A lipid signature of BAK-driven apoptotic pore formation

Apoptotic cell death is regulated by the BCL -2 protein family, with clusters of BAK or BAX homodimers driv ing pore formation in the mitochondrial outer membrane via a poorly understood process. There is growing evidence that, in addition to BAK and BAX, lipids play an important role in pore formation . Towards a better understanding of the lipidic drivers of apoptotic pore formation in isolated mitochondria, two complementary approaches were taken. Firstly, the lipids released during BAK-mediated pore formation were measured with targeted lipidomics, revealing enrichment of long chain polyunsaturated lysophospholipids (LPLs) in the released fraction. In contrast, the BAK protein was not released suggesting that BAK and LPLs locate to distinct microdomains. Secondly, added cholesterol not only prevented pore formation but prevented the clustering of BAK homodimers. Our data lead us to a model in which BAK clustering triggers formation of a separate microdomain rich in LPLs that can progress to lipid shedding and the opening of a lipid-lined pore. Pore stabilisation and growth may be due to BAK dimers then moving to the pore edge. Our BAK-lipid microdomain model supports the heterogeneity of BAK assemblies, and the observed lipid-release signature gives new insight into the genesis of the apoptotic pore. preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.618570doi: bioRxiv preprint Uren et al | A lipid signature of the apoptotic pore 2

A key decision point during apoptotic cell death is the rupture of the mitochondrial outer membrane (MOM) triggered by effector proteins of the BCL -2 family , BAK (BCL-2 homologous antagonist/killer) and BAX (BCL-2-associated protein X) 1, 2, and perhaps BOK (BCL-2-related ovarian killer) 3. MOM rupture releases cytochrome c and other proteins from the mitochondrial intermembrane space into the cytoplasm, leading to the a ctivation of caspases and loss of cell viability. Here we refer to the MOM rupture that is dependent on BAK or BAX as the “apoptotic pore ”. Importantly, the m olecular mechanism of apoptotic pore formation remains unclear (reviewed in 4). BAK and BAX each comprise nine a-helices, with marked structural and sequence homology 5, 6. Initially, non-activated BAK is a globular protein anchored to the MOM by its C-terminal transmembrane domain 7, whereas BAX is largely cytosolic until it translocates to the MOM 8- 10. Upon an apoptotic trigger, initiators of the BCL-2 family (e.g. BID, BIM) are upregulated and then transiently bind to BAK or BAX to trigger a series of conformational changes including dissociation of the N- and C-termini to expose the BAK or BAX BH3 domain. The exposed BH3 domain of a ctivated BAK or BAX can then engage the hydrophobic binding groove of another activated monomer on the membrane to create a closed stable symmetric homodimer 11, 12. Structural studies of truncated BAK and BAX show that a2-a5 core dimers are rigid and present a hydrophobic surface on one face 13-17. The BAK a2-a5 core dimers also showed binding sites for phospholipid head group s and acyl chains , with lipid “bridging” between dimers 15. Related lipid binding sites were not evident in BAX a2-a5 core dimers although association with lipids was inferred as BAX structures were dependent on the lipid environment 17. In s imulations the BAK a2-a5 core dimer perturbs lipid bilayers causing increased membrane curvature and bilayer thinning 15, together with aggregation of triacylglycerides (TAGs) under the BAK a2-a5 core dimer and compensatory changes in the lipid distribution in the surrounding membrane 18. Mitochondrial studies show that beyond the a2-a5 core regions, the N- and C-termini in BAK and BAX homodimers are relatively flexible 7, 11 -14, 19 -22. While the N -terminus makes no contact with protein or membrane, the C-terminal amphipathic helices a6, 7 and 8 lie in-plane with the membrane suggesting that this region may also perturb the membrane bilayer 21. Membrane-embedded BAK dimers adopt disordered clusters on the MOM, as various contact points between dimers have been captured with cysteine linkage or magnetic distance measures but none appear essential for clustering or pore formation 19, 23, 24 . Consistent with multiple contact points, BAK and BAX dimers coalesce into various architectures including clusters, lines, rings or arcs (reviewed in 25). In addition, if cells express both BAK and BAX , the proteins form mixed unordered assemblies of variable dimensions that completely or partially encircle the pore 26. Collectively, these data argue that the apoptotic pore is not a rigid proteinaceous structure, but contains both protein and lipid elements 27. Consistent with the apoptotic pore being at least partly lipidic is the ability of the MOM phospholipid composition to influence apoptosis. Membrane composition can alter s everal steps in BAK and BAX function , including their translocation to mitochondria , activation, dimerisation and pore formation 28-35. Membrane domains with different lipid compositions may also be important , as unsaturated phospholipids were enriched surrounding BAK oligomers 36. Directly measuring how lipid composition affects pore formation in the MOM is especially challenging as the MOM lipid composition is distinct from that of other cellular preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.618570doi: bioRxiv preprint Uren et al | A lipid signature of the apoptotic pore 3 compartments including the inner mitochondrial membrane 37-42, and the MOM is difficult to physically separate from the inner membrane. Given the underlying importance of lipid composition, we examined isolated mitochondria t o further understand the involvement of lipids in apoptotic pore formation. Targeted lipidomics was used to quantify lipids released from mitochondria during pore formation, reasoning that released material may reflect a lipid defect from which the pore emerges. Moreover, studying isolated mitochondria allowed examination of lipid redistribution events in the absence of significant lipid metabolism o r lipid transfer between organelles. Notably, BAK-driven pore formation was associated with release of lysophospholipids but not BAK, suggesting the presence of discrete microdomains of lysophospholipids and of BAK , and a pore opening closest to the site of lysophospholipid aggregation rather than BAK clustering . Exogenous cholesterol was able to prevent BAK clustering and pore formation, indicating the importance of microdomains in apoptotic pore formation.

Apoptotic pore formation in isolated mouse liver mitochondria To explore the interplay between BAK and membrane lipids, mouse liver mitochondria (MLM) were freshly isolated 43. In this system, fractionation removes c ytosolic BAX, allowing BAK function to be studied in the absence of BAX. As previously 43, incubation with caspase-cleaved BID (cBID) released essentially all cytochrome c into the supernatant (Fig. 1a) and activated all BAK (Fig. 1b), which then assembled into higher order complexes (Fig. 1 c). Timecourse experiments demonstrate d significant BAK activation at 2 min (Fig. 1b), high order BAK clusters at 15 min (Fig. 1c) and full cytochrome c release at 20 min (Fig. 1a). BAK is retained in mitochondria following pore formation To examine the release of lipids as well as proteins we employed a two-step centrifugation protocol (Fig. 2a). After pore formation initiated by cBID, MLM were centrifuged at 10,000 g, and the resulting supernatant (SN 10) subjected to 100,000 g, generating a secondary pellet fraction (Pellet100). The Pellet100 fraction was the optimal fraction for lipidomics analysis as it removed buffer constituents (such as HEPES) that interfere with mass spectrometry and also concentrated the sample to improve detection (see Discussion). As a negative control for pore formation, MLM from Bak-/- mice 2 were tested in parallel with MLM from wildtype mic e. As a positive control for liberation of lipids independent of any apoptotic signalling , MLM from both genotypes were separately incubated with the phospholipase A2 (PLA2) from bee venom 44 to digest the mitochondrial membrane. While cBID failed to release cytochrome c from Bak-/- MLM, PLA2 caused cytochrome c release from both genotypes (Fig. 2 b), indicating that the lipase was acting in a BAK - independent manner. PLA2 also partially digested the inner membrane, as shown by the release of the matrix protein HSP60 (Fig. 2 b). Notably, whereas a portion of BAK was liberated by PLA2 treatment, BAK was not released during apoptotic pore formation (see Pellet100, Fig 2c). Thus, BAK is membrane -integrated in healthy mitochondria, and remains integrated and associated with the mitochondrial pellet upon apoptotic pore formation. As bee venom PLA2 is a calcium -dependent enzyme, the calcium chelator EDTA prevented PLA2-mediated release of cytochrome c and HSP60 (Fig. 3). In contrast, EDTA had no effect on BAK-mediated cytochrome c release (Fig. 3) , indicating that the apoptotic pore forms independently of calcium-dependent enzymes like PLA2. preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.618570doi: bioRxiv preprint Uren et al | A lipid signature of the apoptotic pore 4 Figure 1. Apoptotic pore formation in mitochondria as shown by cytochrome c release, BAK activation and BAK clustering. WT or Bak-/- MLM were incubated with cBID at 37oC and sampled at the indicated times. (a) Robust cytochrome c release is dependent on the presence of both cBID and BAK. Pellet and supernatant (SN) samples were immunoblotted for cytochrome c. (b) BAK activation is triggered by cBID as shown by increased susceptibility to proteinase K digestion. Aliquots removed prior to fractionation were treated with proteinase K and immunoblotted with antibody to the BAK BH3 domain (clone 4B5). After proteinase K, non- activated BAK runs as a ~23 kD fragment, while activated BAK runs as ~17 kD and ~15 kD fragments. (c) High molecular weight BAK complexes (“BAK clusters”) form in the presence of both cBID and BAK. The pelleted mitochondrial fractions from (a) were run on Blue Native PAGE and immunoblotted for BAK (aa23 -38). Addition of 5 mM EDTA immediately prior to BNP slightly enhances the detection of mouse BAK under BNP conditions, providing some evidence of dimers (2X species) forming a fter cBID, but decreasing as high molecular weight complexes form (see lane 2, 30 min i ncubation). Data are representative of three independent experiments. Figure 2. BAK is retained in the mitochondrial outer membrane during pore formation. (a) Two-step centrifugation used to fractionate proteins and lipids released from mitochondria during apoptotic pore formation. WT or Bak-/- MLM were incubated with cBID (100 nM) or PLA2 (5 µg/mL) at 37 oC for 30 min. The indicated fractions were collected for western blot and lipidomics. (b) 10,000 g fractions show that pore formation triggered by cBID released cytochrome c but not BAK or HSP60. (c) 100,000 g fractions show only background levels of BAK in Pellet 100 after cBID treatment (compare lanes 1 and 2). Note that Pellet 100 samples are ~30X more concentrated than SN 100 samples. Data are representative of three independent experiments. Figure 3. BAK -mediated pore formation is calcium-independent. Prior to incubation of WT or Bak-/- MLM with cBID (100 nM) or PLA2 (5 or 50 µg/mL), MLM were incubated with EDTA (20 mM) to sequester residual calcium in the reaction buffer, and thus inhibit any calcium -dependent enzymes such as PLA2. Pellet 10 and SN10 fractions were western blotted for cytochrome c or HSP60. Data are representative of three independent experiments. preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.618570doi: bioRxiv preprint Uren et al | A lipid signature of the apoptotic pore 5 Targeted lipidomics to identify lipids liberated from apoptotic mitochondria To test for release of lipids from mitochondria, Pellet100 fractions were collected from six repeat experiments, each experiment comparing liver mitochondria from one WT and one Bak -/- mouse under three different treatments (untreated, cBID, or PLA2) (Table S1). The expected cytochrome c release was confirmed by western blot analysis of the Pellet10 and SN10 fractions (Fig. S1). The Pellet100 samples were then analysed by mass spectrometry using a targeted lipidomics panel of 495 lipid metabolites including 4 internal standards. A total of 348 (70%) metabolites were detected (Table S2; MassHunter, Agilent), demonstrating excellent coverage by this lipid panel. According to the raw lipid counts , the majority of lipids in the Pellet100 fractions from WT Untreated MLM were PC or PE species (Table S3 and S4). PC and PE are the major components of all cellular membranes including the MOM 39, 42, indicating recovery of membrane fragments of some kind in the Pellet 100 fractions. Notably, there was very low abundance of cholesterol (<0.01 %), consistent with the low abundance of cholesterol in the MOM relative to other cellular membranes 39, 42, 45. To summarise changes in lipid release across all samples, a multidimensional scaling (MDS) analysis was performed. A distinct cluster was obtained from PLA2 samples (Fig. S2). PLA2 cleaves di-acyl phospholipids (e.g. PC and PE) at the sn-2 acyl linkage to generate free fatty acids and lysophospholipids (Fig. S3a, and reviewed 46-48). Accordingly, fold-change analysis revealed lysophospholipids (e.g. LPC and LPE) were increased in the Pellet100 fractions of PLA2 treatme nts, denoting a signature of released lysophospholipids ( Fig. S4). A corresponding decrease of substrate phospholipids (such as PC and PE) in the Pellet100 fractions was also evident (Fig. S4). Parallel production of free fatty acids by PLA2 is assumed to occur, but they are not measured by the lipidomics assay. Together, these findings indicate that PLA2 successfully hydrolysed a range of phospholipids in MLM, with expected lysophospholipid products released from MLM and recovered in the Pellet100 fractions. Specific release of long chain polyunsaturated lysophospholipids during pore formation To test for lipids released after BAK activation and pore formation, we next compared the lipidomic profiles of the Pellet100 samples from WT MLM treated with cBID versus WT Untreated controls. The “top ten” over -represented or under -represented lipids were determined by ranking data by raw p-value, and then classifying by log2fold-change as over- represented lipids (positive log 2fold-change) or under -represented lipids (negative log 2fold- change) (Fig. 4, Tables 1 and 2). A full list of fold-changes across all 348 of the detected lipids, for all contrasts, is available in Table S6. P-values were adjusted for multiple testing using the false discovery rate (FDR) method, and those with a FDR < 0.05 were considered differentially abundant. There was one significantly altered lipid, lysophosphatidylcholine (LPC 22:5(104)), FDR = 0.03224) that was significantly increased in Pellet100 fractions from cBID treated WT MLM compared to Untreated controls (Fig. 4a, Table 1 and Table S6). Remarkably, related lysophospholipids accounted for all ten of the most increased lipids liberated from apoptotic mitochondria as all were long chain (>20 carbons) and polyunsaturated (3-5 double bonds ). The top ten included LPC 22:5, an isomer of the top - ranked LPC 22:5(104). This polyunsaturated lysophopholipid signature was not observed when comparing Bak-/- cBID treated samples with Bak-/- Untreated controls (Fig. 4b, Table S6), further supporting the specificity of this signature for BAK -dependent pore formation. An important observation is that the lysophospholipids liberated with apoptosis were longer and more unsaturated than those liberated by PLA2 treatment (Fig. S4, Table S6), again indicating distinct modes of lipid release by these different treatments (see Discussion). preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.618570doi: bioRxiv preprint Uren et al | A lipid signature of the apoptotic pore 6 Lipid log2(fold- change) p-value Adjusted p- value LPC 22:5(104) 0.6779 0.0001 0.0322 LPC 22:4 0.4391 0.0009 0.1081 LPC 22:6 0.5332 0.0028 0.1499 LPC 22:5 0.5124 0.0029 0.1499 LPC 20:5 0.4737 0.0030 0.1499 LPE 22:6 0.5434 0.0030 0.1499 LPE 20:4 0.5315 0.0071 0.2877 LPC 18:3 0.4399 0.0078 0.2877 LPC 20:4 0.4749 0.0083 0.2877 LPC 20:3 0.4541 0.0112 0.3543 Lipid log2(fold- change) p-value Adjusted p- value TG 18:0_18:0_18:0 -0.5944 0.0009 0.1081 Cer(d18:1_26:0) -0.3940 0.0223 0.4755 dhCer 24:1 -0.2280 0.0547 0.9068 Cer(d17:1_24:0) -0.2568 0.0729 0.9957 PE(P-16:0_18:1) -0.1433 0.1104 0.9957 Cer(d16:1_24:0) -0.2529 0.1136 0.9957 Hex1Cer (d18:1_24:0) -0.2091 0.1223 0.9957 PC 34:0 -0.1547 0.1792 0.9957 SM 44:1 -0.1146 0.1961 0.9957 SM 41:1 -0.1272 0.2160 0.9957 Figure 4. Specific release of long chain polyunsaturated lysophospholipids (LPLs) from mitochondria during pore formation. Volcano plots compare Pellet 100 fractions from (a) WT cBID versus WT Untreated samples and (b) Bak-/- cBID versus Bak-/- Untreated samples. Lipids with FDR <0.05 are highlighted in red (increased, i.e. over-represented) or blue ( decreased, i.e. under-represented). Lipid log2(fold-change)(x-axis) is plotted against (-log10(p-value)(y-axis). The 10 most over-represented or under-represented lipids (as ranked by p-value) are labelled. (See also Table S6 for all contrasts, and Tables 1 and 2 for 10 most over-represented or under-represented lipids in the WT cBID versus WT Untreated contrast). Table 1: Over-represented lipids WT cBID versus WT Untreated (ranked by p-value) Table 2: Under-represented lipids WT cBID versus WT Untreated (ranked by p-value) preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.618570doi: bioRxiv preprint Uren et al | A lipid signature of the apoptotic pore 7 To gain a wider view of the behaviour of major lipid classes during apoptotic pore formation, lipid set enrichment analysis of the lipidomics output was p erformed by two different approaches: a specific analysis classified by lipid headgroup (ROAST 49)(Fig. 5a) or by a comprehensive analysis covering the biophysical, chemical and cell biological features of each lipid molecule (LION/web 50, 51 ) (Fig. 5b). ROAST analysis showed e nrichment of lysophospholipids with a phosphatidylcholine headgroup (LPC lipid class) following cBID treatment in the increased category ( p-value = 0.029) (Fig. 5a), consistent with eight of the top ten over-represented lipids being LPC lipids (Table 1 ). After PLA2 treatment there was enrichment for LPC, LPE and LPI headgroup classes amongst increased lipids (Fig. 5a), again highlighting the distinct lipid signatures of apoptotic pore formation and PLA2 digestion (see Discussion). The LION/web analysis confirmed enrichment of lysoglycerophospholipids Figure 5. Enrichment analysis of lipid classes detected with apoptotic signalling. (a) Preference for phosphatidylcholine headgroup in lysophospholipids released during apoptotic pore formation. An enrichment analysis based on the major lipid headgroup classes using ROAST (see Methods) was performed for Pellet100 fractions with contrasts indicated (WT cBID vs WT Untreated or WT PLA2 vs WT Untreated) (see also Table S7 for all contrasts) . ROAST enrichment p-values are shown for each lipid headgroup testing for decreased or increased lipid classes. Significant p-values (<0.05) are highlighted (purple gradient). Also note the distinct signatures for cBID compared to PLA2. (b) Preference for lipids with positive intrinsic curvature in lipids released during apoptotic pore formation. A complementary enrichment analysis encompassing a range of lipid features using the LION/web lipid ontology tool (see methods) was performed. The top 26 lipids (raw p-value < 0.2 ( Supp. Table 8)) with positive log2fold-change from the WT cBID vs WT Untreated contrast were input as the ‘lipid target list’ and the full list of detected lipids (384) were input as the ‘lipid background list’. “Lysoglycerophospholipids” (which encompasses LPC), “lipid -mediated signalling ” and “positive intrinsic curvature” were the top 3 category descriptors. A vertical dotted line indicates the FDR q-value significance cut off of 0.05. preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.618570doi: bioRxiv preprint Uren et al | A lipid signature of the apoptotic pore 8 (which includes LPC) (Fig. 5b). In addition , lipids in the category of “ positive intrinsic curvature” were enriched , consistent with LPC promoting positive curvature 52. Positive intrinsic curvature is a key feature of lipids that can promote membrane budding and vesicle release. In summary, lipidomics identified for the first time a lipid signature of long chain polyunsaturated lysophospholipids (LPLs) specifically released during apoptotic pore formation. The specificity of the lipids being released argues that bulk membrane changes such as lateral lipid redistribution into microdomains (e.g. enriched for lysophosholipids) is occurring during pore formation, and possibly driving one or more steps leading to pore formation. As lysophospholipids are known to position at the edge of lipidic pores, our data support previous evidence that apoptotic pores are lipidic in nature. As BAK is not co-released with the lysophospholipids, BAK does not directly bind to them and may not be present in the microdomain from which these lipids are released. Triacylglycerides are retained in mitochondria following apoptotic pore formation While lysophospholipids were relatively over-represented in the Pellet100 fraction following pore formation , other lipids were relatively under-represented (Fig. 4a, Table 2, Table S6), likely due to enhanced retention in the MOM upon pore formation. The ten lipid species most under-represented in Pellet100 in the cBID/BAK group included tri acylglycerides and ceramides (Table 2). The most under-represented was the triacylglyceride TG 18:0_18:0_18:0 which reached a log2fold-change of -0.59, (FDR = 0.11, ns) . These lipids may be retained in the MOM by being attracted to other domains. Indeed, triacylglycerides were seen to focus beneath BAK a2-a5 core dimers in molecular dynamics simulations 18 (see Discussion). Inhibition of apoptotic pore formation by altering the lipid content of mitochondrial membranes The specific release of lyso phospholipids during pore formation raises the possibility that microdomains enriched for lyso phospholipids contribute to pore formation. For example, lysophospholipids induce positive membrane curvature 52 and can limit the size and distribution of ordered lipid domains 53. These effects are even more pronounced if fatty acid chains are long (>20 residues) and polyunsaturated 52, 53 . Cholesterol is an important regulator of membrane domains 54 and can counteract the positive curvature of lysophospholipids 53, so we next asked if addition of cholesterol could block pore formation. We compared cholesterol with oxysterol 7-ketocholesterol (7KC) (Fig. S3c) 55, 56 that contains an additional ketone group which can lead to preferential expansion of ordered (cholesterol) verus disordered (7KC) domains 54. Both sterols can induce ordered (i.e. tightly packed) membrane domains when used at high concentrations 54. Cholesterol and 7KC were incubated with mitochondria as water -soluble complexes with methyl-b-cyclodextrin (mbCD) (see Methods)55. Pre-incubation of MLM with the vehicle only control (mbCD-EtOH) had no effect on cytochrome c release (Fig. 6a). In contrast, addition of mbCD-cholesterol or mbCD-7KC prevented cytochrome c release, with cholesterol appearing as a slightly more potent inhibitor than 7KC (Figs. 6a, b). To interrogate if cholesterol might be acting directly on BAK en route to pore formation, the impact of cholesterol and 7KC on BAK activation and BAK clustering was assessed. BAK was still activated by cBID, as shown by increased susceptibility to proteolysis by proteinase K (Fig. 6c) and loss of the monomeric nonactivated BAK (1X band; Fig . 6d). Notably, high molecular weight BAK clusters were absent (Fig. 6d) despite evidence of BAK homodimers migrating at 146 kD (Fig. 6d). Thus, cholesterol and 7KC do not interfere with BAK activation preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.618570doi: bioRxiv preprint Uren et al | A lipid signature of the apoptotic pore 9 or dimer formation, but do stop homodimers aggregating into high molecular weight complexes. The ability of cholesterol to prevent the clustering of BAK homodimers (Fig. 6c) may explain its ability to prevent cytochrome c release (Fig s. 6a,b). Due to its wide-ranging effects on membranes, cholesterol may also act downstream e.g. by inhibiting lipid microdomains and bilayer deformation (see Discussion). Nevertheless, the strong inhibition by cholesterol argues the crucial role that interactions between proteins and lipids play in apoptotic pore formation in native mitochondria. Figure 6. Cholesterol inhibits apoptotic pore formation in mouse liver mitochondria. WT MLM were incubated with the lipid carrier m bCD loaded with EtOH, cholesterol or 7KC , prior to incubation with cBID (100 nM) at 37 oC for 30 min, and fractions analysed, as in Fig. 1. (a) Cytochrome c release was blocked by pre - incubation with cholesterol or 7KC. (b) Cholesterol and 7KC exhibit similar concentration- dependent inhibition of cytochrome c release. (c) BAK activation by cBID was not inhibited by cholesterol, as shown by susceptibility to proteinase K. Aliquots removed prior to fractionation were treated with proteinase K and immunoblotted with antibody to the BAK BH3 domain (clone 4B5). (d) BAK dimers failed to assemble into higher order BAK clusters in the presence of cholesterol or 7KC. The pelleted mitochondrial fractions from (a) were run on Blue Native PAGE and immunoblotted for BAK (aa23 - 38). Data are representative of three (a, d) or two (b, c) independent experiments. preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.618570doi: bioRxiv preprint Uren et al | A lipid signature of the apoptotic pore 10

To examine both the protein and lipid changes that drive apoptotic pore formation we used freshly isolated mitochondria as they contain endogenous levels of BAK residing in a native membrane environment. Upon apoptotic pore formation BAK remained in the mitochondrial pellet, while a specific group of lipids, namely long chain polyunsaturated lysophospholipids (LPLs), were released. To our knowledge, t his is the first report of membrane phospholipids being released during apoptotic pore formation. The specificity of the lipid signature suggests sorting of these lipids into microdomains which drive a localised positive membrane curvature and subsequent lipid shedding (possibly as micelles ) in concert with membrane rupture. Furthermore, a causal role for lipid microdomains in pore formation was suggested by the ability of cholesterol to block cytochrome c release and BAK clustering. The lipid particles characterised in this study appear distinct from other recorded types of mitochondrial lipid particles such as structures positive for outer membrane (SPOTs) 57, mitochondrial-derived vesicles (MDVs) 58, 59 and yeast mitochondrial -derived compartments (MDCs) 60-63 that are involved in quality control in response to various stresses. Those particles contain a range of proteins, and MDVs contained a lipid profile similar to that of the outer membrane 58. Note that the isolated mouse liver mitochondria prepared herein have undergone fission, which could preclude the release of large lipid particles such as those that may be observed in cells. The methodology adopted for lipidomics was designed to detect low levels of lipids released from mitochondria. The method involved initial sedimentation of the bulk of the mitochondria and debris by gentle centrifugation (10,000 g), and recovery of released material the supernatant (SN 10) by a second ultracentrifugation step (100,000 g). Lipidomics of this secondary pellet fraction (Pellet100) allowed removal of buffer constituents (e.g. HEPES) that interfere with ion detection during mass spectrometry, and concentrat ed the SN10 samples. While sedimentation at 100,000 g may not capture the smallest and least dense lipid particles, good coverage of the lipids in our targeted panel was achieved as a large proportion of lipids in the targeted lipidomics panel (348/495 lipids) were detected in Pellet100. Importantly, comparison of apoptotic and non -apoptotic samples identified very specific lipid shedding events in response to apoptosis. Most striking was the over -representation of long chain polyunsaturated lysophospholipids (LPLs) upon BAK-driven pore formation. This lipid signature is consistent with BAK activation causing formation of distinctive lipid microdomains, which are unstable and vulnerable to fissure. As a counterpoint to the LPL release driven by BAK activation, PLA2 treatment released lysophospholipids of all chain lengths and all levels of saturation. The PLA2 family of enzymes act at the face of a membrane to cleave di-acyl phospholipids at the sn-2 acyl linkage generating free fatty acids and lysophospholipids ( Fig. S3a and reviewed 46-48). Interestingly, PLA2 promoted microdomain formation and vesiculation in GUVs 64 and degradation of supported lipid bilayers 65. Given the increased representation of lysophospholipids in Pellet100 following PLA2 treatment and BAK-mediated pore formation, there may be mechanistic overlap in lipid release by BAK and PLA2 . However, a key distinction is that PLA2 enzymatic activity increases lysophospholipids, whereas BAK activation appears to drive the redistribution of existing lysophospholipids into microdomains. Lysophospholipids are non-lamellar lipids due to their a cone -shaped architecture (Fig. S3a). This shape confers a preference for adopting positive membrane curvature 52, for their participation at the edge of toroidal pores 66 and for forming micelles in solution 67. In coarse grain simulations asymmetric incorporation of lysophosphatidylcholine induced significant membrane curvature 68. Moreover, simulations show cone-shaped lipids coalesce into domains preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.618570doi: bioRxiv preprint Uren et al | A lipid signature of the apoptotic pore 11 that burst the membrane leaving behind a single toroidal pore lined by cone-shaped lipids 69. Thus, release of lysophospholipids that are long chain and polyunsaturated during apoptosis suggests these lipids become enriched within dedicated microdomains where positive curvature can cause lipid shedding and pore formation (Fig. 7, Movie S1). Lipidomics also revealed a subtle depletion of certain neutral lipids (TAGs and ceramides) in the Pellet100 samples following apoptotic signalling . Neutral lipids like TAGs and ceramides lack polar headgroups (Fig. S3b), and thus reside more centrally in the hydrophobic interleaflet space of a membrane bilayer. For instance, TAGs readily form non-lamellar “lens structures” within membrane bilayers 70. The depletion of these neutral lipids in our Pellet 100 samples suggests their retention in mitochondria during pore formation. Retention of TAGs is consistent with recent molecular dynamics simulations of TAGs aggregating beneath the BAK a2-a5 core dimer in response to membrane deformation by the dimer 18. Although we have no direct measure of TAG association with BAK clusters in our system, recruitment of TAGs under BAK dimers provides a plausible mechanism for BAK dimer -driven lipid redistribution and aligns with retention of BAK in the mitochondrial fraction (Fig. 2). Interestingly, another feature of neutral lipids such as ceramides is their capacity to co - segregate with lysophospholipids by packing closely with the lysophospholipid acyl chains underneath the lysophospholipid headgroup 71. Presuming that neutral lipids ( e.g. ceramides and TAGs) tend to co-segregate with lysophospholipids when distributed in healthy mitochondria, any concentration of neutral lipids under BAK dimers may leave the lysophospholipids “orphaned”, prompting formation of lysophospholipid microdomains which exhibit positive curvature. Together, the effects of neutral lipids segregating with BAK dimers and of lysophospholipids forming new positively curved microdomains could have a dramatic impact on the integrity of the membrane bilayer. To test the impact of altered lipid composition on pore formation mitochondria were pre - incubated with exogenous cholesterol. Cholesterol is also a neutral lipid and has many effects on model membranes, including the ordering of phospholipid acyl chains (formation of “lipid rafts”), lipid phase separation and increased bilayer thickness (reviewed in 72, 73 ). In MLM, sterols blocked BAK dimer clustering and MOMP (Fig. 6) . Given the plei otropic roles of cholesterol in membrane behaviour, we propose that cholesterol treatment of MLM could inhibit various stages of BAK-lipid mediated pore formation. We first considered if BAK activation and dimerization could be inhibited by cholesterol . Previous studies with BAX in liposomes and isolated mitochondria showed that cholesterol inhibited earl y steps in BAX function 34. This early effect on BAX was likely due to the requirement for BAX translocation and insertion into the membrane, which would be impeded by increased membrane order due to elevated cholesterol. We found no such inhibition of BAK activation or dimer formation by cholesterol, and this aligns with our understanding that BAK is constitutively embedded in the membrane, in contrast to BAX. Next we considered how cholesterol might block BAK dimer clustering. Formation of oligomers (i.e. clusters) of BAK (or BAX) dimers is an important pre -requisite for pore formation, yet there is no defined protein:protein interface between dimer subunits. Several lines of evidence suggest that the membrane bilayer itself promotes clustering of BAK dimers 15, 19, 26, and are supported here by the ability of cholesterol to inhibit clustering. According to modelling studies, aggregation of transmembrane proteins may result when the membrane bilayer attempt s to minimise hydrophobic mismatch between transmembrane proteins and surrounding lipids 74. Thus, cholesterol may minimise mismatch by being recruited to the BAK dimers (as per TAGs) or by increasing the membrane thickness surrounding each dimer. Cholesterol’s ability to create ordered domains with reduced lateral mobility in the bilayer may preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.618570doi: bioRxiv preprint Uren et al | A lipid signature of the apoptotic pore 12 also inhibit protein clustering. A complementary phenomenon is that increasing phospholipid unsaturation increases membrane fluidity 75 and promotes MOMP 36. Downstream of BAK clustering, cholesterol may directly inhibit lysophospholipid microdomains and/or their collapse into lipidic toroidal pores . For example, increased cholesterol prevented curvature stress induced by lysophospholipids in simulations and AFM studies 53. As has been shown for ceramides and TAGs, cholesterol and other sterols are neutral lipids and can also co-segregate with lysophospholipids via hydrophobic interactions 76, 77 . Thus, especially considering that cholesterol levels in mitochondria are relatively low 45, a small increase may counter effects of other neutral lipids (e.g. TAGs, ceramides). While there are several possible modes by which cholesterol treatment could disrupt pore formation, our data highlight that cholesterol is a potent inhibitor of this process. Future simulations of membranes containing a wider repertoire of lipids (e.g. lysophospholipids that are saturated or unsaturated, TAGs, ceramides and cholesterol ) may help delineate which mode(s) of action accounts for the robust inhibition of pore formation by cholesterol. Identifying the topology of BAK dimers in relation to the membrane bilayer, to other dimers, and to a toroidal pore, has been of significant interest in terms of understanding the mechanism of pore formation and how to regulate apoptotic cell death. A range of studies indicate that a large part of each BAK dimer interacts with the membrane, including transmembrane insertion of the C -terminal a9-helices at either end of the dimer , and outer leaflet insertion by the amphipathic a2-a5 core and a6-helices (reviewed in 25, Fig. 7, Movie S1). This dimer topology is consistent with retention of BAK in the mitochondrial fraction rather than release with lysophospholipids. Thus, BAK retention in the bulk pellet indicates that apoptotic pore formation is distinct from the “carpet model” of membrane disruption reported for certain types of pore-forming peptides 78, and also excludes a model like the “cookie cutter” pores formed by the lytic cell death regulator NINJ1 in which the protein subunits and the membrane disc enclosed within that ring are both liberated from the membrane 79. The topology of each BAK dimer in relation to other dimers, and to the pore itself, is less clear. Microscopy suggests that within each apoptotic mitochondrion most of BAK forms oligomers (i.e. clusters), but only a portion positions at the pore edge 26, 80, 81. This implies that clustering precedes pore formation, but also implies that biochemical assays will be monitoring a mixture of two (or more) populations of BAK dimers. Distinct topologies may exist for BAK dimers in clusters and at the pore edge based on flexibility in the latch regions (e.g. a5-a6 and a8-a9 linker regions). Dimers clustered on the membrane surface may adopt the in-plane orientation 21, followed by repositioning at the rim of the pore . Further repositioning of the a2-a5 core across pore edge, as in the clamp model, would force one a9 anchor to become inverted 82. Our findings lead us to a BAK-lipid microdomain model of apoptotic pore formation (Fig. 7, Movie S1). As BAK homodimers embed in the membrane surface, they drive membrane thinning and lipid redistribution, causing lipids without polar headgroups (neutral lipids such as TAGs and perhaps ceramides) to focus beneath the dimer . The “orphaned” LPLs then coalesce into a microdomain so as to minimise hydrophobic mismatch at the interface of individual LPLs and the surrounding bilayer . In parallel, BAK dimers coalesce into microdomains (clusters) to minimise hydrophobic mismatch at the interface of individual BAK dimers and the surrounding bilayer. When a critical density of LPLs is attained, the localised positive curvature results in membrane rupture and release of LPLs (possibly as a micelle), leaving behind a lipid-lined toroidal pore. The pore may be stabilised and enlarged or distorted to create a tear in the membrane by two events: flow of lipids from the outer to the inner leaflet around the pore edge to neutralise membrane tension; and BAK homodimers positioning at the pore rim. preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.618570doi: bioRxiv preprint Uren et al | A lipid signature of the apoptotic pore 13 This BAK-lipid microdomain model is consistent with our understanding of BAK mulitimers adopting a variety of architectures (i.e. arcs, lines, rings, clusters) in different contexts and the apparent lack of an essential protein:protein contact between dimers , but highlights the requirement for lipid re organisation to effect pore formation. Future testing for lysophospholipid release from toroidal pores formed by other proteins, and from other native and model membranes, may reveal if the lipid signature we have observed for apoptotic pores in mouse liver mitochondria is a common feature of toroidal pore formation. In conclusion, examining endogenous BAK protein in its native mitochondrial environment has enabled a broad survey of lipids liberated from the membrane upon formation of the apoptotic pore. We have identified for the first time a lipid signature of long chain polyunsaturated lysophospholipids released during apoptotic pore formation. Data showing that exogenous lipids can prevent BAK dimer clustering highlights the important interplay of BAK and lipids en route to pore formation. Collectively, our data support the lipidic nature of BAK-dependent pores, with the BAK-lipid microdomain model providing a possible mechanism for membrane forces driving both BAK clustering and pore formation. preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.618570doi: bioRxiv preprint Uren et al | A lipid signature of the apoptotic pore 14 Figure 7. BAK-lipid microdomain model of apoptotic pore formation a) In healthy mitochondria, BAK monomers are distributed on the MOM surface surrounded by a mixture of lipids (phosphoplipids, grey). Long chain polyunsaturated lysophospholipids (LPLs , red, <1% total ) are distributed in the membrane and associated with neutral lipids (e.g. TAGs and ceramides, yellow). b) After apoptotic signalling , BAK dimers form and sink into the outer leaflet, displacing lipids with large polar headgroups (e.g. phospholipids ) and recruiting neutral lipids under each BAK dimer . Any associated lysophospholipids also become more mobile. c) Separate microdomains of LPLs and BAK dimers form. LPL microdomains drive positive curvature and initial delamination of the two leaflets. d) Growth of microdomains driven by membrane stress. C ontinued LPL aggregation drives severe positive membrane curvature, with non-bilayer LPLs liberated from the membrane (possibly as a micelle), and leaving behind a lipidic pore partially lined with LPLs. e) Pore stabilization and enlargement. Free movement of lipids between leaflet s can relieve membrane tension. In addition, BAK dimers remain anchored in the membrane via their transmembrane domains, with the less restrained core region dragged adjacent to the pore edge . Created with BioRender.com preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.618570doi: bioRxiv preprint Uren et al | A lipid signature of the apoptotic pore 15

Isolation of mouse liver mitochondria Mouse liver mitochondria (MLM) were isolated from male wildtype C57BL/6 or Bak– /– C57BL/6 mice 2 (ranging 74 to 155 days old, wildtype n=6, Bak-/- n=6) that were bred and maintained at the Walter and Eliza Hall Institute of Medical Research Animal Facility. All animal experiments were approved by the WEHI Animal Ethics Committee (WEHI internal ethics number 2022.014) and were conducted in accordance with the Prevention of Cruelty to Animals Act (1986) and the Australian National Health and Medical Research Council Code of Practice for the Care and the Use of Animals for Scientific Purposes (1997). Trained Bioservices staff performed euthanasia of adult mice by cervical dislocation. Mouse livers were collected in chilled sucrose solution (300 mM sucrose, 10 mM Tris HCl, pH7.4, 1 mM EDTA). Isolated mitochondria were prepared from livers as described previously 43. The concentration of MLM was estimated by absorbance at 280 nm. Reagents Recombinant caspase-8 cleaved human BID (cBID) was prepared as per 43 and aliquots stored at -80oC before thawing on ice. Phospholipase A 2 (PLA2) from bee venom (Sigma Aldrich) was prepared as a stock of 1mg/mL and stored at -20 oC. Methyl b cyclodextrin (mbCD) (Sigma Aldrich) is a lipid carrier commonly used at high concentrations to extract cholesterol, or at lower concentrations (as per this study) to solubilise sterols for delivery to membranes 55. Methyl b cyclodextrin (m bCD) was loaded with cholesterol (Sigma Aldrich) or 7 - ketocholesterol (7KC) (Sigma Aldrich) as described 55. Briefly, 5% w/v m bCD (50 mg/mL, 400 µL aliquots) was pre-warmed to 80oC on a heatblock for 5 min. Sterols dissolved in ethanol at 15 mg/mL were added to the heated m bCD, 10 µL added at 5 min intervals mixing with vortex with each addition and returned to heat , and repeated four times for a total of 40 µL sterol added per 400 uL preparation to yield 1.5 mg/mL of each sterol in 50 mg/mL mbCD. To generate a “vehicle-only” control, an ethanol only loading of mbCD was also prepared (mbCD- EtOH) in parallel. Sterol reagents were stored at 4oC protected from light. MLM incubations Freshly prepared MLM (used within 3 hours of isolation) were kept on ice before dilution in mitochondrial assay buffer containing 100 mM KCl, 2.5 mM MgCl2, 100 mM sucrose, 20 mM HEPES/KOH at pH 7.5, protease inhibitor cocktail (Roche) and 8 µg/mL pepstatin A (Sigma Aldrich). For lipidomic studies, samples were incubated for 45 min at 37oC with either 100 nM cBID or 5 µg/mL or 50 µg/mL PLA2. For sterol treatments, 200 µL aliquots were pre-incubated with 1-10 µL 1.5 mg/mL mbCD-cholesterol, mbCD-7KC or mbCD-EtOH at 37oC for 30 min, then treated with 100 nM cBID for 30 min. For EDTA inhibition assays, samples were pre - incubated with 20 mM EDTA at 37oC for 30 min, prior to addition of cBID or PLA2. MLM cytochrome c release assay To monitor mitochondrial permeabilisation, pellet and supernatant fractions from a 10,000 g 5 min fractionation were analysed by SDS PAGE. Fractions were resuspended reducing 2x SDS sample loading buffer (150 mM Tris pH 6.8, 1.2% (w/v) SDS, 30% (v/v) glycerol, 0.018 mg/mL bromophenol blue, 5% v/v 2-mercaptoethanol). Samples were boiled for 3 min, then resolved by SDS PAGE (12% TGX gels, BioRad) and transfered to 0.2 µm nitrocellulose membrane (Bio-RAD) via wet transfer (25 mM Tris, 192 mM Glycine, 20% v/v Methanol) . Membranes were blocked in 5% w/v skim milk in TBS with 0.1% v/v Tween and immunoblotted for cytochrome c (mouse monoclonal antibody, clone 7H8.2C12, #556433, BD Biosciences) or HSP60 (rabbit polyclonal antibody, #A302845A, Thermo Fisher Scientific). preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.618570doi: bioRxiv preprint Uren et al | A lipid signature of the apoptotic pore 16 MLM BAK activation assay with proteinase K proteolysis BAK activation status was monitored via susceptibility of the BAK protein to digestion with proteinase K (PK) as per 83. Whole mitochondrial fractions were pre-chilled on ice for 15 min, prior to addition of proteinase K ( 30 µg/mL final) and incubation for 20 min on ice. The reaction was stopped with the addition of PMSF (2 mM final) , then an equal volume of reducing 2x SDS sample loading buffer added to each sample. Samples were resolved by SDS PAGE (12% TGX gels, BioRad) and transfered to 0.2 micron nitrocellulose via wet transfer (25 mM Tris, 192 mM Glycine, 20% v/v Methanol). Membranes were blocked in 5% w/v skim milk in TBS with 0.1% v/v Tween and immunoblotted for BAK 4B5 (rat monoclonal antibody, clone 4B5, WEHI mAb facility). Blue Native PAGE Higher order complexes of BAK protein were measured using Blue Native PAGE. Samples were analysed by Blue Native PAGE (BNP) to preserve the native interface within BAK dimers and allow assessment of clustering of multiple BAK dimers. Membrane fractions were isolated by centrifugation at 1 0,000 g for 5 min. Supernatants were discarded and membrane pellets were solubilised in 20 mM Bis-Tris pH 7.4, 50 mM NaCl, 10% v/v glycerol, 1% w/v digitonin, 10 mM DTT and incubated on ice for 1 hr. A second centrifugation of 16,000 g for 5 min was used to sediment any i nsoluble material . Native Sample buffer (Life Technologies) and Coomassie Additive (Life Technologies) were added to t he soluble s upernatants. In some instances, detection of BAK dimers was enhanced by supplementation with 5 mM EDTA. Samples were resolved with Novex 4 –16% Native PAGE 1.0 mm 10 well gels (Life Technologies). Western blot transfer to 0.2 µm PVDF membrane (Thermo Fisher Scientific) was performed via wet transfer (25 mM Tris, 192 mM Glycine, 20% v/v Methanol , supplemented with 0.037% w/v SDS). PVDF membranes were de-stained with 10% v/v acetic acid 30% v/v ethanol, then further de -stained in methanol and rinsed thoroughly with dH2O before immunoblotting. Membranes were blocked in 5% w/v skim milk in TBS with 0.1% v/v Tween, and immunoblotted for BAK aa23-38 (clone DF9) (rabbit polyclonal antibody, B5897, Sigma-Aldrich, Castle Hill, NSW, Australia). Western blot secondary antibody detection and image capture For secondary detection, membranes were incubated with Horseradish peroxidase conjugated IgG secondary antibodies; anti -rabbit (4010 –05, Southern Biotech ), anti -rat (3010 –05, Southern Biotech) and anti -mouse (1010 –05, Southern Biotech ). Immobilised horseradish peroxidase was detected with Immobilon Forte Western H RP substrate (WBLUF0500, Millipore, Billerica, MA, USA), images captured with the ChemiDoc MP System (Bio-RAD, Hercules, CA, USA) and signal intensity measured with Image Lab 6.1 software (Bio-RAD). MLM fractionation for lipidomic analysis A small volume of each sample for lipidomic analysis (50 µL from 1.5 mL total) was prepared by centrifugation at 10,000g for 5 min, and pellet and SN fractions collected. Fractions were resolved by SDS -PAGE and cytochrome c release measured (as described above) . The remainder each treatment sample (~1.45 mL) was subjected to a 2-step fractionation procedure. Samples were first separated by centrifugation at 10,000 g (10,000 rpm) 5 min at 4oC, and the supernatant (SN10) retrieved. The SN10 fraction was then applied to pre -weighed tubes (Beckman Coulter, maximum volume 1.5 mL #357448 ). Samples were then separated by a centrifugation at 40,000 rpm (TLA-55 angle rotor S/N 19U1568 in Beckman Coulter Optima Max XP Ultra Centrifuge , rmax 98,600g) for 20 min at 4 oC. Supernatants were removed and pellets weighed and placed on wet ice. The mass of samples ranged from 18.5 mg to 33.9 mg from pooled duplicate pellets (Supp. Table 1). Pellets were stored at -80oC prior to being pooled preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.618570doi: bioRxiv preprint Uren et al | A lipid signature of the apoptotic pore 17 during extraction and analysis by Metabolomics Australia (Bio21 Institute, Melbourne, VIC, Australia). Lipidomic profiling Targeted lipidomic profiling was conducted for samples fractionated from isolated liver mitochondria from age-matched adult male mice (wildtype C57BL/6 n = 6, Bak -/- n = 6, male, ages ranged from 74 to 156 days old, 3 treatment groups per mouse with a total of 36 samples processed for lipidomics (Supp. Table 1). Internal standards comprising a mixture of PC(19:0/19:0), PE-D31, PG (17:0/17:0), TG-D5 (Avanti Polar Lipids, Alabama, USA) were added (10 mg/L) to the stock extraction solution . The extraction solution containing internal standards were used for all sample extractions to assess sample processing and instrument performance. Ice cold 1:9 chloroform:methanol (600 µL) and internal standards were added to sample pellets. Suspensions were transferred to 2 mL microcentrifuge tubes, then 1 mL 100% (v/v) chloroform added to obtain a final ratio of 2:1 chloroform:methanol. Samples were vortexed for 30 sec, then mixed at 950 rpm for 5 min at 10 oC in an Eppendorf Thermomixer (Eppendorf South Pacific Pty Ltd., Macquarie Park, Australia) . Samples were centrifuged at 15,000 rpm for 5 min at 10 oC and supernatants collected. Supernatants were completely dried in a vacuum concentrator to complete dryness (100 µL for 15 cycles , Christ® RVC 2–33, Martin Christ Gefriertrocknungsanlagen GmbH, Osterode am Harz, Germany). Dried samples were reconstituted in methanol:water-saturated butanol (100 uL, 1:9, v/v). To assess instrument performance, pooled biological quality control samples (PBQCs) were prepared by combining 10 µL aliquots of each sample extract. Samples were injected (2 µL volume) in a random order for analysis by LC-QQQ-MS. PBQCs were injected at intervals, after every 4 samples. Extracted lipids were processed and detected by Metabolomics Australia (Bio21 Institute, Melbourne, VIC, Australia) as previously described 84, 85 with an Agilent 1290 liquid chromatography (LC) system and Triple Quadrapople 6490 mass spectrometer (MS, Agilent Technologies Australia, Mulgrave, Australia). A curated list of 348 metabolites (from the targeted 495 metabolites in the Mass Hunter Database/Mass Hunter Quantitative Analysis) was obtained and data imported into MetaboAnalyst 5.0 for initial quality control checks and analysis by Metabolomics Australia (Bio21 Institute, Melbourne, VIC, Australia) . This curated dataset was subject to further downstream statistical analysis as follows. Statistical analysis of lipidomic data Relative abundance of lipids were analysed in R 86 using limma 87,Glimma 88 and edgeR 89. After removing pooled biological quality control (PBQC) samples, quantile normalization 90 of log2-counts was performed. Linear models 91 with sample-specific weights 92 were fitted to summarise the data from each experimental group (WT Untreated, WT cBID, WT PLA2, KO Untreated, KO cBID and KO PLA2) while accounting for variation between sample collection dates. Pairwise contrasts between the experimental groups were estimated, and differential lipid abundance was assessed using moderated t-statistics 91. Lipids were ranked according to their false discovery rate (FDR) 93, and those with a FDR < 0.05 were considered differentially abundant. Figures were plotted with functions in the limma, ggplot2 94 and ggrepel 95 packages. Lipid set enrichment analysis Lipids were classified according to headgroup and enrichment of classes with at least 3 lipids was tested using ROAST 49. LION/web analysis was performed in “target-list mode”. The top 26 lipids (raw p-value < 0.2) from the WT cBID vs WT Untreated contrast with positive preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.618570doi: bioRxiv preprint Uren et al | A lipid signature of the apoptotic pore 18 log2fold-change were input as the “lipid target list” and the full list of lipids detected were input as the “lipid background list”. DATA A V AILABILITY STATEMENT Lipidomic data and analysis code is available via GitHub at https://github.com/mritchie/BAKLipidomicAnalysis/ ANIMAL ETHICS APPROV AL All animal experiments were approved by the WEHI Animal Ethics Committee (WEHI internal ethics number 2022.014) and were conducted in accordance with the Prevention of Cruelty to Animals Act (1986) and the Australian National Health and Medical Research Council Code of Practice for the Care and the Use of Animals for Scientific Purposes (1997). FUNDING RMK was supported by Australian National Health and Medical Research Council (NHMRC) Program Grant GNT1113133 and the Leukemia & Lymphoma Society of America Specialized Centre of Research [SCOR] grant 7015-18. MER is supported by NHMRC Investigator Grant GNT2017257. DS was supported by the NIH Grants HL158305 and NS124477 . Work in the laboratories of the authors was made possible through Victorian State Government Operational Infrastructure Support (OIS) and Australian Government NHMRC Independent Research Institute Infrastructure Support (IRIIS) Scheme. This study used NCRIS -enabled Metabolomics Australia infrastructure at the University of Melbourne and funded through BioPlatforms Australia. CONFLICT OF INTEREST STATEMENT The authors declare no conflict of interest. AUTHOR CONTRIBUTIONS RTU: Conceptualization, Formal analysis, Supervision, Investigation, Visualization, Methodology, Writing—original draft, Writing—review and editing. MER: (Lipidomics) Data curation, Software, Formal analysis, Visualization, Methodology, Writing—review and editing. AWW, JPL: Investigation, Writing—review and editing. EU: Visualization, Writing—review and editing. VKN, DPDS: (Lipidomics) Investigation, Methodology, Writing—review and editing. DS: Writing—review and editing. RMK: Conceptualization, Supervision, Visualization, Methodology, Writing —original draft, Writing—review and editing. SUPPLEMENTARY MATERIAL Supplementary materials are available online. Supplementary Tables S1-S8 Supplementary Movie S1 Supplementary Figures S1-S4 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.618570doi: bioRxiv preprint Uren et al | A lipid signature of the apoptotic pore 19

1. Wei, M.C. et al. Proapoptotic BAX and BAK: a requisite gateway to mitochondrial dysfunction and death. Science 292, 727-730 (2001). 2. Lindsten, T. et al. The combined functions of proapoptotic Bcl-2 family members Bak and Bax are essential for normal development of multiple tissues. Mol. Cell 6, 1389-1399. (2000). 3. Shalaby, R., Diwan, A., Flores-Romero, H., Hertlein, V . & Garcia-Saez, A.J. Visualization of BOK pores independent of BAX and BAK reveals a similar mechanism with differing regulation. Cell Death Differ. 30, 731-741 (2023). 4. King, L.E., Hohorst, L. & Garcia -Saez, A.J. Expanding roles of BCL -2 proteins in apoptosis execution and beyond. J. Cell Sci. 136 (2023). 5. Suzuki, M., Youle, R.J. & Tjandra, N. Structure of Bax: coregulation of dimer formation and intracellular localization. Cell 103, 645-654 (2000). 6. Moldoveanu, T. et al. The x-ray structure of a BAK homodimer reveals an inhibitory zinc binding site. Mol. Cell 24, 677-688 (2006). 7. Iyer, S. et al. Bak apoptotic pores involve a flexible C -terminal region and juxtaposition of the C - terminal transmembrane domains. Cell Death Differ. (2015). 8. Wolter, K.G. et al. Movement of Bax from the cytosol to mitochondria during apoptosis. J. Cell Biol. 139, 1281-1292 (1997). 9. Gahl, R.F., He, Y ., Yu, S. & Tjandra, N. Conformational rearrangements in the pro-apoptotic protein, Bax, as it inserts into mitochondria: a cellular death switch. J. Biol. Chem. 289, 32871-32882 (2014). 10. Robin, A.Y . et al. Ensemble Properties of Bax Determine Its Function. Structure 26, 1346-1359 e1345 (2018). 11. Dewson, G. et al. To trigger apoptosis Bak exposes its BH3 domain and homo -dimerizes via BH3:grooove interactions. Mol. Cell 30, 369-380 (2008). 12. Dewson, G. et al. Bax dimerizes via a symmetric BH3:groove interface during apoptosis. Cell Death Differ. 19, 661-670 (2012). 13. Czabotar, P.E. et al. Bax Crystal Structures Reveal How BH3 Domains Activate Bax and Nucleate Its Oligomerization to Induce Apoptosis. Cell 152, 519-531 (2013). 14. Brouwer, J.M. et al. Bak Core and Latch Domains Separate during Activation, and Freed Core Domains Form Symmetric Homodimers. Mol. Cell 55, 938-946 (2014). 15. Cowan, A.D. et al. BAK core dimers bind lipids and can be bridged by them. Nat. Struct. Mol. Biol. 27, 1024-1031 (2020). 16. Birkinshaw, R.W. et al. Structure of detergent -activated BAK dimers derived from the inert monomer. Mol. Cell (2021). 17. Miller, M.S. et al. Sequence differences between BAX and BAK core domains manifest as differences in their interactions with lipids. FEBS J. (2023). 18. Smith, N.A. et al. The Bak core dimer focuses triacylglycerides in the membrane. Biophys. J. 121, 347-360 (2022). 19. Uren, R.T. et al. Disordered clusters of Bak dimers rupture mitochondria during apoptosis. Elife 6 (2017). 20. Dewson, G. et al. Bak activation for apoptosis involves oligomerization of dimers via their alpha6 helices. Mol. Cell 36, 696-703 (2009). 21. Westphal, D. et al. Apoptotic pore formation is associated with in -plane insertion of Bak or Bax central helices into the mitochondrial outer membrane. Proc Natl Acad Sci U S A 111, E4076-4085 (2014). 22. Alsop, A.E. et al. Dissociation of Bak alpha1 helix from the core and latch domains is required for apoptosis. Nat Commun 6, 6841 (2015). 23. Aluvila, S. et al. Organization of the mitochondrial apoptotic BAK pore: oligomerization of the BAK homodimers. J. Biol. Chem. 289, 2537-2551 (2014). 24. Oh, K.J. et al. Conformational changes in BAK, a pore-forming proapoptotic Bcl-2 family member, upon membrane insertion and direct evidence for the existence of BH3 -BH3 contact interface in BAK homo-oligomers. J. Biol. Chem. 285, 28924-28937 (2010). 25. Uren, R.T., Iyer, S. & Kluck, R.M. Pore formation by dimeric Bak and Bax: an unusual pore? Philos Trans R Soc Lond B Biol Sci 372 (2017). preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.618570doi: bioRxiv preprint Uren et al | A lipid signature of the apoptotic pore 20 26. Schweighofer, S.V . et al. Endogenous BAX and BAK form mosaic rings of variable size and composition on apoptotic mitochondria. Cell Death Differ. 31, 469-478 (2024). 27. Cosentino, K. & Garcia-Saez, A.J. Bax and Bak Pores: Are We Closing the Circle? Trends Cell Biol. 27, 266-275 (2017). 28. Schafer, B. et al. Mitochondrial outer membrane proteins assist Bid in Bax -mediated lipidic pore formation. Mol Biol Cell 20, 2276-2285 (2009). 29. Renault, T.T. & Chipuk, J.E. Death upon a kiss: mitochondrial outer membrane composition and organelle communication govern sensitivity to BAK/BAX -dependent apoptosis. Chem. Biol. 21, 114-123 (2014). 30. Kuwana, T. et al. Bid, Bax, and lipids cooperate to form supramolecular openings in the outer mitochondrial membrane. Cell 111, 331-342 (2002). 31. Kushnareva, Y ., Andreyev, A.Y ., Kuwana, T. & Newmeyer, D.D. Bax activation initiates the assembly of a multimeric catalyst that facilitates Bax pore formation in mitochondrial outer membranes. PLoS Biol. 10, e1001394 (2012). 32. Bozelli, J.C., Jr., Hou, Y .H., Schreier, S. & Epand, R.M. Lipid asymmetry of a model mitochondrial outer membrane affects Bax -dependent permeabilization. Biochim Biophys Acta Biomembr 1862, 183241 (2020). 33. Lucken-Ardjomande, S., Montessuit, S. & Martinou, J.C. Contributions to Bax insertion and oligomerization of lipids of the mitochondrial outer membrane. Cell Death Differ. 15, 929 -937 (2008). 34. Lucken-Ardjomande, S., Montessuit, S. & Martinou, J.C. Bax activation and stress -induced apoptosis delayed by the accumulation of cholesterol in mitochondrial membranes. Cell Death Differ. 15, 484-493 (2008). 35. Basanez, G. et al. Bax-type apoptotic proteins porate pure lipid bilayers through a mechanism sensitive to intrinsic monolayer curvature. J. Biol. Chem. 277, 49360-49365 (2002). 36. Dadsena, S. et al. Lipid unsaturation promotes BAX and BAK pore activity during apoptosis. Nat Commun 15, 4700 (2024). 37. Comte, J., Maisterrena, B. & Gautheron, D.C. Lipid composition and protein profiles of outer and inner membranes from pig heart mitochondria. Comparison with microsomes. Biochim Biophys Acta 419, 271-284 (1976). 38. Colbeau, A., Nachbaur, J. & Vignais, P.M. Enzymic characterization and lipid composition of rat liver subcellular membranes. Biochim Biophys Acta 249, 462-492 (1971). 39. Horvath, S.E. & Daum, G. Lipids of mitochondria. Prog Lipid Res 52, 590-614 (2013). 40. Daum, G. & Vance, J.E. Import of lipids into mitochondria. Prog Lipid Res 36, 103-130 (1997). 41. de Kroon, A.I., Dolis, D., Mayer, A., Lill, R. & de Kruijff, B. Phospholipid composition of highly purified mitochondrial outer membranes of rat liver and Neurospora crassa. Is cardiolipin present in the mitochondrial outer membrane? Biochim Biophys Acta 1325, 108-116 (1997). 42. Ardail, D. et al. Mitochondrial contact sites. Lipid composition and dynamics. J. Biol. Chem. 265, 18797-18802 (1990). 43. Uren, R.T. et al. Mitochondrial Release of Pro-apoptotic Proteins: Electrostatic Interactions can Hold Cytochrome c but not Smac/DIABLO to Mitochondrial Membranes. J. Biol. Chem. 280, 2266-2274 (2005). 44. Scott, D.L., Otwinowski, Z., Gelb, M.H. & Sigler, P.B. Crystal structure of bee -venom phospholipase A2 in a complex with a transition-state analogue. Science 250, 1563-1566 (1990). 45. Elustondo, P., Martin, L.A. & Karten, B. Mitochondrial cholesterol import. Biochim Biophys Acta Mol Cell Biol Lipids 1862, 90-101 (2017). 46. Dennis, E.A. Allosteric regulation by membranes and hydrophobic subsites in phospholipase A(2) enzymes determine their substrate specificity. J. Biol. Chem. 298, 101873 (2022). 47. Murakami, M. The phospholipase A(2) superfamily as a central hub of bioactive lipids and beyond. Pharmacol. Ther. 244, 108382 (2023). 48. Khan, S.A. & Ilies, M.A. The Phospholipase A2 Superfamily: Structure, Isozymes, Catalysis, Physiologic and Pathologic Roles. Int J Mol Sci 24 (2023). 49. Wu, D. et al. ROAST: rotation gene set tests for complex microarray experiments. Bioinformatics 26, 2176-2182 (2010). preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.618570doi: bioRxiv preprint Uren et al | A lipid signature of the apoptotic pore 21 50. Molenaar, M.R., Haaker, M.W., Vaandrager, A.B., Houweling, M. & Helms, J.B. Lipidomic profiling of rat hepatic stellate cells during activation reveals a two-stage process accompanied by increased levels of lysosomal lipids. J. Biol. Chem. 299, 103042 (2023). 51. Molenaar, M.R. et al. LION/web: a web-based ontology enrichment tool for lipidomic data analysis. Gigascience 8 (2019). 52. Fuller, N. & Rand, R.P. The influence of lysolipids on the spontaneous curvature and bending elasticity of phospholipid membranes. Biophys. J. 81, 243-254 (2001). 53. Krasnobaev, V .D., Galimzyanov, T.R., Akimov, S.A. & Batishchev, O.V . Lysolipids regulate raft size distribution. Front Mol Biosci 9, 1021321 (2022). 54. Massey, J.B. & Pownall, H.J. The polar nature of 7 -ketocholesterol determines its location within membrane domains and the kinetics of membrane microsolubilization by apolipoprotein A -I. Biochemistry 44, 10423-10433 (2005). 55. Rentero, C. et al. Functional implications of plasma membrane condensation for T cell activation. PLoS One 3, e2262 (2008). 56. Ayee, M.A.A. & Levitan, I. Lipoprotein -Induced Increases in Cholesterol and 7 -Ketocholesterol

in Opposite Molecular-Scale Biophysical Effects on Membrane Structure. Front Cardiovasc Med 8, 715932 (2021). 57. Li, X. et al. Mitochondria shed their outer membrane in response to infection-induced stress. Science 375, eabi4343 (2022). 58. Konig, T. et al. MIROs and DRP1 drive mitochondrial -derived vesicle biogenesis and promote quality control. Nat. Cell Biol. 23, 1271-1286 (2021). 59. Sugiura, A., McLelland, G.L., Fon, E.A. & McBride, H.M. A new pathway for mitochondrial quality control: mitochondrial-derived vesicles. EMBO J. 33, 2142-2156 (2014). 60. Raghuram, N. & Hughes, A.L. Amino acids trigger MDC-dependent mitochondrial remodeling by altering mitochondrial function. bioRxiv (2024). 61. Hughes, A.L., Hughes, C.E., Henderson, K.A., Yazvenko, N. & Gottschling, D.E. Selective sorting and destruction of mitochondrial membrane proteins in aged yeast. Elife 5 (2016). 62. Schuler, M.H. et al. Mitochondrial-derived compartments facilitate cellular adaptation to amino acid stress. Mol. Cell 81, 3786-3802 e3713 (2021). 63. Wilson, Z.N. et al. Mitochondrial-derived compartments remove surplus proteins from the outer mitochondrial membrane. J. Cell Biol. 223 (2024). 64. Staneva, G., Angelova, M.I. & Koumanov, K. Phospholipase A2 promotes raft budding and fission from giant liposomes. Chem Phys Lipids 129, 53-62 (2004). 65. Kuzmina, N., V olynsky, P., Boldyrev, I. & Alekseeva, A. sPLA2 Wobbles on the Lipid Bilayer between Three Positions, Each Involved in the Hydrolysis Process. Toxins (Basel) 14 (2022). 66. Woo, S.Y. & Lee, H. Effect of lipid shape on toroidal pore formation and peptide orientation in lipid bilayers. Phys Chem Chem Phys 19, 21340-21349 (2017). 67. Tasis, D. et al. Diameter-selective solubilization of carbon nanotubes by lipid micelles. J Nanosci Nanotechnol 8, 420-423 (2008). 68. Yoo, J. & Cui, Q. Curvature generation and pressure profile modulation in membrane by lysolipids: insights from coarse-grained simulations. Biophys. J. 97, 2267-2276 (2009). 69. Sakuma, Y ., Taniguchi, T. & Imai, M. Pore formation in a binary giant vesicle induced by cone - shaped lipids. Biophys. J. 99, 472-479 (2010). 70. Khandelia, H., Duelund, L., Pakkanen, K.I. & Ipsen, J.H. Triglyceride blisters in lipid bilayers: implications for lipid droplet biogenesis and the mobile lipid signal in cancer cell membranes. PLoS One 5, e12811 (2010). 71. Al Sazzad, M.A. et al. Natural Ceramides and Lysophospholipids Cosegregate in Fluid Phosphatidylcholine Bilayers. Biophys. J. 116, 1105-1114 (2019). 72. Marquardt, D., Kucerka, N., Wassall, S.R., Harroun, T.A. & Katsaras, J. Cholesterol's location in lipid bilayers. Chem Phys Lipids 199, 17-25 (2016). 73. Yang, S.T., Kreutzberger, A.J.B., Lee, J., Kiessling, V. & Tamm, L.K. The role of cholesterol in membrane fusion. Chem Phys Lipids 199, 136-143 (2016). 74. Katira, S., Mandadapu, K.K., Vaikuntanathan, S., Smit, B. & Chandler, D. Pre -transition effects mediate forces of assembly between transmembrane proteins. Elife 5, e13150 (2016). preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.618570doi: bioRxiv preprint Uren et al | A lipid signature of the apoptotic pore 22 75. Harayama, T. & Antonny, B. Beyond Fluidity: The Role of Lipid Unsaturation in Membrane Function. Cold Spring Harb Perspect Biol 15 (2023). 76. Ramsammy, L.S. & Brockerhoff, H. Lysophosphatidylcholine-cholesterol complex. J. Biol. Chem. 257, 3570-3574 (1982). 77. Demel, R.A., Bruckdorfer, K.R. & van Deenen, L.L.M. Structural requirements of sterols for the interaction with lecithin at the air-water interface. Biochim. Biophys. Acta 255, 311-320 (1972). 78. Oren, Z. & Shai, Y . Mode of action of linear amphipathic alpha -helical antimicrobial peptides. Biopolymers 47, 451-463 (1998). 79. David, L. et al. NINJ1 mediates plasma membrane rupture by cutting and releasing membrane disks. Cell 187, 2224-2235 e2216 (2024). 80. Garcia-Saez, A.J., Chiantia, S., Salgado, J. & Schwille, P. Pore formation by a Bax-derived peptide: effect on the line tension of the membrane probed by AFM. Biophys. J. 93, 103-112 (2007). 81. Cosentino, K. et al. The interplay between BAX and BAK tunes apoptotic pore growth to control mitochondrial-DNA-mediated inflammation. Mol. Cell 82, 933-949 e939 (2022). 82. Bleicken, S. et al. Structural model of active bax at the membrane. Mol. Cell 56, 496-505 (2014). 83. Iyer, S., Uren, R.T. & Kluck, R.M. Probing BAK and BAX Activation and Pore Assembly with Cytochrome c Release, Limited Proteolysis, and Oxidant-Induced Linkage. Methods Mol Biol 1877, 201-216 (2019). 84. Huynh, K. et al. Lipidomic Profiling of Murine Macrophages Treated with Fatty Acids of Varying Chain Length and Saturation Status. Metabolites 8 (2018). 85. Huynh, K. et al. High-Throughput Plasma Lipidomics: Detailed Mapping of the Associations with Cardiometabolic Risk Factors. Cell Chem Biol 26, 71-84 e74 (2019). 86. R-Core-Team-2023 R: A language and environment for statistical computing. . Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/ (2023). 87. Ritchie, M.E. et al. limma powers differential expression analyses for RNA -sequencing and microarray studies. Nucleic Acids Res. 43, e47 (2015). 88. Su, S. et al. Glimma: interactive graphics for gene expression analysis. Bioinformatics 33, 2050- 2052 (2017). 89. Robinson, M.D., McCarthy, D.J. & Smyth, G.K. edgeR: a Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics 26, 139-140 (2010). 90. Bolstad, B.M., Irizarry, R.A., Astrand, M. & Speed, T.P. A comparison of normalization methods for high density oligonucleotide array data based on variance and bias. Bioinformatics 19, 185 -193 (2003). 91. Smyth, G.K. Linear models and empirical bayes methods for assessing differential expression in microarray experiments. Stat Appl Genet Mol Biol 3, Article3 (2004). 92. Ritchie, M.E. et al. Empirical array quality weights in the analysis of microarray data. BMC Bioinformatics 7, 261 (2006). 93. Benjamini, Y . & Hochberg, Y . Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society: Series B (Methodological) 57, 289-300 (1995). 94. Wickham, H. ggplot2: Elegant Graphics for Data Analysis. (Springer-Verlag New York, 2016). 95. Slowikowski, K. ggrepel: Automatically Position Non -Overlapping Text Labels with 'ggplot2'. https://ggrepel.slowkow.com/, https://github.com/slowkow/ggrepel (2024). preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.618570doi: bioRxiv preprint