Mitochondrial phospholipid metabolism in health and disease

As major constituents of biological membranes, phospholipids define cellular and subcellular environments by providing physical barriers and influencing the structure and function of membrane proteins (Holthuis and Menon, 2014; Lee, 2004). In this Review, we will focus on phospholipids of mitochondria, double membrane-bound organelles that house many fundamental biochemical processes, including energy generation. Mitochondrial membranes are composed of all major classes of phospholipids – phosphatidylcholine (PC), phosphatidylethanolamine (PE), phosphatidylinositol (PI), phosphatidylserine (PS), phosphatidic acid (PA), phosphatidylglycerol (PG) and cardiolipin (CL) – which are either biosynthesized within or imported into mitochondria (Horvath and Daum, 2013). The importance of phospholipids in mitochondrial function and human health is evident from rare but devastating mitochondrial disorders, such as Barth syndrome, which are caused by mutations in phospholipid biosynthetic enzymes, (Lu and Claypool et al., 2015). Here, we describe recent advances in our understanding of mitochondrial phospholipid metabolism and their roles in the pathology of rare mitochondrial disorders.

Phospholipids are amphipathic molecules that consist of a glycerol backbone to which a hydrophilic headgroup and hydrophobic fatty acyl chains are attached (Fig. 1A). Variations in the head group determine the size, shape and charge of the phospholipid (Fig. 1B,C). There are also differences in the length and degree of unsaturation of the fatty acyl chains, which give rise to a staggering array of phospholipid species (Sud et al., 2007). The overall shape of each phospholipid molecule is dependent on the cross-sectional area of the head group and the acyl chains (Fig. 1B) (Basu Ball et al., 2018). For example, in the case of PC, the most abundant mitochondrial phospholipid, the area of the cross-section of the head group and the acyl chains is similar, resulting in an overall cylindrical shape (Fig. 1B). These cylindrically shaped molecules self-assemble into lamellar structures typically seen in phospholipid bilayers. When the cross-sectional area of the head group is smaller than their acyl chains, as in the case of PE, the second most abundant mitochondrial phospholipid (Fig. 1B), it results in a cone-shaped molecule that tends to form non-bilayer structures. The phospholipid composition and relative ratio of bilayer to non-bilayer phospholipids in mitochondrial membranes is evolutionarily conserved (Fig. 1C). The diversity of phospholipid species determines the biophysical and biochemical properties of membranes, which in turn influences the behavior of membrane proteins as described below.

The electrical charge on an individual phospholipid molecule is determined by the type of the head group (Fig. 1C), which in turn dictates the membrane surface charge (Fig. 1D). In the case of mitochondria, surface charge densities on the inner and outer leaflet of the inner mitochondrial membrane (IMM) are likely influenced by asymmetric distribution of phospholipids (Horvath and Daum, 2013) and differences in the pH of the mitochondrial matrix and the intermembrane space (IMS).

Membrane thickness is determined by the acyl chain length and its degree of saturation (Tillman and Cascio, 2003). Because of the high energetic cost of exposing the hydrophobic region of either the fatty acyl chain or the membrane protein to the aqueous environment, the thickness of the bilayer is expected to match the thickness of the hydrophobic regions of the membrane proteins (Lee, 2004) (Fig. 1E). This hydrophobic coupling can occur through distortion of the bilayer, the protein or both. The stability of the human voltage-dependent anion channel (VDAC2), an abundant outer mitochondrial membrane (OMM) protein, depends on membrane thickness provided by dipalmitoyl PC, highlighting the importance of hydrophobic coupling (Srivastava et al., 2018).

Membrane fluidity or viscosity refers to the freedom of movement of a lipid or protein within the bilayer (Tillman and Cascio, 2003). It strongly correlates with the level of unsaturated fatty acids; the higher the proportion of unsaturated fatty acids, the greater the fluidity (Fig. 1F). Mitochondrial membrane fluidity can impact the function of the respiratory chain. For example, isolated mitochondria from yeast mutants engineered to produce a greater amount of unsaturated fatty acids showed increased respiration (Budin et al., 2018). This increased respiration could be attributed to increased mobility of electron carriers, such as ubiquinone, between respiratory enzyme complexes.

When non-bilayer-forming phospholipids are forced to stay in the bilayer, they impart curvature stress, which can be released by positive or negative curvature of the membrane (Fig. 1G). In vitro studies and molecular dynamics simulations have shown that CL, a cone-shaped phospholipid that is almost exclusively found in mitochondrial membranes, has the propensity to accumulate in regions of membranes that exhibit high negative curvature (Beltran-Heredia et al., 2019; Boyd et al., 2017). In the IMM, which exhibits numerous tubular invaginations referred to as cristae, one would expect a higher proportion of CL. Indeed, CL is associated with oxidative phosphorylation complexes present in cristae (Cogliati et al., 2013).

The close packing of different phospholipids in the membrane results in mechanical frustrations, the pressure of which varies across the depth of the membrane and is referred to as lateral pressure profile (π) (van den Brink-van der Laan et al., 2004) (Fig. 1H). The lateral pressure profile depends on other biophysical properties of membranes and has been proposed to influence protein conformation and function (Marsh, 2007). For example, a new study has shown that reactive aldehydes produced by reactive oxygen species covalently modify the PE headgroup, reshaping PE and altering the lateral pressure profile of membranes. This in turn impacts the function of mitochondrial uncoupling protein 1 (UCP1) and adenine nucleotide translocator 1 (ANT1) (Jovanovic et al., 2022).

Although we have described each of the biophysical and biochemical properties of the phospholipids separately, it should be stressed that these properties together determine membrane function (Fig. 1I).

Mitochondria are semi-autonomous in phospholipid biosynthesis – they can biosynthesize PA, PG, CL and PE in situ (Fig. 2), but other phospholipids, such as PC, PS and PI must be imported. Many of the mitochondrial phospholipid biosynthetic enzymes are conserved from yeast to mammals (Table 1), highlighting the conserved nature of these pathways.

As the simplest phospholipid, PA serves as a substrate for the biosynthesis of other phospholipids. PA biosynthesis primarily occurs in the endoplasmic reticulum (ER). Within mitochondria, its biosynthesis occurs via three different pathways (Fig. 2). The first involves sequential acylation of glycerol-3-phosphate (G3P) by glycerophosphate acyltransferases (GPAT1 and GPAT2) and acylglycerophosphate acyltransferases (AGPAT4 and AGPAT5) on the OMM (Bradley et al., 2015; Lewin et al., 2004) (Fig. 2; Table 1). The second pathway involves phosphorylation of diacylglycerol by acylglycerol kinase (AGK) to form PA in the IMM (Bektas et al., 2005). The third pathway for PA biosynthesis is the hydrolysis of CL by MitoPLD (also known as PLD6) in the OMM (Choi et al., 2006).

PG biosynthesis occurs in the IMM from PA (Fig. 2). In the first step, PA is converted into cytidine diphosphate diacylglycerol (CDP-DAG) in a reaction catalyzed by phosphatidate cytidylyltransferase called TAM41 mitochondrial translocator assembly and maintenance homolog (TAMM41) (Blunsom et al., 2018). In the rate-limiting second step, phosphatidylglycerol phosphate synthase (PGS1) converts CDP-DAG into phosphatidylglycerol phosphate (PGP) through the transfer of the phosphatidyl group from CDP-DAG onto the G3P (Cao and Hatch, 1994). Finally, PGP is dephosphorylated to PG by a phosphatase called protein tyrosine phosphatase mitochondrial 1 (PTPMT1) (Zhang et al., 2011).

Cardiolipin is almost exclusively present in mitochondria. Nascent CL (CL_n) is biosynthesized via condensation of PG and CDP-DAG in a reaction catalyzed by cardiolipin synthase (CRLS1) (Houtkooper et al., 2006) (Fig. 2). CL_n, which is characterized by the presence of saturated acyl chains, undergoes remodeling by sequential deacylation and reacylation reactions to form mature CL (CL_m), which is enriched in unsaturated fatty acyl chains (Oemer et al., 2020). In yeast, the phospholipase Cld1 has been shown to preferentially hydrolyze saturated fatty acyl chains from CL_n to form monolyso-CL (MLCL) (Beranek et al., 2009). Analogous phospholipase(s) have not yet been identified in mammals (Table 1). MLCL is then re-acylated by the transacylase TAFAZZIN (Xu et al., 2003), where PC, and to a lesser extent PE, serve as an acyl chain donor (Xu et al., 2006). Apart from TAFAZZIN-mediated transacylation, CL remodeling can also occur via the acyl transferases ALCAT1 or MLCL AT1 (also known as HADHA) (Cao et al., 2004; Taylor and Hatch, 2009). The contribution of these alternative routes to CL remodeling is unclear, but the fact that TAFAZZIN mutations lead to drastic phenotypes in Barth syndrome (BTHS) individuals suggest that they cannot compensate for the loss of TAFAZZIN (Bione et al., 1996). Notably, the fatty acyl chain arrangement in CL is not stochastic. Cardiac mitochondria are mainly dominated by tetralineoyl CL (18:2)₄ species, whereas brain mitochondria contain a multiplicity of CL species. The various CL compositions in different tissues has been attributed to the levels of linoleoyl (18:2) and oleoyl (18:1) fatty acyl species in other phospholipids in those tissues (Oemer et al., 2020).

PS imported from the ER serves as a precursor for mitochondrial PE biosynthesis, which is carried out by the IMM-localized PS decarboxylase (PISD) (Hovius et al., 1992) (Fig. 2). Recently, the yeast PS decarboxylase (Psd1) was shown to localize to both the ER and mitochondria, where it contributes to the formation of functionally distinct pool of PE (Friedman et al., 2018). Notably, PE biosynthesized outside mitochondria can also be imported, as shown in yeast cells lacking Psd1 (Baker et al., 2016; Calzada et al., 2019). Importantly, rescue of the yeast psd1Δ respiratory growth phenotype requires stimulation of the non-mitochondrial pathway of PE biosynthesis via ethanolamine supplementation (Baker et al., 2016). In mammals, a lack of PISD results in embryonic lethality (Steenbergen et al., 2005), suggesting that, under basal conditions, non-mitochondrial PE cannot compensate for the lack of mitochondrial PE biosynthesis during embryonic development.

To achieve mitochondrial phospholipid homeostasis, both inter-organellar phospholipid trafficking between the ER and mitochondria, as well as intraorganellar phospholipid transport between the OMM and the IMM, is required (Fig. 3). Mitochondria must import PS, PC and PI, because biosynthetic enzymes for these phospholipids are not present in situ (Fig. 2). Conversely, PE biosynthesized within mitochondria is exported to other subcellular compartments, although the mechanism of its export remains to be determined (Fig. 3). Phospholipid transport to mitochondria presents a unique challenge because mitochondria are not connected with other subcellular organelles via the vesicular trafficking pathway, and thus cannot acquire phospholipids via vesicles. Therefore, alternative mechanisms involving specialized membrane contact site (MCS) complexes (Scorrano et al., 2019; Tamura et al., 2019) and lipid transport proteins (LTPs) (Wong et al., 2019) have been proposed. Although there are many parallels, differences exist between yeast and mammalian mitochondrial phospholipid trafficking proteins (Fig. 3). Given that many excellent review articles have described mitochondrial phospholipid trafficking pathways in great detail (Acoba et al., 2020; Tamura et al., 2014), here, we mainly focus on our current state of knowledge.

The ER is the major site of phospholipid biosynthesis in eukaryotes. Early microscopy and cell fractionation studies have shown that certain regions of ER membranes are in close association with mitochondria and these were named mitochondria-associated membranes (MAMs), which are important sites of lipid exchange between these organelles (Achleitner et al., 1999; Shiao et al., 1995). Once the molecular basis of MAMs was identified, they were commonly referred to as ER–mitochondrial membrane contact sites (ERMCS) (Wang et al., 2021). One of the first molecular tethers identified to hold these membranes together in yeast was the ER–mitochondrial encounter structure (ERMES) (Kornmann et al., 2009) (Fig. 3). Owing to the lipid-binding properties of subunits of ERMES, it was suggested to be an ideal candidate for lipid transport from the ER to mitochondria (AhYoung et al., 2015); indeed, several in vitro studies have supported the phospholipid transport activity of ERMES (Jeong et al., 2017; Kawano et al., 2018; Kojima et al., 2016). These studies have suggested a critical role of ERMES complex in PS transport from the ER to mitochondria. If true, loss of PS transport to mitochondria in ERMES mutants should prevent biosynthesis of mitochondrial PE via Psd1 (Fig. 2). However, PS to PE conversion is not abrogated in ERMES mutants and the phenotypes of ERMES mutants are not identical to yeast psd1Δ cells (Baker et al., 2016; Burgess et al., 1994; Nguyen et al., 2012; Tamura et al., 2012). To explain this conundrum, it was postulated that other proteins take over the function of ERMES in its absence (Lang et al., 2015). Indeed, multiple suppressors of ERMES phenotypes have been reported. For example, overexpression of mitochondrial proteins Mcp1 and Mcp2, vacuole fusion factor vacuolar protein sorting 39 (Vps39), or dominant mutations in an evolutionarily conserved LTP, Vps13, can rescue phenotypes of ERMES mutants (Honscher et al., 2014; Lang et al., 2015; Park et al., 2016; Tan et al., 2013). Importantly, combined loss of ERMES with Vps13 or Vps39 is synthetic lethal, suggesting that they have an overlapping function (Elbaz-Alon et al., 2014; Lang et al., 2015). More recently it has been shown in vivo that, ERMES could be involved in PC transport (John Peter et al., 2022). Thus, the in vivo role of ERMES is much more complex than originally thought (Kornmann, 2020).

More recently, molecular tethers that form ERMCS, analogous to the yeast ERMES complex, have been discovered in mammals (Fig. 3). Among the membrane-tethering protein complexes in mammals, PTPIP51, a mitochondrial protein, was shown to bind and transfer PA in vitro, and its deletion resulted in reduced mitochondrial CL levels in vivo (Yeo et al., 2021). The crystal structure of PTPIP51 revealed that its phospholipid-binding motif interacts with vesicle-associated membrane protein-associated protein B (VAPB), an ER membrane protein, to form a tether (Yeo et al., 2021). Recently, mitoguardin-2 (MIGA2), an OMM protein that forms an ERMCS in higher eukaryotes, was suggested to mediate lipid transport (Hong et al., 2022; Kim et al., 2022). Indeed, the crystal structure of zebrafish mitoguardin-2 (zMiga2; PDB ID: 7X15) revealed a hydrophobic cavity for lipid binding (Hong et al., 2022; Kim et al., 2022). The biochemical analysis of MIGA2 demonstrated that it can transfer various phospholipids with a ∼6–12 times preference for PS as compared to PC, PA, PE or ceramides (Kim et al., 2022).

Another mechanism of inter-organelle phospholipid transport is LTPs, which extract lipids from donor membrane and deliver them to acceptor membrane. The oxysterol receptor domain-containing proteins ORP5 and ORP8 are examples of LTPs (Maeda et al., 2013; Moser von Filseck et al., 2015) that are also localized at MAMs (Galmes et al., 2016) (Fig. 3). ORP5 and ORP8 cooperates with SAM50 (also known as SAMM50), a component of the mitochondrial intermembrane space bridging (MIB) complex, and Mic60 (also known as IMMT), a subunit of the mitochondrial cristae organizing structure (MICOS) complex, which anchors the IMM and OMM at the cristae junction and plays a role in the transport of PS from the ER to mitochondria (Monteiro-Cardoso et al., 2022). A direct comparison between Osh6, a yeast homolog of ORP5 and ORP8, and MIGA2 showed that MIGA2 largely contributes to PS transport at the ERMCS in mammalian cells (Kim et al., 2022). An evolutionarily conserved protein VPS13A was shown to localize at the MAMs and act as an LTP in transporting glycerophospholipids in vitro, though its specificity for a given phospholipid in vivo remains to be determined (Kumar et al., 2018). Transport of PC from the ER to the OMM and IMM via the STAR family protein StarD7 represents another example of LTP-mediated phospholipid transport (Horibata et al., 2020, 2017; Horibata and Sugimoto, 2010; Saita et al., 2018).

Recently, using a yeast genetic approach we have shown that Vps39, a subunit of vacuole mitochondria tethering complex (vCLAMP) and the homotypic fusion and protein sorting (HOPS) complex, is required to rescue PE-dependent mitochondrial functions in psd1Δ mutants (Iadarola et al., 2020, 2021). These findings implicate Vps39 in the transport of extramitochondrial PE to mitochondria (Fig. 3). Notably, vCLAMP and HOPS complexes have been previously shown to be involved in the transport of phospholipids and membrane fusion, respectively (Elbaz-Alon et al., 2014; Wurmser et al., 2000). However, the role of Vps39 in PE trafficking to mitochondria was independent of its previously described function as a part of the vCLAMP and HOPS complex (Elbaz-Alon et al., 2014; Honscher et al., 2014; Iadarola et al., 2020; Wurmser et al., 2000). It remains to be determined whether Vps39 acts as an LTP or indirectly regulates intracellular PE trafficking.

Despite this progress, we currently do not know how PI, PG and PE are transported between mitochondrial and ER membranes. Based on our current understanding of mitochondrial phospholipid trafficking, it is very likely that multiple mechanisms are at play in this process.

Phospholipids imported from the ER or biosynthesized on the OMM are transported to the IMM. The Ups (yeast) and PRELID (mammals) family of proteins play a crucial role in phospholipid transport across the IMS. Ups1 is an LTP that was shown to shuttle PA between mitochondrial membranes (Fig. 3) (Connerth et al., 2012; Potting et al., 2010). Loss of Ups1 leads to decreased CL and increased PA in mitochondria as expected (Connerth et al., 2012). During PA transport, Ups1 dynamically associates with Mdm35 and this interaction is evolutionarily conserved in human homologs of these proteins, PRELID1 and TRIAP1, which also assemble as a heterodimeric complex in the IMS (Connerth et al., 2012; Potting et al., 2013). The other members of the Ups/PRELID family, such as Ups2 and Ups3 in yeast, and PRELI2, SLMO1 and SLMO2 in mammals, have also been implicated in phospholipid transfer activities between the IMM and OMM, with the Ups2–Mdm35 complex (Aaltonen et al., 2016; Miyata et al., 2016) and its human homolog PRELID3b–TRIAP1 being involved in PS transport (Potting et al., 2010; Tamura et al., 2009).

Currently, we still do not understand the mechanism by which CL, PG, PI and PE are trafficked between the two mitochondrial membranes, possibly because we lack a robust in vivo technique to annotate specific phospholipids and link transport functions to certain molecular tethers or proteins. However, in the case of CL, a mitochondrial phospholipid scramblase (PLS3; also known as PLSCR3) and nucleoside diphosphate kinase D (NDPK-D; also known as NME4) have been suggested to promote transport of CL from the IMM to OMM in mammalian systems (Kagan et al., 2016; Liu et al., 2008).

Phospholipids can impact mitochondrial structure and function either through specific molecular interactions with membrane proteins or by modulating the bulk properties of membranes. Given that several comprehensive reviews are available on this topic (Acoba et al., 2020; Colina-Tenorio et al., 2020; Ji and Greenberg, 2022), here we focus on selected recent studies.

CL is essential for optimal mitochondrial bioenergetics (Liang et al., 2022) owing to its role in stabilizing IMM proteins (Musatov and Sedlak, 2017). One of the specialized functions of CL is to provide stability to the respiratory chain supercomplexes (RSCs) (Baker et al., 2016; Pfeiffer et al., 2003; Zhang et al., 2005). The relationship between CL and RSCs is reciprocal given that the increased function and expression of RSCs stimulates CL biosynthesis and remodeling (Gohil et al., 2004; Xu et al., 2019; Xu et al., 2021). CL has also been shown to be essential for the stability, assembly and activity of mitochondrial carrier proteins and ion channels. For example, CL is crucial for the conformation and the tertiary and quaternary assembly of ADP/ATP carrier (AAC2; also known as SLC25A5) protein (Senoo et al., 2020; Yi et al., 2022). Additionally, CL is essential for the assembly and activity of the mitochondrial Ca²⁺ uniporter (MCU) complex, which coordinates cellular energy demands with mitochondrial bioenergetics via Ca²⁺ signaling (Ghosh et al., 2020, 2022). Here, even a partial depletion of CL leads to reduced mitochondrial Ca²⁺ uptake, which disrupts the Ca²⁺-induced increase in mitochondrial respiration (Ghosh et al., 2020, 2022). Consistent with these genetic and bioenergetic studies, cryoEM structures of the MCU holo-complex have shown that MCU tetramers are stabilized by eight CL molecules (Zhuo et al., 2021). CL is also required for the activation of other mitochondrial proteins including pyruvate dehydrogenase (Li et al., 2019) and frataxin, a protein necessary for the biogenesis of mitochondrial iron-sulfur clusters (Li et al., 2020). Moreover, CL deficiency in a TAFAZZIN mutant HEK293 cell line caused an increased expression of an IMM protease, presenilin-associated rhomboid-like protein (PARL), which leads to defective mitochondrial quality control (Anzmann et al., 2021). Together, these studies have uncovered CL as a crucial regulator of the mitochondrial matrix, as well as IMM proteins.

PE is required for the activities of respiratory complexes III and IV in yeast and complexes I and IV in mammals (Baker et al., 2016; Tasseva et al., 2013). In mammals, exercise and exposure to cold temperature results in increased mitochondrial PE and CL levels, with elevated mitochondrial PE deemed sufficient to increase oxidative phosphorylation capacity without an increase in abundance of the respiratory chain enzymes (Heden et al., 2019; Johnson et al., 2023). Genetic ablation of mitochondrial PE biosynthesis lowered UCP1-dependent respiration in mice (Johnson et al., 2023). Similarly, CL biosynthesis is induced in brown and beige fat upon cold exposure of mice, and loss of CL biosynthetic enzymes in these tissues diminishes mitochondrial uncoupling and thermogenesis (Sustarsic et al., 2018). These data reinforce findings from the yeast system that first identified the specific requirement of PE and CL for mitochondrial bioenergetics (Baker et al., 2016).

The morphology of mitochondria in cells is dynamic; mitochondria can exist in a tubular network or a more fragmented state, which is mediated by two opposing processes – fusion and fission (Shaw and Nunnari, 2002). Two non-bilayer phospholipids, PA and CL, play a crucial role in the maintenance of mitochondrial dynamics by regulating the fusion and fission machinery (Kameoka et al., 2018). For example, PA induces fusion of the OMM via two dynamin-related GTPases, mitofusin 1 and 2 (Choi et al., 2006), and at the same time inhibits fission by binding to Drp1 (also known as DNM1L), the fission-inducing GTPase (Adachi et al., 2016). CL mediates IMM fusion through direct interaction with the dynamin-related GTPase Opa1 or its yeast homolog Mgm1 (Ban et al., 2017; Yan et al., 2020). PE has also been shown to promote mitochondrial fusion by regulating topogenesis of Mgm1 and, in this regard, has overlapping function with CL (Chan and McQuibban, 2012; Joshi et al., 2012). Recently, nuclear magnetic resonance-based studies have uncovered CL-binding motifs (CBMs) on the variable domain of Drp1 (Mahajan et al., 2021). The electrostatic association of the negatively charged head group of CL on the OMM with amino acid residues in one CBM, along with hydrophobic interactions of acyl chains of CL with the other CBM of Drp1, leads to structural rearrangements that promote membrane constriction and thus mitochondrial membrane fission (Mahajan et al., 2021).

Mitophagy is a type of autophagy where the cell selectively degrades damaged mitochondria (Killackey et al., 2020). Mitochondrial damage triggers the externalization of CL to the OMM, which recruits the proteins of the LC3 autophagy family to initiate mitophagy (Chu et al., 2013). Variability in the N-terminal region of the LC3 protein subfamily determines its specificity for CL, where LC3A (also known as MAP1LC3A) has high affinity and specificity for CL and participates in CL-mediated mitophagy (Iriondo et al., 2022). Interestingly, another member of the LC3 family, LC3C (MAP1LC3C), has the highest affinity for CL but does not participate in CL-mediated-mitophagy (Iriondo et al., 2022). Instead, it has been suggested that binding of LC3C to PA present on the OMM helps to maintain the integrity of the mitochondrial network under basal growth conditions by selective degradation of mitochondrial proteins (Iriondo et al., 2022; Le Guerroue et al., 2017).

Mitochondria import nearly 1000 proteins via dedicated protein import machineries such as the translocase of the outer membrane (TOM) complex, translocase of the inner membrane (TIM) complex and the sorting and assembly machinery (SAM) present on the OMM (Busch et al., 2023). Early studies uncovered the specific requirement for phospholipids in the stability and function of mitochondrial protein import machineries. For example, CL is required for the stability of TOM and SAM complexes (Gebert et al., 2009), PE is required for the import of β-barrel proteins through the TOM complex (Becker et al., 2013) and PC is required for the stability and activity of the TIM23 complex (Schuler et al., 2016). Molecular insights on the specific role of phospholipids in mitochondrial protein import are emerging from recent cryoEM-based structural studies. For example, the structure of yeast and human TOM complexes show that the protein-conducting channel contains a phospholipid molecule (Araiso et al., 2019), which was shown to be PC (Wang et al., 2020). Through structural and biophysical analysis, it was shown that the soluble domain of Tim50 (also known as TIMM50), a subunit of TIM23 complex, interacts with the IMS domain of another subunit, Tim23 (TIMM23), in a CL-dependent manner (Malhotra et al., 2017).

The overwhelming majority of studies to date have described the functional roles of PC, PE and CL, the three most abundant mitochondrial membrane phospholipids, but the roles of less abundant phospholipids, such as PI and PS, remain understudied. Although recent studies have shown a requirement of PS for mitochondrial protein import and integrity (Park et al., 2021; Yang et al., 2019), future studies on less abundant phospholipids are necessary to reveal their specific role(s) in mitochondrial biology.

Owing to the crucial role of phospholipids in various mitochondrial processes, it is of no surprise that disruptions in their biosynthesis and remodeling lead to disorders with a wide array of clinical pathologies. This section of the Review will describe the genetics, clinical presentations and biochemical underpinnings of rare genetic disorders of mitochondrial phospholipid metabolism (Table 2).

BTHS was first described in 1983 by Peter Barth as an X-linked disease characterized by neutropenia, cardiomyopathy, skeletal myopathy and a high rate of infant mortality (Barth et al., 1983). The causal gene for BTHS, TAFAZZIN, was identified in 1996 (Bione et al., 1996). A recent study has estimated the prevalence of BTHS at one in a million in the male population (Miller et al., 2020). With proper management of this disorder, affected individuals have an expected lifespan of ∼40 years (Rigaud et al., 2013). Owing to variations in clinical presentations, diagnosis based solely on the classical symptoms can be highly unreliable. Recently, an ultra-high performance liquid chromatography-mass spectrometry-based diagnostic method that measures the elevated ratio of MLCL:CL in blood samples from BTHS individuals has been developed and is considerably more reliable for diagnostic purposes (Vaz et al., 2022).

At the cellular and biochemical level, BTHS mitochondria contain perturbed CL species, with an overall decrease in CL levels and an increase in MLCL levels, which results in defective mitophagy, disorganized and disarrayed cristae, ‘leaky’ IMMs with reduced membrane potential, destabilized RSCs, increased oxidative stress and lower production of ATP (Ghosh et al., 2019; Lou et al., 2018; Saric et al., 2016). Furthermore, it has recently been shown that the MCU complex is destabilized in BTHS models and cardiac tissue from BTHS individuals (Bertero et al., 2021; Ghosh et al., 2020, 2022). Together, these bioenergetic defects explain cardio-skeletal myopathies; however, the biochemical basis of neutropenia, another characteristic feature of BTHS pathology, has not yet been elucidated. Although there is no existing treatment for this disorder, cardiac arrhythmia, heart failure and neutropenia are managed pharmacologically (Zegallai and Hatch, 2021). We highlight potential therapies for BTHS in Box 1.

Sengers syndrome is an autosomal recessive disorder originally described in 1975, but not recognized as a disorder of phospholipid metabolism until 2012 (Mayr et al., 2012; Sengers et al., 1975). Fewer than 1000 people in the USA have this disease (https://rarediseases.info.nih.gov/diseases/1142/sengers-syndrome; accessed 14/04/2023), which can present in either a severe or mild form. The severe form of this disorder results in infant mortality due to the early onset of cardiomyopathy, whereas individuals suffering from a milder form can survive through the fourth decade of life (Wortmann et al., 2015). The neonatal form of this disorder was also described as mitochondrial encephalomyopathy because of the involvement of the central nervous system (Perry and Sladky, 2008).

This disorder is caused by the loss of function of AGK, which catalyzes the phosphorylation of diacylglycerol and monoacylglycerol to form PA and lysophosphatidic acid, respectively (Mayr et al., 2012). Since PA is a precursor of CL (Fig. 2), loss of AGK function could perturb the biosynthesis of CL and contribute to disease pathology, including cardiomyopathy and skeletal myopathy, among other complex symptoms (Table 2). Perturbation in CL biosynthesis could also explain complex I, III and IV deficiencies and severe mtDNA depletion, which has been reported in two unrelated individuals with AGK mutations (Calvo et al., 2012). However, the extent to which CL biosynthesis relies on AGK-derived PA is unknown. AGK has also been identified as a constituent of the TIM22 protein import complex, which facilitates the import of mitochondrial carrier proteins such as ATP:ADP carrier (Vukotic et al., 2017). It has been suggested that destabilization of the TIM22 complex could also contribute to the pathology of Sengers syndrome (Vukotic et al., 2017). Indeed, cardiac and skeletal muscle of Sengers syndrome individuals exhibit strikingly reduced levels of ATP:ADP carrier protein in the IMM (Haghighi et al., 2014; Mayr et al., 2012). Currently, no specific treatment is available for this fatal disorder.

3-Methylglutaconic aciduria, dystonia–deafness, encephalopathy, Leigh-like (MEGDEL) syndrome is an autosomal neurodegenerative disorder that was first described in 2006 (Wortmann et al., 2006). The age of onset can be early (infantile) or delayed (juvenile), with varying degrees of severity (Table 2) (Finsterer et al., 2020). The genetic basis of this disorder was first described in 2012 with biallelic variants in the SERAC1 gene (Wortmann et al., 2012). Since its initial description in 2006, 102 cases have been reported worldwide (Finsterer et al., 2020), and it is estimated that 27 individuals will be born with MEGDEL syndrome each year (Maas et al., 2017).

The precise biochemical function of SERAC1 is not yet known; however, this protein has been reported to be essential for PG remodeling, such that loss of SERAC1 function leads to the accumulation of 34:1 PG instead of 36:1, which results in an altered CL composition (Wortmann et al., 2012). A nonsense mutation in SERAC1 was shown to result in impaired mitochondrial liver function with a decreased level of respiratory complexes I, III and IV (Fellman et al., 2022). Consistent with this, reduced activity of the respiratory chain complexes has been observed in liver biopsies of other MEGDEL syndrome individuals (Sarig et al., 2013). Further research is necessary to unravel the precise connection between the biochemical basis of the disease and its pathology. Currently, treatment for MEGDEL syndrome is mostly palliative, and an early death typically follows brain degeneration (Maas et al., 2017).

This rare genetic disorder was first reported clinically in 1986 (Liberfarb et al., 1986), but the underlying molecular cause was not described until 2019 (Peter et al., 2019). It is an autosomal recessive disorder of unknown prevalence caused by a partial loss of function of mitochondrial PISD (Peter et al., 2019), which is required for conversion of PS into PE in mitochondria. The small number of Liberfarb syndrome patients that are positively diagnosed show a wide spectrum of clinical presentations (Table 2) (Peter et al., 2019; Zhao et al., 2019).

The cellular and biochemical phenotypes of PISD mutations include fragmentation of the mitochondrial network, reduced oxygen consumption and impaired activity of mitochondrial proteases. Restoration of these phenotypes can be achieved by lyso-PE or ethanolamine supplementation in fibroblasts from affected individuals, consistent with the notion that the aforementioned phenotypes are a result of reduced mitochondrial PE (Girisha et al., 2019; Zhao et al., 2019). These findings also provide potential treatment options for this disease.

The most recently identified disorder of mitochondrial phospholipid metabolism is caused by homozygous recessive mutations in CRLS1 gene, which encodes CL synthase (Lee et al., 2022). These mutations have been detected in four individuals from three different families (Lee et al., 2022). These individuals have compromised CL biosynthesis, but exhibit distinct symptoms as compared to the other disorders of CL deficiency discussed here (Table 2). Investigation of fibroblasts derived from affected individuals identified impaired mitochondrial morphology and biogenesis, with reduced CL levels and increased levels of PG, which is a substrate for CL (Lee et al., 2022).

Owing to the different biophysical and biochemical properties of mitochondrial membrane phospholipids, each class of phospholipid plays a unique role in mitochondrial biogenesis, bioenergetics and dynamics. Thus, maintaining phospholipid homeostasis is crucial, and in this regard, we anticipate the discovery of novel proteins for the transport of PI, PG and PE to, from and within mitochondria. In addition to the requirements of the absolute levels of non-bilayer forming phospholipids CL and PE, their enrichment in membrane microdomains by scaffolding factors, such as prohibitins and stomatin-like protein-2, are likely to impact specific mitochondrial functions (Christie et al., 2011; Tatsuta and Langer, 2017). Determining the functions of mitochondrial phospholipids is of high biomedical significance because advances in human genomics have identified rare but debilitating genetic disorders that occur due to defects in phospholipid metabolism (Lu and Claypool, 2015). Currently, the major challenge in understanding the biochemical basis of rare genetic disorders of mitochondrial phospholipid metabolism is the lack of correlation between genotype and phenotype (Ronvelia et al., 2012). This could be due to the presence of modifier genes in the genome, as has been recently shown in a TAFAZZIN-knockout mouse model (Wang et al., 2023). Identifying these modifier gene(s) could pave the way to therapeutic approaches for these orphan disorders.

We thank Dr Benoit Kornmann for clarifying the function of ERMES and MAM in his personal communications to the authors.