7-Dehydrocholesterol is an endogenous suppressor of ferroptosis

Ferroptosis is a form of cell death that has received considerable attention not only as a means to eradicate dened tumour entities but also because it provides unforeseen insights into the metabolic adaptation exploited by tumours to counteract phospholipid oxidation. Here, we identify a pro-ferroptotic activity of 7-dehydrocholesterol reductase (DHCR7) and an unexpected pro-survival function of its substrate, 7-dehydrocholesterol (7-DHC). Although previous studies suggested that high levels of 7-DHC are cytotoxic to developing neurons and favour lipid peroxidation, we now demonstrate that 7-DHC accumulation confers a robust pro-survival function in cancer cells. 7-DHC, due to its far superior reactivity towards peroxyl radicals, is shown here to effectively shield (phospho)lipids from autoxidation and subsequent fragmentation. We further demonstrate in a subset of ferroptosis-sensitive Burkitt lymphomas - where DHCR7 mutations have been reported - that the accumulation of 7-DHC is sucient to suppress the basal sensitivity of cells toward ferroptosis, thereby translating into an unexpected growth advantage. Conclusively, our ndings provide compelling evidence of a yet-unrecognised anti-ferroptotic activity of 7-DHC as a cell-intrinsic mechanism that could be exploited by cancer cells to escape ferroptosis. The present study uncovered and characterised an unexpected role for 7-dehydrocholesterol reductase (DHCR7) in the ferroptotic process. DHCR7 catalyses the nal step in cholesterol biosynthesis, and its inhibition leads to the accumulation of 7-dehydrocholesterol (7-DHC). 7-DHC was initially reported to accumulate in preputial gland tumours by Kandutsh and Russel 17 and whose function, at that time, was only assumed to function as a spare capacity for cholesterol synthesis. We show that the accumulation of 7-DHC translates into an increased tolerance towards (phospho)lipid peroxidation, thus providing a robust and unexpected resistance to ferroptosis. This mechanism could be potentially exploited by Burkitt ´s lymphomas (BL) to overcome the metabolic stress characteristic of their lineage specic low system x c- activity 18 . Moreover, the detailed characterisation of the unique protective effect of 7-DHC on (phospho)lipid peroxidation allows us to provide evidence that ferroptotic cell death is a consequence of the accumulation of oxidatively truncated (phospho)lipid species rather than solely lipid hydroperoxide accumulation and that these species are integral players in ferroptosis execution.

to suppress the basal sensitivity of cells toward ferroptosis, thereby translating into an unexpected growth advantage. Conclusively, our ndings provide compelling evidence of a yet-unrecognised antiferroptotic activity of 7-DHC as a cell-intrinsic mechanism that could be exploited by cancer cells to escape ferroptosis.

Main Text
Ferroptosis has attracted considerable attention in recent years 1 , and a detailed characterisation of this pathway has uncovered it's implication in a series of pathological conditions ranging from tissue ischaemia/reperfusion injury to infection 2,3 . Moreover, the modulation of ferroptosis is increasingly recognised as a potential avenue for developing therapeutics against various diseases 4 . At the molecular level, ferroptosis was initially characterised as the cell death process induced by cysteine starvation, usually caused by lack or insu cient activity of the cystine-glutamate antiporter (designated system x c -) 5 .
Low system system x c activity directly impacts substrate availability for the key enzyme regulating ferroptosis, namely glutathione peroxidase 4 (GPX4), leading to lipid hydroperoxide accumualtion and cell death 6 . Early works have established the central role played by the enzymatic activity of GPX4 in suppressing the process of ferroptosis [7][8][9] . GPX4 is the sole enzyme responsible for reducing peroxidised phospholipids 10 and can be inhibited by a series of alkylating small molecules, such as RSL3 and ML210 8 , leading to cell death in ferroptosis-sensitive cancer cell lines. The initial characterisation of this pathway demonstrated the critical role of esteri ed polyunsaturated fatty acid (PUFA) oxidation downstream of GPX4 inhibition as a driver of ferroptosis 11,12 . Speci cally, we and others have shown that the activity of the acyl-CoA-synthetase long-chain family 4 (ACSL4) is required for the enrichment of phospholipids with PUFAs 11 . The enrichment of phospholipids with PUFAs results in a marked dependency on GPX4 activity 11,13 . Accordingly, the inhibition of GPX4 in ferroptosis-prone cell lines leads to the characteristic oxidation ngerprint entailing the accumulation of peroxidised products of phosphatidylethanolamine (PE) containing arachidonic acid (AA) and adrenic acid (AdA) 12 . It has been further demonstrated that the sole accumulation of peroxidised fatty acids is not su cient to induce ferroptosis, and a central role in the free radical-mediated propagation step de ned 14 . This process has been shown to contribute to the formation of pore-like structures of ill-de ned identity 15 that drive the osmotic lysis of the cells 16 .
The present study uncovered and characterised an unexpected role for 7-dehydrocholesterol reductase (DHCR7) in the ferroptotic process. DHCR7 catalyses the nal step in cholesterol biosynthesis, and its inhibition leads to the accumulation of 7-dehydrocholesterol (7-DHC). 7-DHC was initially reported to accumulate in preputial gland tumours by Kandutsh and Russel 17 and whose function, at that time, was only assumed to function as a spare capacity for cholesterol synthesis. We show that the accumulation of 7-DHC translates into an increased tolerance towards (phospho)lipid peroxidation, thus providing a robust and unexpected resistance to ferroptosis. This mechanism could be potentially exploited by Burkitt s lymphomas (BL) to overcome the metabolic stress characteristic of their lineage speci c low system x c activity 18 . Moreover, the detailed characterisation of the unique protective effect of 7-DHC on (phospho)lipid peroxidation allows us to provide evidence that ferroptotic cell death is a consequence of the accumulation of oxidatively truncated (phospho)lipid species rather than solely lipid hydroperoxide accumulation and that these species are integral players in ferroptosis execution.
DHCR7 is a pro-ferroptotic gene Spurred by the still incomplete understanding of the ferroptotic process and the development of nextgeneration single guide RNAs (sgRNAs) 19 , we performed a genome-wide reverse genetic CRISPR screen using second-generation CRISPR libraries to identify genes that may confer robust protection against ferroptosis. To this end, the Pfa1 cell line 6 was transduced with a CRISPR library covering 18,424 genes with a total representation of 90,230 sgRNAs followed by a stringent selection for fourteen days using 200 nM of the GPX4 inhibitor (1S,3R)-RSL3, from now on only RSL3 (Fig. 1a). Consistent with the results of our previous screen and those of other groups, Acsl4 emerged as the highest-scoring hit 11,13,[20][21][22] . The second top-scoring gene with multiple sgRNAs enriched was Dhcr7 (Fig. 1b). The identi cation of Dhcr7 as a potential pro-ferroptotic gene was unexpected in light of a number of studies indicating that loss or inhibition of DHCR7 is associated with increased lipid peroxidation 23,24 . Intrigued by this nding, we set out to explore the basis of this unanticipated discovery. Using the bona de ferroptosis cell line model HT1080, we generated DHCR7-de cient cell lines using two independent sgRNAs. The successful loss of DHCR7 (Fig. 1c) was mirrored by the accumulation of its direct substrate 7-DHC. Notably, cholesterol depletion was not observed as most of it stems from the uptake of cholesterol present in the serum.
Importantly, knockout of DHCR7 did not affect the expression of known ferroptosis regulators (Fig. 1d). In support of this conclusion, we could con rm the screening results showing that DHCR7-de cient HT1080 cells present a marked resistance towards ferroptosis inducing compounds (Fig. 1e). Similar results were obtained with three independent clonal cell lines derived from Pfa1, HT1080 and MDA-MB-435 cells, supporting the general impact of this system in preventing cell death (Extended Data Fig. 1). We further show that DHCR7 loss does not modulate sensitivity to a panel of cytotoxic compounds, thus highlighting its speci city to the ferroptotic process (Extended Data Fig. 1d, f). We further corroborated these ndings with studies of a clonal cell line derived from the HT1080 DHCR7 knockout (KO) pool to avoid confounding results from non-edited cells. Using this cell line, we could unequivocally show the proferroptotic activity of DHCR7 as genetic reconstitution of a sgRNA resistant DHCR7 variant abolished 7-DHC levels and re-sensitised cells to ferroptosis without impacting on the cell's response to other cytotoxic agents (Fig. 1f, g and Extended Data Fig. 1e, f).

7-DHC is a bona de anti-ferroptotic metabolite
In the penultimate step of the cholesterol biosynthesis pathway, lathosterol, through lathosterol oxidase (SC5D), is converted to 7-DHC, which, in turn, is reduced to cholesterol by DHCR7 in the nal step of the cholesterol pathway (Fig. 2a). Several earlier studies have pointed to a toxic effect of 7-DHC accumulation via an increased susceptibility toward lipid peroxidation due to the very high inherent reactivity of 7-DHC 24 . To shed light on these seemingly contradictory ndings, we generated a DHCR7/SC5D double mutant cell line to address whether 7-DHC accumulation indeed mediates the protective effects induced by the loss of DHCR7. In agreement with a protective effect of 7-DHC, the loss of SC5D in the DHCR7 knockout cell line completely abolished the protective effect conferred by the single loss of DHCR7 (Fig. 2b). As expected, deletion of both genes led to a detectable accumulation of lathosterol and completely abolished the formation of 7-DHC (Fig. 2c). Subsequently, the serial reconstitution of DHCR7 and SC5D in a DHCR7/SC5D knockout background demonstrated that only the re-expression of SC5D resulted in an accumulation of 7-DHC (Extended Data Fig. 2a), and consequently a marked resistance to ferroptosis induced by GPX4 inhibition (Fig. 2d), but not to other cytotoxic agents (Extended Data Fig. 2b). Similarly, pharmacological inhibition of several upstream steps of the cholesterol biosynthetic process also resulted in the complete loss of the protective effect conferred by loss of DHCR7 (Extended Data Fig. 2c, d). Using the DHCR7 and the DHCR7/SC5D de cient cell lines in a series of sterol supplementation experiments, we further demonstrate that exogenous supplementation of 7-DHC protected all cell lines from ferroptosis (Fig. 2e). Additionally, lathosterol only increased ferroptosis resistance in cell lines able to accumulate 7-DHC. Interestingly, the high levels of cholesterol assayed here blunted the protective effect in all genotypes, likely due to feedback inhibition of upstream steps of the mevalonate pathway (Fig. 2e). To further corroborate the anti-ferroptotic role of 7-DHC in ferroptosis, we could show that the protective effect was also observed in the Pfa1/TAM system, a genetic model of Gpx4 de ciency 6 (Extended Data Fig. 2e). In line with these observations, 7-DHC was the only sterol able to suppress BODIPY-C11 oxidation, a marker of lipid peroxidation (Fig. 2f). These results rmly establish a yet-unrecognised role of the endogenous metabolite, 7-DHC, in preventing (phospho)lipid oxidation and associated death by ferroptosis.

7-DHC protects phospholipids from autoxidation
7-DHC is reported to be among the most autoxidisable lipid components in vivo 24 . To investigate the impact of this sterol in a well-de ned (phospho)lipid autoxidation model, we prepared unilamellar liposomes of soy phosphatidylcholine (PC) loaded with 7-DHC (Fig. 3a). We used the recently developed FENIX assay 25 , which employs the lipophilic radical initiator (E)-1,2-bis((2-methyldecan-2-yl)oxy)diazene (DTUN) to speci cally generate lipid peroxyl radicals. STY-BODIPY competes with phospholipids for propagating lipid peroxyl radicals, and the uorescence of its oxidised product(s), STY-BODIPY ox , can be monitored by uorescence (Fig. 3a, b and c). Typical radical trapping antioxidants (RTA), such astocopherol or its truncated form, 2, 2, 5, 7, 8-pentamethyl-6-hydroxychromane (PMC), inhibit autoxidation and thus retard STY-BODIPY oxidation until the RTA is consumed (Fig. 3A). Interestingly, 7-DHC-enriched liposomes resulted in a dose-dependent suppression of the rate of STY-BODIPY oxidation, albeit without the evident in exion point characteristic of good RTAs, exempli ed by PMC ( Fig. 3b and c). Since the suppression of STY-BODIPY oxidation could arise from dilution of the pool of autoxidisable phospholipids upon supplementation of the liposomes with 7-DHC, similar experiments wherein nonoxidisable DPPC were incorporated in place of 7-DHC were performed, allowing us to demonstrate no difference from the native soy PC liposomes (Fig. 3b). Furthermore, since sterols alter membrane uidity and could confer protection through dynamic parameters 26 that could impact lipid peroxidation 27 , corresponding experiments were carried out on cholesterol-loaded liposomes (Fig. 3c). Yet again, there was no effect on the rate of STY-BODIPY oxidation -even beyond concentrations of 7-DHC used (Extended Data Fig. 3a) -suggesting that physical changes in the bilayer imparted by the sterol framework do not impact the oxidation rates in our model system, neither do impact their integrity (Extended Data Fig. 3b). Given the indirect nature of the assay, we also directly measured the impact of 7-DHC on soy PC peroxidation, i.e. PLPC-OOH, DLPC-OOH and DLPC-2OOH, by LC-MS/MS (Fig. 3e, f). While supplementation of the liposomes with DPPC (up to 32 mol%) had no effect on the rate of PLPC and DLPC oxidation, cholesterol (at 8 mol%) had a modest effect on the accumulation of PLPC-OOH, DLPC-OOH and DLPC-2OOH. Entirely consistent with the FENIX results, 7-DHC supplementation led to a dosedependent suppression in the rate of PLPC and DLPC oxidation. To demonstrate that this suppression corresponded with the intervention of 7-DHC in the radical chain reaction, the consumption of 7-DHC was monitored spectrophotometrically via its characteristic absorbance (Fig. 3f, g and Extended Data Fig. 3c). This data suggests that the oxidation of 7-DHC in vitro is responsible for the inhibition of (phospho)lipid peroxidation a notion we could further validate in a model using iron/ascorbate as the source of oxidation (Extended Data Fig. 3d). Hence, if this hypothesis were correct, 7-DHC oxidation should lead to the accumulation of these products during the course of ferroptosis, and by doing so, it could spare phospholipids from accumulating oxidative damage. Fig. 3h illustrates the major detectable products formed upon 7-DHC oxidation, where the major quanti able product is the oxysterol 3β,5αdihydroxycholest-7-en-6-one (DHCEO). To assess if 7-DHC oxidation products also accumulate upon triggering ferroptosis in cells, we treated the HT1080 DHCR7/SC5D double-knockout cell line expressing SC5D and an empty vector with the GPX4 inhibitor RSL3. While no major loss in the total content of 7-DHC was noticeable (Extended Data Fig. 3e), the exact quanti cation of the two major non-enzymatic oxidation products of 7-DHC, namely DHCEO and 4a-OH 7-DHC and, revealed a signi cant increase ( Fig.  3i and Extended data 3e). To demonstrate that the 7-DHC products originate from the peroxyl radicalmediated oxidation of 7-DHC, we further incubated these cells with the RTA and ferroptosis inhibitor liproxstatin-1 (Lip1) 9 . In agreement with the free radical-mediated formation of DHCEO and 4a-OH 7-DHC, Lip1 fully inhibited their accumulation (Fig. 3i). Hence, these results strongly suggest that due to its inherent reactivity, 7-DHC autoxidises preferentially, thereby suppressing the propagation of peroxyl radical-mediated (phosphor)lipid damage.

7-DHC suppresses the formation of truncated phospholipids and membrane rupture
Following these results, we reasoned that the presence of 7-DHC in phospholipid bilayers generates a strong pro-survival effect by increasing the resistance of membranes to oxidation-mediated permeabilisation. Therefore, a model system was employed that consists of 5(6)-carboxy uorescein (CF) encapsulated in liposomes allowing for the detection of a uorescent signal upon membrane permeabilisation (Extended Data Fig. 4a). Using the iron/ascorbate couple as a well-established oxidation model, we now show that liposomes containing 7-DHC are remarkably resistant to oxidation-mediated membrane permeabilisation (Extended Data Fig. 4b).
To further support the relevance of this simpli ed system for ferroptosis, we could show that the process of vesicle rupture could be prevented entirely by the ferroptosis inhibitor Lip1 (Extended Data Fig. 4c), indicating that Lip1 acts similarly to prevent membrane permeabilisation in cells. Recent reports studying the relative contribution of different photosensitisation mechanisms to membrane permeabilisation suggested that truncated phospholipid species rather than phospholipid hydroperoxide are key in generating membrane pores and consequently mediating the loss of membrane integrity 28 . Therefore, we reasoned that a similar mechanism could be at play during iron-induced permeabilisation and ferroptosis execution 29 . As such, we next explored the feasibility of this mechanism using our cellular models treated with a GPX4 inhibitor. In-depth epilipidomics analysis indeed detected a substantial accumulation of PE and plasmalogen PE truncated products in cells undergoing ferroptosis (Fig. 4a). Notably, cell permeabilisation, monitored as PI-positive cells, was only detectable in conditions marked by an increase in these oxidised and truncated species (Fig. 4a). We also show that Lip1 fully inhibited the formation of these species, thus con rming their origin from the autocatalytic lipid peroxidation process (Fig. 4a). In accordance, cells accumulating 7-DHC behaved similarly to Lip1-treated cells and the speci c oxidation product of 7-DHC, DHCEO, accumulated in these cells. This thus demonstrates that 7-DHC is preferentially oxidised in cells, thereby sparing phospholipids and preventing the formation of oxidised and truncated species (Fig. 4a). To establish the functional link between truncated lipids and ferroptosis execution, we assayed a panel of different truncated species regarding their capacity to destabilise membranes in model systems and in cells (assayed structures are depicted in Extended Data Fig. 4d). Accordingly, all tested truncated lipids were able to permeabilise liposomal membranes and to kill cells more e ciently than the parental lipid and the corresponding hydroperoxide (Extended Data Fig. 4e). In line with the proposed mechanism 7-DHC did not affect permeabilization mediated by truncated phospholipid species (Extended Data Fig. 4eh). We reasoned that the extent of the bilayer packing alterations induced by the truncated lipids, and thereby their potency to destabilise the membrane in order to generate pores and induce cell death, would depend on their acyl chain length, with shorter truncated chains being more cytotoxic. Although no apparent relationship with the length of the truncated tail was evident in the above experiments, it was clear that exogenous addition of the lipids could result in less e cient membrane incorporation of the shorter and more hydrophilic species. To circumvent this issue, a system in which the species are formed in situ would be required. We took advantage of the cell's own fatty acid incorporation machinery to achieve this goal. ACSL4-de cient cells have a profound loss of PUFA content in membranes, resulting in a marked resistance to ferroptosis due to the lack of oxidisable substrates. Sensitivity to ferroptosis in this setting can be regained by feeding exogenous PUFAs 11 . This feature should facilitate a better control of the substrates utilised for ferroptosis execution. Using this model, we compared side-by-side the sensitisation provided by α-linolenic acid (αLNN) and γ-linolenic acid (γLNN). Both fatty acids have an identical structure in length and number of double bonds leading to a similar propensity to be oxidised, yet the position of the last double bond determines the structure of the resulting truncated product. Analysis of the lipidomic changes of ACSL4 wildtype (WT) and KO cells treated with αLNN and γLNN con rmed that both lipids are directly and e ciently esteri ed into PE suffering limited metabolisation to longer and esteri ed species, likely a re ex of the loss of ACSL4 and its requirement for the e cient metabolisation of these species (Fig. 4b, c). The supplementation restored the oxidisable pool of PUFA to a similar extent as in WT cells (Fig. 4b, d). Remarkably, despite their equal abundance and propensity to undergo oxidation, γLNN appeared to be a superior ferroptosis executing substrate ( Fig. 4d and Extended Data Fig. 4i), in line with its potential to generate shorter truncated phospholipid products. These results are remarkable because they indicate that the product formed determines cell death rather than solely its capacity to autoxidise. Together, these observations provide compelling evidence for the role of truncated products in contributing to ferroptosis execution and that 7-DHC and other ferroptosis inhibitors such as Lip1, directly suppress their formation.

7-DHC accumulation increase lymphoma cell tness
Having characterised the molecular underpinnings by which 7-DHC prevents ferroptosis execution, we next asked if this protective effect could have a potential role in supporting tumour growth under conditions where ferroptosis inhibition is critical. To our initial surprise, DHCR7 mutations, despite being rare, have been described in Burkitt's Lymphoma (BL) patients 30 . BL is a tumour entity characterised by MYC translocations and is considered the prototypic ferroptosis cancer entity. The reason for this traces back to their inherent low activity of system x c -18,31 , likely re ecting metabolic adaptation required to spare glutamine whose dependency is increased in cell expressing high MYC levels 32 . Accordingly, the requirement of thiol donating compounds, such as ß-mercaptoethanol (ßMe), to support the growth of murine leukemic b-cell and human BL cell lines has been known for many decades 18,33 . Accordingly, we rst set out to explore the function of the reported mutations described for DHCR7 30 and assess if they could, in principle, increase the tness of BL. Brie y, mutations N274K and L306R have been reported in two BL patients, and a second mutation, A24S, was reported in two different BL cell lines (Raji and BL58); we additionally included another mutation identi ed in a MM cell line (L317V). We created a model for the DHCR7 structure using a homologous structure (pdb id 4QUV, sequence identity 37%, similarity 51%) to gain insights into the molecular consequences of these mutations. L306R, N274K and L317V are predicted to be located in the transmembrane domain (Fig. 5a). While the substitution of the hydrophobic amino acid leucine by another hydrophobic amino acid, valine, is predicted to be tolerable in the hydrophobic membrane interior, the introduction of a positively charged amino acid (K or R) is highly disfavoured thermodynamically in transmembrane regions of proteins and could result in misfolding of the protein. Re-expression of DHCR7-Flag-tagged version of the four corresponding mutants in the DHCR7-de cient HT1080 cell line allowed us to validate these predictions experimentally. Fig. 5b shows that except for mutation N274K all are generally well expressed compared to the WT. The potential misfolded nature of mutant N274K is likely, in addition of been thermodynamically unfavourable, a consequence of the disruption of the helix-helix interactions established between N252 and L253 (Extended Data Fig. 5a). Next, we addressed the functionality of these mutations and, in agreement with the predictions, the A24S and L317V mutations appear to lead to functional enzymes able to metabolize 7-DHC when overexpressed (Fig. 5c) and re-sensitise the DHCR7 de cient cells to ferroptosis like the WT enzyme (Fig. 5d). On the other hand, the two mutations reported in patients, N274K and the L306R are not functional, cannot metabolise 7-DHC ( Fig. 5c) and cannot restore sensitivity to ferroptosis (Fig. 5d). Of notice, the A24S variant which behaved similarly to WT DHCR7 introduces a serine at the N-terminus, which has three reported phosphorylation sites, S5, S14 and S25 34 . We, therefore, subjected the WT and A24S mutant sequence to the NetPhos server 35 , which was able to correctly identify the three known sites with scores of 0.556, 0.606 and 0.643, respectively (typical con dence threshold set at 0.5). Intriguingly, this analysis predicted A24S as an additional phosphorylation site, with an extremely high con dence score (0.991), and further increased the con dence of the neighbour S25 to be phosphorylated (0.774). Thus, it is tentative to speculate that the A24S mutation could be a neo-phosphosite that, in context speci c situation would be able to modulate DHCR7 function/levels and ultimately impact on 7-DHC levels. In-depth studies to explore this possibility are certainly warranted.
Given the lack of available cell lines with the loss of function mutations we further study the role of 7-DHC in suppressing ferroptosis by deleting DHCR7 in the BL cell line BL41 and in the multiple myeloma (MM) cell line KMS26, which shares the BL thiol dependency for growth. Following our previous results, genetic loss of DHCR7 in these cell lines conferred robust protection towards GPX4 inhibitors, and to some extent, this effect appeared to be even more pronounced (Fig. 5e). Strikingly, loss of DHCR7 abolished the characteristic thiol dependent growth as both cell lines could proliferate in the absence of thiol donating compound (Fig. 5f). Highlighting the speci city of this effect, the inhibition of DHCR7, using the highly speci c DHCR7 inhibitor RB38 36 , could also bypass the dependency on thiol donating compound as well as Lip1 (Extended Data Fig. 5b). Accordingly, the genetic and pharmacological effects were blunted by inhibiting upstream step in the biosynthesis of 7-DHC, speci cally by targeting lathosterol biosynthesis using the emopamil binding protein (EBP) inhibitor Tasin-1 (Fig. 5f). Using these pharmacological tools, we expanded this observation to a larger panel of cell lines where we could show that only cells accumulating 7-DHC are able to grow in the absence of thiol donating compounds, with the exception of the BL2 cell line, which was unable to accumulate 7-DHC upon inhibition and did not show any growth limitations (Extended Data Fig. 5c, d). Having established a pro-survival role of 7-DHC in a subset of B-cell lymphomas we provide an initial assessment of the in vivo relevance of this nding using a xenograft de cient for DHCR7. For this, we used the KMS26 cell line de cient for DHCR7 and could report a signi cant growth advantage, suggesting that lipid peroxidation is a metabolic hurdle for these xenografts and that the accumulation of 7-DHC can mitigate this metabolic stress in vivo.

Discussion
Conclusively, our work adds to expanding biological activities of 5,7-unsaturated sterol metabolites as recent data have suggested a potential role for this class of metabolites in mitochondrial quality control 37 and immunity 38,39 . Speci cally, we identify that 7-DHC is an endogenous metabolite that robustly protects cellular membranes from (phospho)lipid peroxidation and associated ferroptotic cell death. We also demonstrate that, by preventing (phospho)lipid peroxidation, 7-DHC suppresses the formation of (phospho)lipid-truncated species, which are likely the most downstream executors of ferroptosis and, upon reaching a certain threshold, could be considered as the point of "no return" in ferroptosis. Furthermore, we provide compelling evidence that the accumulation of 7-DHC increases the tness of BL cells and could compensate for their intrinsic low system x c activity and increased dependency on GPX4.
This recognition is critical as recent reports have indicated that high MYCN levels, a close homologue of the BL driving oncogene MYC 30 , increases cancer cells dependency on GPX4 to supress ferroptosis 40,41 . Given the already reported multiple levels of posttranslational regulation of DHCR7 via ubiquitination and phosphorylation 34,42-44 , our work should stimulate a better understanding of the events that can disrupt DHCR7 activity beyond the obvious loss of function identi ed in the mutations assayed here.
Finally, analogously to our observations, ergosterol, the major sterol found in cell membranes of fungi and protozoa and contains an indistinguishable sterol ring from 7-DHC, has been repeatedly associated with increased tolerance to oxidative stress 36,37,[45][46][47][48] . These studies suggest that the mechanism protecting membranes from (phospho)lipid peroxidation described here could be an overlooked and general tolerance mechanism kept across multiple species and highjacked by cancer cells to evade ferroptosis.      Determination of cell numbers: 50,000 BL cells were seeded on a 6-well plate in triplicates at density of 25,000 per ml. The cell number was determined for a period of 28 days using a Neubauer improved chamber. The cells were kept at a constant split ration of 1 to 2 every third day.
Preparation of lentiviral particles. HEK 293T cells were used to produce replication-incompetent lentiviral particles pseudotyped with the ecotropic envelope protein of the murine leukaemia virus (MLV) or the pantropic envelope protein VSV-G. A third generation lentiviral packaging system consisting of a transfer plasmid, pEcoEnv-IRES-puro (ecotropic particles) or pMD2.G (pantropic particles), pMDLg_pRRE and pRSV_Rev was co-lipofected into HEK 293T cells using HiPerFect (Roche  processed by smoothing (scan window 2, 20 smooths, method: mean) and taking the ratio of PLPC-OOH peak integration / IS peak integration. Each reaction was repeated at least twice and is reported as the mean ± standard deviation for the kinetic plot or mean ± standard error for relative rates derived from linear regression.

UV-Vis Analysis of Soy PC Autoxidations
A 3 mL quartz cuvette was equilibrated in a Cary 100 spectrophotometer at 37°C with PBS (2.38 mL) for 5 mins, and then baselined. Liposomes were added (83.3 μL of 30 mM) and the cuvette inverted 5 times to mix before an initial spectrum was recorded. The reactions were then initiated with addition of DTUN (41.7 μL of 12 mM in EtOH), the cuvette inverted 5 times to mix, and spectra from 260 to 300 nm were recorded every 10 mins. The spectra were processed by subtracting each spectrum of the 7-DHC + DTUN loaded liposomes by the rst spectrum of vehicle liposomes + DTUN. A standard curve for 7-DHC in liposomes was prepared in a similar manner using the spectra obtained with liposomes prepared with non-puri ed soy PC that contained inhibitor to minimise 7-DHC autoxidation. The 7-DHC was quanti ed at 294 nm to minimise interference by lipid conjugated diene formation. The resulting kinetic traces eventually begin to increase due to these products and the formation of 7-DHC derived oxidation products. For this reason, the loss of absorbance at 294 nm plateaus at ca. 60% of the expected conversion of 7-DHC initially in the liposome sample.

Iron mediated liposomal oxidation
Lipid oxidation analysis through Ultra High Performance Liquid Chromatography (UHPLC).
All reagents and lipid standards were purchased from Sigma Aldrich (St. Louis, US) or Avanti Polar Lipids (Alabaster, US). Organic solvents were purchased from Supelco/Merck KGaA (Darmstadt, Germany).
Quanti cation of lipid substrates and oxidation products via UHPLC coupled to UV detection Collected aliquots were analysed through reversed-phase HPLC (Nexera UHPLC, Shimadzu, Kyoto, Japan) coupled to UV detection (scan from 190 to 370 nm) using a "Luna 5u C8 (2)  After this step, cells were washed with PBS (600 g for 5 min), and ressuspended in 2,2 mL of PBS. Two aliquots of 1 mL were pelleted and frozen in liquid N 2 . 50 µL of cell suspension was kept for PI analysis as described previously. DHCEO, 4α-OH-7-DHC and 4β-OH-7-DHC were analysed by LC-MS/MS using an APCI source in the positive ion mode as described previously 12 . Brie y, lipid content from cell lysate was extracted and the neutral lipids fraction was puri ed by SPE chromatography. Puri ed content was resuspended in methanol and 10

Bioinformatics
Homology models were generated using SWISS-MODEL 18 via its integrated web-based service available at https://swissmodel.expasy.org/. We used the target-template alignment function of swiss model to match the human DHCR7 sequence with the Methylomicrobium alcaliphilum sequence, and modelled the DHCR7 structure using pdb le 4QUV. Prediction of phosphorylation sites was carried out using the NetPhos 3.1 server 19 using default parameters. Protein structures were visualised using PyMOL (version-2.3.4, Schrodinger, LLC). Amino acid neighbors were identi ed using a cut-off distance of 5Å. The DHCR7 transmembrane boundaries were predicted based on the 4QUV positioning in a lipid bilayer that had been predicted by minimising its transfer energy from water to the membrane and stored in the Orientations of Proteins in Membranes (OPM) database 20 .

Xenograft experiments
Animal studies were approved by the district government of lower Franconia (protocol number 55.2-2532-2-335) and were conducted in accordance with the US National Institutes of Health Guide for the Care and Use of Laboratory Animals. Brie y, female NOD.Cg-Prkdcscid Il2rgtm1WjI/SzJ (NSG)-mice (8 to 12 weeks old) were purchased from Charles River, Sulzfeld. A mixture of 50 µL ECM gel (Merck, Darmstadt, Germany) and 50 µL RPMI-1640 medium containing 5x10 5 cells was injected, subcutaneously on the right and left anks of the mice, genotypes of the cells were kept blinded. Four to ve weeks after injections, animals were euthanized, the tumour explanted and its mass determined.
Data presentation and statistical analyses. Data are presented as mean ± s.d. unless stated otherwise. As a general rule for cell-based experiments, graphs show the mean ± s.d. of n = x wells (x values are given in the gure legends) representative of a single experiment performed independently y times (y value is given in gure legends) for reproducibility. Statistical analysis was performed using GraphPad Prism 5.0 software. and GPX4 in DHCR7 de cient cells. e, Dose-dependent toxicity of the ferroptosis inducers RSL3, ML210 and FIN56 in HT1080 cell lines stably transduced with a vector expressing Cas9 and an sgRNA targeting DHCR7 and EGFP as a control. Data are the mean ± s.d. of n = 3 wells of a 96-well plate from one representative of two independent experiments. f, Relative quanti cation of 7-DHC levels in HT1080 levels WT and DHCR7-Knockout cells re-expressing DHCR7 and an empty vector. Data are the mean ± s.d. of n = 3 wells of a 6-well plate from one representative experiment. g, dose-dependent toxicity of RSL3 and FIN56 in HT1080 Cas9 DHCR7-KO clone, and overexpressing DHCR7 or mock. Cell viability was monitored using Alamar blue. Data are the mean ± s.d. of n = 3 wells of a 96-well plate from one representative of two independent experiments; *P < 0.05; two-way analysis of variance (ANOVA).