The topology and regulation of cardiolipin biosynthesis and remodeling in yeast

Abstract

The signature mitochondrial phospholipid cardiolipin plays an important role in mitochondrial function, and alterations in cardiolipin metabolism are associated with human disease. Topologically, cardiolipin biosynthesis and remodeling is complex. Precursor phospholipids must be transported from the ER, across the mitochondrial outer membrane to the matrix-facing leaflet of the inner membrane, where cardiolipin biosynthesis commences. Post-synthesis, cardiolipin undergoes acyl chain remodeling, requiring additional trafficking steps, before it achieves its final distribution within both mitochondrial membranes. This process is regulated at several points via multiple independent mechanisms. Here, we review the regulation and topology of cardiolipin biosynthesis and remodeling in the yeast Saccharomyces cerevisiae. Although cardiolipin metabolism is more complicated in mammals, yeast have been an invaluable model for dissecting the steps required for this process.

1. Introduction

The unique phospholipid cardiolipin (CL) is required for the efficiency of a number of mitochondrial processes (Claypool and Koehler, 2012). CL is unique for a number of reasons: 1) unlike most other phospholipids which are synthesized in one or a few cellular locations, then disseminated throughout a cell's membranes, CL by and large remains in the mitochondrion, its site of synthesis; 2) CL is essentially a lipid dimer; it consists of two phosphate headgroups, which are attached by a glycerol moiety, and four acyl chains; and 3) after its synthesis, CL undergoes acyl chain remodeling, where acyl chains are removed by a lipase and replaced by a transacylase or acyltransferase, resulting in the establishment of only a few molecular forms of CL in a cell or tissue. Surprisingly, the acyl chain specificity of the lipase has never been demonstrated (Beranek et al., 2009), and the transacylase tafazzin has no acyl chain specificity (Schlame, 2012), although tafazzin from Drosophila has been shown to preferentially catalyze transacylation reactions on curved membranes leading to the establishment of CL with unsaturated acyl chains, which were proposed to decrease lipid disorder in areas of high curvature (Schlame et al., 2012a). Curiously, the final molecular form of CL varies between organisms and even between cell types within the same organism.

CL serves the cell in multiple capacities: it associates with all the major proteins of the mitochondrial respiratory chain and thereby increases the efficiency of electron flow and ADP/ATP exchange (Acehan et al., 2011; Bazan et al., 2013; Claypool et al., 2008b; Fry and Green, 1981; Jiang et al., 2000; Schagger et al., 1990; Schwall et al., 2012; Yu and Yu, 1980), modulates the catalytic activities and stability of interacting proteins (Claypool et al., 2008b; Gomez and Robinson, 1999; Jiang et al., 2000; Pfeiffer et al., 2003; Wenz et al., 2009), is critical for the biogenesis of mitochondrial proteins (Gebert et al., 2009; Jiang et al., 2000; Joshi et al., 2009), facilitates mitochondrial fission/fusion (Ban et al., 2010; DeVay et al., 2009; Joshi et al., 2012), and is involved in the maintenance and plenitude of cristae morphology (Acehan et al., 2009; Acehan et al., 2007; Mileykovskaya and Dowhan, 2009).

In addition to the importance of CL in promoting and maintaining normal mitochondrial function, alterations in CL metabolism have been associated with ischemia and reperfusion, heart failure, diabetic cardiomyopathy, and Barth syndrome (Chicco and Sparagna, 2007; Claypool et al., 2006; Gu et al., 2004; Paradies et al., 1997; Schlame and Ren, 2006). Barth syndrome is caused by mutations in tafazzin (TAZ1), and patients present with cardio- and skeletal myopathy, neutropenia, 3-methylglutaconic aciduria, and abnormal mitochondria (Barth et al., 1983; Schlame and Ren, 2006).

Much of the knowledge of CL biosynthesis and remodeling comes from studies in yeast. In addition to the “usual” advantages of using yeast as a model system (Baile and Claypool, 2013; Botstein and Fink, 2011), yeast are viable in the absence of CL and CL precursor phospholipids (Chang et al., 1998a; Chang et al., 1998b; Jiang et al., 1997; Osman et al., 2010; Tuller et al., 1998) whereas in higher eukaryotes CL is required for life (Zhang et al., 2011). Although CL biosynthesis and remodeling are highly conserved between yeast and higher eukaryotes, there are still a few differences. There are no orthologs of Gep4p, the phosphatidylglycerolphosphate (PGP) phosphatase, or Cld1p, a CL lipase, in higher eukaryotes (Beranek et al., 2009; Osman et al., 2010). However, the phylogenetically unrelated PTPMT1 performs the same function as Gep4p (Zhang et al., 2011); and a calcium-independent phospholipase A₂ has been implicated as a CL lipase (Malhotra et al., 2009; Mancuso et al., 2007; Schlame et al., 2012b), although its exact role in CL remodeling remains nebulous (Kiebish et al., 2013). Additionally, only the TAZ1-mediated CL remodeling pathway exists in yeast, while additional remodeling enzymes have been identified in mammals (reviewed in Claypool and Koehler, 2012). Thus, while yeast have been useful in dissecting this process, the complexity of and multitude of players in mammalian CL remodeling suggest that there is still much to discover.

With the recent characterizations of Cld1p, Gep4p and Tam41p (Beranek et al., 2009; Osman et al., 2010; Tamura et al., 2013), it is likely that all of the proteins catalyzing CL synthetic or remodeling reactions have been identified in yeast; however, many questions regarding the regulation of this process, as well as the topology and trafficking of CL and its precursors, remain (Fig. 1).

FIGURE 1. The topology of CL biosythesis and remodeling.

Phosphatidic acid (PA) is synthesized in the ER and translocates to mitochondria in a process that is influenced by the ERMES (ER-mitochondria encounter structure) complex. Ups1/Mdm35p heterodimers transport PA from the OM to the IM, potentially at contact sites (established by MINOS/MICOS/MitOS complexes). PA is converted to CDP-diacylglycerol (CDP-DAG) by Tam41p on the matrix-facing leaflet of the IM. CDP-DAG is used to generate phosphatidylglycerolphosphate (PGP) by Pgs1p. PGP is dephosphorylated to phosphatidylglycerol (PG) by Gep4p. PG and another CDP-DAG are condensed to form unremodeled CL by Crd1p. CL is deacylated by Cld1p on the matrix-facing leaflet of the IM, forming MLCL. Via an unknown mechanism, MLCL must flip to the IMS-facing leaflet of the IM or be transported to the OM to gain access to the transacylase Taz1p, which regenerates CL. Multiple rounds of deacylation/reacylation result in remodeled CL which is enriched in unsaturated acyl chains. CL achieves its final distribution on both leaflets of the IM and OM through currently ill-defined mechanisms. The depicted topology of Pgs1p has not been experimentally verified. Solid lines indicate known pathways. Dashed lines delineate potential but currently unknown phospholipid transport processes.

2. Delivering precursor phospholipids to the IM

CL biosynthesis requires CDP-diacylglycerol (CDP-DAG), which is formed from phosphatidic acid (PA) and CTP by a CDP-DAG synthase (Shen et al., 1996). Yeast contain two CDP-DAG synthases: Cds1p in the ER (Kuchler et al., 1986), and the recently characterized Tam41p in the mitochondrial inner membrane (IM; Tamura et al., 2013).

Although CDP-DAG (containing an NBD moiety) is able to be translocated from the ER to the IM in vitro, this process is inefficient (Tamura et al., 2013). The very low abundance of CL in Δtam41 yeast (Kutik et al., 2008; Tamura et al., 2012) suggests that if Cds1p-derived CDP-DAG contributes to CL biosynthesis, its role is very minor. Tam41p is peripherally associated with the matrix side of the IM (Table 1) (Gallas et al., 2006; Tamura et al., 2013). Thus, Tam41p activity requires that its substrate, PA, be transported from the ER to the matrix-facing leaflet of the IM. Phospholipid transport between the ER and mitochondrial outer membrane (OM) was suggested to be mediated by the ER-mitochondria encounter structure (ERMES) complex which physically tethers the two organelles (Kornmann et al., 2009). Indeed, loss of any ERMES complex subunit (Mdm10p, Mdm34, Mdm12p, or Mmm1p) alters the mitochondrial phospholipid profile, including reducing CL (Kornmann et al., 2009; Stroud et al., 2011; Tamura et al., 2012). However, its direct role in phospholipid transport has recently been challenged (Nguyen et al., 2012; Voss et al., 2012). Further, defects caused by the loss of a functional ERMES complex can be rescued by expressing an artificial ER-mitochondria tether, suggesting that the ERMES complex facilitates phospholipid transport by forming close contact sites between the two membranes, rather than directly transporting phospholipids (Kornmann et al., 2009; Nguyen et al., 2012; Voss et al., 2012). Notably, these studies focused on the transport of phosphatidylserine from the ER to mitochondria (and phosphatidylethanolamine to the ER after is decarboxylation in mitochondria). Thus, the mechanisms of PA and CDP-DAG transport from the ER to mitochondria, and the players involved, including a direct assessment of the role of the ERMES complex, remain to be discovered.

Table 1.

Topology of CL synthesis and remodeling enzymes

Protein	Location/membrane association	Predicted transmembrane domainsa	Biochemical experiments	References
Tam41p	Matrix leaflet of IM/peripheral	0	Protected from protease in mitoplasts, extracted with carbonate	(Tamura et al., 2006)
Pgs1p	Matrix leaflet of IM/peripheral	0, 1, 2b	N-terminal presequence imports LacZ to mitochondrial matrix	(Dzugasova et al., 1998)
Gep4p	Matrix leaflet of IM/peripheral	0	Protected from protease in mitoplasts; extracted with carbonate	(Osman et al., 2010)
Crd1p	Active site faces the matrix of the IM/integral	2, 3, 4, 5	Protected from protease in mitoplasts; blocking divalent cation entry into the matrix inhibits CL synthesisc	(Schlame and Haldar, 1993)
Cld1p	Matrix leaflet of IM/non-integral	0, 1, 2	Protected from protease in mitoplasts; partially extracted with carbonate; extracted with high salt concentrations	(Baile et al., 2013)
Taz1p	IMS-facing leaflet of IM and OM/non-integral	0, 1, 2	Degraded by protease in mitoplasts; partially extracted with carbonate; epitope tags throughout the polypeptide face the IMS	(Claypool et al., 2006)

^aTransmembrane predictions were determined using the DAS-TMfilter prediction server (Cserzo et al., 2004), TMpred (Hofmann and Stoffel, 1993), HMMTOP (Tusnady and Simon, 1998), TMHMM (Krogh et al., 2001), and SPLIT (Juretic et al., 2002)

^bMost programs predicted Pgs1p to have 0 transmembrane domains, except TMpred which predicted either 1 or 2 transmembrane segments

^cBiochemical experiments have not been performed on yeast Crd1p. The experiments here analyzed the rat Crd1p homolog

To reach to IM, CL precursor phospholipids must traverse the OM, but little is known about this process. Phospholipid exchange between leaflets of purified OM vesicles is rapid, suggesting that proteins mediate this process. However, treatment with proteases or with sulfhydrul reactive compounds does not inhibit transbilayer movement across the OM (Janssen et al., 1999).

PA is transported from the OM to the IM by the intermembrane space (IMS) resident protein, Ups1p (Connerth et al., 2012). Mdm35p binds Ups1p, facilitating its import into the IMS and preventing its proteolytic degradation (Potting et al., 2010; Tamura et al., 2010). Although Ups1p/Mdm35p dimers can bind negatively charged phospholipids, only PA is transported in vitro, demonstrating the specificity of its transport activity. Ups1p is unable to dissociate from membranes containing physiological levels of CL. Thus, the higher amount of CL in the IM is modeled to confer directionality of PA transport and the ability to limit CL accumulation (Connerth et al., 2012). Once delivered to the IM, PA must traverse to the matrix side of the IM. This could be accomplished by an unidentified protein or alternatively, PA may redistribute to both leaflets of the IM based on the transmembrane pH gradient (Gallet et al., 1999; Hope et al., 1989).

3. Synthesizing CL

The first committed step of CL biosynthesis is the formation of PGP from CDP-DAG and glycerol-3-phosphate by Pgs1p (Chang et al., 1998a). While the topology of Pgs1p has never been formally investigated (Table 1), the presence of an N-terminal presequence, which is able to import the lacZ gene product to the matrix (Dzugasova et al., 1998), suggests that Pgs1p is localized on the matrix side of the IM. PGP is then dephosphorylated to phosphatidylglycerol (PG) by Gep4p, a protein peripherally attached to the matrix side of the IM (Osman et al., 2010). In the final step of CL biosynthesis, PG and another CDP-DAG are condensed to form CL by Crd1p (Chang et al., 1998b; Jiang et al., 1997). Characterization of the rat Crd1p homolog from liver indicates that it is an integral membrane protein and that its active site faces the matrix (Gallet et al., 1997; Schlame and Haldar, 1993).

Exogenous inositol downregulates phosphatidylcholine and phosphatidylinositol biosynthesis through transcriptional repression via an inositol sensitive upstream activating sequence (UAS_INO) (Henry et al., 2012). Pgs1p activity is similarly reduced in the presence of inositol, but it contains a mutated, nonfunctional UAS_INO sequence (Bachhawat et al., 1995) and PGS1 mRNA levels are unchanged in the presence of inositol (Zhong and Greenberg, 2003). Further, deletion of the UAS_INO-binding genes INO2, INO4, or OPI1 does not affect Pgs1p activity (Greenberg et al., 1988), suggesting its inositol-mediated regulation is independent of the INO2-INO4-OPI1 circuit. Indeed, inositol increases Pgs1p phosphorylation, leading to its repressed activity, although the kinase(s) involved has yet to be identified (He and Greenberg, 2004). Independent from its inositol-mediated regulation, Pgs1p activity is increased under conditions indicative of mitochondrial biogenesis; its mRNA abundance is highest when cells enter stationary phase, and its activity is higher in the presence of non-fermentable carbon sources and when cells contain functional mtDNA (Gaynor et al., 1991; Shen and Dowhan, 1998; Zhong and Greenberg, 2003).

Crd1p activity is similarly increased during stationary growth, in the presence of mtDNA, and in the presence of non-fermentable carbon sources, leading to increased CL levels (Baile et al., 2013; Claypool et al., 2008a; Gaynor et al., 1991; Jakovcic et al., 1971; Jiang et al., 1999; Su and Dowhan, 2006) This is not surprising considering the importance of CL in a myriad of mitochondrial functions (Claypool and Koehler, 2012); as mitochondrial biogenesis increases, CL levels concurrently increase.

Crd1p activity can be additionally regulated by the matrix pH (Gohil et al., 2004). Treatment of yeast with the protonophore CCCP (which disrupts the pH and electrical gradients across the IM), but not the K⁺ ionophopre valinomycin (which disrupts the electrical gradient but does not affect the matrix pH), decreases Crd1p activity. A decrease in the matrix pH is indicative of less robust electron transport chain activity, coordinating the mitochondrion's energetic requirements with CL biosynthesis. This is further exemplified by the decreases in CL levels that result from defects in respiratory complexes and/or bioenergetic function (Gohil et al., 2004; Zhao et al., 1998).

Interestingly, while steady state CL levels were reduced in Δtaz1 yeast, synthesis of CL was actually increased in the mutant, concurrent with MLCL accumulation (Gu et al., 2004). These observations led the group to suggest that de novo CL biosynthesis might be regulated by downstream CL acylation/remodeling. Crd1p's activity might therefore be negatively regulated by its own product, and under conditions where CL is decreased or aberrantly acylated, the cell compensates by promoting CL biogenesis.

CL biosynthesis is thus regulated via multiple independent mechanisms: by inositol, which regulates Pgs1p; by mitochondrial biogenesis, which affects Pgs1p and Crd1p activity; by CL, which may inhibit Crd1p activity; and by the capacity for oxidative phosphorylation, which affects Crd1p.

4. Remodeling CL

CL remodeling is initiated by the lipase Cld1p in yeast (Beranek et al., 2009), which removes an acyl chain from CL, generating MLCL. Taz1p performs an acyl-CoA independent transacylation reaction, transferring an acyl chain from a phospholipid to a lysophospholipid (generated by a phospholipase), regenerating CL (Gu et al., 2004; Xu et al., 2003; Xu et al., 2006).

Cld1p is located on the matrix-facing leaflet of the IM and does not traverse the membrane (Baile et al., 2013). Surprisingly, the localization of Taz1p is not the same as enzymes upstream in the pathway. In yeast, Taz1p was originally localized to the mitochondrial OM (Brandner et al., 2005), but was later shown to be present on both the inner and outer membrane, on leaflets facing the IMS (Claypool et al., 2006; Gebert et al., 2009). Taz1p is an interfacial membrane protein; it contains residues that are embedded in, but not through, the membrane (Claypool 2006).

Cld1p and Taz1p are localized to different sides of the IM, and neither contains transmembrane domains, suggesting that an as yet unidentified protein(s) transports MLCL generated by Cld1p to the opposite side of the IM and/or to the OM. This trafficking of MLCL is expected to occur rapidly after CL deacylation as MLCL does not accumulate in yeast with a functional Taz1p (Baile et al., 2013; Gu et al., 2004). That both Cld1p and Taz1p assemble into higher-order complexes (Baile et al., 2013; Claypool et al., 2008a; Claypool et al., 2006), and that their binding partners have been, at best, partially defined, raises the exciting possibility that the protein(s) mediating MLCL translocation physically interacts with Cld1p, Taz1p, or both enzymes, although this has yet to be tested. While proteins mediating phospholipid redistribution between membrane leaflets have been identified for the plasma membrane, Golgi, and endosomes (van Meer et al., 2008), considerably less is known about this process in the mitochondrion. CL redistribution between IM leaflets has been observed (Gallet et al., 1997; Gallet et al., 1999), but the protein(s) responsible has not been identified. So far, phospholipid scramblase 3 (PLS3) is the only mitochondrial protein suggested to facilitate transbilayer lipid trafficking (Liu et al., 2003), but this has not been formally demonstrated. Further, deletion of its predicted yeast ortholog, AIM25, in yeast does not result in MLCL accumulation (Baile, M.G., Lu, Y., Claypool, S.M. unpublished data). Importantly, the translocation of phospholipids between membrane leaflets may not be facilitated by specific proteins, but instead non-specifically by the presence of numerous transmembrane proteins, as has been suggested for bacterial membranes and the ER (Kol et al., 2004; Kol et al., 2001; van Meer et al., 2008).

Similar to Crd1p, Cld1p activity is upregulated in the presence of non-fermentable carbon sources. Cld1p expression increases in lactate-containing media, and is repressed in dextrose-containing media (Baile et al., 2013), suggesting that CL remodeling is coordinately regulated with biosynthesis. Cld1p is also regulated by changes in the electrochemical gradient, but through a different mechanism and with a different functional outcome than Crd1p regulation. While reduction of the matrix pH decreases Crd1p activity (Gohil et al., 2004), Cld1p activity increases upon dissipation of the electrical potential (Baile et al., 2013). These differences may provide a mechanism by which CL biosynthesis and remodeling activity can be independently adjusted to fit the requirements of the mitochondrion.

Cld1p is the only protein in the CL remodeling pathway whose activity is known to be regulated (Taz1p expression increases when yeast are grown in the presence of non-fermentable carbon sources; Baile et al., 2013). Despite the fact that the spatial separation of Cld1p and Taz1p provides a potential point of regulation, MLCL levels remain unchanged and very low/absent unless Taz1p is non-functional (Baile et al., 2013), suggesting that the activity of the MLCL flippase and Taz1p is never limiting.

5. Establishing the final distribution of CL

CL is enriched in the IM, but is also present on the OM (Gebert et al., 2009). How it achieves its final distribution in yeast is still unclear. Intriguingly, the presence of a subpopulation of Taz1p on the OM opens up the possibility that MLCL may be the lipid species trafficked from the IM to the OM, where it is then reacylated to form CL (Claypool et al., 2006; Gebert et al., 2009).

Phospholipid transfer between the OM and IM has been suggested to occur at contact sites between the two membranes (Blok et al., 1971; Simbeni et al., 1990). Recently, the proteins comprising this complex (termed MINOS, mitochondrial inner membrane organizing system, MitOS, or MICOS) have been identified (Harner et al., 2011; Hoppins et al., 2011; von der Malsburg et al., 2011). Loss of this complex results in abnormal cristae morphology and loss of cristae junctions. However, the effects on phospholipid transport, the import of the CL precursor PA, or the final distribution of CL, have yet to be studied in contact site mutant yeast strains.

Currently, three proteins have been described that have the ability to traffic/redistribute CL in mammals: the mitochondrial creatine kinase (MtCK; Epand et al., 2007), the mitochondrial nucleoside diphosphate kinase (NDPK-D; Epand et al., 2007; Schlattner et al., 2013), and phospholipid scramblase 3 (PLS3; Liu et al., 2003). However, MtCK has no ortholog in yeast. Yeast also do not contain a mitochondrial-specific ortholog of NDPK-D, although a small portion of the yeast nucleoside diphosphate kinase, Ynk1p, localizes to mitochondria (Amutha and Pain, 2003). However, if Ynk1p can transport CL between membranes remains to be determined. PLS3 was shown to redistribute CL between the IM and OM in mammalian mitochondria (Liu et al., 2003), but phospholipid transport between membranes is inconsistent with its role as a scramblase. Thus, it is likely that PLS3 instead coordinates with a CL transport protein, and the altered distribution of CL between the OM and IM when PLS3 is overexpressed reflects the increased availability of CL on the IMS-facing leaflet of the IM.

Three UPS isoforms exist in yeast, although a concrete function has yet to be assigned to Ups2p or Ups3p. It is tempting to speculate that, like Ups1p and PA (Connerth et al., 2012), either of these proteins can transport CL (or MLCL) between the IM and OM. Interestingly, total CL levels in UPS2 and UPS3 mutants remain largely unaffected (Osman et al., 2009; Tamura et al., 2009), but the relative distribution of CL between the IM and OM has never been analyzed.

6. Perspectives

While most, if not all, of the enzymes involved in CL biosynthesis and remodeling have been identified (at least in yeast), many questions remain. How CL precursors are trafficked, and how CL achieves its final distribution, remains incompletely resolved. The ERMES complex is undoubtedly important for the trafficking of phospholipids between the ER and OM, but whether it directly transports phospholipids or simply mediates the apposition of the two membranes remains to be answered. OM/IM contact sites are potentially important for the movement of both CL precursors and CL itself between mitochondrial membranes, but this is currently an understudied aspect of CL metabolism. Additionally, the proteins mediating CL/MLCL movement, both between membranes and between leaflets of the same membrane, await identification. The regulation of CL precursor trafficking, except for the potential ability of CL to inhibit Ups1p/Mdm35p-mediated PA import, is unknown. Trafficking steps have the potential to regulate the flux of precursors through the CL pathway, but whether this is the case has yet to be determined. Further, compared to yeast (Fig. 2 summarizes modes of regulation identified in yeast), knowledge of these processes and their regulation in mammals is lacking.

FIGURE 2. Regulatory mechanisms of CL biosynthesis and remodeling.

The CL biosynthetic pathway is upregulated under conditions favoring mitochondrial biogenesis. In contrast, deficiencies in ERMES (ER-OM), MINOS/MitOS/MICOS (OM-IM contact sites) complexes, and components of the electron transport chain, as well as increased levels of inositol and reduced matrix pH, can all lead to a down-regulation of CL biosynthesis. Additionally, CL levels can be modulated by Ups1p/Mdm35p-mediated PA transport. Similar to CL biosynthesis, growth of yeast on respiratory media also promotes CL remodeling by upregulating the activity and/or expression of enzymes in the remodeling pathway. Distinct from CL biosynthesis, dissipation of the electrical potential across the IM, indicative of reduced OXPHOS capacity, increases Cld1p activity. Green boxes indicate conditions that promote CL biosynthesis and remodeling while red boxes indicate conditions that repress CL biosynthesis.

Highlights.

Cardiolipin is required for numerous mitochondrial functions

Yeast provide an excellent model to dissect cardiolipin metabolism

The topology of cardiolipin biosynthesis and remodeling is complex

Cardiolipin metabolism is regulated by multiple independent mechanisms

ABBREVIATIONS

CCCP

carbonyl cyanide 3-chlorophenylhydrazone

CDP-DAG

CDP-diacylglycerol

CL

cardiolipin

ERMES

ER-mitochondrial encounter structure

IM

inner membrane

IMS

intermembrane space

OM

outer membrane

PA

phosphatidic acid

PG

phosphatidylglycerol

PGP

phosphatidylglycerolphosphate

UAS_INO

inositol sensitive upstream activating sequence