Human Haploid Cell Genetics Reveals Roles for Lipid Metabolism Genes in Nonapoptotic Cell Death

Several nonapoptotic, regulated cell death (RCD) pathways have been discovered in recent years, including necroptosis, pyroptosis and ferroptosis.¹ These pathways have been implicated in a variety of pathological conditions, and small molecules that can activate or inhibit nonapoptotic cell death have emerged as important therapeutic leads for a variety of diseases.¹⁻³

Ferroptosis is an iron-dependent form of RCD observed in cancer cells, kidney cells and neurons.⁴⁻⁸ Ferroptosis can be caused by inhibition of system x_c^– or glutathione peroxidase 4 (GPX4).^9,10 System x_c^– is required to import cystine into cells for glutathione synthesis, and GPX4 uses glutathione to detoxify lipid peroxides. Erastin is a system x_c^– inhibitor, while (1S,3R)-RSL3 (hereafter RSL3) and ML162 (also known as DPI7) inactivate GPX4.^9,10 The genetic networks regulating sensitivity to these and other death-inducing compounds are not well-defined.

Defining small molecule mechanism of action is challenging.^11,12 One potential approach employs a nearly haploid chronic myeloid leukemia cell line, KBM7, coupled with retroviral-mediated insertional mutagenesis and sequencing to identify genes required for lethality.¹³⁻¹⁵ This approach has been used to investigate the lethal mechanism of action of tunicamycin, 3-bromopyruvate and YM155 and, in each case, has identified a specific transporter protein necessary for the cellular uptake of each of these lethal compounds.¹⁵⁻¹⁷ We sought to apply this approach to identify genes involved in ferroptosis and other forms of nonapoptotic RCD.

We examined the mechanism of action of five small molecules that trigger nonapoptotic cell death: erastin, RSL3, ML162, PK-3,¹⁸ and a novel compound, CIL56²⁷ (Figure 1a). As a control, we examined the apoptosis-inducing small molecule BH3 mimetic ABT-263. Based on previous results using a related analogue, ABT-737,¹⁴ we expected to identify insertions into BCL2-associated X protein (BAX) and phorbol-12-myristate-13-acetate-induced protein 1 (PMAIP1/NOXA) as suppressors of ABT-263-induced death if our screening system was performing as desired.

All six test compounds were lethal to KBM7 cells, with erastin, RSL3, and ML162 inducing ferroptosis, as judged by the ability of the lipophilic antioxidant ferrostatin-1 (Fer-1) and the iron chelator ciclopirox olamine (CPX) to prevent death⁴ (Figure 1b). The lethality of erastin was also suppressed by the reducing agent beta-mercaptoethanol (β-ME), as expected for a system x_c^– inhibitor⁴ (Figure 1b). The lethality of ABT-263 was not suppressed by Fer-1, CPX, or β-ME, as expected for an apoptosis-inducing agent. In these experiments, neither PK-3 nor CIL-56 induced ferroptosis; other results^18,27 suggest that these compounds induce unknown forms of nonapoptotic cell death. Indeed, an important goal of these studies was to identify genes that might distinguish the lethal pathways activated by these compounds.

Screening trials identified compound concentrations yielding resistant colonies for PK-3 (55 nM), ML162 (9 μM), RSL3 (4 μM), CIL56 (5.5 μM), and ABT-263 (2.5 μM). For erastin, none of the tested lethal concentrations yielded any resistant colonies, suggesting that cystine import via system x_c^– is unconditionally essential for colony formation and can not be compensated for by the deletion of any single gene; this finding suggests that tumors might find it difficult to evolve resistance to system x_c^– inhibitors through gene deletion.

The five remaining compounds were individually added to an initial pool of 1 × 10⁸ KBM7 cells that were heavily mutagenized using a retroviral genetrap vector, as described previously.¹⁴ After compound incubation for 10 to 21 days, resistant colonies were collected and retroviral insertion sites mapped to individual genes¹⁴ (Figure 2). Not shown are insertions into three genes: NF1, WT1 and SPI1, which were commonly identified in most screens, suggesting that they may confer a general drug-resistant phenotype when mutated in KBM7 cells.

As expected,¹⁴ ABT-263-resistant cells had a significant overrepresentation of insertions in BAX (P < 10^–50) and PMAIP1/NOXA (P = 5.5 × 10^–39) (Figure 2a). We confirmed that clonal KBM7 cell lines lacking either gene were resistant to ABT-263 (Supplemental Figure 1). These results provided confidence in our ability to identify genes required for small molecule-induced cell death using this approach. Unexpectedly, the analysis of PK-3-resistant colonies yielded no significantly enriched genes (Figure 2b). PK-3 binds tubulin and disrupts the intracellular microtubule network, which is necessary for mitosis.¹⁸ We speculate that PK-3-resistant colonies could arise in the absence of specific mutations from a transient subpopulation of nondividing (i.e., quiescent, G₀) cells. Transcriptional upregulation of genes encoding xenobiotic detoxification enzymes or drug pumps in these cells could then allow for survival and subsequent proliferation in the presence of PK-3. Further analysis of additional microtubule-disrupting agents in KBM-7 cells would help resolve this issue.

We next analyzed RSL3 and ML162, two structurally distinct compounds that both inhibit GPX4.¹⁰ Analysis of RSL3-resistant cells identified four genes significantly enriched for gene trap insertions: acyl-CoA synthetase long-chain family member 4 (ACSL4, P = 3.5 × 10^–259), lysophosphatidylcholine acyltransferase 3 (LPCAT3, P = 3.0 × 10^–107), 1-acylglycerol-3-phosphate O-acyltransferase 3 (AGPAT3, P = 6.2 × 10^–16), and hydroxysteroid (17-beta) dehydrogenase 11 (HSD17B11, P = 5.8 × 10^–9) (Figure 2c). Analysis of ML162-resistant cells identified three genes significantly enriched for insertions, two of which were shared with RSL3 (ACSL4 and LPCAT3, P values of 5.7 × 10^–20 and 7.1 × 10^–11), and one of which was unique, but less significantly enriched (SET domain containing 1B, SETD1B, P = 1.9 × 10^–4) (Figure 2d). The overlap of ACSL4 and LPCAT3 between RSL3 and ML162 provides confidence that these two genes are essential for the execution of ferroptosis induced by inhibition of GPX4.

Cells undergoing ferroptosis are depleted of arachidonic acid (AA, 20:4) and other polyunsaturated fatty acids (PUFAs).^5,7 The prevailing model is that the oxidative destruction of PUFAs, following GPX4 inactivation, is essential for the execution of ferroptosis.^5,7 ACSL4 preferentially acylates AA,¹⁹ while LPCAT3 preferentially catalyzes the insertion of acylated AA into membrane phospholipids.²⁰ Thus, deletion of ACSL4 and LPCAT3 likely suppress ferroptosis by limiting the membrane-resident pool of oxidation-sensitive fatty acids. Further studies will be required to elucidate the roles of AGPAT3, HSD17B11, and SETD1B and determine if the identification of these genes in only the RLS3 (HSD17B11, AGPAT3) or the ML162 (SETD1B) screens reflects the existence of false negatives unique to each screen, or to true biological differences in the lethal effects of these two compounds.

Finally, we analyzed results obtained with CIL56 and identified a unique set of significantly enriched genes compared to ABT-263, RSL3, and ML162, including zinc finger, DHHC domain containing 5 (ZDHHC5, P = 6.3 × 10^–218), trans-2,3-enoyl-CoA reductase (TECR, P = 1.2 × 10^–131), acetyl-CoA carboxylase alpha (ACACA, P = 3.6 × 10^–56, individual insertions sites into this gene shown in Supplemental Figure 3a), and NAD kinase (NADK, P = 8.8 × 10^–31) (Figure 2e). ACACA (encoding ACC1) is a key regulator of tumor cell survival,²¹ and we therefore focused on the role of ACC1 in CIL56-induced death.

ACC1 catalyzes the conversion of acetyl-CoA to malonyl-CoA, the rate-limiting step in fatty acid synthesis (Figure 3a). Two clonal ACACA null HT-1080 cells lines (H and P), which we generated using CRISPR/Cas9 technology, lacked ACC1 expression and exhibited 5-fold resistance to CIL56, but no consistent resistance to RSL3 or the apoptosis-inducing agents hydrogen peroxide (H₂O₂) or staurosporine (STS) (Figure 3b–f, Supplemental Figure 2b). Silencing of ACACA in HT-1080 cells using two independent siRNAs likewise resulted in resistance toward CIL56, but not RSL3 (Supplemental Figure 2c–e). Furthermore, the lethality of CIL56, but not RSL3 or erastin, was suppressed by the small molecule ACC1 inhibitor 5-tetradecyloxy-2-furonic acid (TOFA)^21,22 (Figure 3g). Conversely, cell death induced by erastin and RSL3, but not CIL56, was suppressed by Fer-1 (Figure 3g). These results suggest that ACC1 activity sensitizes cells to CIL56-induced cell death.

Using mass spectrometry, we analyzed the metabolome of HT-1080 cells treated with CIL56 (6.5 μM) ± TOFA (4 μM), compared to vehicle-treated controls. Among the 298 polar and nonpolar metabolites identified in this analysis, the levels of 141 metabolites were significantly altered by CIL56 treatment, with 82 metabolites significantly increased and 59 significantly decreased (FDR q < 0.01) (Figure 3h, Supplemental Table 1). In cells cotreated with CIL56 and TOFA, we detected no significantly altered metabolites compared to vehicle-treated cells (at FDR q < 0.01) (Figure 3h, Supplemental Table 1). This suggested that the effects of CIL56 on metabolism were completely reversed by TOFA. Of note, CIL56 triggered the striking TOFA-sensitive accumulation of all detectable long chain saturated and monounsaturated fatty acids and all detectable polyunsaturated fatty acids (Figure 3i,j and Supplemental Table 1). A plausible model to account for the accumulation of both nonessential and essential fatty acids species is that CIL56 inhibits the normal breakdown of fatty acids by mitochondrial β-oxidation. This could occur if, for example, CIL56 accelerated the ACC1-dependent production of malonyl-CoA, a metabolite that acts as a negative regulator of this process.²³ While it remains possible that CIL56 could engage more than one target, our results are consistent with the model that CIL56 stimulates the activity of ACC1 itself, or an upstream component of the fatty acid synthesis pathway, to cause cell death. One prediction of this model is that overexpression of ACC1 might sensitize to CIL56-induced death, although regulation of ACC1 and downstream lipid synthetic enzymes may be sufficiently complex to preclude this.²⁴

Together, these results implicate the lipid metabolic genes ACSL4 and LPCAT3 in ferroptosis and ACACA in the novel cell death program triggered by CIL56. It remains unclear how CIL56 treatment causes cell death, although the accumulation of saturated fatty acids²⁵ or the inhibition of mitochondrial β-oxidation could be involved. An important aim of future studies will be to pinpoint the specific molecular changes that are responsible for CIL56-induced cell death. Future studies using clonal cell lines lacking the lipid metabolic enzymes identified here should greatly facilitate this process.