Phosphatidylethanolamine Deficiency in Mammalian Mitochondria Impairs Oxidative Phosphorylation and Alters Mitochondrial Morphology*

Background: The contribution of phosphatidylethanolamine (PE) to mammalian mitochondrial function was unknown. Results: A decrease in mitochondrial PE impairs cell growth, respiratory capacity, and ATP production and profoundly alters mitochondrial morphology. Conclusion: Mitochondrial PE is required for normal morphology and function of mammalian mitochondria. Significance: Modest reduction of mitochondrial PE might contribute to mitochondrial dysfunction in some disease states. Mitochondrial dysfunction is implicated in neurodegenerative, cardiovascular, and metabolic disorders, but the role of phospholipids, particularly the nonbilayer-forming lipid phosphatidylethanolamine (PE), in mitochondrial function is poorly understood. Elimination of mitochondrial PE (mtPE) synthesis via phosphatidylserine decarboxylase in mice profoundly alters mitochondrial morphology and is embryonic lethal (Steenbergen, R., Nanowski, T. S., Beigneux, A., Kulinski, A., Young, S. G., and Vance, J. E. (2005) J. Biol. Chem. 280, 40032–40040). We now report that moderate <30% depletion of mtPE alters mitochondrial morphology and function and impairs cell growth. Acute reduction of mtPE by RNAi silencing of phosphatidylserine decarboxylase and chronic reduction of mtPE in PSB-2 cells that have only 5% of normal phosphatidylserine synthesis decreased respiratory capacity, ATP production, and activities of electron transport chain complexes (C) I and CIV but not CV. Blue native-PAGE analysis revealed defects in the organization of CI and CIV into supercomplexes in PE-deficient mitochondria, correlated with reduced amounts of CI and CIV proteins. Thus, mtPE deficiency impairs formation and/or membrane integration of respiratory supercomplexes. Despite normal or increased levels of mitochondrial fusion proteins in mtPE-deficient cells, and no reduction in mitochondrial membrane potential, mitochondria were extensively fragmented, and mitochondrial ultrastructure was grossly aberrant. In general, chronic reduction of mtPE caused more pronounced mitochondrial defects than did acute mtPE depletion. The functional and morphological changes in PSB-2 cells were largely reversed by normalization of mtPE content by supplementation with lyso-PE, a mtPE precursor. These studies demonstrate that even a modest reduction of mtPE in mammalian cells profoundly alters mitochondrial functions.


Mitochondrial dysfunction is implicated in neurodegenerative, cardiovascular, and metabolic disorders, but the role of phospholipids, particularly the nonbilayer-forming lipid phosphatidylethanolamine (PE), in mitochondrial function is poorly understood. Elimination of mitochondrial PE (mtPE) synthesis via phosphatidylserine decarboxylase in mice profoundly alters mitochondrial morphology and is embryonic lethal (Steen
. We now report that moderate <30% depletion of mtPE alters mitochondrial morphology and function and impairs cell growth. Acute reduction of mtPE by RNAi silencing of phosphatidylserine decarboxylase and chronic reduction of mtPE in PSB-2 cells that have only 5% of normal phosphatidylserine synthesis decreased respiratory capacity, ATP production, and activities of electron transport chain complexes (C) I and CIV but not CV. Blue native-PAGE analysis revealed defects in the organization of CI and CIV into supercomplexes in PE-deficient mitochondria, correlated with reduced amounts of CI and CIV proteins. Thus, mtPE deficiency impairs formation and/or membrane integration of respiratory supercomplexes. Despite normal or increased levels of mitochondrial fusion proteins in mtPE-deficient cells, and no reduction in mitochondrial membrane potential, mitochondria were extensively fragmented, and mitochondrial ultrastructure was grossly aberrant. In general, chronic reduction of mtPE caused more pronounced mitochondrial defects than did acute mtPE depletion. The functional and morphological changes in PSB-2 cells were largely reversed by normalization of mtPE content by supplementation with lyso-PE, a mtPE precursor. These studies demonstrate that even a modest reduction of mtPE in mammalian cells profoundly alters mitochondrial functions.
Phosphatidylethanolamine (PE) is more abundant in mammalian mitochondria than in other organelles and contributes ϳ30% of mitochondrial phospholipids. PE participates in contractile ring disassembly during cytokinesis (20,21) and regulates the topology of integral membrane proteins in Escherichia coli (22,23). In addition, PE is a precursor of the ϳ30% of hepatic phosphatidylcholine made via PE N-methyltransferase (24). PE is also a precursor of anandamide (N-arachidonoylethanolamine) (25,26), the ethanolamine moiety of glycosylphosphatidylinositol anchors of many cell surface signaling proteins (27), and the ethanolamine linked to eukaryotic elongation factor eEF1A (28). Furthermore, PE appears to be required for autophagosome formation (12,29) and prion propagation in the brain (30). Although recent in vitro studies suggest that mitochondrial PE (mtPE) regulates mitochondrial outer membrane permeability (31), the role of mtPE in mammalian mitochondrial function is poorly understood.
The majority of PE in mammalian cells is synthesized by two spatially separated pathways (32,33) as follows: the CDP-ethanolamine pathway (the final step of which occurs on the endoplasmic reticulum (ER) (34,35)), and the phosphatidylserine (PS) decarboxylase (PSD) pathway in mitochondrial inner membranes (36,37). In addition, small amounts of PE are made by base-exchange between ethanolamine and PS (38). In yeast, PE can also be made by acylation of lyso-PE by the acyltransferase, Ale1p (39,40). Although a family of genes related to Ale1 is present in mammals, the contribution of this pathway to PE synthesis in mammalian cells is unknown. Importantly, the majority of mtPE in mammalian and yeast cells is made in mitochondria from PSD, whereas only a small fraction of mtPE is imported from the ER (41)(42)(43)(44). For PS to be decarboxylated to PE, PS is imported into mitochondria from its site of synthesis in mitochondrionassociated membranes (MAM) (45), an ER domain that becomes transiently tethered to mitochondria (35, 46 -48).
In mammalian cells, a single gene (Pisd) encodes PSD activity (1,49), whereas yeast PSD activity is encoded by two distinct genes (50,51). Reduction of PE synthesis via PSD alters mitochondrial morphology in yeast (52) and Trypanosoma brucei (53), and depletion of PE to Ͻ4% of yeast phospholipids impairs survival (51,54). PE synthesis from PSD is essential in yeast mutants lacking CL synthase, suggesting that CL and PE have overlapping functions (55). Remarkably, however, E. coli cells survive when PE, normally the most abundant phospholipid, is reduced to 0.007% of total phospholipids (56). PSD is required for mouse development because disruption of the mouse Pisd gene was embryonic lethal and profoundly altered mitochondrial morphology, even when PE was actively made from CDPethanolamine (1,57). Moreover, global disruption of the CDPethanolamine pathway for PE synthesis in mice is also embryonic lethal (58) demonstrating that both major PE biosynthetic pathways are required for mouse viability.
We have now investigated mechanisms underlying mitochondrial defects in mammalian cell models in which mtPE was reduced either chronically or acutely. Depletion of mtPE by only 20 -30% reduced cell growth, respiratory capacity, and ATP production, consistent with altered respirasome organization and defects in ETC complexes I and IV. In addition, mitochondrial ultrastructure was profoundly aberrant, and mitochondria were extensively fragmented. Supplementation of mtPE-deficient cells with lyso-PE normalized mtPE content and mitochondrial morphology, promoted cell growth, and increased ATP production. Thus, even a modest decrease in mtPE is detrimental for mitochondrial function and dynamics.

EXPERIMENTAL PROCEDURES
Cell Culture-CHO-K1 cells were purchased from the American Type Tissue Culture Collection (Manassas, VA). The mutant CHO cell line PSB-2 (59) was a gift from Dr. M. Nishijima (Tokyo, Japan). All cells were cultured at 37°C in Ham's F-12 medium (Invitrogen) containing 10% fetal bovine serum (FBS; Invitrogen).
Mitochondrial DNA-Genomic DNA was isolated using the DNeasy kit (Qiagen). Mitochondrial DNA was amplified with primers complementary to ATP synthase subunit ATP6 as follows: forward 5Ј-ACC GCA GGA CAT CTG CTA AT-3Ј and reverse 5Ј-GGT AGG CTG ATT GAG GTT AGG A-3Ј. Nuclear DNA was amplified with primers complementary to Pisd as follows: forward 5Ј-CTG GAC CTT TGG GGT AAA CA-3Ј and reverse 5Ј-AGG TTT CGG TAG TGG TGC AG-3Ј.
Metabolic Labeling of PS and PE-Cells were grown to 80% confluence in 100-mm dishes and then incubated in Ham's F-12 medium containing 5 Ci/ml L-[G-3 H]serine (Perkin-Elmer Life Sciences). Lipids were extracted (60) and separated by thin layer chromatography with the developing solvent chloroform/methanol/acetic acid/formic acid/water (70:30:12:4:2, v/v). Phospholipids were visualized by exposure to iodine vapor and PS and PE identified by comparison with authentic PS and PE (Avanti Polar Lipids). Incorporation of [ 3 H] into PS and PE was quantified by scintillation counting.
Isolation of Mitochondria, MAM, and Microsomes-Subcellular fractions were isolated as described previously (35,41). Briefly, cells were homogenized in isolation medium (225 mM mannitol, 75 mM sucrose, 10 mM Tris-HCl (pH 7.2), 1 mM EGTA, 0.1% fatty acid-free bovine serum albumin) with 15 strokes in a Balch homogenizer (Isobiotec Precision Engineering) followed by 20 strokes in a glass-Teflon homogenizer. The homogenate was centrifuged at 600 ϫ g for 5 min at 4°C, and the supernatant was centrifuged at 10,300 ϫ g for 10 min to pellet crude mitochondria. Microsomes were pelleted by centrifugation of the resultant supernatant at 100,000 ϫ g for 1 h. Crude mitochondria were applied to the top of tubes containing 30% Percoll (v/v) in isolation medium and centrifuged at 95,000 ϫ g for 30 min. Mitochondria were removed and washed with isolation medium (without albumin). MAM were removed from the Percoll gradient, diluted with isolation medium, and centrifuged for 10 min at 6,300 ϫ g. The supernatant was centrifuged at 100,000 ϫ g for 1 h to pellet MAM.
Measurement of Phospholipid Mass-Phospholipids were extracted from cell lysates and subcellular fractions (60), isolated by thin layer chromatography as described above, and quantified as P i (61). CL was isolated by thin layer chromatography in the developing solvent chloroform/hexane/methanol/ acetic acid (50:30:10:5, v/v).
Confocal Microscopy-Cells were imaged using an LSM 510 multiphoton confocal microscope (Zeiss) and analyzed with Zeiss software. Fixed cells were plated on coverslips coated with poly-D-lysine (Sigma). For visualization of mitochondria, cells were stained for 30 min with 100 nM MitoTracker Red CMX (Molecular Probes), fixed for 15 min at room temperature in 4% paraformaldehyde, and then washed with phosphate-buffered saline. Nuclei were stained with DAPI and cells mounted on slides using ProlongGold (Molecular Probes). For immunocytochemistry, cells were fixed for 15 min at room temperature in 4% paraformaldehyde and then permeabilized for 5 min with 0.25% (v/v) Triton X-100 in phosphate-buffered saline. After blocking for 30 min with Image-iT FX Signal Enhancer (Molecular Probes), cells were incubated overnight at 4°C with mouse monoclonal primary antibodies in blocking solution (Dako) as follows: anti-Drp1 (catalog no. 61112, dilution 1:200; BD Transduction Laboratories); anti-Mfn1 (catalog no. ab57602, dilution 1:500; Abcam); anti-Mfn2 (catalog no. ab56889, dilution 1:500; Abcam), and subsequently with FITC-conjugated secondary antibody (dilution 1:1000; Molecular Probes). All experiments included negative controls with secondary antibody alone.
Mitochondrial membrane potential was measured in live cells using tetramethylrhodamine methyl ester (TMRM) (10 nM, 30 min at 37°C). Mitochondrial mass was assessed by staining live cells for 30 min at 37°C with 100 nM 10-N-nonylacridine orange (membrane potential-insensitive dye; Molecular Probes). Fluorescence intensities were measured in five random fields/cell line. Nuclei were stained for 10 min with 500 nM Hoechst 33342 (Molecular Probes).
Electron Microscopy-Cells were fixed in phosphate-buffered saline containing 4% paraformaldehyde and 2% glutaraldehyde. Ultra-thin sections were stained with uranyl acetate and osmium tetroxide. Transmission electron microscopy images were obtained with a JEOL 1200 EX microscope and analyzed by ImageJ software.
Cellular ATP and ATP Production-Trichloroacetic acid (2.5%) was added to cells, and the mixture was neutralized with NaHCO 3 /Tris (pH 7.0). ATP was quantified with an ATP bioluminescence kit (Molecular Probes) and normalized to protein concentration.
For measurement of the rate of ATP synthesis, 1.85 ϫ 10 6 wild-type cells/15-cm dish and 3.7 ϫ 10 6 PSB-2 cells/15-cm dish were plated 48 h prior to experiments so that 70 -80% confluency was attained. Pisd KD#1 and NegCtrl cells were cultured as above. Cells were washed with 25 mM Tris-HCl (pH 7.4), 140 mM KCl, 10 mM NaCl, 2.5 mM MgCl 2 , 0.1 M CaCl 2 and then permeabilized for 5 min at 37°C in medium containing 40 g/ml digitonin, 150 mM KCl, 25 mM Tris-HCl (pH 7.4), 2 mM EDTA, 10 mM KH 2 PO 4 , and 0.1% fatty acid-free bovine serum albumin. The cells were then incubated for 30 min at 37°C in the same medium containing 1 mM ADP Ϯ respiratory substrates. Reactions were terminated with 1% trichloroacetic acid and neutralized with NaHCO 3 . Samples were centrifuged for 10 min at 13,000 ϫ g, and nanomoles of ATP/min/mg of protein were quantified in supernatants.
ETC Enzymatic Activities-Specific activities of complex I (NADH-ubiquinone oxidoreductase), complex II (succinate dehydrogenase), and complex IV (cytochrome c oxidase) were measured spectrophotometrically in mitochondria exposed to one freeze-thaw cycle to permeabilize inner membranes to NADH. Mitochondria (20 g of protein/ml) were incubated in buffer containing 10 mM Tris-HCl (pH 7.4), 50 mM KCl, 1 mM EDTA, 2 mM KCN, 2 M antimycin A, and 100 M NADH. The reaction was started with 50 M coenzyme Q 1 . Complex I activity (NADH oxidation) was measured by decreased absorbance at 340 nm. Complex II activity was measured with 30 g of protein/ml in 10 mM Tris-HCl (pH 7.2), 50 mM KCl, 2 mM KCN, and 32 mM succinate as reduction of 2,6-dichloroindophenol (absorbance at 600 nm). The sample was preincubated for 20 min at room temperature with 8 M dichloroindophenol, and then the reaction was started with 50 M dichloroindophenol and 100 M coenzyme Q 1 . Complex IV activity was measured as oxidation of reduced cytochrome c (at 550 nm) in buffer containing 2-5 g of protein/ml, 10 mM KH 2 PO 4 (pH 7.2), and 0.015% Triton X-100. Mitochondrial membranes were permeabilized by incubation for 5 min at 37°C, and the reaction was started with 0.04% (w/v) reduced cytochrome c.
In-gel Assay of OXPHOS Activities-Duplicate BN-polyacrylamide gels were used for immunoblotting and in-gel assays. For complex I activity (NADH dehydrogenase), the gel was incubated for 1 h at room temperature with 5 mM Tris-HCl (pH 7.4), 0.1 mg/ml NADH ϩ 2.5 mg/ml nitrotetrazolium blue (Sigma), then stained overnight at 4°C, and washed, and band intensity was quantified. For complex II, the gel was incubated for 1 h at room temperature with 5 mM Tris-HCl (pH 7.4), 0.25 mM phenazine methosulfate, 0.5 M succinate, and 2.5 mg/ml nitrotetrazolium blue and then washed, and band intensity was immediately quantified. For complex IV, the gel was incubated with 1 mg/ml 3,3Ј-diaminobenzidine, 50 mM phosphate buffer (pH 7.0), 75 mg/ml sucrose, 34 mg/ml catalase, and 10 mg/ml cyto-chrome c. For complex V, the gel was incubated for 3 h in 270 mM glycine, 35  Other Methods-Protein concentrations were determined by the BCA method (Pierce) with bovine serum albumin as standard.
Statistical Analysis-Data are means Ϯ S.D. or S.E. as indicated. Differences between means were evaluated by the twotailed Student's t test. p values Ͻ0.05 were considered significant.

RESULTS
Cells with Reduced mtPE-The requirement of mtPE for mitochondrial function was investigated in two cell models with reduced mtPE synthesis as follows: (i) PSB-2 cells, mutant CHO cells in which PS synthesis was only 5% of that in wildtype (WT) cells (chronically reduced mtPE synthesis) (63), and (ii) CHO cells in which PSD expression was silenced with DsiRNA (acutely reduced mtPE synthesis). Duplexes targeting distinct regions of hamster Pisd mRNA were transiently transfected into CHO cells to generate Pisd knockdown (KD) cells (Pisd KD#1, KD#2, and KD#3) in which Pisd transcript levels 48 h after transfection were 80 -94% lower than in cells transfected with nontargeting DsiRNA (NegCtrl) (Fig. 1A). Transfection with fluorophore-labeled DsiRNA (Cy5-KD#1) for 48 h yielded a signal in 80% of cells, consistent with 78% reduction of Pisd mRNA and confirming efficient transfection (Fig. 1A). In vitro PSD activity in Pisd KD#1 and KD#2 cells 48 h after transfection was 58 and 44% lower, respectively, than in NegCtrl cells (Fig. 1B) (Fig. 1F). Thus, PE production from PS was diminished in both PSB-2 cells (chronically reduced PS supply) and Pisd KD#1 cells (acutely reduced PSD activity).
Consistent with attenuated production of PE from PS in Pisd KD#1 and PSB-2 cells, the mass of mtPE was 21.5 and 27% lower than in NegCtrl cells and WT cells, respectively ( Fig. 2A), with no significant change in mitochondrial PS (Fig. 2B). The amount of mtPE as percent of total mitochondrial phospholipids was 14 and 24% lower in Pisd KD#1 and PSB-2 cells, respec-tively, than in control cells (Fig. 2C). Interestingly, the PS content of MAM from PSB-2 cells was 50% lower than in WT MAM with no reduction in PE (Fig. 2D). The PE content of cell lysates and microsomes was not reduced by mtPE deficiency; PSB-2 and WT cell lysates contained 68.3 and 71.4 nmol of PE/mg of protein, respectively, and lysates of Pisd KD#1 and NegCtrl cells contained 69.8 and 71.4 nmol of PE/mg of protein, respectively. Amounts of PS and phosphatidylcholine were not lower in lysates or microsomes of Pisd KD#1 cells than in Neg-Ctrl cells, nor was phosphatidylcholine reduced in PSB-2 lysates or microsomes (data not shown). Thus, decreased mtPE production reduced the PE content of mitochondria but not microsomes or cell lysates. Importantly, the amount of cardiolipin (CL), a mitochondrion-specific phospholipid required for normal mitochondrial functions (52,64,65), was not reduced in either PSB-2 or Pisd KD#1 cells (Fig. 2E).
Mitochondrial biogenesis was assessed by quantification of mtDNA. The mtDNA content was equivalent in Pisd KD#1 and NegCtrl cells 72 h after transfection, whereas PSB-2 cells contained 41% less mtDNA than did WT cells (Fig. 3D). Thus, chronic, but not acute, mtPE depletion decreases mitochondrial biogenesis. In addition, as an indicator of mitochondrial mass, cells were stained with 10-N-nonylacridine orange, a mitochondrial membrane potential-insensitive dye. According to this indicator, mitochondrial mass was similar in PSB-2 and WT cells and was also similar in Pisd KD#1 and NegCtrl cells both 48 and 72 h after transfection (Fig. 4, A and B).
Mitochondrial membrane potential, a key indicator of mitochondrial function and cell viability, was assessed with the membrane potential-sensitive dye TMRM. Mitochondria in PSB-2 and Pisd KD#1 cells (48 and 72 h after transfection) were significantly hyperpolarized (increased fluorescence) compared with control mitochondria (Fig. 4, C and D). Thus, mitochondrial membrane potential was increased, not decreased, by mtPE deficiency.

Cell Size and Mitochondrial Morphology of mtPE-deficient
Cells-The average area of Pisd KD#1 cells was 22% larger than that of NegCtrl cells (337 m 2 versus 278 m 2 , p Ͻ 0.05) (Fig. 5,  A and B). Fig. 5B shows enlarged Pisd KD cells in a population of cells of different sizes. Approximately 18% of Pisd KD cells were 3-4-fold larger than NegCtrl cells and contained enlarged (Fig.  5B, upper panel) and/or multiple (Fig. 5B, lower panel) nuclei. Moreover, PSB-2 cells were markedly more elongated than WT cells (Figs. 4C and 5C).
The balance between fusion and fission of mitochondria regulates mitochondrial morphology and function (66,67). Because mitochondria in Pisd Ϫ/Ϫ mouse embryos are highly fragmented (1), we hypothesized that mtPE deficiency would increase mitochondrial fragmentation. Confocal microscopy using MitoTracker Red revealed that filamentous mitochon-dria were present in Ͼ95% of WT and NegCtrl cells (Fig. 5, A and C), whereas Ͼ80% of Pisd KD#1 cells contained many small, fragmented mitochondria (Fig. 5, A and B). No tubular mitochondria were detected in Ͼ95% of PSB-2 cells (Fig. 5C). These observations indicate that reduction in mtPE induces mitochondrial fragmentation. Mitochondrial fusion is promoted by two dynamin-related, outer mitochondrial membrane proteins, mitofusin 1 and 2 (Mfn1 and Mfn2) (66,68). To determine whether reduction in the mitofusins could have contributed to mitochondrial fragmentation, we compared amounts of Mfn1 and Mfn2 by immunoblotting of mitochondria from mtPE-deficient cells and control cells. Mfn1 and Mfn2 levels were not lower in Pisd KD#1 cells than NegCtrl cells, and Mfn1 was not reduced in PSB-2 cells (Fig. 6A). The small (22%) increase in Mfn2 in PSB-2 cells is consistent with greater immunofluorescence staining for Mfn2 in PSB-2 cells, compared with WT cells (data not shown). Opa1 (optic atrophy protein 1) is another protein required for mitochondrial fusion (69 -71). The amount of Opa1 in Pisd KD#1 mitochondria was the same as in NegCtrl cells. Thus, increased mitochondrial fragmentation in Pisd KD#1 cells did not correlate with reduced expression of Mfn1, Mfn2, or Opa1. However, PSB-2 cells contained 25% less Opa1 than did WT cells (Fig. 6A) possibly contributing to increased mitochondrial fragmentation.
Mitochondrial morphology is also regulated by mitochondrial fission. Recruitment of the dynamin-related protein 1 (Drp1) to mitochondria enhances mitochondrial fission (72,73). Confocal immunofluorescence microscopy showed no significant difference in co-localization of Drp1 with mitochondria between Pisd KD#1 and NegCtrl cells (Fig. 6, B and D), whereas more Drp1 co-localized with mitochondria in PSB-2 than in WT cells (Fig. 6, C and D). Thus, increased Drp1-mediated fission might contribute to increased mitochondrial fragmentation in PSB-2 cells.  Mitochondrial morphology and ultrastructure were also examined by electron microscopy. Mitochondria in NegCtrl and WT cells contained intact cristae and had a normal elongated shape (Fig. 7, A and B) with shape factors (ratio of major and minor axes) of 2.7 Ϯ 0.1 and 3.2 Ϯ 0.2, respectively (Fig.  7C). In contrast, mitochondria in Pisd KD#1 and PSB-2 cells were more rounded (Fig. 7, A and B) with reduced shape factors (1.9 Ϯ 0.1 and 2.0 Ϯ 0.1, respectively) (Fig. 7C). Correspondingly, mean mitochondrial circularity was significantly higher in Pisd KD#1 and PSB-2 cells than control cells (Fig. 7C). Abnormal, swollen mitochondria were abundant in Pisd KD#1 and PSB-2 cells but not in NegCtrl or WT cells (Fig. 7, A and B), and mitochondrial matrix density in mtPE-deficient cells was less than in control cells (Fig. 7, A and B). The abundance of small, rounded mitochondria in Pisd KD#1 cells was consistent with an ϳ50% reduction in mean area/mitochondrion compared with that in NegCtrl cells (data not shown). Opa1 deficiency is associated with cristae disorganization (69,74,75) so that the small reduction in Opa1 in PSB-2 mitochondria (Fig.  6A) is consistent with the higher proportion of PSB-2 mitochondria lacking distinct cristae (Fig. 7). Thus, even a modest reduction of mtPE in Pisd KD#1 and PSB-2 cells profoundly alters mitochondrial morphology and ultrastructure.
MtPE Deficiency Decreases Mitochondrial Respiration and ATP Synthesis-Mitochondria generate ATP by OXPHOS coupled to O 2 consumption via the ETC. In intact PSB-2 cells, basal respiration (O 2 consumption from endogenous substrates) was 33% lower than in WT cells, and maximal O 2 consumption (in the presence of an uncoupling agent) was similarly reduced (Fig. 8A). In Pisd KD#1 cells maximum, but not basal, O 2 consumption was significantly lower than in NegCtrl cells (Fig. 8B). Rotenone (a CI inhibitor) decreased oxygen consumption in both mtPE-deficient cell lines by 80% (data not shown) con-  firming that CHO cells rely primarily on CI for basal respiration. Furthermore, the rate of substrate-driven respiration through ETC CI and CII (from pyruvate/malate or succinate, respectively) was markedly attenuated in digitonin-permeabilized PSB-2 cells compared with WT cells, and respiration with CI, but not CII, was lower in Pisd KD#1 than in NegCtrl cells (Fig. 8, C and D). Oxygen consumption via CIV from an artificial electron donor was not reduced by mtPE deficiency (Fig. 8,   C and D). These observations indicate that mtPE deficiency inhibits respiration and ETC activity.
In PSB-2 cells, cellular ATP was Ͼ2-fold higher than in WT cells (Fig. 9A), whereas in Pisd KD#1 cells (after 72 h but not 48 h) ATP was reduced by 80% (Fig. 9A). The reason for the divergent responses of ATP levels to acute and chronic mtPE deficiency is unclear, but long term compensatory mechanisms might have been induced in PSB-2 cells. Because the steady- state level of ATP reflects a balance between ATP synthesis and degradation, we measured the rate of ATP synthesis driven by respiratory substrates. In Pisd KD#1 and PSB-2 cells, the rate of ATP production via CI was 33 and 53% lower, respectively, than in control cells (Fig. 9B) and was 40 and 46% lower via CII (Fig.  9B). Furthermore, ATP production through CIV, from the artificial electron donor N,N,NЈ,NЈ-tetramethyl-p-phenylenediamine in the presence of ascorbate, was also lower in PSB-2 cells   than in control cells (Fig. 9B). Thus, multiple ETC components are dysfunctional in mtPE-deficient cells, leading to diminished ATP production.
MtPE Deficiency Inhibits ETC Activity and Supercomplex Formation-As additional evidence that ETC activity is compromised by mtPE deficiency, we measured enzymatic activities of individual ETC components. The specific activities of CI (NADH dehydrogenase) and CIV (cytochrome c oxidase) were significantly lower in mitochondria from both mtPE-deficient cell lines than in control cells (Fig. 10, A and B). In addition, CII activity (succinate dehydrogenase) was 45% lower in PSB-2 cells than in WT cells (Fig. 10B). Consistent with these observations, immunoblotting of Percoll-purified mitochondria showed that Pisd KD#1, KD#2, and PSB-2 cells contained less CIV protein than did NegCtrl cells (Fig. 11). Moreover, amounts of CI and CII were significantly reduced in PSB-2 cells but not in Pisd KD#1 cells (Fig. 11). However, the amount of CV (ATPase) was not changed by mtPE deficiency (Fig. 11). Although we detected CIII in rat heart mitochondria by immunoblotting, we did not detect CIII in CHO cells using the same antibodies. Consequently, we could not assess amounts of CIII. These data show that the abundance and activities of multiple ETC proteins, but not CV, were reduced by mtPE deficiency.
Decreased levels of ETC proteins can be associated with impaired assembly and/or stability of ETC supercomplexes. In mammals, these supercomplexes contain CI monomers, CIII dimers, and up to four copies of CIV; association with CIV enhances the activity of CI and CIII (76). We assessed amounts of mitochondrial ETC supercomplexes by BN-PAGE analysis (76). Location of complexes on the gel was established by immunoblotting with a mixture of CI-CV antibodies (Fig. 12A). Identity of CI, CII, CIV, and CV was confirmed by in-gel assays (Fig. 12, B-D). CI is present in a lower molecular mass complex in CHO mitochondria (750 -800 kDa, Fig. 12A) than in rat/ bovine heart mitochondria (ϳ1000 kDa) (77) but is similar in size to rat muscle CI (78). The strong in-gel activity of this band (I 1 in Fig. 12B) confirmed its presence as functional CI. Higher molecular mass assemblies of CI were observed only in control cells (arrow, Fig. 12B). CII was also identified by in-gel assays (data not shown) as a rapidly migrating band on BN-polyacryl-  amide gels (II 1 , Fig. 12A). PSB-2 cells contained less II 1 than WT cells, but the amount of II 1 was only slightly lower in Pisd KD#1 cells than NegCtrl cells (Fig 12A). Several CIV oligomers and complexes with dimeric CIII are present in rat heart mitochondria (77); CHO cells also contain CIV oligomers (IV n ) and complexes of CIV with CIII and CI (Fig. 12, A and C). Immunoblotting of CIV (data not shown) and in-gel assays (Fig. 12C) revealed fewer CIV oligomers and supercomplexes in Pisd KD#1 and PSB-2 cells than in control cells. In-gel assays showed that, as in rat and bovine heart (77), CV was present in an ϳ700-kDa complex corresponding to CV monomers (V 1 , Fig. 12D). Monomeric ATP synthase is active (77,79), whereas dimeric CV has been reported to be either active (80) or inactive (79); no CV dimers were detected in CHO cells. Mitochondria from Pisd KD#1 and NegCtrl cells had similar in-gel CV activity, but this activity was significantly lower in PSB-2 cells than WT cells (Fig. 12D). Thus, immunoblotting and in-gel assays indicate that mtPE plays an important role in formation and/or stability of multiple ETC proteins and supercomplexes.
Restoration of mtPE with Exogenous Lyso-PE-To confirm that mitochondrial defects in PSB-2 cells were specifically due to reduction of mtPE rather than reduction of PS (63), we supplemented the cells with 100 M lyso-PE, which in Saccharomyces cerevisiae is preferentially converted to mtPE (39,40). Lyso-PE supplementation of PSB-2 cells for 96 h increased mtPE by 45% (Fig. 13A) to approximately the WT level ( Fig. 2A) without significantly altering the amount of PS or other phospholipids in mitochondria or microsomes (Fig. 13A). The mitochondrial phospholipid composition of WT cells was unaltered by lyso-PE supplementation (data not shown). Lyso-PE supplementation increased the rate of cell growth and approximately doubled the number of PSB-2 cells (Fig. 13B), and it also increased the rate of ATP synthesis via CI and CII (Fig. 13C). Furthermore, the number of punctate mitochondria was reduced, and filamentous mitochondria were present in Ͼ80% of lyso-PE-supplemented PSB-2 cells (Fig. 13D). Thus, replenishment of mtPE in PSB-2 cells markedly improved the defects in cell growth, ATP production, and mitochondrial morphology, without normalizing the PS content of mitochondria or microsomes. We conclude that these mitochondrial functions critically depend on the level of mtPE.
In conclusion, cell growth, mitochondrial morphology, respiratory capacity, ATP production, and ETC activity are impaired in mammalian cells with chronic (PSB-2 cells) or acute (Pisd KD cells) mtPE deficiency. Importantly, mitochondrial defects in PSB-2 cells were largely reversed when mtPE content was normalized by lyso-PE supplementation.

DISCUSSION
Previous studies showed that in CHO cells (41) and yeast (42,43) most mtPE is synthesized in mitochondria from PSD and that import of PE synthesized from CDP-ethanolamine cannot maintain a normal complement of mtPE. Moreover, elimination of PSD in mice, in which PE synthesis from CDP-ethanolamine was active, was embryonic lethal establishing an absolute requirement for PSD in mice (1). However, Pisd ϩ/Ϫ mice that expressed 50% of normal Pisd mRNA and PSD activity were viable, and the mtPE content of tissues was normal (1,57). We therefore tested the hypothesis that reduction of mtPE synthesis by Ͼ50% in CHO cells would decrease mtPE content and impair mitochondrial morphology and function.
mtPE Deficiency Reduces Cell Growth-In PSB-2 cells (chronically reduced mtPE) and Pisd KD cells (acutely reduced mtPE), mtPE mass was reduced by 22-27% with no reduction in CL, a lipid required for normal mitochondrial function. Although mtPE was only modestly reduced, the growth of PSB-2 and Pisd KD cells was compromised, and attenuation of growth of Pisd KD cells correlated with the extent of Pisd The following complexes and supercomplexes were identified: CI monomer (I 1 ); CI associated with CIII and/or CIV (I 1 III 2 ; I 1 IV n ; I 1 III 2 IV n ); CIV monomer or multimer and associated with CIII (III 2 IV 1 ) and CI; supercomplexes of I 1 III 2 IV n , with monomeric CI (I 1 ), dimeric CIII (III 2 ), and n copies of CIV (IV n , n ϭ 1-4); CV monomer (V 1 ). CII was identified as a complex (II 1 ) by immunoblotting and in-gel activity (data not shown). Data are representative of 3-4 independent experiments with similar results.
knockdown. Because PE was reduced in mitochondria but not lysates, we conclude that the mitochondrial abnormalities and growth defects in the mtPE-deficient cells are caused by depletion of PE specifically in mitochondria. Interestingly, extra-mitochondrial PE levels were normal but did not compensate for decreased mtPE synthesis and did not normalize mtPE content. These observations are consistent with the embryonic lethality of Pisd knock-out mice in which PE synthesis from CDP-ethanolamine was normal or increased (1).
Reduction of mtPE and the ETC-Respiratory capacity, ETC enzymatic activities, and ATP synthesis are compromised in cells with reduced mtPE content. The rate of ATP synthesis was attenuated in mtPE-deficient cells despite normal amount and enzymatic activity of CV (ATP synthase) probably because the assembly/stability of several supercomplexes, including CI and CIV, were reduced. We could not discriminate between altered formation and stability of these complexes using BN-PAGE because Coomassie Blue disrupts protein interactions. However, because the amounts of CI and CIV in purified mitochondria were reduced, integration of supercomplexes into the mitochondrial inner membrane was probably impaired. We therefore attribute the decreased activities of CI and CIV primarily to reduced amounts of normally active CI and CIV rather than decreased intrinsic activity of these complexes. Similarly, the reduced activity of CII in PSB-2 cells appears to be due to lower abundance of normally active CII. Consistent with these conclusions, x-ray crystallography revealed that PE is tightly associated with bovine heart mitochondrial CIV (81) and that CL and PE are bound to yeast CIII (8). In addition, the activity of CI depends on bound PE (82). Moreover, inclusion of PE in proteoliposomes reconstituted with CIII increased the respiratory control ratio (83). Thus, a threshold level of mtPE appears to be crucial for formation and/or integration of supercomplexes into mitochondrial membranes.
Inhibition of the ETC often correlates with decreased mitochondrial membrane potential. However, although OXPHOS activity and ATP production were decreased in Pisd KD#1 cells, the mitochondria were more highly polarized than those of NegCtrl cells, according to TMRM staining. The enhanced membrane potential was not due to increased mitochondrial mass because nonylacridine orange staining and quantification of mtDNA indicated that mitochondrial mass and biogenesis were not increased in Pisd KD#1 or PSB-2 cells. These data suggest that ETC function is uncoupled from ATP synthesis in Pisd KD#1 and PSB-2 cells probably because of decreased phosphorylation rather than increased proton leakage. Thus, protons would accumulate in the inter-membrane space and membrane potential would rise. The idea that mtPE deficiency uncouples ETC function and ATP synthesis is consistent with the lower maximum rate of oxygen consumption in mtPE-de- pleted cells compared with control cells. Thus, organization of ETC complexes and ATP generation are both compromised, perhaps because of limited availability of ADP for ATP production or because CV is bound to an inhibitor (84).
Mitochondrial Morphology in mtPE-deficient Cells-Electron microscopy highlighted the grossly aberrant morphology of PE-deficient mitochondria. Furthermore, confocal fluorescence microscopy showed that mitochondrial fragmentation was greatly increased in mtPE-deficient cells compared with control cells, as was previously observed in Pisd Ϫ/Ϫ embryonic fibroblasts (1). In yeast, elimination of mtPE synthesis (85) or concomitant loss of synthesis of both CL and PE also promotes mitochondrial fragmentation (52). Mitochondrial morphology and function are closely linked, and mitochondrial dysfunction can induce autophagy/mitophagy and mitochondrial fragmentation (86 -88). Defects in mitochondrial fusion, ATP synthesis, and cell growth occur in mammalian cells lacking the mitochondrial fusion factors Mfn1 and Mfn2, and in cells with reduced amounts of the fusion factor Opa1 (66,85). It is unlikely, however, that the extensive mitochondrial fragmentation in Pisd KD#1 cells is primarily due to deficiency of fusion factors because amounts of Mfn1, Mfn2, and Opa1 are equivalent in Pisd KD#1 and NegCtrl cells, nor is it likely that the widespread mitochondrial fragmentation in PSB-2 cells is due to reduced amounts of Mfn1 or Mfn2 because Mfn1 is equally abundant in PSB-2 and WT cells, and Mfn2 is slightly increased. Mfn2 is required for tethering mitochondria to ER/MAM (48) and mediating PS import from the ER to mitochondrial PSD (12). Thus, the small increase in Mfn2 in PSB-2 cells might reflect induction of a long term compensatory mechanism for promoting PS import into the mitochondria, particularly because the PS content of MAM was reduced by ϳ50%. PSB-2 cells also contained less Opa1 than did WT cells, which might contribute to the increased mitochondrial fragmentation in PSB-2 cells.
Mitochondrial fusion is required for inheritance and maintenance of mtDNA (66, 89 -91). Thus, the reduction in mtDNA content in PSB-2 cells might be caused by defective mitochondrial fusion. Moreover, our finding that mitochondrial mass was not decreased in either model of mtPE deficiency, despite increased mitochondrial fragmentation, is consistent with the report that hyperpolarized mitochondria fail to be eliminated by autophagy (88,92).
When mitochondrial fusion is inhibited, unopposed fission can lead to increased mitochondrial fragmentation. Drp1-dependent mitochondrial fission can be enhanced by inhibition of either the ETC or the permeability transition pore (93). The increased co-localization of the fission protein Drp1 with mitochondria of PSB2, compared with WT, cells suggests that the enhanced fragmentation of PSB-2 mitochondria might be partially due to increased Drp1-mediated fission. However, because mitochondrial fragmentation was similarly increased in both PSB-2 cells and Pisd KD cells (in which Opa1 was not reduced and Drp1 was not increased), it is unlikely that the small changes in these factors in PSB-2 cells are primarily responsible for the increased mitochondrial fragmentation. We speculate that a change in physical properties of mitochondrial membrane, as a result of mtPE deficiency (a feature of both PSB-2 and Pisd KD cells), is more likely the primary cause of morphological changes in mitochondria. PE is a cone-shaped lipid that can generate nonlamellar membrane structures and induce negative membrane curvature and membrane fusion (85,94,95). Thus, mtPE deficiency might inhibit mitochondrial fusion (85). In general, defects in mitochondrial function and morphology were more pronounced in PSB-2 cells (chronic mtPE depletion) than in cells with acute reduction in mtPE (Pisd KD cells). In addition, some long term compensatory changes appear to have been induced in PSB-2 cells.
Reversal of mtPE Deficiency with Lyso-PE-To confirm that mitochondrial defects in PSB-2 cells could be ascribed specifically to reduction in mtPE, we supplemented PSB-2 cells with lyso-PE thereby restoring a normal mtPE content without increasing PE, or normalizing the PS content, of cell lysates or microsomes. Thus, lyso-PE is efficiently used for mtPE synthesis in mammalian cells as well as yeast (39). Normalization of mtPE in PSB-2 cells largely restored cell growth, increased ATP synthesis, and re-generated a filamentous mitochondrial network. In cell lysates, the content of neither PE nor PS content was increased by lyso-PE supplementation confirming that the mitochondrial defects in PSB-2 cells are caused specifically by mtPE depletion.
A limitation of the models used in these studies is that the sequence of events by which depletion of mtPE impairs mitochondrial properties remains unclear. A graded reduction in mtPE might establish the sequence in which mitochondrial targets (bioenergetics, cell growth, and morphology) are affected. Our results suggest, however, that altered mitochondrial morphology in mtPE-deficient cells is not a consequence of impaired mitochondrial function. These studies represent the first demonstration that attenuation of mtPE biosynthesis can reversibly and profoundly alter mitochondrial morphology and function in mammalian cells. Manipulation of the mtPE content could therefore be useful for controlling mitochondrial dynamics and bioenergetics. The findings raise the possibility that mitochondrial defects in some common human disorders might, at least in part, be attributable to mtPE deficiency.