Intramitochondrial Transport of Phosphatidic Acid in Yeast by a Lipid Transfer Protein

See allHide authors and affiliations

Science  09 Nov 2012:
Vol. 338, Issue 6108, pp. 815-818
DOI: 10.1126/science.1225625


Mitochondria are dynamic organelles whose function depends on intramitochondrial phospholipid synthesis and the supply of membrane lipids from the endoplasmic reticulum. How phospholipids are transported to and in-between mitochondrial membranes remained unclear. We identified Ups1, a yeast member of a conserved family of intermembrane space proteins, as a lipid transfer protein that can shuttle phosphatidic acid between mitochondrial membranes. Lipid transfer required the dynamic assembly of Ups1 with Mdm35 and allowed conversion of phosphatidic acid to cardiolipin in the inner membrane. High cardiolipin concentrations prevented membrane dissociation of Ups1, leading to its proteolysis and inhibiting transport of phosphatidic acid and cardiolipin synthesis. Thus, intramitochondrial lipid trafficking may involve a regulatory feedback mechanism that limits the accumulation of cardiolipin in mitochondria.

Each cellular membrane has a characteristic lipid composition that is required for its function (1, 2). Phospholipids are synthesized predominantly in the endoplasmic reticulum (ER) and must be redistributed to all cellular membranes (3). Extensive exchange of phospholipids between the ER and mitochondria is important in the synthesis of specific lipids (4, 5). These include cardiolipin (CL), a signature phospholipid of mitochondrial membranes, which is required for mitochondrial function and morphogenesis (5). CL is synthesized along an enzymatic cascade in the mitochondrial inner membrane (IM) from phosphatidic acid (PA) that is imported from the ER (Fig. 1A). However, how phospholipids shuttle from the ER to mitochondria, across the outer membrane (OM) to the IM and back, is not understood.

Fig. 1

CL synthesis in Δups1 mitochondria. (A) Enzymatic cascade mediating CL synthesis. The function of Tam41 at early stages of CL synthesis is unclear. MLCL, monolysocardiolipin; LPC, lysophosphatidylcholine; G3P, glycerol-3-phosphate; FA, fatty acid; CMP, cytidine 5′-monophosphate; CTP, cytidine 5′-triphosphate; Pi, inorganic phosphate; PPi, inorganic pyrophosphate. Other abbreviations are used as in the text. (B) Impaired CL synthesis in Δups1 mitochondria. Incorporation of 32P into CL, PE, or PC was monitored at the indicated times in wild-type (wt), Δups1, and Δcrd1 cells. 32P incorporation in wt was set to 1. (C) Phospholipidome of sucrose gradient–purified wt and Δups1 mitochondria determined by MS. Data represent mean values ± SD, n = 4: *P < 0.05, **P < 0.005 (unpaired t test, two-tailed).

In yeast, Ups1 and Ups2, members of the conserved MSF1′/PRELI protein family localized in the mitochondrial intermembrane space (IMS), are required for the mitochondrial accumulation of CL and phosphatidylethanolamine (PE), respectively (6, 7). Incorporation of 32P into CL was impaired in Δups1 mitochondria and completely inhibited in the absence of the cardiolipin synthase Crd1 (Fig. 1B). To define the role of Ups1 in CL synthesis, we determined the phospholipidome of Δups1 mitochondria by quantitative mass spectrometry (MS) (Fig. 1C). CL was reduced in Δups1 mitochondria, whereas other main phospholipids remained unaffected (Fig. 1C and fig. S1) (6, 8). The low-abundance precursor phospholipid PA accumulated in these mitochondria (Fig. 1C), suggesting that Ups1 acts early during CL synthesis.

To substantiate this conclusion, we performed a genetic epistasis analysis of UPS1 with genes affecting early steps of CL synthesis such as PGS1 and TAM41. Pgs1 catalyzes the formation of phosphatidylglycerolphosphate (PGP) (9, 10), whereas Tam41 acts before Pgs1, exerting an unknown function (1113). Cell growth was severely impaired upon deletion of PGS1, TAM41, or both, but restored upon deletion of UPS1 (Fig. 2A and fig. S2). Ultrastructurally, mitochondria lacking Pgs1 contained extremely elongated cristae sheets, which remained connected to the inner boundary membrane and frequently formed IM septae or onionlike structures (Fig. 2B). Deletion of UPS1 did not affect the mitochondrial ultrastructure but restored mitochondrial cristae morphology in Δpgs1 cells (Fig. 2B).

Fig. 2

Ups1 acts early during CL synthesis. (A) Deletion of UPS1 restored growth of Δpgs1 and Δtam41 cells. Serial dilutions of cell suspensions were spotted on YPD, YPGal, or YPG and incubated at 30°C. (B) Deletion of UPS1 restored mitochondrial ultrastructure in Δpgs1 cells. Electron micrographs of mitochondria in cells grown on YPGal are shown. Scale bar, 200 nm. (C) Deletion of UPS1 in Δpgs1 and Δtam41 cells restored CL and PA levels. Mitochondrial phospholipidome of indicated strains determined by MS (see also fig. S3). Data represent mean values ± SD, n = 3 (wt and Δups1, n = 4).

We further determined by MS the mitochondrial phospholipidome of cells lacking Ups1 in combination with Pgs1 or Tam41 (Fig. 2C and fig. S3). CL was reduced or absent in mitochondria lacking either protein, but PA, phosphatidylinositol (PI), and cytidine 5′-diphosphate-diacylglycerol (CDP-DAG) accumulated in mitochondria lacking Pgs1 (Fig. 2C and fig. S3). Thus, the different membrane lipid composition rather than merely the absence of CL affects cristae morphology in Δpgs1 mitochondria. Deletion of TAM41 resulted in accumulation of PA but not CDP-DAG, suggesting that Tam41 affects CDP-DAG synthesis (Fig. 2C). The phospholipid composition of Pgs1- and Tam41-deficient mitochondria lacking Ups1 was similar to that of Δups1 mitochondria. Thus, Ups1 is epistatic to Pgs1 and Tam41.

The requirement of Ups1 in the IM for early steps of CL synthesis suggests that Ups1 may facilitate the transport of PA to the IM (Fig. 1A). To define the molecular function of Ups1, we carried out in vitro experiments with purified Ups1. Ups1 assembles with Mdm35, which ensures its accumulation in the IMS and protects Ups1 against degradation by Yme1 (14). Consistently, Ups1 was prone to aggregation upon expression in Escherichia coli and recovered in the soluble fraction only when coexpressed with Mdm35 (fig. S4). This allowed purification of heterodimeric Ups1-Mdm35 complexes to homogeneity (fig. S5).

To examine whether Ups1-Mdm35 binds phospholipids, we performed flotation experiments using liposomes composed of phosphatidylcholine (PC) (80%) and another phospholipid (20%) (Fig. 3A). Ups1 exclusively bound to liposomes containing negatively charged phospholipids such as CL, PA, phosphatidylglycerol (PG), phosphatidylserine (PS), PI, or CDP-DAG but did not interact with PE or PC (Fig. 3, A and B). Purified Mdm35 did not bind any phospholipid tested (fig. S6), suggesting liposome binding via Ups1. The low amount of Mdm35 recovered with floated membranes in the presence of Ups1 indicated that liposome binding destabilizes Ups1-Mdm35 complexes (Fig. 3B).

Fig. 3

Lipid transfer by Ups1-Mdm35 complexes in vitro. (A) Phospholipid binding. Purified Ups1-Mdm35 complexes were incubated with liposomes composed of PC (80%) and the indicated phospholipid (20%), and binding was assessed by flotation of liposomes in a sucrose gradient. (B) Quantification of (A). (C) Lipid transport. Donor liposomes (50 μM; PC/PE/PI/CL/PA/PG/PS/CDP-DAG/NBD-PE = 40/24.6/10/5/5/5/5/5/0.4%) and acceptor liposomes (200 μM; PC/PE/PI/CL/Rhod-PE = 42.2/25.4/16.2/16.1/0.1%) were incubated with Ups1–Mdm35 complexes for 10 min. Acceptor liposomes were analyzed by MS. (D) Lipid specificity. Donor liposomes composed of PC (50%), PE (40%), and 10% of the indicated lipid were incubated with acceptor liposomes (PC/PE = 50/50%) and Ups1-Mdm35 complexes (66.7 nM). Acceptor liposomes were analyzed by MS. (E) PA transport. Donor liposomes (50 μM; PC/PE/PI/lactosyl-PE/PS/CL/PA/NBD-PE = 40/17.1/15/10/5/5/5/0.4%) containing 14C-PA or 14C-PC and acceptor liposomes (200 μM; PC/PE/PI/CL/Rhod-PE = 42.2/25.4/16.2/16.1/0.1%) were incubated for 10 min with the indicated proteins. Data represent mean values ± SD, n = 3.

Next, we investigated whether the Ups1-Mdm35 complex transfers phospholipids between liposomes in vitro. Donor liposomes, whose lipid composition resembles that of the OM, were incubated with Ups1-Mdm35 complexes and acceptor liposomes resembling the IM but lacking PA, PS, PG, or CDP-DAG. We isolated acceptor membranes after flotation and determined their phospholipid composition by MS (Fig. 3C and fig. S7). Ups1-Mdm35 complexes, but not Mdm35, facilitated PA transfer (Fig. 3C). To further define the transport specificity, we used PC-PE donor liposomes containing one negatively charged phospholipid only (Fig. 3D). Transfer of PA but not of PI, CDP-DAG, CL, PS, or PG occurred, substantiating the high selectivity of Ups1. Lipid transfer assays using 14C-PA and 14C-PC confirmed the protein-dependent and lipid-specific transport of PA by Ups1-Mdm35 complexes (Fig. 3E).

Because negatively charged phospholipids like CL, PG, or CDP-DAG were bound by Ups1 but not efficiently transported, we tested whether they influence PA transport when present in acceptor membranes (Fig. 4A). We observed slow PA transport in the absence of any negatively charged phospholipid. However, transport was accelerated when CL, PG, or CDP-DAG was present in the acceptor membrane (10%) (Fig. 4A), indicating that Ups1 binding is rate-limiting for transport under these conditions. Although binding to Ups1 with similar efficiency (Fig. 3A), the presence of PA in acceptor membranes stimulated the transport more than other negatively charged phospholipids (Fig. 4A).

Fig. 4

Characteristics of PA transfer by Ups1-Mdm35 complexes. (A) Negatively charged phospholipids in the acceptor membrane facilitated PA transport. Ups1-Mdm35 complexes (20 nM) were incubated with donor liposomes (25 μM; PC/PE/PA/lactosyl-PE/NBD-PE = 50/29.6/10/10/0.4%) containing 14C-PA and acceptor liposomes (100 μM) composed of PC and PE (50/40%) and the indicated phospholipid (10%). Data represent mean values ± SEM, n = 3. (B) Bidirectional transport of PA. Ups1-Mdm35 complexes (20 nM) were incubated with heavy liposomes [(50 μM; PC/PE/DOPA(di-18:1PA)/NBD-PE = 50/39.9/10/0.1%] and light liposomes [(50 μM; PC/PE/POPA(16:1/18:1PA)/Rhod-PE = 50/39.9/10/0.1%]. Liposomes were separated by flotation and lipids analyzed by MS. (C) CL in the acceptor membrane inhibits PA transport. Donor liposomes (25 μM; PC/PE/PA/lactosyl-PE/NBD-PE = 50/29.6/10/10/0.4%) containing 14C-PA and acceptor liposomes (100 μM; PC/PE/PA/Rhod-PE = 50/39.9/10/0.1%) containing increasing concentrations of CL (replacing PE) were incubated with Ups1-Mdm35 complexes (20 nM). Data represent mean values ± SEM, n = 3. (D) Ups1 remains bound to CL-rich liposomes. Ups1-Mdm35 complexes (20 nM) were incubated with liposomes (100 μM) of the indicated composition. Liposome-associated proteins were analyzed by SDS–polyacrylamide gel electrophoresis and immunoblotting.

We reasoned that Ups1-Mdm35 complexes mediated PA transport in a bidirectional manner, similar to that of other lipid transfer proteins (15, 16). Accordingly, PA in acceptor liposomes could facilitate membrane dissociation of Ups1 after PA delivery. To obtain evidence for bidirectional transport, we included PA with different acyl chains in donor 1,2-dioleoyl-sn-glycero-3-phosphate (DOPA) and acceptor 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphate (POPA) liposomes and monitored the redistribution of PA in the presence of Ups1-Mdm35 complexes by MS (Fig. 4B). DOPA accumulated in acceptor membranes at levels similar to those of POPA in donor membranes, suggesting that PA transport can occur bidirectionally and independent of its acyl chain composition (Fig. 4B).

The IM contains higher concentrations of CL than the OM (17). We thus performed lipid transfer assays at optimal PA and increasing CL concentrations in acceptor membranes (Fig. 4C). CL inhibited PA transfer when present in acceptor membranes at concentrations that mimic the IM (10 to 20%). Similarly, PG exerted an inhibitory effect on PA transport (fig. S8). At high CL concentrations, Ups1 remained associated with liposomes (Fig. 4D), indicating that CL impairs the dissociation of Ups1 from the membrane. Similarly, Ups1 accumulated at the IM in mitochondria that contain similar concentrations of CL but lack Yme1, which is responsible for its rapid turnover (14) (fig. S9). Thus, CL present at physiological concentrations traps Ups1 irreversibly at membranes, where it is degraded by Yme1, rendering PA transport irreversible.

Here, we identify Ups1 as a lipid transfer protein in the IMS that acts early during CL biosynthesis. Our in vitro results suggest that Ups1 mediates PA transport between mitochondrial membranes in distinct steps (fig. S10): Upon PA binding to Ups1 and PA extraction from the membrane, Ups1 assembles with Mdm35, which stabilizes Ups1 in a transfer-competent conformation. Negatively charged phospholipids facilitate the interaction of Ups1-Mdm35 complexes with the acceptor membrane, which is accompanied by the dissociation of Mdm35 and the release of PA. The enzymatic conversion of PA into CL in the IM provides directionality to the transport reaction.

Ups1 binds but does not transfer negatively charged phospholipids like CL, which is enriched at contact sites between the IM and OM (18). PA accumulates at these sites in Δups1 mitochondria (fig. S11). CL binding may thus recruit Ups1 to contact sites and facilitate PA transfer at sites of close membrane apposition in vivo, although direct membrane contacts are not required for Ups1-mediated PA transport in vitro (figs. S12 and S13). Moreover, high CL concentrations impair the dissociation of Ups1 from the IM and inhibit PA transfer, offering an intriguing possibility to limit CL accumulation in the IM.

Our epistasis analysis revealed an alternative route for CL synthesis that is activated in the absence of Ups1 and does not depend on PGP synthesis by Pgs1. These observations indicate that CL precursor lipids other than PA can reach the IM by other means. Although this pathway remains to be defined, the positive genetic interaction of Ups1 with Pgs1 and Tam41 suggests deleterious effects of PA accumulating in the IM in Δpgs1 and Δtam41 mitochondria.

The mechanism of PA transport by Ups1 is reminiscent of other known lipid transfer proteins that shuttle lipids between cellular membranes (2, 19). Although not related at the sequence level, structural modeling using template-based comparative modeling [i-TASSER (20)] suggests a fold for Ups1 similar to that of phosphatidylinositol transfer proteins (fig. S14). Ups1 is functionally conserved from yeast to human and a member of a conserved family of mitochondrial proteins (21). Other family members may serve as lipid transfer proteins with different lipid specificity within mitochondria.

Supplementary Materials

Materials and Methods

Figs. S1 to S16

Table S1

References (2231)

References and Notes

  1. Acknowledgments: We thank G. Zimmer for technical assistance, S. Geimer for advice on electron microscopy, T. Endo for sharing unpublished results, and J. Nunnari for yeast strains. This work was supported by grants from the Deutsche Forschungsgemeinschaft (SFB635, FOR885) and the European Research Council (AdG No. 233078) to T.L.

Stay Connected to Science

Navigate This Article