Ferroptosis in Neurons and Cancer Cells Is Similar But Differentially Regulated by Histone Deacetylase Inhibitors

Visual Abstract


Introduction
Cancer is among the leading causes of death worldwide, and its incidence is expected to increase within the next decades. Accordingly, there are intense efforts to develop novel small molecules that can induce death of cancer cells. In an attempt to target tumor cells with activating mutations of RAS, which are present in around a third of all cancers, Dixon et al. (2012) identified the small molecule chemotherapeutic agent erastin. The authors characterized the mechanisms of erastin-induced death and found that cells die in an iron-dependent form of non-apoptotic cell death they called ferroptosis. They further showed that erastin induces ferroptosis by selectively inhibiting the plasma membrane transport of cyst(e)ine via a well-characterized cystine/glutamate exchanger (X c Ϫ ). Cyst(e)ine depletion leads to decreased levels of glutathione (Bridges et al., 2012;Lewerenz et al., 2013), culminating in cell death due to the production of reactive lipid species (Tan et al., 1998;Yang and Stockwell, 2016).
A major challenge in effective cancer chemotherapy is to reduce side effects on the central nervous system (e.g., chemo brain-deficits in cognitive function, memory, and attention) or peripheral nervous system (e.g., neuropathic pain) toxicities (Banach et al., 2017;Vitali et al., 2017). These toxicities can reduce the quality of life and functional status even in circumstances where chemotherapy effectively neutralizes the tumor (Kerckhove et al., 2017).
To understand the potential toxicities of erastin in CNS neurons, we here examine the signaling pathways engaged by erastin in post-mitotic cortical neurons and compare them to those activated in HT1080 fibrosarcoma cells.

Animals
All animal procedures were approved by the Weill Cornell Medicine Institutional Animal Care and Use Committee (Approval #0707-633A) and conducted in accordance with the NIH Guide for the Care and Use of Laboratory Animals and ARRIVE guidelines. Mice were purchased from Charles River Laboratories and housed at 20 -22°C, 30 -70% humidity, under a 12 h light/dark cycle, with food (PicoLab Rodent diet 5053, LabDiet) and water ad libitum.

Cell culture
Primary cortical neurons were obtained from CD-1/ICR mice of either sex at embryonic day 14.5. Briefly, cortices were dissected, homogenized, and plated in poly-Dlysine-coated plates in minimum essential medium containing 10% fetal bovine serum, 5% horse serum, and 1% penicillin/streptomycin (1,000,000 cells/ml).
Primary dorsal root ganglia (DRG) neurons were obtained from C57BL/6 mice of either sex at 6 weeks of age. Briefly, the spinal column was isolated by dissection and the spinal cord removed by hydraulic extrusion. The spinal cord was split longitudinally and the DRGs located and removed. DRGs were dissociated with collagenase, plated onto poly-L-lysine/laminin-coated plates in DMEM/ nutrient mixture F-12 supplemented with 1ϫ N1 and 10% horse serum (1,000,000 cells/ml).
Immortalized hippocampal neuroblasts (HT22 cells) were cultured in DMEM containing 10% fetal bovine serum and 1% penicillin/streptomycin (50,000 cells/ml). HT1080 cells were obtained from American type Culture Collection and cultured in DMEM containing 10% fetal bovine serum, 1% penicillin/streptomycin, and 1% nonessential amino acids (25,000 cells/ml). Hep3B cells were obtained from American type Culture Collection and cultured in Eagle's minimum essential medium containing 10% fetal bovine serum and 1% penicillin/streptomycin (50,000 cells/ml). SH-SY5Y cells were also obtained from American type Culture Collection and cultured in DMEM/ nutrient mixture F-12 containing 10% fetal bovine serum and 1% penicillin/streptomycin (50,000 cells/ml). Cell lines were treated at 24 h when density reached 70% confluency.
All cells were cultured at 37°C in a humidified 5% CO 2 atmosphere.

Cell viability
Cell viability was determined at 22-26 h following erastin or glutamate analog homocysteate (HCA) exposure using MTT assay, a colorimetric assay of cell metabolic activity. We measured the plates at SpectraMax Plus Microplate Reader using SoftMax Pro v4.7.1 (both Molecular Devices). The results of population, quantitative assays of cell viability (MTT) were verified by qualitative LIVE/DEAD assay and fluorescence microscopy at Nikon Eclipse TS100 microscope using Nikon DS-L3 (Nikon Instruments).

Statistical analysis
All data represent biological replicates. For the MTT assay, each biological replicate is the mean of four technical replicates. Normality was evaluated with the Kolmogorov-Smirnov test and variance homogeneity using the Levené test. When data were normally distributed and variance was homogeneous, we performed one-way ANOVA followed by the post hoc Bonferroni test. In case one of the criteria was not met, the Kruskal-Wallis test was performed followed by the post hoc Mann-Whitney U test with ␣ correction according to Bonferroni to adjust for the inflation of type I error due to multiple testing. Data are represented as mean Ϯ SD except for nonparametric data, where medians are given. A value of p Ͻ 0.05 was considered statistically significant. For the Kruskal-Wallis test followed by Mann-Whitney U, p ϭ 0.05/k was used, with k as the number of single hypotheses. K ϭ 2 for gene expression experiments (comparison of 2 different concentrations vs vehicle-treated cells), k ϭ 4 (comparison of 3 different concentrations vs vehicle-treated cells) for all nonparametric data of drug treatments, except for Necrostatin-1, Scriptaid, and U0126, where k ϭ 12 (comparison of 4 different concentrations vs vehicle-treated cells and additional four comparisons vs inactive analog), and pRIP1, where k ϭ 9 (all vs 0 h treatment and Necrostatin-1 vs same condition without Necrostatin-1). Thus ␣ ϭ 0.025 for two comparisons, ␣ ϭ 0.0125 for four comparisons, ␣ ϭ 0.0056 for 9 comparisons, and ␣ ϭ 0.0042 for 12 comparisons was considered statistically significant. To analyze contingency tables, Fisher's exact test was used. Detailed statistical analyses can be found in the extended data (Figs. 3-1, 5-1, 7-1, 9-1, 10-1, 13-1, 13-2, 13-3, 13-4, and 14-1). All statistical analyses were performed with IBM SPSS v23 (RRID:SCR_002865).

Erastin-induced ferroptosis in cancer cells is similar to erastin-and glutamate-induced toxicity in neurons
Ferroptosis has been shown to be induced by cyst(e)ine deprivation ( Fig. 1A; Bridges et al., 2012;Lewerenz et al., 2013) and is operationally defined by sensitivity to a panel of inhibitors targeting macromolecular synthesis (e.g., cycloheximide), reactive lipids (e.g., ferrostatin-1, N-acetylcysteine, Trolox), iron (e.g., DFO), and ERK signaling (e.g., U0126;Dixon et al., 2012Dixon et al., , 2014. The initial goal of the current studies was to determine whether erastin-induced ferroptosis in cancer cells occurs via mechanisms that are similar or distinct from those induced by erastin in primary neurons. Incubation of HT1080 fibrosarcoma cells with increasing concentrations of erastin revealed that doses of erastin used previously (10 M) to study the mechanisms of death led to a 95% reduction in cell survival (Fig. 1B, left), a higher degree of cell death than the LD 50 usually targeted for toxicological studies. Studies of chemotherapeutic agents at their LD 50 allows one to discern the protective mechanism from the detrimental effects of the chemical and molecular manipulations of cellular pathways. Accordingly, we adjusted the density of HT1080 cells to 70% confluence before the treatment with different doses of erastin, and found that 1 M is around the LD 50 in this particular cell line (Fig. 1B,right). In primary neurons, we found the LD 50 at around 5 M (Fig. 1C, left). Although cortical neuron dysfunction/death may be responsible for "chemo-brain", DRG sensory neurons dysfunction/death is likely responsible for chemotherapyinduced neuropathy. Of note, we also found that erastininduced death of DRG sensory neurons (Fig. 1D). Because it is better characterized, we focused on erastininduced death in cortical neurons for the remainder of our studies.
We performed a systematic analysis of ferroptosis inhibitors of erastin-induced death and confirmed that they prevent erastin-induced toxicity in both HT1080 cells and primary neurons (Figs. 2, 3). Moreover, live/dead assays, representing a visible, fluorescence microscopic measure of cell death in single cells, revealed results that were similar to quantitative measurements of cell death performed with MTT assays (Fig. 2).
Interestingly, cyst(e)ine or glutathione depletion has been elucidated as an in vitro model of neuronal death in the late 1980s, where glutamate or its analogs were used to induce cell death in cultured neurons (at 2 d in vitro) via a non-receptor-mediated mechanism involving inhibition of the System X c Ϫ (cystine/glutamate) antiporter (Fig. 1A). This model has been leveraged to understand how cystine deprivation leads to death via oxidative stress (Murphy et al., 1989;Ratan et al., 1994a, b). To determine whether glutamate analog (HCA) and erastin induce neuronal death via similar pathways, we examined the ability of ferroptosis inhibitors to abrogate glutamate analog (HCA)-induced death. We found that glutamate analog (HCA)-induced death (at the LD 50 of 5 mM; Fig. 1C, right) was abrogated by inhibitors of ferroptosis (Figs. 2, 3).

Ferroptosis in cancer cells and neurons is abrogated by inhibitors of autophagy and necroptosis
There are many other modes of cell death (Galluzzi et al., 2018). Apoptosis is a caspase-dependent mode of regulated cell death initiated by perturbations of the intracellular (intrinsic apoptosis) or extracellular (extrinsic ap-New Research optosis) microenvironment, whereas parthanatos is pathway-dependent on apoptosis-inducing factor that is induced by poly(ADP-ribose) polymerase 1 hyperactivation. We found that erastin-induced death of HT1080 fibrosarcoma cells or primary neurons, as well as glutamate analog (HCA)-induced death of primary neurons, was not altered by inhibitors of apoptosis or parthanatos (Figs. 4-7). Unexpectedly, several modulators of autophagy (Bafilomycin A1, Chloroquine, and rapamycin; (Pasquier, 2016) abrogated erastin-induced cell death in HT1080 fibrosarcoma cells. In primary neurons, the autophagy inducer rapamycin decreased both erastin-and glutamate analog (HCA)-induced toxicity. In addition, we found a significant increase in viability with the autophagy inhibitors 3methyladenine and Bafilomycin A1 in glutamate analog (HCA)-induced neuronal death (Figs. 8,9).
Necroptosis is another mode of regulated cell death that depends on the activation of receptor-interacting proteins (RIPs;Degterev et al., 2005;Sun et al., 2012). The RIP1 inhibitor Necrostatin-1 abrogated erastin-induced cell death in HT1080 cells and primary neurons as well as glutamate analog (HCA)-induced neuronal toxicity. In addition, the RIP3 inhibitor GSK872 blocked neuronal cell death from erastin and glutamate analog (HCA) and the inhibitor of mixed lineage kinase domain-like protein (MLKL), necrosulfonamide, prevented erastin toxicity in HT1080 cells (Figs. 6,7).
The specificity of Necrostatin-1 against RIP1 was suggested by a structural analog (Necrostatin-1i), with no activity toward RIP1, that did not block erastin-or glutamate analog (HCA)-induced toxicity. Moreover, as RIP1 kinase activity is required to execute necroptosis (Berger et al., 2014), we assessed the ability of erastin to induce active RIP kinase using an antibody against a known RIP1 autophosphorylation site at serine 166 . We found a Necrostatin-1-sensitive increase in phospho-RIP1 following erastin and glutamate analog (HCA) treatment in HT22 mouse hippocampal neuronal cells (Fig. 10). These cells also died in response to both inducers of cyst(e)ine/glutathione depletion.
To elucidate whether erastin-and glutamate analog (HCA)-induced toxicities are similar or distinct modes of cell death (Fig. 11A), we performed statistical comparisons between the inhibitor profiles (Fig. 11B). First, we found that erastin toxicity in HT1080 fibrosarcoma cells can be considered ferroptosis as previously described (Dixon et al., 2012; Fisher's exact test, two-tailed, p ϭ 0.202), despite the ability of the inducers of autophagy or the inhibitors of necroptosis to prevent "ferroptosis" at the lower dose (1 M) in HT1080 cells. Second, erastin toxicity in HT1080 fibrosarcoma cells and in neurons can be considered mechanistically similar (Fisher's exact test, two-tailed, p ϭ 0.350). Similarly, erastin and glutamate analog (HCA) toxicity in neurons can be considered mechanistically similar (Fisher's exact test, two-tailed, p ϭ 0.758).
Collectively, these findings suggest that erastin induces cell death in both cancer cells and neurons through ferroptotic, autophagic, and necroptotic pathways and that they are not different from what is known from the established glutathione depletion model referred to incorrectly as apoptosis (Ratan et al., 1994b) or correctly as oxytosis (Lewerenz et al., 2018) in prior studies ( Fig. 11B; Fisher's exact test, two-tailed, p ϭ 0.751). Convergence of these distinct modes of cell death in a single cell death paradigm has not been described so far.

Promoting cancer cell death while protecting neurons
Because erastin toxicity in HT1080 cells and glutamate analog (HCA)-induced glutathione depletion in neurons induce ferroptotic cell death, we hypothesized that prior knowledge about glutamate analog (HCA)-induced ferroptosis in neurons should inform the points of convergence and divergence with erastin-induced ferroptosis in HT1080 fibrosarcoma cells. We therefore investigated whether erastin-induced ferroptosis in HT1080 can be abolished or exacerbated by the chemical inhibitors known to be effective in glutamate analog (HCA)-induced neuronal ferroptosis (Figs. 12, 13A).
Glutamate analog (HCA)-induced ferroptosis in neurons requires de novo transcription via the leucine zipper transcription factor ATF4 leading to upregulation of putative pro-death genes such as Trib3, Chop, and Chac1 (Lange et al., 2008;Karuppagounder et al., 2016). Recently, it was shown that HIF prolyl hydroxylases (HIF PHDs) are required for pro-death ATF4 transcription, and a selective small molecule inhibitor of the HIF PHDs, Adaptaquin, abrogates glutamate analog (HCA)-induced ferroptosis and improves functional recovery after intracerebral hemorrhage, where cell death has also been defined as ferroptotic (Karuppagounder et al., 2016). As expected from these findings, Adaptaquin also protected against erastininduced toxicity in HT1080 cells (Figs. 12, 13A).
Prior studies have shown that, like Adaptaquin, the aureolic acid antitumor agent Mithramycin can also act in the nucleus to abrogate glutamate analog (HCA)-induced ferroptosis in vitro and extend survival in vivo in an HD model, where ferroptosis has also been implicated continued nonspecific inhibitors of the System X cϪ transporter by counteracting the glutamate gradient resulting in reduced uptake of cystine into the cells. This leads to a decrease in glutathione synthesis, which is essential for the endogenous antioxidant defense. Ferroptosis inhibitors are indicated in green. NAC, N-acetylcysteine. B, HT1080 fibrosarcoma cells were treated with increasing dose of erastin at previously described density (left; 100,000 cells/ml) and adjusted density to reach 70% confluency before treatment (right; 25,000 cells/ml) to determine the toxicologically meaningful dose of erastin. Representative live/dead staining are shown, green indicating live cells (calcein AM) and red indicating dead cells (ethidium homodimer-1). Scale bar, 50 m. C, Dose-response of erastin (left) and HCA (right) in primary cortical neurons. Representative live/dead staining are shown. Scale bar, 50 m. D, Dose-response of erastin in DRG neurons and representative live/dead staining. Scale bar, 100 m.  (Figs. 12, 13A). We confirmed that Mithramycin was able to induce p21 gene expression in HT1080 cells, but failed to reduce basal c-Myc levels suggesting the possibility that augmentation of tumor suppression genes rather than suppression of proto-oncogenes may mediate Mith-ramycin's effects in HT1080 cells ( Fig. 13B; Sleiman et al., 2011b).
Transglutaminases are established transcriptional targets for Huntington's disease (McConoughey et al., 2010) and have been shown to mediate glutamate analog  et al., 2003). By contrast to Adaptaquin, Mithramycin A, and transglutaminase inhibitors, class I HDAC inhibitors failed to protect HT1080 cells from erastin toxicity, in fact, they exacerbated cell death (Figs. 12, Fig. 13A). In contrast, Scriptaid, which was able to protect neurons from ferroptosis (Fig. 13C), did not exacerbate ferroptosis in cancer cells. We confirmed that Scriptaid but not its structurally similar, inactive analog Nullscript, increased the acetylation of histones using Western blotting (Fig. 13D).
We then hypothesized that the expression of distinct HDAC isoforms in neurons versus cancer cells explains this difference. We therefore assessed the gene expres- sion of HDACs from class I and found that HT1080 cells and primary neurons express HDAC1, 2, and 3. Surprisingly, only HT1080 cells express HDAC8, raising the intriguing possibility that HDAC 8 expression may explain the differential effects of HDAC inhibitors between HT1080 cells and primary neurons (Fig. 13E).
To determine whether the effect is specific to HT1080 cells or applicable to other cancer cells, we evaluated whether Scriptaid protected against erastin-induced toxicity in the neuroblastoma cell line SH-SY5Y and the hepatocellular carcinoma cell line Hep3B (at the LD 50 of erastin for each cell line). We found that 2.5-10 M Scriptaid exacerbated erastin-induced cell death in SH-SY5Y cells, whereas the negative control did not. However, it also decreased viability of SH-SY5Y cells without erastin treatment (Fig. 14A). In contrast, 10 M Scriptaid also enhanced erastin toxicity in Hep3B cells, although it was not toxic for the cells at that dose (Fig. 14B). Altogether, these findings suggest that HDAC inhibitors could be good adjunctive treatment for erastin or erastin-like drugs in fibrosarcomas, neuroblastoma, and hepatomas.

Discussion
The objective of this study was to understand the mechanisms by which erastin (an X c Ϫ transporting agency inhibitor) induces death in cultured CNS neurons compared with erastin-induced death in cancer cells. The X c Ϫ transporting agency is a chloride-dependent transporter that exchanges cystine for glutamate. Cystine is converted intracellularly to cysteine, which is used for protein synthesis and synthesis of the versatile antioxidant glutathione. Depletion of cyst(e)ine leads to depletion of glutathione and reactive lipid species-induced death now defined as ferroptosis.
Our ultimate goal in pursuing these studies is to develop combinatorial therapeutic approaches that on the one hand enhance erastin-induced death of tumor cells, and on the other, prevent its potential toxicities in the nervous system mediated via erastin-induced neuronal death. Our results suggest that cell death pathways induced by erastin in post-mitotic cortical neurons and HT1080 fibrosarcoma are similar to each other and to oxidative glutamate toxicity in neurons (Figs. 2-9, 11). Both glutamate and erastin can inhibit the system X c Ϫ to induce death in immature neurons where functional, ionotropic glutamate receptors are not expressed (Murphy et al., 1989;Ratan et al., 1994a, b,). We leveraged prior studies of cystine deprivation in neurons to identify HDAC inhibitors as agents that enhance the chemotherapyinduced ferroptosis of tumors while inhibiting ferroptosis in neurons. These findings suggest a combinatorial approach to cancer chemotherapy (erastin plus a class I HDAC inhibitor) designed to maximize tumor cell death and minimize neuronal toxicity. Our studies also identified agents that prevent ferroptotic death in cancer cells and neurons that would not be good candidates for cancer treatment (e.g., HIF PHD inhibitors, transglutaminase inhibitors, and the gene selective Sp1 inhibitor Mithramycin).
Our studies systemically evaluated whether erastininduced ferroptosis in cancer cells occurs via mechanisms that are similar to or distinct from those induced by erastin in primary neurons. We found that erastin-induced toxicity in tumor cells can be abolished not only by previously published ferroptosis inhibitors (Dixon et al., 2012;Figs. 2, 3) but also by inhibitors of necroptosis (Figs. 6-7, 10) and autophagy (Figs. 8,9) when erastin was used at a toxicologically relevant dose.
With respect to autophagy, Torii et al. (2016) previously reported that Bafilomycin A1, a vacuolar ATPase inhibitor that prevents endosomal acidification, abolished cell death induced by erastin (all doses tested) as well as another inducer of ferroptosis, RSL3, in HT1080 and Calu-1 cells. Similarly, knock-out of autophagy-related genes Atg5 and Atg7 in mouse embryonic fibroblasts and knockdown in HT1080 cells partially inhibited erastininduced cell death (Hou et al., 2016). These data suggest that cyst(e)ine deprivation may not only be interpreted by the cell as uncompensated redox but also nutrient dyshomeostasis, thereby triggering multiple parallel pathways (e.g., reactive lipid species, autophagy) that collectively elevate the threshold of the cell toward cell demise (Zille et al., 2017).   With respect to necroptosis, Yu and colleagues demonstrated that Necrostatin-1 as well as knockdown of RIP3 abrogated erastin toxicity in acute myeloid leukemia cells (Yu et al., 2015). Interestingly, MLKL deficiency increased sensitivity to ferroptosis stimuli in mouse embryonic fibroblast. In turn, the loss of acyl-CoA synthetase long-chain family member 4, that suppresses ferroptosis by limiting the membrane-resident pool of oxidationsensitive fatty acids, predisposed cells to necroptosis (Muller et al., 2017). This suggests that necroptosis and ferroptosis are differentially regulated cell death pathways. It has previously been demonstrated that ferroptosis and necroptosis are recruited independently in neurons exposed to intracerebral hemorrhage and that they may then converge at a yet to be identified common denominator leading to a necrotic morphology (Zille et al., 2017). Here, we provide evidence that erastin is able to recruit the necroptotic machinery in diverse cell types (Fig. 10).
However, HT1080 fibrosarcoma cells have been reported not to express RIP3 (de Almagro et al., 2017) and were therefore not responsive to the inhibition by GSK872 in our study. This is interesting, because Necrostatin-1 and necrosulfonamide abolished erastin-induced toxicity. Although the activation of RIP1, RIP3, and MLKL have been considered to be indispensable for necroptosis, recent studies suggest possible necroptosis-independent roles of each protein, including the induction of apoptosis and the inflammasome (Mandal et al., 2014;. In addition to the mentioned differences in the expression and post-translational regulation of necrosome complex proteins and caspase-8 in various cell types (Newton et al., 2016;de Almagro et al., 2017), they also differ in their sensitivity to erastin (Yang et al., 2014;Yu et al., 2015). Thus, the relative contribution of ferroptosis and necroptosis during regulated cell death needs further investigation, especially with respect to cell type, tissue, and disease. This will greatly influence the effectiveness and toxicity of chosen chemotherapeutic approaches.
To investigate the ability of erastin to cause unwanted toxicities in the nervous system, we examined whether erastin engages a similar pattern of cell death signaling in primary neurons compared with HT1080 cells. Indeed, the exposure to erastin led to the death of primary cortical neurons, which was reversed by a statistically similar profile (Fig. 11) of inhibitors of cell death, including autophagy (Figs. 8, 9), ferroptosis (Figs. 2, 3), and necrop- tosis (Figs. 6,7,10). We also found that in an established glutathione depletion model involving non-receptormediated toxicity by glutamate (HCA), a panoply of chemical inhibitors also blocked erastin-induced death 11). In addition, both erastin-and glutamate (HCA)induced ferroptosis recruited the necrosome by activating RIP1 (Fig. 10). This is further substantiated by recent findings of Neitemeier et al. (2017) who demonstrated mitochondrial dysfunction in glutamate and erastin toxicity in HT22 hippocampal neuroblast cells, and this effect was abolished by the loss of BH3 Interacting Domain Death Agonist (BID).
Finally, we sought to demonstrate whether the knowledge about glutamate analog (HCA)-induced ferroptosis/ oxytosis in neurons can be applied to ferroptosis in cancer cells (Figs. 12, 13). Indeed, we found that Adaptaquin, transglutaminase inhibitors, and Mithramycin all protect cancer cells from erastin similar to their protection in neuronal toxicity. However, HDAC inhibitors exacerbated erastin toxicity in cancer cells while promoting survival in neurons. Collectively, our data indicate that a combination of anticancer drugs may reverse the effect of single agents (as it is the case for erastin and Mithramycin), and that selectively targeting cancer cells while sparing neurons may be a more promising therapeutic approach.
It is important to note that sulfasalazine, which functions as an X c Ϫ transport inhibitor like erastin, has been shown to not only inhibit glial-derived tumor (glioma) growth but also to inhibit neuronal excitotoxic death and seizures associated with these tumors (Chung et al., 2005;Robert et al., 2015). Under these circumstances, sulfasalazine confers salutary effects on neurons by preventing the release of glutamate from gliomas. Indeed, glioma growth depends on glutamate release, and this may be a mechanism by which tumor cells remove neurons that compete for space and nutrients. These results suggest that erastin may actually prevent some toxicities to the CNS that are derived from tumors, rather than creating toxicities on its own. However, several observations potentially reconcile our results with seminal studies from Sontheimer et al. First, they evaluated the effects of X c Ϫ transport inhibition with only one drug concentration (sulfasalazine) where there is clear sensitivity of the glioma tumor to glutathione depletion but no effect on neurons or astrocytes. Indeed, studies here clearly demonstrate that cortical neurons are sensitive to X c Ϫ transport inhibition with an LD 50 5-fold higher than those required to kill 50% of the HT1080 cells. Second, they did not evaluate toxicity to peripheral neurons or central neurons directly, still leaving the question open whether X c Ϫ inhibition systemically can lead to chemotherapy-induced neuropathy or chemo brain. Accordingly, apparent toxicity may occur under circumstances where drug concentrations used to kill tumor cells are higher and begin to approach the susceptibility of neurons. HDAC inhibitors in combination with erastin appear to be a viable strategy to minimize toxicity as they would reduce the concentration of erastin or sulfasalazine required for killing but would provide for neuroprotection from ferroptosis in their own right. In conclusion, cell death pathways in cancer cells and neurons exposed to glutathione depletion are similar as judged by the Fisher's tests. Our data suggest that HDAC inhibitors are promising therapeutic agents for combinatorial cancer chemotherapy to enhance chemotherapyinduced ferroptosis of tumors while inhibiting ferroptosis in neurons. The current study further identifies agents, some of which are known antitumor drugs, which would prevent erastin-induced toxicity in neurons (e.g., Adaptaquin, Mithramycin, transglutaminase inhibitors) and in cancer cells suggesting that these are not good combinatorial therapies to optimally kill tumor cells while preserving CNS and PNS neurons.