Blockade of the lysophosphatidylserine lipase ABHD12 potentiates ferroptosis in cancer cells

Abstract

Ferroptosis is a type of cell death caused by the pathogenic accumulation of lipid hydroperoxides. Pharmacological mechanisms to induce ferroptosis may provide a way to kill cancer cells that are resistant to other forms of cell death like apoptosis. Nonetheless, the proteins that regulate ferroptotic sensitivity in cancer cells remain incompletely understood. Here, we screened a panel of inhibitors of serine hydrolases – an enzyme class important for regulating lipid metabolism – for potentiation of ferroptosis in HT1080 fibrosarcoma cells. We found that DO264, a selective inhibitor of the lyso- and ox-phosphatidylserine (PS) lipase ABHD12, enhances ferroptotic death caused by RSL3, an inhibitor of the lipid peroxidase GPX4. RSL3-induced ferroptosis was also potentiated by genetic disruption of ABHD12. Metabolomic experiments revealed that, in addition to elevated lyso-PS, ABHD12-inactivated cells show higher quantities of arachidonate (C20:4)-containing PS and 2-arachidonoyl glycerol, pointing to potential oxidation-sensitive lipid mediators of ferroptosis regulated by ABHD12.

Ferroptosis is a form of cell death defined by runaway peroxidation of membrane phospholipids.¹ This lipid peroxidation is dependent on iron, and ferroptotic cell death can be inhibited by iron chelators.¹ Ferroptosis is implicated in several types of pathological cell death, often related to cell and/or organ damage and associated degenerative disorders.² Ferroptosis has more recently garnered interest as a way to promote cancer cell death independent of the apoptotic pathways that are often defective in transformed cells.³ For instance, drug-resistant, or persister, cancer cells⁴ and those that have undergone an epithelial-to-mesenchymal transition⁵ exhibit high susceptibility to ferroptosis.

Ferroptosis can be induced by at least two known pharmacological mechanisms: blockade of the cellular import of cystine by inhibitors of the cystine/glutamate transporter SCL7A11 (or system x_c⁻), such as erastin, or by direct inhibition of the phospholipid hydroperoxide glutathione peroxidase 4 (GPX4).^{6, 7} The lipid peroxidation step of ferroptosis occurs on phospholipids bearing polyunsaturated fatty acids (PUFAs), such as linoleic (C18:2), arachidonic (C20:4), and adrenic (C22:4) acids,⁸ and it is either catalyzed by lipoxygenases⁹ or iron-mediated autoxidation.¹⁰ Accordingly, prevention of PUFA incorporation into phospholipids by disruption of enzymes like ACSL4¹¹ or LPCAT3¹² attenuates ferroptosis.

Whether inhibitors of SLC7A11 or GPX4 can induce ferroptosis in cancer cells with sufficient selectivity to avoid general cell or organ toxicity in vivo remains unclear, and it would therefore be of interest to identify other proteins that modulate the ferroptotic potential of cancer cells. Two recent studies used genetic screens of ferroptosis-resistance cancer cells to identify the coenzyme Q oxidoreductase FSP1 (or AIFM2) as a suppressor of ferroptosis in cells that lack dependency on GPX4.^{13, 14} Here, we considered an alternative, chemical screen to discover potentiators of ferroptosis. Recognizing that the hydrolysis of the sn-1 and sn-2 acyl chains of phospholipids constitutes a principal route for metabolism of these lipids in both unoxidized and oxidized states, and that this reaction is often catalyzed by phospholipase enzymes from the serine hydrolase class¹⁵, we reasoned that a screen of serine hydrolase inhibitors may identify enzymes that modulate GPX4-dependent ferroptosis. This screen identified the lyso- and oxidized-phosphatidylserine (PS) lipase ABHD12 as a ferroptosis regulatory enzyme, where pharmacological or genetic disruption of this enzyme enhances GPX4-mediated ferroptosis concurrently with elevating lyso-PS, arachidonoyl-PS, and 2-arachidonoyl glycerol (2-AG).

Serine hydrolases are one of the largest and most diverse enzyme families in Nature, comprising ~1% of all proteins in humans.¹⁶ The substrates of serine hydrolases are diverse and include proteins, peptides, and lipids.¹⁵ Owing to a shared catalytic mechanism that involves an activated serine nucleophile, serine hydrolases are amenable to functional analysis by the chemical proteomic method activity-based protein profiling (ABPP).¹⁷ We and others have used ABPP to generate selective inhibitors for several serine hydrolases, including many that act on lipid substrates, including neutral lipids,^18-25 (lyso)phospholipids,^26-28 and oxidized phospholipids²⁹. We assembled a set of ~25 serine hydrolase-directed inhibitors for screening in the ferroptosis-sensitive fibrosarcoma cell line HT1080 (see Table S1 for list of serine hydrolase-directed inhibitors). The screen involved pre-treatment for 1 h with individual inhibitors (tested at a concentration that exceeds the EC₉₀ value for the preferred serine hydrolase target) or DMSO control, followed by exposure to a concentration range of the GPX4 inhibitor RSL3 for 24 h, from which EC₅₀ values for ferroptosis induction were calculated. The screen was performed twice and the reported EC₅₀ values reflect average values for the two independent replicate experiments. Ferroptosis-enhancing compounds were defined as those that produced EC₅₀ values for RSL3 that were > 1.5-fold of DMSO control values.

Our screen identified two compounds that substantially enhanced ferroptosis – the ABHD12 inhibitor DO264²⁶ and PAFAH2 inhibitor AA39-2²⁹ (Figure 1A). Even though we were primarily interested in identifying mechanisms that potentiated ferroptosis, we also noted that one compound – the DAGLA/DAGLB inhibitor DO34²⁴ – substantially suppressed this process by ~50% (blue bar, Figure 1A). We confirmed the activity of each compound in an independent set of ferroptosis assays (Figure 1B) and proceeded to focus on the ferroptosis-enhancing compounds for further characterization. We first wanted to verify the target profiles of the active compounds, since we have not previously performed ABPP experiments in HT1080 cells. We treated HT1080 cells with DMSO or AA39-2 or DO264 (1 μM each) for 2 h, after which the cells were lysed, and proteomes exposed to the serine hydrolase-directed activity-based probe fluorophosphonate (FP)-biotin³⁰ (10 μM, 1 h) and processed for streptavidin enrichment and quantitative mass spectrometry (MS)-based proteomics. Protein quantitation was performed by isotopic labeling with reductive dimethylation, using heavy and light formaldehyde, as described.³¹ These MS-based ABPP experiments verified that DO264 selectively inhibited ABHD12 in HT1080 cells (Figure 1C and Table S2; > 90% reductions in ABDH12 activity, with only a single partial cross-reactivity with ABHD13 (~50% inhibition) being observed across 50+ quantified serine hydrolases), but revealed that AA39-2 inhibited several enzymes in addition to PAFAH2 (Figure 1D and Table S2). These data indicate that, while AA39-2 – an electrophilic 1,2,3-triazole urea – may selectively inhibit PAFAH2 at lower concentrations (e.g., < 10 nM in cells),²⁹ the compound is more promiscuous at the screening concentration of 1 μM used in this current study. We therefore chose to mechanistically characterize DO264 due to its selectivity for inhibiting ABHD12.

Figure 1.

A screen of serine hydrolase-directed inhibitors to identify ferroptosis-modulating compounds. (A) Ferroptosis screen. HT1080 cells were treated with individual serine hydrolase inhibitors (all compounds tested at 1 μM, except: DO18 (0.2 μM), THL (10 μM); see Supplementary Table 1 for inhibitor structures and associated serine hydrolase targets) for 1 h before induction of ferroptosis with a concentration range of RSL3 (10 μM with threefold serial dilution, 24 h). Compounds were considered to substantially enhance ferroptosis if they increased the potency of RSL3 (EC₅₀ value) by at least 50% compared to DMSO control. Two compounds met this criteria and are shown in red. Blue marks a compound (DO34) that substantially reduced RSL3-mediated ferroptosis by ~50%. Data represent mean values ± s.d. for two independent experiments. (B) Confirmatory cell viability curves for the hit compounds found to enhance RSL3-dependent ferroptosis from (A). EC₅₀ value shifts for each compound (tested at 1 μM) reported below the curves. Data represent mean values ± s.e.m for four independent experiments. (C, D) Competitive ABPP data for serine hydrolases from HT1080 cells treated with ferroptosis-enhancing compounds DO264 (C) and AA39-2 (D) Cells were treated with each compound at 1 μM for 2 h and then harvested, lysed, treated with the serine hydrolase-directed probe FP-biotin (10 μM), and processed and analyzed following previously described quantitative MS-ABPP protocols³¹. Ratios represent compound/DMSO (light/heavy), such that low values correspond to inhibition of serine hydrolases. Data represent mean values ± s.d. for median aggregate peptide ratios for each serine hydrolase from two independent experiments.(E) DO264, but not its inactive structural analog (S)-DO271 (1 μM each, 1 h), enhances RSL3-dependent ferroptosis in HT1080 cells. (F) Structures of DO264 and (S)-DO271. (G) DO264 enhances RSL3-dependent ferroptosis in a concentration-dependent manner in HT1080 cells. Data represent mean values ± s.e.m for four independent experiments.

We first verified that (S)-DO271, a structurally related analogue to DO264 that does not inhibit ABHD12,²⁶ did not alter GPX4-dependent ferroptosis (Figure 1E). We also found that DO264, but not (S)-DO271, enhanced ferroptosis in a second cancer cell line, the B cell lymphoma line SU-DHL-5 (Figure S1A). The ferroptosis-potentiating activity of DO264 showed concentration-dependence that generally matched the previously reported in situ potency of the compound for inhibiting ABHD12 (IC₅₀ < 1 μM)²⁶ (Figure 1F and Figure S1B) and was blocked by co-treatment with the ferroptosis inhibitor ferrostatin-1 (Figure S2). DO264 did not potentiate other forms of cell death in HT1080 cells, such as apoptosis induced by staurosporine or necrosis induced by H₂O₂ (Figure S3), and was not independently cytotoxic in HT1080 or SU-DHL-5 cells at concentrations below 10 μM (Figure S2). These data, taken together, indicate that ABHD12 inhibiton by DO264 specifically potentiates ferroptotic cell death.

To provide further evidence that the ferroptosis-potentiating effects of DO264 occurred through inhibition of ABHD12, we generated ABHD12-knockout (ABHD12-KO) HT1080 cells using CRISPR-Cas9 genome editing. We performed these studies on a population level rather than by selecting clones to avoid potential clonality effects on the basal ferroptotic sensitivity of cells. We confirmed substantial loss of ABHD12 in gene-edited HT1080 cells using a tailored activity-based probe JJH350²⁶ (Figure 2A). The ABHD12-KO cells displayed heightened sensitivity to GPX4-dependent ferroptosis compared to cells treated with control sgRNAs (Figure 2B). The impact of genetic disruption of ABHD12 was lower in magnitude than the effect of DO264 (Figure 2B), which could indicate a residual amount of ABHD12 expression in the ABHD12-KO cell population or that part of the ferroptotic-enhancing effects of DO264 occur through an additional target. Regardless, these data demonstrate that the pharmacological or genetic impairment of ABHD12 enhances the ferroptotic sensitivity of cancer cells.

Figure 2.

Genetic disruption of ABHD12 enhances ferroptosis. (A) Loss of ABHD12 in CRISPR-Cas9-edited HT1080 cells (ABHD12-KO cells), as measured using the ABHD12-directed activity-based probe JJH350 (2 μM, 45 min, 37 °C), following described protocols²⁶. Populations of HT1080 cells were separately generated with three ABHD12-targeting sgRNAs (KO1, KO2, KO3) and two non-targeting control sgRNAs (Ctrl 1, Ctrl 2). (B) Relative RSL3 EC₅₀ values for control (Ctrl) and ABHD12-KO populations treated with DMSO or DO264 (1 μM). Data are normalized to DO264-treated control cells and reflect three independent experiments performed with each control (Ctrl 1, Ctrl 2) and ABHD12-KO (KO1, KO2, KO3) population. **P < 0.01; ***P < 0.001 (two-sided Student’s t-test performed relative to control sgRNA cells treated with DMSO).

ABHD12 has been shown to have at least three metabolic functions in mammals – 1) a lysophospholipase activity that acts on lyso-phosphatidylserine (lysoPS) lipids;³² 2) a phospholipase activity that acts on oxidized variants of PUFA PS lipids;³³ and 3) a neutral lipase activity that acts on monoacylglycerols (MAGs), including the endocannabinoid 2-arachidonoylglycerol, or 2-AG.³⁴ Genetic or pharmacological blockade of ABHD12 has been shown to increase both the lyso-PS and PUFA (C20:4) PS content of mammalian cells and tissues,^{26, 32} with the latter change likely reflecting the action of lysophospholipid acyltransferases on elevated lyso-PS. Using targeted lipidomics we found that both DO264-treated cells and ABHD12-KO cells showed substantially elevated 18:0 lysoPS and 18:0/C20:4 PS (Figure 3A, B; Figure S4) The inactive control compound (S)-DO271 did not alter 18:0 lysoPS or 18:0/C20:4 PS lipids (Figure S4). Other phospholipids were largely unchanged in ABHD12-disrupted cells (Table S3), consistent with previous studies pointing to a specialized role for ABHD12 as a lyso-PS lipase.³² Interestingly, exposure to RSL3 further increased 18:0 lyso-PS and dramatically lowered the 18:0/C20:4 PS content of DO264-treated cells (Figure 3C, D). RSL3 had a similar effect on 18:0/C20:4 PS, but did not alter lyso-PS, in DMSO-treated control cells (Figure 3C, D). These data suggest that the heightened PUFA content of PS lipids resulting from ABHD12 blockade can serve as a substrate for lipid peroxidation during ferroptosis, pointing to one potential mechanism for sensitization to this form of cell death. We attempted to identify oxidized PS lipids in DO264/RSL3-treated cells, but were unable to detect these lipids, possibly due to their low abundance and diversified structures.

Figure 3.

Genetic or chemical disruption of ABHD12 elevates lyso-PS and C20:4 PS content of HT1080 cells. (A,B) Quantification of 18:0 lysoPS and 18:0/20:4 PS content of control (Ctrl) or ABHD12-KO HT1080 cells treated with DMSO or DO264 (1 μM, 6 h). Data represent mean values ± s.e.m for three independent experiments for each sgRNA population (two control, three ABHD12-KO). *P < 0.05; **P < 0.01; ***P < 0.001 (two-sided Student’s t-test performed relative to control sgRNA cells treated with DMSO). (C,D) Quantification of 18:0 lysoPS and 18:0/20:4 PS content of ferroptotic HT1080 cells. HT1080 cells were treated with DMSO or DO264 (1 μM) for 1 h, followed by DMSO or RSL3 (2 μM) for 4 h. Data represent mean values ± s.e.m for four independent experiments per group. *P < 0.05; **P < 0.01; ***P < 0.001 (two-sided Student’s t-test performed relative to DMSO control, unless otherwise indicated).

As noted above, ABHD12 can also serve as a 2-AG hydrolase, although this function is primarily performed by MAG lipase (MGLL) in biological systems expressing both enzymes.³² We noticed, however, that MGLL was not detected in our ABPP experiments (Figure 1C, D), indicating that HT1080 cells do not express, or express very low levels, of this enzyme. Consistent with ABHD12 serving as a principal 2-AG lipase in HT1080 cells, ABHD12-KO cells showed substantially lower 2-AG hydrolysis rates compared to control cells (Figure S5). Additionally, both ABHD12-KO and DO264-treated cells displayed heightened 2-AG content compared to control cells (Figure 4A, B). Exposure to RSL3 lowered the 2-AG content of both control and DO264-treated cells (Figure 4C), indicating the potential for 2-AG to undergo peroxidation in ferroptotic cells. Finally, we note that DO34, an inhibitor of the 2-AG biosynthetic enzymes DAGLA and DAGLB, which we found to in our initial screen to protect cells from ferroptosis (Figure 1A, B), substantially lowered the 2-AG content of HT1080 cells (Figure 4D). These data, taken together, indicate that 2-AG and its presumed peroxidation products may contribute to the ferroptotic sensitivity of cancer cells.

Figure 4:

Genetic or chemical disruption of ABHD12 elevates 2-arachidonoylglycerol (2-AG) content of HT1080 cells. (A) Quantification of 2-AG content of HT1080 cells treated with DMSO or DO264 or the inactive control (S)-DO271 (1 μM for each compound, 24 h). Data represent mean values ± s.e.m. for four independent experiments per group. (B) 2-AG content of control (Ctrl) or ABHD12-KO HT1080 cells treated with DMSO or DO264 (1 μM, 24 h). Data represent mean values ± s.e.m for three independent experiments for each sgRNA population (two control, three ABHD12-KO). *P < 0.05; **P < 0.01; ***P < 0.001 (two-sided Student’s t-test performed relative to control sgRNA cells treated with DMSO). (C) Quantification of 2-AG content of ferroptotic HT1080 cells. HT1080 cells were treated with DMSO or DO264 (1 μM, 24 h), followed by DMSO or RSL3 (2 μM, 4 h). Data represent mean values ± s.e.m for three independent experiments per group. *P < 0.05; **P < 0.01; ***P < 0.001 (two-sided Student’s t-test performed relative to DMSO control, unless otherwise indicated). (D) DO34 lowers 2-AG content of HT1080 cells. Quantification of 2-AG content of HT1080 cells treated with DMSO or DO34 (1 μM, 24 h). Data represent mean values ± s.e.m for four independent experiments per group. *P < 0.05; **P < 0.01; ***P < 0.001 (two-sided Student’s t-test performed relative to DMSO control).

Here, we have performed a focused screen of serine hydrolase inhibitors to identify lipid metabolic enzymes that modulate ferroptotic sensitivity of human cancer cells. That the enzymes discovered to either potentiate (ABHD12) or suppress (DAGLA/B) ferroptosis both regulate C20:4 lipids supports the model that ferroptotic potential of cells is coupled to their PUFA content. While past work has emphasized the role of PUFA-containing phosphatidylethanolamine lipids in promoting ferroptosis,⁸ our data suggest that additional classes of PUFA lipids (e.g. C20:4 PS, C20:4 MAG) may contribute as well. We also cannot exclude that some of the lipid changes caused by ABHD12 (and DAGLA/B) blockade may contribute to ferroptosis through a signaling mechanism that does not involve peroxidation of C20:4 acyl chains. Both lyso-PS and 2-AG, for instance, act on GPCRs, for instance, to regulate diverse cellular processes,^{35, 36} and it will be important, in the future, to evaluate the potential role of these signaling pathways in ferroptosis.

In considering the potential pathophysiological and translational implications of our findings, we note that loss of ABHD12 in humans leads to an age-dependent polyneuropathy disorder termed PHARC.³⁷ It is therefore interesting to speculate whether ferroptosis in the CNS, where ABHD12 is highly expressed, may contribute to the development of PHARC. ABHD12 is also found at high levels in innate immune cells, and pharmacological or genetic disruption of this enzyme has immunostimulatory effects in vivo.³⁸ Recent work points to an important role for GPX4-dependent lipid peroxidation in macrophage pyroptotic cell death,³⁸ and understanding how ABHD12 and other PUFA-related metabolic enzymes modulate this process in innate immune cells represents an interesting future objective. Finally, we should qualify the extent to which our findings can be interpreted for cancer biology and therapy. First, we have only evaluated the contributions of ABHD12 in two ferroptosis-sensitive human cancer cell lines – HT1080 and SU-DHL-5 – and the magnitude of potentiation of GPX4-induced ferroptosis caused by ABHD12 blockade was relatively modest in these cells (~1.5-fold), precluding assessment of the potential for synergy. Going forward, it will be important to establish a broader understanding of the role that ABHD12 plays in ferroptosis through, for instance, the analysis of additional cancer cell types both in vitro and in vivo. Also, some of the metabolic functions performed by ABHD12 in HT1080 cells, such as the hydrolysis of 2-AG, likely reflects the absence of other enzymes (e.g., MGLL) in this cell type. To the extent that a ferroptosis-relevant lipid species is subject to regulation by different enzymes, it is possible that this context-dependent metabolism could offer a way to tailor ferroptotic outcomes to specific cell types and avoid broader systemic effects that may occur if targeting a central ferroptosis regulator like GPX4. We therefore encourage the continued pursuit of enzymes that can modulate ferroptotic potential in a cell- and/or state-dependent manner.