsin using a pipet. The cells were then incubated for 10 min in an incubator at 37°C with 5% CO₂ before being further resuspended into single cell suspension. The cells were then spun down at 1500 RPM for 5 min to remove the trypsin, after which they were resuspended in 10mL of fresh media and transferred to a 10cm dish. The cells were passaged once before treating for 24 h with 3 μg/mL polybrene to enhance infection efficiency and 1:1000 adenovirus Cre-Recombinase or adenovirus null to generate Pten KO MEFs and Pten WT MEFs, respectively. After two additional passages and confirmation of PTEN KO via Western blot, the MEF pair was used for experiments up to passage 8. Prior to the cross, mice were housed in groups of the same gender from the same litter with access to food, water, and nesting material. Cages were kept in a temperature controlled room (20–26°C) with a semi-natural light cycle of 12 : 12 h of light: dark. Mice were randomly selected for the cross.

Cell lines

HCC1419, HCC1395, NCIH226, NCIH446, ZR-75-1, HCC1937, HCC1187, HCC1806, BT549, and NCIH520 cells were grown in RPMI-1640 media supplemented with 10% fetal bovine serum (FBS), and 1% penicillin-streptomycin (pen/strep). H4, MDA-MB-468, SKBR3, and MDA-MB-231 cells were grown in DMEM media supplemented with 10% FBS and 1% pen/strep. U87 MG, Daoy, and T98G cells were grown in MEM media supplemented with 10% FBS and 1% pen/strep. SUM149 cells were grown in Ham’s F12 media supplemented with 5% FBS, 1% pen/strep, 500 ng/mL hydrocortisone, and 0.01 mg/mL insulin. MX1 cells were grown in DMEM/F-12 media supplemented with 10% FBS and 1% pen/strep. MEFs were grown in DMEM media supplemented with 10% FBS, 1% pen strep and 2mM L-glutamine for a total of 6 mM LN18 and LN229 cells were grown in DMEM supplemented with 5% FBS and 1% pen/strep. NCIH2085 cells was grown in RPMI-1640 media supplemented with 5% FBS and 1% pen/strep. NCIH2066 and NCIH2126 cells was grown in DMEM/F-12 media supplemented with 5% FBS, 1% pen/strep, 0.005 mg/mL insulin, 0.01 mg/mL transferrin, 30nM sodium selenite, 10 nM hydrocortisone, 10nM beta-estradiol, and 2 mM L-glutamine for a total of 4.5 mM. Cell lines listed above were obtained directly from ATCC and tested monthly for mycoplasma. All of the above cell lines were cultured in a 37°C incubator with 5% CO₂ and humidity, and determined mycoplasma negative by luminescence-based Lonza kit. Freezing method: Cell lines were frozen down in 1 mL of their respective media with 5–10% of DMSO in a cryogenic tube and then placed in a container with isopropanol for 24 h in −80°C before being transferred into liquid nitrogen for long term storage. Thawing method: cells were thawed rapidly in a 37°C water bath and then transferred into a t25 flask with their respective media. 24 h later the media was replaced and the cells were then cultured as above.

METHOD DETAILS

Immunoblotting

Cells were collected and lysed using 2X Laemmli sample buffer supplemented with 10% β-mercaptoethanol, sonicated, and boiled at 100°C for 5 min before separation via SDS-PAGE on NuPAGE 4–12% Bis-Tris gels at 100 V. Time varied by molecular weight of protein being examined. Subsequently the gels were subjected to wet-transfer to a PVDF membrane at 360 mA for 1 h. Membranes were blocked in 10% non-fat milk for 30 min in TBST before being incubated overnight at 4°C in primary antibody. They were then washed 3 times in TBST for 10 min each and then incubated at room temperature for an hour in secondary antibody in 5% non-fat milk in TBST. Membranes were washed 3 more times for 5 min each before being developed using ECL and autoradiography film. Antibody dilutions were done as directed by the manufacturer. N = 3 unless otherwise noted. Quantification was performed using ImageJ via the densitom etry macro.

Proliferation assay

Cells were seeded at 10% confluency into 96-well plates and confluency readings over time were monitored on an Incucyte Zoom to track growth and generate growth curves. The hardware took images of each well every 6 h using a 43 objective and the software, using a unique processing definition for each cell line, distinguished cell from background and calculated the confluency of the wells at each time point. GI 50s were calculated by determining the dose of drug that resulted in a 50% reduction in growth, using the non-linear regression, dose response inhibition feature on GraphPad PRISM9. EC50s were calculated by determining the concentration of metabolite needed to achieve 50% of maximum growth, using the non-linear regression, dose response - stimulation feature on GraphPad PRISM9. N = 3.

DRAQ7 cell death assay

Cells were seeded at 80% confluency into 96-well plates (corning 720,089) and treated with 1.5μM DRAQ7 to monitor cell death and placed in an Incucyte Zoom (Essen Biosciences). The hardware took images of each well every 6 h using a 4× objective and scanned for red fluorescence, while the software, using a unique processing definition for each cell line, distinguished cell from background to calculate the confluency of the wells at each time point and identify the number of dead cells in the well. The number of dead cells in a well was reported as the “red dot count”. Using the confluency readout at time 0 when plates were first seeded with a known cell number, the number of cells per percent confluence was calculated and used to determine total cell count for the wells at any time point. Percent cell death was calculated by red dot count/total number of cells (calculated from confluency data). LD50 and LD90 were determined by calculating the dose of a drug that resulted in 50% and 90%, cell death, respectively, using the non-linear regression, dose response - inhibition feature on GraphPad PRISM9. N = 3.

Dose response titrationsS

ensitivity to various drugs or metabolite deprivation was determined by dose-response titrations and monitored in an Incucyte zoom. If cell death was being investigated, cells would be seeded at 80% confluency in a 96-well plate; if growth was being monitored then cells were seeded at 10% confluency in a 96-well plate. The titration was conducted as follows: 300μL of drug or metabolite-containing media was added to one of the wells and the remainder of the wells contained 150μL of media with DMSO or no metabolite. Next, 1:1 serial dilutions were then conducted, resulting in a gradient with each well having ½ of the concentration of drug or metabolite as the one proceeding it in the titration. LD50s, EC50s, and GI 50s were then calculated as described in the DRAQ7 Cell Death Assay section or Proliferation Assay section of the STAR Methods. All drugs were dissolved in DMSO except NAC, which was dissolved in water. N = 3.

Crystal violet Cell Death Assay

Cells were seeded 24 h before treatment so that they would be 100% confluent the next day. After 48–96 h of treatment depending on the drug type, cells were then stained with crystal violet for 30 min to monitor cell death that occurred. Any loss of confluency as indicated by lack of crystal violet staining is indicative of cell death. N = 1 unless otherwise noted. Quantification was performed using the ImageJ plugin ColonyArea as described in Guzman et al.¹⁰⁶

Microarray

This microarray data was previously generated by Steinbach et al.⁷⁹ 250 ng of RNA was extracted from Pten WT and Pten KO MEFs and processed according to the Ambion Whole Transcript (WT) Expression Array (Life Technologies) protocol. The samples were analyzed by Applied Biosystems utilizing Affymetrix GeneChips Mouse Gene 2.0 ST Arrays. Results were normalized and checked for quality control, including signal intensity and consistency of probe hybridization. N = 4.

RNA extraction and RT-qPCR

RNA was extracted using the Qiagen RNeasy Plus Micro Kit according to the manufacturer’s protocol. RNA was then converted to cDNA using the iScript cDNA Synthesis Kit following the manufacturer’s protocol. qPCR reaction was set up as follows: 10 μL 2× Bimake SYBR Green Master Mix, 0.4 μL ROX reference dye, 1 μL of 5uM forward primer, 1 μL of 5uM reverse primer, 10ng of cDNA, and add distilled water up to a final volume of 20mL. Mouse SLC7A11 forward primer 5’ - TGGGTGGAACTGCTCGTAAT-3′, Mouse SLC7A11 reverse primer: 5’ - AGGATGTAGCGTCCAAATGC-3′, Mouse GAPDH forward primer: 5’ - TCACCAGGGCTGCTTTTAAC-3′, Mouse GAPDH reverse primer: 5’ - AATGAAGGGGTCATTGATGG-3′. Results were analyzed by calculating the relative expression from the C_T value (ΔΔ C_T) and GAPDH was used as the control gene. N = 3.

Analysis of patient tumor samples for differential SLC7A11 expression

The SLC7A11 expression data (determined by RNAseq and reported in log₂) from patient tumor samples from breast, brain, and lung cohorts were generated by The Cancer Genome Atlas (TCGA) Research Network (https://www.cancer.gov/tcga) and analyzed using cBio Portal. The SLC7A11 expression data from the pan-cancer patient tumor sample cohort was generated by the ICGC/TCGA Pan-Cancer Analysis of Whole Genomes Consortium and analyzed using cBio Portal. Within each cohort, patient tumor samples were categorized by mutation status: Unaltered - wild-type PIK3CA AKT1 AKT2 and PTEN, PTEN Mut - PTEN mutation, PTEN HetLoss - loss of one allele, PTEN HomDel - loss of both alleles, PIK3CA Mut - PIK3CA mutation, PIK3CA Amp - PIK3CA amplification, PIK3CA gain - PIK3CA gain of function, PIK3CA AKT1 or AKT2 Altered - mutation amplification or gain of function, or PTEN Altered - mutation heterozygous loss or homozygous deletion. The mean expression of SLC7A11 for each group was compared and a difference in mean SLC7A11 expression was determined by a Kruskal-Wallis test. N varied by dataset. The GPX4 expression data from the pan-cancer patient tumor sample cohort was generated by the ICGC/TCGA Pan-Cancer Analysis of Whole Genomes Consortium and analyzed using cBio Portal. Within each cohort, patient tumor samples were categorized by mutation status: Unaltered – wild-type PIK3CA AKT1 AKT2 and PTEN, PTEN Mut - PTEN mutation, PTEN HetLoss - loss of one allele, or PTEN HomDel - loss of both alleles. The mean expression of GPX4 for each group was compared and a difference in mean SLC7A11 expression was determined by a Kruskal-Wallis test. N varied by dataset.

Cloning of dox-inducible xCT MDA-MB-231 cells

Vector design: The stably expressed Myc-DDK-tagged xCT pLVX-Puro vector was a gift from the Paul Mischel lab at the University of California San Diego. The xCT segment of this plasmid was amplified by PCR and cloned into a dox-inducible backbone (TRE3G-MYC-Puro, a gift from the David Dominguez-Sola Lab at Mount Sinai) using the Gibson Assembly kit and restriction enzymes MluI-HF and Notl-HF. NEB stable bacteria were transformed to harvest the plasmid and HEK293T cells were utilized for plasmid transfection using 6 μg psPAX2, 1.5 μg pCMV VSV-G, and 7.5 μg plasmid with lipofectamine 2000 as directed by the manufacturer’s protocol. The supernatant was collected 48 h after transfection and filtered with a 0.45 μm syringe filter before transducing parental MDA-MB-231 cells. MDA-MB-231 cells were seeded at 30% confluency and transduced for 48 h with the dox-inducible xCT lentiviral expression vector (1:10,000) with 10 μg/mL polybrene and subjected to 1 μg/mL puromycin selection for an additional 48 h. The stable cell line was collected and expanded as a heterogeneous pool of puromycin-resistant cells. PCR primers: Myc-DDK-xCT forward 5′ - TCGAGCTTGCGTTGGATTGCACCGGTGAGGAGATCTGCCGCCGC-3′, Myc-DDK-xCT reverse 5′ - GAGGCCAGATCTGGAATTCATTAAACCTTATCGTCGTCATCCTTGTAATCC-3′. Snapgene was used for vector design.

siRNA transfection of MDA-MB-468 cells

Cells were seeded at 90% confluency and transfected for 24 h with 25pmol of siRNA Control or siRNA xCT using 7.5 uL of Lipofectamine RNAiMAX. siRNA was purchased from Millipore Sigma (SIC002 and SASI_Hs01_00158008).

Cloning of U87 MG cells

To generate PTEN and EV lentiviruses, HEK293T cells were transfected with 6 μg psPAX2, 1.5 μg pCMV VSV-G, and either 7.5 μg pTRI-Bla V5/6xHis EV or PTEN plasmid with Lipofectamine 2000 according to the manufacturer’s instructions. The supernatant was collected 48 h after transfection and filtered with a 0.45 μm syringe filter. Parental U87 MG cells were then transduced with 1:8000 lentivirus with 10 μg/mL polybrene. After 48 h transduction, stably-infected cells were selected with 4 μg/mL Blasticidin S HCl for 7 days. Stable cell lines were collected and expanded as a heterogeneous pool of blasticidin-resistant cells. The PTEN cDNA insert was cloned into the doxycycline-inducible pTRI-blas vector, a gift from Dr. Poulikos Poulikakos and Dr. Stuart Aaronson. A carboxyterminus V5 peptide and 6x His tag (GKPIPNPLLGLDSTRTGHHHHHH) was cloned between the EcoRI and MluI restriction sites of the pTRI-blast backbone, resulting in the pTRI-Bla V5/6xHis vector used for this study. The parental backbone was digested with EcoRI and MluI and the tag sequence was ligated from dsDNA oligos 5′-AATTCAAGGGCAATGGCGGCTTCGAAGGTAAGCC TATCCCTAACCCTCTCCTCGGTCTCGATTCTACGCGTACCGGTCATCATCACCATCACCATTGAA-3’. The PTEN insert was cloned in-frame with V5/6xHis tag from pcDNA3.1 PTEN plasmid that was generated in the lab and amplified with primers “PTEN_Fwd” (5′-GAATCTCAGGATCCCCACCATGACAGCGATCATCAAAGAGATCGTTAG-3′) and “PTEN_Rev”(5′GTCCTGAATTCGACTTTTGTAATTTGTGTATGCTGATCTTCATCAAAAGGTTCATTCTCTGGATCAGAGTCAG-3′) using the Platinum SuperFi DNA polymerase.¹⁰⁷ Both the PTEN insert and the TRI-Bla V5/6xHis plasmid were digested with and ligated with T4 DNA ligase to yield pTRI-Bla V5/6xHis PTEN. The insert sequence was confirmed with Sanger sequencing with GENEWIZ (South Plainfield, NJ) with the following primers: “pTRI-Bla_F_new” (5′-CCGAGGTTCTAGACGAGTTTAC-3′) and “pTRIPZ_seq_R′′ 5′-TCTGACGTGGCAGCGCTCGCC-3′. The pTRI-Bla V5/6xHis plasmid was used as the EV. Stbl3 bacteria were transformed to harvest the plasmid.

Metabolite labeling

For cystine flux metabolomics, media void of cystine was supplemented with 3,3′−¹³C2-cystine to 63 mg/L. On day 0, Pten WT and Pten KO MEFs were seeded at 80% confluency in standard media in 10cm plates. After 24 h, the standard media was replaced with the heavy isotope labeled media for 12 h before metabolites were extracted as described in the Metabolite Extraction section in Star Methods. N = 5.

Metabolite Extraction

As previously described by Yuan, et al. in 2012, metabolites were extracted as follows: Media was fully removed from the plates, and cells were incubated in 2.5 mL of ice-cold 80% methanol (stored at −80°C before use) for 20 min at −80°C to ensure the unstable metabolites were not degraded.¹⁰⁸ Cells were then scraped into falcon tubes and centrifuged. The soluble supernatant was collected and two more extractions on the insoluble pellet using 500 μL of ice-cold methanol each were performed over dry ice. Each time the soluble supernatant was collected after centrifuging. The pooled extractions were then dried to completion by speed-vac and over-night shipped on dry ice to the BIDMC Mass Spectrometry Core for LC-MS/MS. N = 3–5.

Targeted mass spectrometry

Beth Israel Deaconess Medical Center at Harvard University, led by Dr. John Asara, performed the mass spectrometry on the metabolite extracts. First, 20μL of HPLC grade water was used to resuspend the samples for mass spec before injecting 5 μL into a hybrid 5500 QTRAP triple quadruple mass spectrometer that is paired to a Shimadzu Prominence UFLC HPLC system. Selective reaction monitoring, SRM, of 259 endogenous water-soluble metabolites was performed for steady state analysis. The positive ion mode ESI voltage was +4900 while the negative ion mode was −4500. Dwell time was set to 3ms per SRM transition and 1.55 s was the total cycle time. HILIC (hydrophilic interaction chromatography) delivered samples to the mass spectrometer using a Waters 4.6mm i.d × 10 cm Amide XBridge column at a rate of 400 μL/min. Gradients were run as the following: 85% buffer B – HPLC grade acetonitrile – to 30% buffer B from 0 – 3 min, then 30% buffer B to 2% buffer B from minute 3 to 12, then 2% buffer B held from minute 12 to 15, followed by 2% buffer B to 85% buffer B from minute 15 to 16, then 85% buffer B held from minute 16 to 23 in order to re-equilibrate the column. Buffer A – 20mM ammonium hydroxide/20mM ammonium acetate, pH – 9.0, in 95:5 water:acetonitrile. Retention time 15–20 s. MultiQuant v2.1 software was used to in integrate peak areas for each metabolite SRM transition. For cystine flux, percent accumulation of heavy isotope labeled metabolite was calculated as follows: labeled metabolite abundance/total metabolite abundance. N = 3–5. For more detailed methods for the targeted mass spectrometry performed by the mass spec core at Beth Israel Deaconess Medical Center please see their original methods paper on this protocol published by John Asara.¹⁰⁸

CRISPR/Cas9 PTEN KO lentivirus generation and cloning of PTEN KO MDA-MB-231 cells

The sgRNA guide sequence 5′-ATTCTTCATACCAGGACCAG-3′, from Human Brunello Genome-Wide Library, targeting Exon 8 of PTEN (Transcript:NM_001304718.1) was subcloned into the lentiviral transfer plasmid lentiCRISPRv2 blast using the restriction enzyme BsmBI and following the protocol published by Sanjana et al. and Shalem et al..^109,110 Stbl3 bacteria were transformed to harvest the plasmid. Lentiviral particles were generated in HEK293T cells from the co-transfection of the resultant lentiCRISPRv2 blast plasmid with the packaging plasmids pMD2.G and psPAX2 with Lipofectamine 2000 according to the manufacturer’s instructions. The supernatant was collected 48 h after transfection and filtered with a 0.45 μm syringe filter. Parental MDA-MB-231 cells were then transduced with 1:100,000 lentivirus with 10 mg/mL polybrene. After 48 h transduction, stably-infected cells were selected with 4 μg/mL Blasticidin S HCl for 7 days. Stable cell lines were collected and expanded as a heterogeneous pool of blasticidin-resistant cells. LentiCRISPRv2 blast was a gift from Brett Stringer (Addgene plasmid # 98293; http://n2t.net/addgene:98293; RRID:Addgene_98293). pMD2.G was a gift from Didier Trono (Addgene plasmid # 12259; http://n2t.net/addgene:12259; RRID: Addgene_12259). psPAX2 was a gift from Didier Trono (Addgene plasmid # 12260; http://n2t.net/addgene:12260; RRID: Addgene_12260).

Gene Set Enrichment Analysis (GSEA)

Pathway analyses were performed using pre-ranked Gene Set Enrichment Analysis (GSEA) 4.1.0 software. Log₂ fold change microarray gene expression between Pten KO and Pten WT MEFs was analyzed. N = 4. Default settings were used: 1000 permutations, phenotype permutation type, meandiv normalization mode, no_balance randomization mode, weighted enrichment statistic, Signal2Noise, max size 500, min size 15. The previously published NFE2L2 (NRF2) gene set was accessed from: https://pubmed.ncbi.nlm.nih.gov/27088724/. E2F and mTORC1 gene signatures were from the Hallmarks of Cancer gene set group. For further inquiry into the development, design, and use of the GSEA software please refer to https://www.pnas.org/content/102/43/15545.

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical analyses were conducted in GraphPad Prism using its built-in statistics tools and the statistical details of every experiment can be found in the figure legends, including the statistical test, error bars, and value of n. Significance was defined as a p value <0.5 for comparison between groups and is indicated in the figures.

Highlights.

• PTEN’s ability to enhance ferroptosis contributes to its tumor suppressor function

• PTEN antagonizes cysteine metabolism by restraining expression of xCT

• Activated AKT because of PTEN loss upregulates xCT through the GSK3β-NRF2 axis

• PTEN-mutant cells are found to be resistant to ferroptosis because of elevated xCT