A hypomorphic PERK variant inhibits ER stress‐induced mitochondrial elongation

Pharmacologic inhibition of PERK signaling, but not other arms of the UPR, blocks mitochondrial elongation induced by ER stress (Lebeau et al, 2018). Here, we further probed the dependence of ER stress induced mitochondrial elongation on PERK activity in Perk ^−/− MEFs. We transfected Perk ^+/+ or Perk ^−/− MEFs with mitochondrial targeted GFP (^mtGFP) and monitored mitochondrial morphology in cells treated with or without the ER stressor thapsigargin (Tg; a SERCA inhibitor). We then scored cells based on the presence of fragmented, tubular, or elongated mitochondria (see Fig EV1A and B for representative examples). Perk ^−/− MEFs showed increases in fragmented mitochondria in the absence of treatment (Figs 1B and C, and EV1C). This corresponds with reductions in the mitochondrial membrane potential in Perk‐deficient cells, as measured by tetramethylrhodamine ethyl ester (TMRE) staining (Fig EV1D). This suggests that the increase of fragmentation in these cells can be attributed to mitochondrial depolarization. Tg‐induced mitochondrial elongation was also impaired in Perk‐deficient cells (Figs 1B and C, and EV1C). However, treatment with cycloheximide (CHX), which induces mitochondrial elongation independent of PERK signaling (Tondera et al, 2009; Lebeau et al, 2018), reduced the population of fragmented mitochondria in Perk ^−/− MEFs (Fig EV1E). This indicates that these cells are not deficient in their ability to induce elongation in response to reduced translation. Reconstitution of Perk ^−/− MEFs with wild‐type PERK restored basal mitochondrial morphology and rescued Tg‐induced mitochondrial elongation (Figs 1B and C, and EV1C). In contrast, reconstitution of Perk‐deficient cells with a hypomorphic PERK haplotype implicated in progressive supranuclear palsy (PSP; PERK^PSP; Hoglinger et al, 2011; Yuan et al, 2018) did not impact basally fragmented mitochondria or rescue Tg‐induced mitochondrial elongation. However, CHX increased mitochondrial length in Perk‐deficient cells expressing PERK^PSP (Fig EV1E). We confirmed similar expression of PERK^WT and PERK^PSP in Perk ^−/− MEFs by immunoblotting (Fig EV1F). These results implicate PERK signaling in ER stress induced mitochondrial elongation and demonstrate that genetic disruptions in PERK activity impair the regulation of mitochondrial morphology in response to ER stress.

Overexpression of cytosolic PA lipases inhibits ER stress induced mitochondrial elongation

Mitochondrial morphology is defined by the relative activities of GTPases localized to the inner and outer mitochondrial membranes that regulate organellar fission and fusion. These include the pro‐fission GTPase DRP1 of the outer mitochondrial membrane (OMM) and the pro‐fusion GTPases MFN1 and MFN2 of the OMM and OPA1 of the inner mitochondrial membrane (IMM; Mishra & Chan, 2016; Chan, 2020; Fenton et al, 2020; Giacomello et al, 2020; Sabouny & Shutt, 2020). Stress‐induced changes in mitochondrial shape can be dictated through posttranslational regulation of these GTPases to alter the relative activities of fusion and fission pathways (Mishra & Chan, 2016; Chan, 2020; Fenton et al, 2020; Giacomello et al, 2020; Sabouny & Shutt, 2020). Previous results indicate that PERK signaling does not influence the posttranslational regulation of these GTPases (Lebeau et al, 2018), suggesting that ER stress‐induced mitochondrial elongation proceeds through an alternative mechanism.

Mitochondrial elongation can be induced by the accumulation of saturated PA on the OMM through mechanisms including PA‐dependent inhibition of the pro‐fission GTPase DRP1 (Baba et al, 2014; Ha & Frohman, 2014; Adachi et al, 2016; Kameoka et al, 2018; Acoba et al, 2020). PERK was previously shown to increase cellular PA during ER stress through a mechanism dependent on PERK kinase activity but independent of signaling downstream of eIF2α phosphorylation (Bobrovnikova‐Marjon et al, 2012). We found that treatment with Tg increases PA in mitochondria‐enriched fractions and whole‐cell extracts from both MEF or HeLa cells using mass spectrometry, biochemical assays, and ELISA (Figs 2A–C and EV2A–F). Phosphatidylcholine (PC) was not affected in enriched mitochondria (Fig 2A). Co‐treatment with the PERK inhibitor GSK2656157, a compound that directly inhibits PERK kinase activity (Fig 1A; Axten et al, 2013), reduced Tg‐dependent increases of PA in both MEF and HeLa cells (Figs 2B and EV2A–D). This indicates that ER stress‐dependent increases in PA require PERK kinase activity, as previously reported (Bobrovnikova‐Marjon et al, 2012). However, co‐treatment of MEFs with Tg and ISRIB, a compound that blocks PERK signaling downstream of eIF2α phosphorylation (Fig 1A; Sidrauski et al, 2013), did not appear to mitigate ER stress induced PA increases in either mitochondria enriched fractions or whole‐cell extracts (Figs 2B, and EV2B and E). This is consistent with previous results suggesting that ER stress increases PA through a mechanism selectively dependent on PERK kinase activity, but not signaling downstream of eIF2α phosphorylation (Bobrovnikova‐Marjon et al, 2012).

We next determined the dependence of PERK‐regulated mitochondrial elongation on PA by monitoring mitochondrial morphology in Tg‐treated HeLa cells co‐overexpressing ^mtGFP and Lipin1—a cytosolic PA lipase that catalyzes the conversion of PA to diacylglycerol (DAG; Baba et al, 2014; Tatsuta & Langer, 2017; Kameoka et al, 2018; Tamura et al, 2020). We showed that Lipin1 overexpression reduced cellular PA and prevented Tg‐dependent increases of PA (Figs 2C, and EV2F and G). Overexpression of wild‐type Lipin1 increased basal mitochondrial fragmentation and inhibited Tg‐induced mitochondrial elongation (Figs 2D and E, and EV2H). Similar results were observed in cells treated with CHX. Lipin1 overexpression did not significantly impact the expression of ATF4 target genes (e.g., Asns and Chop) or increases of ATF4 protein in Tg‐treated cells (Fig EV2I and J). Furthermore, overexpression of a catalytically inactive Lipin1 did not influence basal mitochondrial morphology or mitochondrial elongation induced by Tg or CHX (Fig EV2K). Overexpression of PA‐PLA1—a cytosolic lipase that converts PA to lysophosphatidic acid (LPA; Baba et al, 2014)—similarly inhibited mitochondrial elongation in cells treated with Tg or CHX without impacting other aspects of PERK signaling (Fig EV2L–O). The sensitivity of Tg‐ and CHX‐induced mitochondrial elongation to PA‐PLA1 also suggests that this process is not mediated through increased LPA—a phospholipid that promotes mitochondrial elongation during starvation through a MTCH2‐dependent mechanism (Labbe et al, 2021). Collectively, our results show that depletion of PA afforded by overexpression of two distinct PA lipases blocks ER stress‐induced mitochondrial elongation, implicating PA in this process.

ER stress prevents DRP1‐dependent mitochondrial fragmentation

Mitochondrial elongation can be induced in response to stress through mechanisms involving posttranslational regulation of the pro‐fission GTPase DRP1. DRP1 phosphorylation at residue S637 promotes mitochondrial elongation by inhibiting DRP1 GTPase activity, while the pro‐fission phosphorylation of DRP1 at S616 increases DRP1 localization to mitochondria and subsequent activity (Chang & Blackstone, 2007; Taguchi et al, 2007; Kar et al, 2017). Pharmacologic mTOR inhibition can induce mitochondrial elongation through a mechanism involving both increased DRP1 phosphorylation at S637 and reduced phosphorylation at S616 (Morita et al, 2017). However, as reported previously (Lebeau et al, 2018), Tg did not influence DRP1 phosphorylation at either S637 or S616 (Fig EV3A) or alter the amount of DRP1 enriched in mitochondrial fractions from MEF^mtGFP cells (Fig EV3B). PERK‐dependent increases in PA can activate mTOR during ER stress (Bobrovnikova‐Marjon et al, 2012). Consistent with this, we observe Tg‐dependent increases in mTOR‐dependent S6K phosphorylation in MEF^mtGFP cells (Fig EV3C). However, despite increasing PA and promoting mitochondrial elongation, Tg did not increase S6K phosphorylation in HeLa cells (Fig EV3D). These results suggest that PERK‐dependent alterations in mTOR activity are unlikely to be primary contributors to ER stress induced mitochondrial elongation across cell types.

Accumulation of PA on the OMM can also promote mitochondrial elongation by inhibiting DRP1 activity (Adachi et al, 2016). This was previously demonstrated by showing that genetically increasing PA on the OMM by overexpressing mitoPLD—an OMM lipase that converts cardiolipin to PA—basally increased mitochondrial elongation and inhibited DRP1‐dependent mitochondrial fragmentation induced by the uncoupler carbonyl cyanide m‐chlorophenylhydrazone (CCCP; Li et al, 2015; Adachi et al, 2016). Consistent with this, we observed that mitoPLD overexpression in HeLa cells increased basal mitochondrial elongation and inhibited CCCP‐induced mitochondrial fragmentation (Fig 3A and B). We found that mitoPLD overexpression increased cellular PA to levels similar to that observed in Tg‐treated cells and did not significantly influence PERK signaling (Fig EV3E and F). Pretreatment with Tg also reduced CCCP‐induced mitochondrial fragmentation. However, Tg pretreatment inhibited CCCP‐induced proteolytic cleavage of the inner membrane GTPase OPA1 (Fig 3C)—a biological process upstream of DRP1 in mitochondrial fragmentation induced by membrane uncoupling (Mishra & Chan, 2016; Jones et al, 2017; Chan, 2020; Fenton et al, 2020; Giacomello et al, 2020; Sabouny & Shutt, 2020). This appears to result from Tg‐dependent increases in mitochondrial membrane polarity (Fig EV3G), preventing efficient uncoupling in CCCP‐treated cells and precluding our ability to determine whether Tg pretreatment directly impairs DRP1 activity under these conditions.

To circumvent this problem, we monitored mitochondria morphology in MEF^mtGFP cells pretreated with Tg and then challenged with ionomycin—a Ca²⁺ ionophore that increases cytosolic Ca²⁺ (Ji et al, 2015; Mishra & Chan, 2016; Chan, 2020; Fenton et al, 2020; Giacomello et al, 2020; Sabouny & Shutt, 2020). Increases in cytosolic Ca²⁺ induced by short (< 30 min) treatment with ionomycin promotes DRP1‐dependent mitochondrial fragmentation through a mechanism independent of membrane uncoupling or OPA1 processing (Ji et al, 2015). Pretreatment for 3 h with Tg—a time point sufficient to increase PA and induce mitochondrial elongation—inhibits ionomycin‐induced mitochondrial fragmentation (Fig 3D and E). This inhibition is reversed by co‐treatment with ISRIB, a small molecule that blocks eIF2α phosphorylation‐dependent signaling downstream of PERK (Fig 1A). This indicates that this inhibition of ionomycin‐induced fragmentation can be attributed to PERK signaling and not dysregulation of intracellular Ca²⁺ induced by the combined treatment of Tg and ionomycin. These results are consistent with a model whereby ER stress promotes mitochondrial elongation through a mechanism involving PA‐dependent inhibition of DRP1‐mediated fission, as reported previously for mitoPLD overexpression (Adachi et al, 2016).

PERK signaling leads to reductions in the intramitochondrial PA transporter PRELID1 during ER stress

ER stress induced mitochondrial elongation is inhibited by shRNA‐depletion of YME1L in HeLa cells (Lebeau et al, 2018). We further confirmed the dependence of Tg‐induced mitochondrial elongation on YME1L in MEF^mtGFP cells where Yme1l was deleted by CRISPR (Fig EV4A and B). PRELID1, an intermembrane space protein that transports PA from the OMM to the IMM (Fig 4A; Tatsuta & Langer, 2017; Tamura et al, 2020), is a known substrate of YME1L (Potting et al, 2010; Tamura et al, 2012; MacVicar et al, 2019). Interestingly, PRELID1 is a short‐lived protein whose levels are highly sensitive to translational attenuation (Li et al, 2021). Consistent with this, PRELID1 levels are reduced in MEF^mtGFP cells treated with CHX for 3 h (Fig 4B). This CHX‐dependent reduction in PRELID1 was blocked in Yme1l‐deficient cells (Fig 4B), confirming that PRELID1 is degraded by YME1L under these conditions. Identical results were observed for TIM17A, another short‐lived mitochondrial protein degraded by YME1L downstream of translation inhibition (Fig 4B; Rainbolt et al, 2013).

The sensitivity of PRELID1 to reductions in protein translation suggests that this protein could be decreased in response to PERK‐dependent translational attenuation. As expected, PRELID1 was rapidly decreased in MEF^mtGFP cells treated with the ER stressor Tg (Fig EV4C). Tg‐dependent reductions in PRELID1 were inhibited in cells deficient in Yme1l, indicating that YME1L was required for this process (Fig 4C). Co‐treatment with either the PERK kinase inhibitor GSK2656157 or the PERK signaling inhibitor ISRIB (Fig 1A) blocked Tg‐dependent reductions in PRELID1 (Fig 4D). Similar results were observed for TIM17A. Tg‐dependent reductions in PRELID1 and TIM17A were also inhibited in Perk ^−/− MEFs (Fig 4E). Reconstitution of Perk ^−/− cells with PERK^WT, but not hypomorphic PERK^PSP, restored Tg‐induced degradation of these proteins. Importantly, CHX reduced PRELID1 and TIM17A in all genotypes, confirming that these proteins remain sensitive to translational attenuation even when PERK signaling is impaired (Fig 4E). Tg‐dependent reductions in PRELID1 were not inhibited in cells deficient in Atf4 (Fig EV4D; Harding et al, 2003), a primary upstream transcription factor in the PERK pathway (Fig 1A). This indicates that this phenotype is independent of PERK‐regulated transcriptional signaling. Collectively, these results suggest that PRELID1, like TIM17A (Rainbolt et al, 2013), is reduced during ER stress through a YME1L‐dependent mechanism downstream of PERK‐dependent translational attenuation.

PERK‐dependent PRELID1 degradation remodels mitochondrial membranes during ER stress

PRELID1 traffics PA from the outer to inner mitochondrial membrane, where it serves as a precursor to the formation of cardiolipin (Potting et al, 2010; Tamura et al, 2012; Tatsuta & Langer, 2017). Thus, reductions in PRELID1 should decrease cardiolipin. To test this, we shRNA‐depleted Prelid1 from MEF^mGFP cells and monitored cardiolipin in isolated mitochondria in the presence or absence of ER stress. We confirmed efficient PRELID1 knockdown by immunoblotting (Fig EV5A). Importantly, Prelid1 depletion did not alter Tg‐induced reductions of TIM17A or increases of ATF4. Furthermore, Tg‐dependent increases in PA were observed in Prelid1‐depleted MEF^mtGFP cells (Fig EV5B). These results indicate that loss of PRELID1 does not impair PERK signaling in these cells. Prelid1 depletion reduced cardiolipin in mitochondria isolated from MEF^mtGFP cells (Fig EV5C). Treatment of MEF^mtGFP cells expressing nonsilencing shRNA with Tg for 3 h also reduced cardiolipin to levels similar to those observed in Prelid1‐deficient cells. However, Tg did not further reduce cardiolipin in Prelid‐depleted cells. These results are consistent with a model whereby ER stress‐dependent reductions in PRELID1 limit PA trafficking across the inner mitochondrial membrane and contribute to reductions in cardiolipin during acute ER stress.

In combination with ER stress‐dependent increases in PA (Figs 2A–C and EV2A–F), reductions in PRELID1‐mediated PA trafficking across mitochondrial membranes should lead to the accumulation of PA on the OMM where it could promote mitochondrial elongation by inhibiting mitochondrial fission (Adachi et al, 2016). To test this, we monitored mitochondrial morphology in Prelid1‐depleted MEF^mtGFP cells in the presence and absence of Tg. Interestingly, Prelid1 depletion did not basally influence mitochondrial morphology or inhibit Tg‐induced mitochondrial elongation (Figs 5A and B, and EV5D). This indicates that reduction of PRELID1, on its own, is not sufficient to increase mitochondrial elongation, likely reflecting the importance of PERK kinase‐dependent increases in PA in this process (Bobrovnikova‐Marjon et al, 2012). Consistent with this model, co‐treatment with the PERK kinase inhibitor GSK2656157 blocked Tg‐induced mitochondrial elongation in Prelid1‐deficient cells (Figs 5A and B, and EV5D). However, we found that Prelid1 depletion partially rescued the Tg‐induced mitochondrial elongation in cells co‐treated with ISRIB—a compound that blocks PERK‐dependent PRELID1 degradation (Fig 4D), but not PERK kinase‐dependent increases in PA (Figs 2B, and EV2B and E). Co‐treatment with ISRIB completely blocked Tg‐induced mitochondrial elongation in MEF^mtGFP cells expressing non‐silencing shRNA (Figs 5A and B, and EV5D). Identical results were observed in HeLa cells depleted of PRELID1 and treated with Tg, GSK2656157, and/or ISRIB (Figs 5C and D, and EV5E and F). These results indicate PERK‐dependent reductions in PRELID1 contribute to the mitochondrial elongation observed during ER stress.

Cell culture, transfections, lentiviral transduction, and CRISPR deletion

MEF^mtGFP (a kind gift from Peter Schultz, TSRI; Wang et al, 2012), Perk ^+/+ and Perk ^−/− MEFs (Harding et al, 2000), Atf4 ^+/+ and Atf4 ^−/− MEFs (Harding et al, 2003; kind gifts from David Ron, Cambridge), HeLa cells (purchased from the ATCC), or HEK293T cells were cultured in DMEM (Corning‐Cellgro) supplemented with 10% fetal bovine serum (FBS; Omega Scientific), 2 mM L‐glutamine (GIBCO), 100 U/ml penicillin, and 100 mg/ml streptomycin (GIBCO). Cells were maintained at 37°C and 5% CO₂. Nonessential amino acids (GIBCO) and 2‐mercaptoethanol (ThermoFisher) were added to culture media of Atf4 ^+/+ and Atf4 ^−/− MEFs and Perk ^+/+ and Perk ^+/+ MEFs. HeLa cells were transfected by calcium phosphate precipitation, as previously described (Lebeau et al, 2018). MEF cells were transfected with MEF Avalanche Transfection Reagent (EZ Biosystems) according to the manufacturer's protocol. Lentivirus were prepared by transfecting HEK293T cells with pRSV‐rev (Addgene #12253), pMDL‐RRE (Addgene, #12251), pMD2.6 (Addgene #12259), and the indicated shRNA in the pLKO.1 vector (Sigma) using calcium phosphate precipitation. After 24 h, the transfection media was removed and replaced with complete DMEM and incubated overnight for viral production. Virus containing media was removed the following day and filtered with a 0.45‐μm syringe filter (Genessee Scientific). Polybrene (ThermoFisher) was added to the virus containing media at a concentration of 10 μg/ml and the media was then added to HeLa or MEF^mtGFP cells. Stable pools of cells expressing nonsilencing or gene‐specific knockdowns were then generated through selection with puromycin (3 mg/ml for MEF cells and 1 mg/ml for HeLa). Knockdown was confirmed by immunoblotting. Yme1l was deleted from MEF^mtGFP cells using CRISPR/Cas9. Briefly, cells were transfected with pSpCas9(BB)‐2A‐Puro (PX459; Addgene, #62988) containing sgRNA against Yme1l (GATCCAATATGAGATGTATGCCAAC AAACGTTGGCATACATCTCATATT) using MEF Avalanche, following manufacturers protocols. After transfection, cells were selected with puromycin and single clones were screened for YME1L depletion by qPCR and immunoblotting.

Plasmids, shRNAs, and compounds

HA‐LIPIN1^WT, HA‐LIPIN^Mut, and mitoPLD‐GFP overexpression constructs were kind gifts from Hiromi Sesaki (Johns Hopkins) and described previously (Adachi et al, 2016). The PA‐PLA1‐GFP overexpression construct was purchased from Addgene (#162880). Perk ^WT and Perk ^PSP overexpression plasmids were kind gifts from Jonathan Lin (Stanford; Yuan et al, 2018). Plasmids containing shRNA were purchased from Sigma in the pLKO.1 vector: mouse Prelid1 shRNA (TRCN0000345802), human PRELID1 shRNA (TRCN0000130829). All compounds used in this study were purchased: thapsigargin (Tg; Fisher Scientific), GSK2656157 (BioVision Inc.), ISRIB (Sigma), CCCP (Sigma), rapamycin (Selleckchem), and ionomycin (Sigma).

Fluorescence microscopy

HeLa cells transfected with ^mtGFP or MEF^mtGFP cells were seeded at a density of 100,000 cells/well on glass‐bottom dishes (MatTek) coated with poly‐D‐lysine (Sigma) or rat tail collagen 1 (GIBCO). Cells were then treated as indicated and images were recorded with an Olympus IX71 microscope with 60× oil objective (Olympus), a Hamamatsu C8484 camera (Hamamatsu Photonics), and HCI image software (Hamamatsu Photonics). Quantification was performed by blinding the images and then scoring cells based on the presence of primarily fragmented, tubular, or elongated mitochondria, as before (Lebeau et al, 2018). At least three different researchers scored each set of images and these scores were averaged for each individual experiment and all quantifications shown were performed for at least three independent experiments quantifying a total of > 60 cells/condition across all experiments. The data were then prepared in PRISM (GraphPad, San Diego, CA) and plotted on a stacked bar plot to show the average morphology and standard error of the mean across all experiments. Statistical comparisons were performed using a two‐way ANOVA in PRISM, comparing the relative amounts of fragmented, tubular, or elongated mitochondria across different conditions.

Phospholipid quantification

For untargeted MS analysis of PA, whole‐cell pellets were resuspended in 500 μl of a cold hypotonic buffer consisting of 1 mM PBS, pH 7.4. The material was then homogenized on ice using a glass Dounce homogenizer (30 strokes). The homogenized sample was centrifuged at 500 × g for 4 min then the supernatant was transferred to a 1.5‐ml microfuge tube and lyophilized overnight. The lyophilized material was weighed and normalized by total mass prior to performing a modified Bligh and Dyer extraction (Bligh & Dyer, 1959). The proceeding steps were carried out with glass pipettes and tubes to avoid plastic contamination. PA was extracted by the addition of 100 μl/mg of cold methanol containing an internal PA standard (Splash Lipidomix 330707, Avanti) at a dilution of (1:50), followed by 50 μl/mg of cold chloroform (CHCl₃) with occasional vortex mixing. Milli‐Q H₂0 containing 5 mM erythorbate was added at a volume of 50 μl/mg. The sample was agitated and centrifuged in glass test tubes at 200 × g for 10 min. The bottom phase was collected in a clean test tube, while the upper phase was re‐extracted two additional times with CH₃OH:CHCl₃ (1:1, v/v) containing HCl at final concentration of 10 mM. The organic phases were combined and dried under vacuum to afford a lipid film that was stored at −80°C until MS analysis. Mitochondria‐enriched fractions were processed similar to whole‐cell samples except they were normalized via protein concentration as determined by the Pierce™ BCA Protein Assay kit (Thermo Scientific). In Brief, PA was extracted from the mitochondria enriched fractions using 10 μl/μg methanol, 5 μl/μg CHCl₃ and 5 μl/μg milli‐Q H₂0 containing 5 mM erythorbate as previously described above.

Extracted lipid samples and external standards (SPLASH Lipidomix 330707, 18:0 CL 710334p, Avanti) were removed from the −80°C freezer after drying and stored under nitrogen were resuspended in 100 μl of methanol. Negative mode LC–MS analysis was performed on an Agilent 6230 ESI‐TOF‐MS System calibrated with a reference solution at m/z 1,033.9881. A XBridge BEH C8 XP Column (Waters, 2.5 μm, 4.6 mm × 150 mm) was used at a flow rate of 500 μl/min, employing the following gradient: 30 to 100% solvent B over 30 min, 100% isocratic B for 10 min followed by a return to 30% B for 5 min. Solvents A consisted of MilliQ water and methanol (9:1, v:v). Solvents B consisted of acetonitrile: 2‐propanol (5:3, v:v), and both contained 10 mM Piperidine, 10 mM ammonium acetate (or 10 mM sodium acetate), and 0.1% formic acid. Prior to processing, raw.d files were converted to the open format mzXML using MSConvert software, which is part of the ProteoWizard software toolkit (Chambers et al, 2012). Mass detection was achieved using mzmine 2 wavelet algorithm, ADAP chromatogram builder and ADAP deconvolution, which are part of the MZmine 2 software package (Pluskal et al, 2010). Initial lipid identifications were achieved using lipidmaps database using [M − H]⁻ for PA with a m/z tolerance of 15 ppm, subsequently detected lipids were filtered out for further processing. The putative PA peaks were validated by aligning to the internal and external standards followed by graphical identification of PA lipids by plotting the Kendrick mass defect plot employing CH₂ as the repeating unit. The quantification of all PA lipid classes was normalized based on the abundance of the internal standard PA (15:0–18:1‐d7‐PA), which factors in extraction efficiency and sample handling. Total PA levels were then normalized to vehicle for the indicated number of independent experiments.

Targeted lipidomics was performed on abundant lipid species for both PA and CL, (34:1) and (66:2, 68:2, 68:3), respectively. The relative abundance of these individual lipid species were quantified using a 1,260 Infinity II LC System (G6125BW) in selected ion monitoring (SIM) mode outfitted with an ESI source. All of the lipid species were observed as [M – H] ⁻ ions. Subsequently, the replicates were pooled together and injected onto a high‐resolution, time‐of‐flight (Q‐TOF) MS/MS (Agilent model 6546) using the identical RP platform as the targeted MS approach for the purpose of molecular identification. For the MS/MS PA, 34:1 was observed as a [M – H] ⁻ peak at 673.4802 with mass fragments of 281.2490, 255.2335, and 152.9963 m/z. For Cl 68:2, we observed the [M – H] ⁻ peak at 1,403.9983 with the MS/MS spectra comprising primarily of 673.4805, 417.2421, 281.2485, and 255.2324 m/z. The RP platform used was the same as previously described above for our untargeted approach except the mobile phase was changed to, 0–10 min 40% solvent B isocratic, 10–40 min 40–100% solvent B as a gradient. Solvent A is MilliQ water:methanol (1:1, v/v) and solvent B is methyl tert‐butyl ether (MTBE): 2‐propanol (1:4, v:v) both containing 2 mM ammonium acetate. To generate the required sensitivity for identification in the MS/MS we extracted the lipids as described previously with these notable exceptions. The dried cell lysate was comprised from a single 15‐cm tissue culture plate affording 5–6 mg of dried material per sample. The total cellular lipids of this material were extracted using four sequential extractions differing in solvent compositions. The first extraction was achieved by adding 200 μl MilliQ water, 400 μl methanol and 400 μl chloroform. The second subsequent extraction was performed by adding an additional 200 μl methanol and 400 μl chloroform to the remaining aqueous layer. The third extraction uses an addition of 500 μl MTBE, this fraction does not contain many enriched lipids, but functions to remove residual methanol/chloroform that will prevent separation of the last butanol extraction. The last extraction uses 100 μl butanol and 400 μl MTBE both of which are water saturated before their addition to the remaining aqueous layer. The organic phases were combined and dried under vacuum to afford a lipid film that was stored at −80°C until targeted MS and MS/MS analysis.

For quantification of PA by ELISA, MEF or HeLa cells were treated as indicated and collected on ice and then lysed with 20 mM Hepes (Sigma, H4034) pH 7.4, 100 mM NaCl (Sigma, S7653), 1 mM EDTA (Sigma, E9884), 1% Triton X100 (Sigma, 9036‐19‐5) supplemented with Pierce protease inhibitor (ThermoFisher, A32963). Protein concentrations for each sample were then quantified using the Bio‐Rad Bradford assay. PA was then measured using the Human Phosphatidic Acid Antibody IgM ELISA Kit (Lifeome Bioloabs) following the manufacturers protocol and monitoring fluorescence on a Tecan F250Pro microplate reader (Tecan).

For quantification of PA by fluorometric biochemical assay, MEF or HeLa cells were treated as indicated and collected on ice and then centrifuged and washed with cold PBS three times. Samples were then sonicated using the Misonix S‐4000 sonicator then processed and PA was measured according the manufacturer's protocol for the Total Phosphatidic Assay Kit (CellBio Labs).

Immunoblotting and antibodies

Whole cells were lysed at room temperature in HEPES lysis buffer (20 mM Hepes pH 7.4, 100 mM NaCl, 1 mM EDTA, 1% Triton X100) supplemented with 1× Pierce protease inhibitor (ThermoFisher). Total protein concentrations of lysates were then normalized using the Bio‐Rad protein assay and lysates were combined with 1× Laemmli buffer supplemented with 100 mM DTT and boiled for 5 min. Samples (100 μg protein) were then separated using 10 or 12% SDS–PAGE gels and transferred to nitrocellulose membranes (Bio‐Rad). Membranes were then blocked with 5% milk in tris‐buffered saline (TBS) and then incubated overnight at 4°C with the indicated primary antibody. The next day, membranes were washed in TBS supplemented with Tween, incubated with the species appropriate secondary antibody conjugated to IR‐Dye (LICOR Biosciences) and then imaged using an Odyssey Infrared Imaging System (LICOR Biosciences). Quantification was then carried out using the LICOR Imaging Studio software.

Primary antibodies were acquired from commercial sources and used in the indicated dilutions in antibody buffer (50 mM Tris [pH 7.5], 150 mM NaCl supplemented with 5% BSA (w/v) and 0.1% NaN₃ (w/v)): TIM17A (Thermo Scientific, PA5‐21925; 1:1,000), PRELID1 [aa27‐54] (LS Bio, LC‐C158729; 1:1,000), YME1L (Proteintech, 11510‐1‐AP; 1:1,000), ATF4 (Cell Signaling, 11815; 1:500), Tubulin [B‐5‐1‐2] (Sigma, T6074; 1:5,000), TIM23 (BD Transduction Labs, 611222; 1:1,000), HSP60 [LK1] (Thermo Scientific, MA1‐35434; 1:1,000), PERK (Cell Signaling, 3192S; 1:1,000), HA [Clone: 16B12] (Biolegend, 901501; 1:1,000), GFP (B2; Santa Crux, sc9996; 1:1,000), OPA1 (BD Biosciences, 612606; 1:2,000), Phospho‐DRP1 (Ser616) (ThermoFisher, PA5‐64821; 1:1,000), Phospho‐DNM1L (Ser637; ThermoFisher, PA5; 101038; 1:1,000), DNM1L (ThermoFisher, MA5‐26255; 1:1,000), p70 S6 Kinase p(Ser389; Cell Signaling, 9206S; 1:1,000), p70 S6 Kinase (Cell Signaling, 9202S; 1:1,000).

Quantitative polymerase chain reaction (qPCR)

The relative mRNA expression of target genes was measured using quantitative RT‐PCR. Cells were treated as indicated and then washed with phosphate buffered saline (PBS; Gibco). RNA was extracted using Quick‐RNA MiniPrepKit (Zymo Research) according to the manufacturers protocol. RNA (500 ng) was then converted to cDNA using the QuantiTect Reverse Transcription Kit (Qiagen). qPCRs were prepared using Power SYBR Green PCR Master Mix (Applied Biosystems), and primers (below) were obtained from Integrated DNA Technologies. Amplification reactions were run in an ABI 7900HT Fast Real Time PCR machine with an initial melting period of 95°C for 5 min and then 45 cycles of 10 s at 95°C, 30 s at 60°C.

Primers used in this study

Human ASNS: forward: GCAGCTGAAAGAAGCCCAAG; reverse: AGCCTGAATGCCTTCCTCAG.

Human CHOP/DDIT3: forward: ACCAAGGGAGAACCAGGAAACG; reverse: TCACCATTCGGTCAATCAGAGC.

Human HSPA5/BIP: forward: GCCTGTATTTCTAGACCTGCC; reverse: TTCATCTTGCCAGCCAGTTG.

Human RIBOP: forward: CGT CGC CTC CTA CCT GCT; reverse: CCA TTC AGC TCA CTG ATA ACC TTG.

Membrane depolarization

Cells were seeded at a density of 85,000 cells/well of a six‐well plate and treated with 500 nM Tg for 3 h prior to collection. CCCP (10 μM) was added 50 min before collection, followed by 200 nM TMRE (Thermofisher) 20 min before collection. Samples were collected using TrypLE Express and cell culture media. Following a brief centrifugation, cell pellets were washed in DPBS (Gibco) and resuspended in DPBS supplemented with 5% BSA. Fluorescence intensity of TMRE for 20,000 cells/condition was recorded on the PE channel of a BD Biosciences LSR II analytical flow cytometer. Data are presented as geometric mean of the fluorescence intensity from three experiments normalized to vehicle‐treated cells.

Mitochondrial isolation

Whole cells were collected on ice from at least 3 × 10 cm plates then pelleted. Cells were mixed with mitochondrial lysis buffer (220 mM sorbitol; 70 mM sucrose; 50 mM MOPS pH 7.4; 5 mM EGTA) supplemented with 1× Pierce protease inhibitor (ThermoFisher) and lysed by passing through a 21 guage needle 10–20 times. Lysed cells were spun down at 1,000 × g for 10 min to pellet the nuclei and unlysed cells. The supernatant was transferred into a new tube and spun down at 9,500 × g for 10 min to pellet the mitochondria. The supernatant was saved as a cytosolic control. For immunoblotting, mitochondrial pellets were lysed on ice for 10 min in mitochondrial wash buffer (220 mM sorbitol; 70 mM sucrose; 50 mM MOPS pH 7.4) supplemented with 1% Triton and protein concentration was determined using the Bio‐rad protein assay and prepared as described above.

Statistical analysis

Statistics were calculated in PRISM 9 (GraphPad, San Diego, CA). Data are presented as mean ± SEM and were analyzed by two‐way ANOVA with Tukey's multiple correction test, one‐way ANOVA, or the appropriate Student's t‐tests, as indicated in the accompanying figure legends. Indications of nonsignificant interactions from ANOVA were generally omitted for clarity.