Remodelling of mitochondrial function by import of speciﬁc lipids at multiple membrane-contact sites

Organelles form physical and functional contact between each other to exchange information, metabolic intermediates, and signaling molecules. Tethering factors and contact site complexes bring partnering organelles into close spatial proximity to establish membrane contact sites (MCSs), which specialize in unique functions like lipid transport or Ca 2 + signaling. Here, we discuss how MCSs form dynamic platforms that are important for lipid metabolism. We provide a perspective on how import of speciﬁc lipids from the ER and other organelles may contribute to remodeling of mitochondria during nutrient starvation. We speculate that mitochondrial adaptation is achieved by connecting several compartments into a highly dynamic organelle network. The lipid droplet appears to be a central hub in coordinating the function of these organelle neighborhoods.

Keywords: autophagy; endoplasmic reticulum (ER); lipid droplets (LDs); membrane contact sites; metabolic adaptation; mitochondria; mitochondrial shape; organelle-network; starvation Membrane contact sites are regulatory platforms controlling organelle function Eukaryotic cells compartmentalize their interior into organelles with specific biochemical functions.While the spatial separation is useful in creating bespoke subcellular environments, it inevitably introduces discontinuity into biochemical and cell biological pathways which are often distributed across several organelles.To integrate these pathways, organelles communicate intensely with one another using classical membrane traffic and non-vesicular transport routes.Accumulating evidence suggests that exchange of information and metabolic intermediates is particularly dependent on physical contact between organelles which is established at membrane contact sites (MCSs) [1][2][3][4].
Membrane contact sites are specialized regions within a cell where membrane-bound organelles are actively held into close apposition, often only at a few nanometers apart [5,6].Their formation depends on tethering complexes that connect partnering membranes across the contact site.MCSs mainly serve as platforms for inter-organelle communication through exchange of biochemical intermediates and signaling molecules, such as lipids and Ca 2+ [5,7].
Initially, it was thought that most organelles interact only at one generic contact site and that these interfaces would be sufficiently 'versatile' to serve all the functions required for inter-organelle crosstalk.However, the ever-expanding repertoire of contact site factors, tethers, MCS-complexes, and transport proteins provides evidence that this initial view was probably too simplistic.There are often several molecularly distinct contact sites between any two organelles in the cell suggesting that MCSs carry out much more specialized function than initially anticipated.Additionally, accumulating evidence indicates that MCSs are individually regulated by signaling through phosphorylation or other post-translational modifications [8,9].
A revised model of organelle interface biology therefore posits that MCSs carry out highly specialized functions and that the activity of each MCS can be individually controlled possibly through signaling.
The specific regulation of cellular activities at individual MCSs might even have been a key selectable advantage that drove the diversification of contact site factors during eukaryotic evolution.Rapid and specific tuning of MCS activity tremendously improves the adaptability of organelle functions in response to internal needs or external cues, for instance allowing the switching of cellular metabolism between anabolic and catabolic states during fasting-feeding cycles.
Some organelles like mitochondria have additionally been recognized to carry out functions beyond their core metabolic tasks, for instance playing a role in the functional implementation of transcriptional instructions from differentiation programs [10][11][12][13][14]. Controlled changes in mitochondrial activity and function seem to be an integral part of 'fate switches', driving cells down a particular differentiation trajectory [12].Despite the central importance in several cell biological contexts, the mechanisms by which MCSs are regulated and how they are involved in controlling or remodeling organelle function are not well understood.
One possibility is that MCSs influence organelle composition by direct import of specific cargoes which may fine-tune or remodel the function of an organelle.For example, non-vesicular transport of lipids at MCSs by lipid transfer proteins (LTPs) is known to be important in creating and maintaining the specific lipid compositions of different organelles [4,[15][16][17] which in turn influences the biochemical activity and cell biological function of the respective compartment or even the whole cell.
A second possibility is that organelle interfaces may control flux through metabolic pathways, which are split between cellular compartments [18].For example, several lipid synthesis pathways are distributed between the endoplasmic reticulum (ER) and mitochondria [19,20], which requires the exchange of metabolites between these two organelles (Fig. 1).
Additionally, MCSs specialize in the exchange of signaling molecules like Ca 2+ , intermediary metabolites, and signaling lipids, which all can function in the regulation of metabolic enzymes and may thereby control flux through specific pathways.The fact that numerous metabolic enzymes are allosterically regulated by Ca 2+ or require specific lipids in their membrane environment for stability and activity further suggests that the different modes of organelle crosstalk might be tightly entangled and influence each other.
Precise and local regulation of the relevant transport and tethering factors by signaling would have a strong impact on the specific activity of each contact site.In doing so, MCSs have the potential to be powerful regulatory platforms integrating signals from internal and external cues adapting the metabolic and cell biological functions of cells as they change e.g., during the cell cycle, differentiation, or feeding fasting cycles.

Mitochondrial contact sites are important for lipid metabolism
Membrane contact sites were first discovered in the 1950s when electron microscopy images of rat liver revealed regions of close apposition between the ER and mitochondria [21,22].Interestingly, it was already apparent in this early work that the association of the two organelles changed reversibly under different metabolic conditions supporting the idea for a function of MCSs in metabolic pathways and providing first evidence for possible regulation of inter-organelle crosstalk by metabolic signaling.Fig. 1.A lipid metabolism map of the ER-mitochondrial contact site (ERMC).Lipid synthesis in the mitochondria depends on import of lipidprecursors and substrates from other organelles like the endoplasmic reticulum (ER).This pathway map focuses on the interdependence of lipid synthesis in the ER and the production of cardiolipins (CL) in the inner mitochondrial membrane (IMM).De novo lipogenesis is outlined in the gray box, which is connected to the synthesis of phosphatidic acid (PA) in the ER and triacylglycerol (TAG) production at the interface between the ER and the lipid droplet (yellow).Membrane lipid synthesis starts in the cytoplasm by preparing activated headgroup precursors such as cytidine-diphospho-choline (CDP-Cho) or CDP-ethanolamine (CDP-Eth).They are subsequently linked to diacylglcyerol (DAG) in the ER yielding phosphatidylcholine (PC) and phosphatidylethanolamine (PE), respectively.These can serve as precursors for the synthesis of phosphatidylserine (PS).A specialized ER domain within ERMCs is called mitochondrial associated membranes (MAM).MAMs are hotspots for PS and PC synthesis.MAMs are closely apposed to the outer mitochondrial membrane (OMM).PS is imported into mitchondria by a lipid transfer proteins (potentially MIGA2 in mammalian cells, ERMES in yeast (not shown)) and decarboxylated into PE.PE may then reside in the mitochondria or is transported back to the ER whereespecially in liver -3 subsequent methylation reactions catalyzed by the PE metyl transferase PEMT convert it into PC.PA can be dephosphorylated into DAG which is then used in the ER for TAG or phosphtidylinositol (PI) synthesis.Alternatively, it can be shunted to the mitochondria (likely by RMDN3 or Prelid3b/TRIAP).Here, it may be scrambled across the membrane by MTCH2 and used for CL synthesis in the IMM.CL, PE and PC are then likely used to expand the IMM and increase the area of cristae, which may stimulate oxidative phosphorylation (OXPHOS).Moreover, mitochondria elongate in response to these changes in lipid metabolism.List of abbreviations used in the map: ACSL1, long-chain-fatty acid-CoA ligase 1; ACSL3, long-chain-fatty acid-CoA ligase 3; AGPAT2, 1-acylglycerol-3-phosphate-O-acyltransferase 2; AGPAT4, 1-acylglycerol-3-phosphate-O-acyltransferase 4; AGPAT5, 1acylglycerol-3-phosphate-O-acyltransferase 5; ATGL1, adipose triglyceride lipase 1; CDS1/2, cytidine diphosphate-diacylglycerol synthase; CEPT1, choline/ethanolaminephosphotransferase 1; CLS, cardiolipin synthase; DGAT1, diacylglycerol O-acyltransferase 1; DGAT2, diacylglycerol Oacyltransferase 2; EKI1 & CKI1, ethanolamine kinase 1, choline kinase 1; ETC, electron transport chain; FATP4, long-chain fatty acid transport protein 4; GPAT2, glycerol-3-phosphate acyltransferase 2; GPD1, glycerol-3-phosphate dehydrogenase 1; GPT1/GPAM, glycerol-3-phosphate acyltransferase, mitochondrial; iPLA2, Ca 2+ phospholipase A2; LCLAT1, lysocardiolipin acyltransferase; LPIN1/2, phosphatidate phosphatase; PCYT1a/b, choline-phosphate cytidylyltransferase 1 a/b; PEMT, phosphatidylethanolamine methyltransferase; PGS, phosphatidylglycerol synthase; PIS, phosphatidylinositol synthase; PSD, phosphatidylserine decarboxylase; PSS1, phosphatidylserine synthase 1; PSS2, phosphatidylserine synthase 2; PTPMT1, phosphatidylglycerophosphatase; PYCT2, choline-phosphate cytidylyltransferase 2; Taffazzin, NA; TAMM41, mitochondrial translocator assembly and maintenance homolog.The molecular study of ER-mitochondrial MCSs was opened when differential centrifugation experiments revealed a fraction of ER membranes that cosedimented with mitochondria [23][24][25].Seminal work by Jean Vance showed that these ER subdomains could be isolated and were then fittingly termed mitochondriaassociated membranes (MAMs, Fig. 1) [26][27][28].
The compositional analysis uncovered that MAMs are hotspots for lipid metabolism.Enzymes that produce phosphatidylserine (PS) called PSS1/2 and PEMT which methylates phosphatidylethanolamine (PE) to produce phosphatidylcholine (PC) are enriched in MAMs [29,30] (Fig. 1).Local PS synthesis at the contact site would be consistent with biochemical data that already existed at the time, showing that ERderived PS is imported into mitochondria and subsequently decarboxylated by the PS-decarboxylase (PISD) to produce phosphatidylethanolamine (PE, Fig. 1) [27,28,31].
Although the cell synthesizes PE by at least another three additional pathways [32], mitochondrial PE is essential and cannot be replaced with the other PE pools in the cell.These observations immediately suggest that MAMs may have an important regulatory role in controlling mitochondrial biology through lipids [30,31,33].Specifically, this scenario raises the interesting possibility that lipid import at MAMs may play a critical role in controlling the mitochondrial membrane composition and thereby mitochondrial function in a regulated fashion.
It is conceivable that PS transport at MAMs may directly determine the abundance of PE in the mitochondria.Modulation of MAM function by regulating the activity of tethering factors or PS transport proteins may allow the cell to control the percentage of PE relative to other lipid classes in the mitochondrial membranes.Since the amount of PE in mitochondrial membranes affects many mitochondrial functions, such as the electron transfer chain (ETC) complexes or the mitochondrial fission and fusion machinery, the lipid transfer proteins (LTPs) which import PS at MAMs may have a decisive impact on the regulation of mitochondrial function.An exciting possibility is that transport of PS across MAMs is selective for specific lipid species [34].PS species, i.e.PS lipids that contain fatty acids with precisely defined number of C-atoms and double bonds would then give rise to specific mitochondrial PE species that may fine-tune mitochondrial functions for instance by regulating specific membrane-bound enzymes allosterically in a fatty acid-dependent manner.
As a 'non-bilayer' lipid, PE is further known to locally alter membrane structure [35], which affects several biochemical activities and even influences the overall shape of mitochondria.In fact, PE controls specific membrane properties which are indispensable for the synthesis of another critical lipid class found in mitochondria called cardiolipin (CL).CLs are exclusively found in mitochondria [36] and preferentially localize to the inner mitochondrial membrane (IMM) [37] where they make up around 10-15% of all lipids in the membrane [38][39][40].CLs uniquely possess four instead of the conventional two fatty acids found in other membrane lipids and contain two instead of one glycerolphosphate in the backbone.Most of the enzymes involved in CL-production reside in the inner mitochondrial membrane (IMM, Fig. 1), but like in the PE synthesis pathway, the provision of the initial precursors depends on import from the ER [17,41,42].
One pathway uses phosphatidic acid (PA) from the ER.Transport of PA to the IMM depends on TRIAP1/PRELID3b [43] where it is converted into CDP-DAG by TAMM41 [41,44].An alternative potential pathway starts with the conversion of PA into CDP-DAG in the ER which is mediated by CDS1/2 [45,46].Usually the CDP-DAG pool that is made in the ER is used for the synthesis of phosphatidylinositol (PI) [47].However, should it also serve as a precursor for CL synthesis in the IMM it must be imported into mitochondria [48].Whether this pathway exists, and which transfer protein mediate the import step is currently unknown.Downstream of CDP-DAG, the two alternative routes would converge again, and synthesize phosphatidyl-glycerol-phosphate (PGP) [49] which is then readily dephosphorylated to form phosphatidylglycerol (PG) [50].Next, the CL synthase (CLS) produces CL from PG and another molecule of CDP-DAG [51].
The newly made CL is subsequently subjected to several remodeling reactions exchanging the FAs with different length and saturation.It is thought that this transacylation process begins with the removal of one fatty acid from CL by a calcium-independent phospholipase A2 (iPLA2), yielding monolyso-CL (MLCL) [52][53][54].The freed-up fatty acid is esterified to a lysophosphatidylcholine (LPC) producing phosphatidylcholine (PC).
The following transacylation reaction is catalyzed by Tafazzin and moves a specific FA from yet another PC to MLCL resulting in the remodeled CL species (Fig. 1) [40,55,56].The mature CLs contain preferentially long chain, polyunsaturated FAs (e.g.linoleic acid FA18:2, which is 18 C-atoms long and contains 2 double bonds) [51,57].Interestingly, it is thought that Taffazin is only selective for specific FAs when the substrates are provided in membranes containing the right amount of PE [57].Without PE, Taffazin is a non-selective enzyme and does not discriminate reliably between different types of fatty acids, at least in vitro.
Although it is known that STARD7 is important for transport of PC from the ER to mitochondria [58], it is currently not clear whether cells have control over the import of a specific PC species delivering suitable FAs for CL remodeling.Evidence from monogenetic diseases such as Barth syndrome, and Charcot Marie Tooth Disease suggests that aberrant PE and CL acyl-chain composition, i.e. the 'wrong' PE and CL-species lead to drastic changes in mitochondrial and cellular fitness.
Taken together these various lines of evidence strongly support a working model whereby selective import of lipids from other organelles determines the membrane lipid composition of mitochondria and thereby strongly influences the biochemical activity and cellular function of the organelle.However, the mechanisms by which mitochondrial lipid metabolism is modulated in different tissues or in response to signaling events is not well understood.The machinery importing lipids into the mitochondria is an active area of research and its regulation may prove to be an essential aspect of metabolic homeostasis.

The machinery forming ER-mitochondria contact sites (ERMCs)
The molecular basis of the machinery connecting the ER and mitochondria remained elusive for a long time, until a genetic screen in baker's yeast uncovered the ER-mitochondrial encounter structure (ERMES) [79].ERMES is a tetrameric complex which tethers the ER to mitochondria: Mdm10 and Mdm34 which localize to mitochondria as integral and peripheral membrane proteins, respectively, bind via a cytosolic bridging factor called Mdm12 to Mmm1, which is integral to the ER.Although ERMES function was debated for a long time, substantial evidence backs up the initial claim that it is indeed involved in lipid transport between the two organelles.The likeliest substrate is PS which is imported from the ER to mitochondria for PE synthesis.ERMES may perhaps even exchange the newly synthesized PE back to the ER where it is converted to PC production through the methylation pathway catalyzed by the PEmethyltransferases Pem1 and Pem2, the baker's yeast orthologues to mammalian PEMT (mammalian components shown in Fig. 1) [80].
While the LTPs that mediate PS (yeast, Ups1-Mdm35; mammals, PRELID1-TRIAP) and PA transport (yeast, Ups3-Mdm35; mammals, PRELID3b-TRIAP) to the IMM are evolutionary conserved [81], mammalian cells do not possess an ERMES orthologue.Candidate LTPs for PS transport at ER-mitochondrial contact sites (ERMCs) in mammalian cells are mitoguardin-2 (MIGA2) [82][83][84][85][86] and the oxysterol binding proteins ORP5/8 [87][88][89][90], which are both completely unrelated to any of the ERMES components.Recent structural evidence suggests that ERMES transports lipids through a Z-shaped tunnel composed of several SMP (synaptotagmin-like mitochondrial lipid binding)domains, which are also found in several other lipid shuttling proteins [91].In contrast, the ORP-proteins are lipid exchangers transporting sterols and phosphoinositides (PIPs) or PS by a counterflow mechanism [4].How the ORPs would engage in lipid transport between ER and mitochondria which do not contain appreciable amounts of sterols is currently unknown.Whether they use PIPs instead of sterols in the exchange reactions with PS is also not completely clear and it remains open whether and how cells transport PIPs to mitochondria.
MIGA2 exhibits yet another completely different lipid binding fold representing the first lipid transfer domain that is helical.MIGA2 does not possess any b sheet elements, which are usually found in most other lipid binding domains [84,86].In vitro experiments seem to suggest that MIGA2 alone is sufficient for transport of lipids.In fact, the lipid transfer domain at the C terminal end of the protein is sufficient for efficient exchange of lipids between artificial membranes [84,86].Binding to its receptor in the ER has likely a regulatory function and may accelerate the transfer rates (see below).
Although the cellular function of the mammalian LTPs is even less understood than ERMES, it seems that the machinery involved in mitochondrial lipid import across ERMCs has diversified significantly, and ERMES was lost during evolution.It appears that the molecular solutions for PS transport in yeast and mammals are completely different.As discussed above, the diversification of PS transport across ERMCs by different LTPs may allow the specific regulation of lipid import into mitochondria determining e.g. the amount of specific mitochondrial lipid classes or it may even control their fatty acid composition during different metabolic conditions.
The translocation of lipids across the OMM by scramblases: a role for MTCH2?
An important intermediate step in the import of lipids to mitochondria is mediated by the machinery that transports lipids across the OMM (Fig. 1).It was recently suggested that the outer multi-pass membrane protein MTCH2 is involved in transporting lipids into mitochondria [92].
A function for MTCH2 in lipid transport was inferred from cell biology experiments showing that mitochondrial shape is under control of this OMM protein.Moreover, Co-IP proteomics experiments showed that the Acyl-glycerophosphate acyl transferase AGPAT5 is a prominent MTCH2 interactor (Fig. 1).AGPAT5 substrates such as lyso-PA may indeed be flipped across the OMM in a MTCH2dependent manner and contribute to PA synthesis.Due to its small headgroup, PA is sometimes considered a fusogenic lipid that may directly promote mitochondrial fusion.This may contribute to the regulation of mitochondrial shape and function through lipids.
MTCH2 knockout cells clearly show fragmented mitochondria and overexpression promotes fusion.Interestingly, MTCH2 was also characterized as a protein insertase, which specializes in the integration of tail-anchored membrane proteins to the OMM [93].Both activities may even be coordinated.

VAP proteinsversatile adaptor proteins in the ERare core components of ERMCs
Most LTPs that interact with the ER employ a highly conserved peptide motif called 'two-phenylalanines in an acidic tract' (FFAT).FFAT motifs are found in about 10-15% of the eukaryotic proteome.FFAT peptides bind to the major sperm protein (MSP) domain within MOSPD and the versatile ER adaptor proteins of the VAMP-associated proteins, VAP-A and VAP-B [94,95].
Interaction between VAPs and the binding partners in neighboring organelles leads to membrane tethering and formation of MCSs (Fig. 2).The affinity between VAP proteins and their cognate tethers is high.The binding mechanism involves a two-stage process starting with a weak 'scanning' phase during which acidic residues around the core FFAT motif are identified followed by the high affinity interaction mode which locks the FFAT-peptide in at the twin phenylalanines (FF) [96][97][98][99].
While the acidic residues in canonical FFAT motifs e.g. in the oxysterol binding protein OSBP1 are mostly aspartates and glutamates (DEDDENEFFDAPEII with the twin FF motif at position 259 and 260 of the human OSPBP1), recent work suggested that FFAT motifs are much more diverse than initially anticipated.In a specific variant called the phospho-FFAT motif, the acidic residues flanking the twin FFs are replaced by serine residues which regulate the tethering function in a kinasedependent fashion.Indeed, some evidence exists that the phospho-FFAT must be phosphorylated to bind VAP or MOSPD proteins [8,100,101].
Phospho-FFAT peptides are computationally predicted in about 50% of all FFAT motifs in the human proteome with MIGA2 at the top of the list with the highest prediction score (SLTSEDSFFSATELF with the twin FF motif at position 293 and 294 of the human MIGA2).Interestingly, the phospho-FFAT motif is evolutionarily conserved and can also be found in the MIGA2 fly-orthologue called Miga [83].
Fly Miga is not only phosphorylated at the FFAT peptide but contains several additional phosphoserines.A serine-rich motif (SRM) close to the mitochondrial membrane-anchor appears to cause a major change in electrophoretic mobility and may be under the control of metabolic signaling mediated by casein kinase (CKI) and Ca 2+ -modulated kinase of the CamKII family [83].While the phosphorylation of the phospho-FFAT motif is necessary for binding to VAP proteins, it remains to be shown how the hyper-phosphorylation of Miga regulates its function at contact sites between the ER and mitochondria.The position of the SRM close to the mitochondrial membrane anchor is conserved in mammalian MIGA2 but computational predictions suggest that the kinases targeting this domain are different in fly and mammals.
The phosphorylation of MCS tethering factors and LTPs appears to be a general mechanism.Apart from MIGA2 another ER-mitochondrial membrane tether called RMDN3/PTPIP51 (Figs 1 and 2) also contains a phospho-FFAT motif and has at least one SRM which may be hyperphosphorylated, too [83].Beyond regulating organelle tethering by FFAT phosphorylation, several other LTPs like the ceramide transfer protein CERT and OSBP1 have SRMs, which appear to cause the same electrophoretic shift as observed for MIGA2 and Miga, suggesting that they are hyperphosphorylated as well [102][103][104].CERT and OSBP1 are found at an MCS between the ER and the Golgi apparatus.In both cases, the SRM is found adjacent to the pleckstrin homology (PH) domain which binds phosphatidylinositol-4-phosphate in the membranes of the Golgi network.The hyperphosphorylation of the SRM in CERT seemingly regulates lipid binding activity in the lipid transfer domain.Apart from phosphorylation, it was suggested that ubiquitination may be involved in promoting lipid binding affinity as shown for RMDN3 [105].
In summary, these data support a new perspective on the regulation of MCSs: the emerging picture suggests Fig. 2. The Lipid droplet-autophagosomes-ER-mitochondria network.When cells starve, they rewire their metabolism to focus on oxidative phosphorylation (OXPHOS) for ATP synthesis.The electron transport chain (ETC) components in the inner mitochondrial membrane (IMM) increase and the cristae expand (brown).Substrates for ATP synthesis are derived from autophagy, which digests cellular organelles and cytoplasmic components in auto-lysosomes (pink).Amino acids are transported to the mitochondria for oxidation in the tricarboxylic acid (TCA) cycle.Fatty acids (FAs) are released from the auto-lysosome and reach the ER.A possible tethering complex composed of Rab7 (lysosome)-Protrudin (ER)-VAPA/B (ER) and PDZD8 (ER) may transport the lipids.Another possible route is via VPS13 proteins.In the ER, FAs are then activated into FA-CoA and esterified with glycerol-3-phosphate (G3P) yielding PA.This can be dephosphorylated into DAG.DAG can then be a substrate for TAGs which are stored in Lipid droplets (yellow) or be converted into several membrane lipids (see Fig. 1).TAGs can be mobilized by the adipose triglyceride lipase (ATGL) and DAG can be hydrolysed by the hormone sensitive lipase (HSL) to release FAs, which are transported across LD-mitochondrial contact sites as substrates for b-oxidation.The TCA-cycle extracts the high energy electrons to produce reducing equivalents (NADH), which are then oxidized in the ETC to build up a membrane potential across the IMM fuelling the ATP-synthesis.Membrane lipids can enter the mitochondria from the ER at a third contact site called the ER-mitochondrial contact site (ERMC).PA, PC and PS may all be transported by their dedicated lipid transfer proteins, feeding them into membrane synthesis in the IMM as described in Fig. 1. that the tethering factors and the LTPs like RMDN3, MIGA2, CERT and OSBP1 are regulated by a variety of distinct mechanisms.Tethering is clearly dependent on active FFAT motifs and lipid transport is likely governed by additional modifications such as the phosphorylation of SRMs.How the different phospho-states influence transport activity is currently unknown, but the existing evidence strongly indicates that metabolic signaling may regulate biochemical activities at MCS through phosphorylation [9].An interesting future line of research will be to elucidate the mechanisms that generate the regulatory input for LTPs.It will be necessary to identify the kinases and ubiquitin ligases that are important in the regulation of LTPs.The insights will significantly improve our understanding of how LTPs control downstream metabolic pathways in the target organelles.
In the case of MIGA2, a specific question is how mitochondrial functions are controlled through lipid import at MCSs.A fascinating possibility is that LTPs like MIGA2 may be able to discriminate between different lipid-species as substrates.This in turn would determine mitochondrial activity through membrane organization or allosteric control of membrane proteins like the components of the respiratory system.
It is further possible that LTPs switch between different partner organelles to import lipids from specific cellular compartments.For instance, MIGA2 has dual affinity to the LDs and the ER (Fig. 2) [82], and might be able to switch between different contact sites in response to metabolic cues.Depending on which organelle the LTPs interact with, they might selectively import specific lipids.For example, it is possible that the LD provides different lipids than the ER.It is conceivable that the various phospho-states have different affinity for specific partner organelles and may show preference for a specific lipid class or species.

Remodeling of mitochondrial function during starvation requires multiple MCSs
The change of mitochondrial metabolism and shape in response to changing lipid composition is probably best characterized during starvation.When cells are deprived of nutrients, they must switch to different pathways of ATP production to meet their metabolic requirements.This involves the rewiring of cellular metabolism to activate respiration and oxidative phosphorylation (OXPHOS) in mitochondria, leading to massive rearrangement of several MCSs between the ER, mitochondria, and other partner organelles (Fig. 2).The switch to OXPHOS coincides with a pronounced change of mitochondrial morphology [106,107].
Upon nutrient starvation, mitochondria fuse into a highly interconnected network.The elongation appears to be strongly correlated with the metabolic adaptation and seems to require import of specific lipids.It is likely that several MCSs between mitochondria and several other organelles like the ER and LDs are a central part of the machinery that is required for the remodeling of mitochondrial shape and function during this process.How the change in mitochondrial shape is coupled to the change in metabolic activity is still not completely understood but it appears that shifting away from carbohydrates as principal energy source seems to be a decisive step in the underlying sequence of events.
Work by several labs indicated that the 'freed-up' fatty acids are re-esterified into triacylglycerols (TAGs) and stored in a specific population of lipid droplets (LDs, Fig. 2), which appear in close contact with the ER and mitochondria [16,113,116].The function of autophagy-induced LDs is, however, not completely understood.One view is that they are consumed by the adipose triglyceride lipase (ATGL) to provide FAs as precursors for b-oxidation and ATP production through oxidative phosphorylation (OXPHOS) [16,116].Other models indicate that the mitochondrialassociated LDs are not necessary for FA delivery to boxidation [113], but instead protect mitochondria from toxic overload with FA-carnitines which may build up during overload of the mitochondrial FA import pathway (Fig. 2).In this scenario the LDs would be a lipid buffer organelle that either neutralizes toxic pathway intermediates or pre-emptively slows down transport rates, tuning FA release from the LD to the consumption rate in the oxidation reactions.
MIGA2 and RMDN3 are candidate OMM proteins that may function as starvation induced tethers between the mitochondria and the ER and might mediate lipid transport between the two organelles.They may even be important in implementing the functional and morphological changes of mitochondria seen during starvation (Fig. 2) [82,117].Given that MIGA2 has been shown to interact with LDs in adipocytes, it may additionally be an important factor that links the LDs and the mitochondria during the starvation response.
Since the ER-tethering step is dependent on phosphorylated phospho-FFAT motifs, it is highly likely that the activation of the ERMC function in lipid transport is under the control of starvation signaling.
It may be the case that the activation of the phospho-FFAT motifs in these two proteins leads to the recruitment of the ER to the mitochondria, forming a starvation-induced contact site through VAP proteins.The resulting MCS may be a platform that facilitates the coordinated exchange of lipids between the involved organelles, which may even be important to trigger mitochondrial fusion.A major question is which functions would be dependent on mitochondrial fusion during starvation.
One aspect is that fused mitochondria are protected from autophagy [107,118].Another well-established fact is that hyperfused mitochondria contain pronounced cristae [119], which may be crucial for the switch to OXPHOS.The increase of the IMM area would allow the accommodation of an expanded respiratory system.Mitochondrial fusion is perhaps required for the efficient expansion of the IMM, which is necessary for the incorporation of the upregulated ETC components.
Elongation of mitochondria requires the concerted activation of membrane fusion reactions [107,118] mediated at the OMM by the dynamin-like integral membrane GTPases called Mitofusin 1 and 2 (MFN1/2) and at the IMM by the GTPase OPA1 [120][121][122][123][124][125].At the same time mitochondrial fission, which is under the control of a dynamin-related protein called DRP1 and its binding partners must be inhibited [126][127][128].Additional factors that regulate the balance between fission and fusion include a specialized vesicle trafficking pathway from the Golgi apparatus to mitochondria and the correct subcellular localization of the small GTPase-sARF1 [129][130][131] (Fig. 2).

Mechanisms by which starvation initiates mitochondrial elongation
Interestingly, the mitochondrial shape-changes appear to occur in a biphasic process: an acute phase of nutrient stress is characterized by elevated adenosinemonophosphate (AMP) which in turn activates AMPactivated protein kinase (AMPK).AMPK activation results in the recruitment of DRP1 to the mitochondria through phosphorylation of its OMM-receptor MFF and the presence of auxiliary factors like MTFRL1 and MiD49 & 51 [132], which triggers a transient fragmentation of mitochondria.
In a second phase, the sustained activation of AMPK prompts a delayed starvation response during which the transcription factor EB (TFEB) and TFE3 activate lysosome production and upregulate autophagy [133,134].TFEB and TFE3 further activate transcription of peroxisome proliferator activated receptor gamma co-activator-a (PGC1a) which stimulates the biogenesis of mitochondria and prepares the cell for ATP production through autophagy and OXPHOS.
While it is easy to understand how cells activate autophagy during prolonged starvation, it is not immediately clear how this catabolic program is coordinated with an anabolic process like mitochondrial biogenesis.One central question is how cells provide the necessary lipid-precursors to increase the membrane area during fusion of mitochondria and how this is coordinated with the seemingly competing tasks of upregulating catabolism and OXPHOS.The situation cannot be solved by simply activating the de novo synthesis of membrane lipids because the critical FA-precursor malonyl-CoA would allosterically inhibit the carnitine palmitoyl transferase CPT1/2 (Figs 1 and 2) [135][136][137] and therefore block the import of FAs into mitochondria.Another possibility is that mitochondria undergo fusion without the need for active increase of their membrane area.However, it is well known that the activation of PGC1a increases the synthesis and thereby the amount of ETC complexes [138,139].To accommodate the expanding respiratory system, the inner mitochondrial membrane area must enlarge, and its lipid composition has to be remodeled [140].
The only remaining alternative to solve this problem is to import lipids from other organelles.An elegant solution would be if membrane lipids that are freed up during breakdown of other organelles during autophagy are directly available for membrane synthesis in the mitochondria.However, since membrane lipids are hydrolysed in the lysosome, autophagy rather frees up FAs [113] (Fig. 2).
The 'autophagic' FAs have two destinations: they can either serve as a source of substrates for lipid synthesis and expand mitochondrial membranes or they are fed into b-oxidation.How the FAs that exit the lysosome are channeled into lipid synthesis or enter the road to degradation is not clear.MCSs might aid in the decision-making process.A possible scenario is that lysosomes contact the ER and funnel FAs through a specialized MCS towards lipid synthesis in the ER (Fig. 2).The newly synthesized membrane lipids might then be transported into mitochondria through a second MCS at the ER-mitochondrial interface and facilitate membrane expansion to accommodate the growing respiratory system (Fig. 2).
It is conceivable that FAs that are earmarked for boxidation reach mitochondria either through a more conventional cytoplasmic transport protein or another MCS directly linking the lysosomes and the mitochondria.Alternatively, they take a 'detour' through LDs from where they are mobilized by lipolysis and then transported to mitochondria for oxidation [16,116,131].
Recent work has shown how Rab7 on lysosomes binds through its GTPase activating protein (GAP) TBC1D15 to FIS1 at the mitochondria, which establishes direct contact between these two organelles [141,142].However, this contact site rather appears to induce mitochondrial fission and may therefore not support mitochondrial fusion in the starvation response.
Another Rab7-based tethering complex links lysosomes also to the ER.The ER membrane protein Protrudin binds through its FYVE domain phosphatidyl-3P (PI3P) on late endosomes and lysosomes in a Rab7dependent manner [143][144][145].Protrudin further interacts with VAP proteins through its conventional FFAT motif and recruits the LTP PDZD8 to this MCS [143].PDZD8 is an ER membrane protein, which was initially discovered by virtue of its SMPdomain and was therefore thought to be the mammalian orthologue of ERMES [146].However, the suggested localization to the ERmitochondrial MCS was later challenged by work that showed how PDZD8 localizes to the interface between ER and lysosomes [147].Interestingly though, the MCS with lysosomes was often seen close to ER-mitochondrial contacts, suggesting a functional link between the two MCSs.Whether this occurred preferentially during starvation was not assessed.
Lastly, a third pathway may mediate lipid transport between lysosomes and the ER which depends on VPS13C, a member of evolutionarily conserved RGB-Motif containing LTPs acting like bridges at several MCSs (Fig. 2) [148][149][150].
Baker's yeast establishes contact between the vacuole/lysosome and the mitochondria by three different MCSs.One of them involves yeast Vps13, which binds to Mcp1 on mitochondria and is redundant with the ERMES complex [151].A second MCS called vCLAMP connects the vacuole to mitochondria by Vps39, a subunit of the HOPS complex and the central component of the outer mitochondrial membrane import machinery Tom40.Vps39 is recruited to the vacuole by the small GTPase Ypt7 [152][153][154].A recent set of preprints further discovered a third interface between these organelles called vCLIP [155,156].Although the function of these interfaces is not yet completely clear they all seem to be under the control of nutrient signaling and are probably involved in metabolic homeostasis, which is especially important during nutrient starvation.Whether mitochondrial morphology is dependent on any of these MCSs remains to be investigated.However, the cristae morphology seems to depend on the vCLAMP which may organize similar functional specialization of mitochondria during nutrient stress as seen when mammalian mitochondria undergo IMM expansion and hyperfusion [152].
In mammalian cells, the most likely scenario for inter-organelle lipid transport that leads to starvationinduced mitochondrial fusion is that autophagic FAs are first released from lysosomes and are transported to the ER.In the ER, they enter lipid synthesis pathways.The MCS may perhaps involve PDZD8 as an LTP and Protrudin as a Rab7-dependent tether.Next, the ER uses the incoming FAs as precursors for PA production which then can enter PC and PS synthesis.RMDN3 may then recruit ORP5 and 8 to the interface between the ER and mitochondria and initiate lipid exchange between the two organelles transporting PS to mitochondria.Additionally, cells may employ MIGA2 at the ERMCs to transport newly synthesized PS to the mitochondria where it is converted into PE contributing to the expansion of the IMM (Fig. 2).The PE is possibly used in two different downstream reactions: it may contribute to creating the correct membrane environment for the CL transacylation reaction catalyzed by Taffazin, and it may additionally regulate ETC components in the IMM, as well as the fission-fusion machinery.
PE and CLs together destabilize a rhomboid protease within the IMM called OMA1 [17], which usually cleaves full length, membrane anchored OPA1 into a shorter soluble form that is released from the IMM into the inter-mitochondrial membrane space (IMS).Regulation of OMA1 activity is essential to control mitochondrial fusion.The starvation-induced hyperfusion of mitochondria may directly be linked to inhibition of OMA1 activity by lipids like PE and CL that are produced from precursors imported through an MCS with the ER.The same lipids that destabilize OMA1 are known to allosterically activate OPA1 and therefore promote mitochondrial fusion.
There is some evidence from in vitro work that RMDN3 is a PA transfer protein [157], which would fit the need for import of PA as CL precursor (Figs 1  and 2).PC may be imported through STARD7 which probably delivers both, lipid for bulk membrane expansion and provides specific PC species for CL remodeling [58].Interestingly, STARD7 is also proteolytically processed in a nutrient stress dependent fashion by an inner mitochondrial rhomboid protease called PARL.When the membrane potential in the mitochondria drops, STARD7 import stalls and processing through PARL produces a soluble STARD7 cleavage product that slides back into the cytoplasm which appears to import PC from the ER into the mitochondria.PC transport across the IMM is also mediated by STARD7.The IMS form is produced by successive cleavage by the metalloprotease MPP in the mitochondrial matrix which removes the mitochondrial targeting sequence from STARD7 [58].At this stage, the C terminal portion of STARD7 has already slid through the TOM complex in the OMM, and the following cleavage by PARL releases a different form of soluble STARD7 into the IMS.Here, it transports PC between the IMM and OMM (Fig. 2).
In summary, STARD7, MIGA2, RMDN3 and ORP5/8 may orchestrate the necessary lipid transport from the ER to mitochondria for membrane expansion during the fusion into longer mitochondria.
The mitochondrial associated LD as a nutrient source or savior?
Once the mitochondria have expanded their respiratory system and are fused into long organelles, they are well prepared for efficient OXPHOS.The most logical next step would be to store the remaining autophagic FAs as triacylglycerols (TAGs) within LDs.A prominent model in the field posits that the lipase ATGL mobilizes FAs from the LDs and makes them available for b-oxidation in the mitochondria (Fig. 2).Transport of FAs from the LD to the OMM is envisioned to depend on PLIN5, which is a LD surface protein and doubles up as a LTP and a mitochondrial tether [158].PLIN5-delivered FAs are then thought to be activated by the CoA-synthetase FATP4, which was recently shown to localize to mitochondria [158].This work identifies FATP4 as a PLIN5 interaction partner and suggests that it is the mitochondrial factor in this MCS.It is currently unclear how FATP4 localizes to the mitochondria.A detailed study that has investigated the subcellular localization of FATPs has shown that FATP4 colocalizes to the ER and excludes the possibility that it is recruited to the mitochondria [159].However, these experiments were carried out with GFP-tagged fusion protein and the localization of the endogenous protein might be different.Perhaps the localization of the endogenously expressed protein is under the control of metabolic signaling and starvation may shift FATP4 from the ER to mitochondria.
Phosphorylation of PLIN5 by cAMP-dependent protein kinase (PKA) appears to stimulate FA transport towards mitochondria, fitting the idea that starvation-induced signaling regulates lipid transport at MCSs.FATP4 would then convert the delivered FAs into FA-CoA which is handed over to CPT1 and imported into mitochondria for b-oxidation.This is an interesting model because most of the literature suggests that FATP4 would be an anabolic CoAsynthetase important for activation of FAs that are used for membrane lipid and TAG synthesis [160] and would not be involved in activating FAs for oxidation.The classical catabolic CoA-synthetase on mitochondria is ACSL1 [161,162].A possible explanation for the 're-purposing' of FATP4 may be found in its localization at the LD-mitochondrial MCS (Fig. 2).ACSL1 may be more specialized in activating fatty acids that reach mitochondria through soluble FA-binding proteins.FATP4 on the other hand may only activate FAs for b-oxidation when they reach mitochondria through the MCS with LDs which is specifically formed during starvation conditions.It would be interesting to investigate whether PLIN5 or FATP4 knockout cells would still activate the stress-induced hyperfusion of mitochondria.
As mentioned above another function of the LDs in stress-induced hyperfusion of mitochondria is possibly found in their ability to neutralize toxic lipids such as accumulating FA-carnitines.A situation where oxidation rates slow down would perhaps lead to a backlog of these import intermediates.As discussed above, LDs may prevent an overflow of the import machinery and safeguard the cell from toxic accumulation of the FA-carnitines (Fig. 2).

Conclusions
Organelle-networks as dynamic cellular neighborhoods that define cellular states Regardless of the actual function of cellular compartments like LDs, ER, mitochondria, and the lysosomes during the nutrient-stress response, it is becoming increasingly evident that organelles not only adapt their functions dynamically to changing environmental and internal cues but achieve this primarily through intense inter-organelle crosstalk enabled by highly regulated organelle partnerships.
Accumulating evidence supports the notion that organelles rarely function in isolation and that they often exchange information and metabolites as well as signaling molecules with several other organelles.MCSs establish the necessary interfaces.The machinery that brings organelles into close apposition and mediates the exchange of metabolites, information and building blocks is highly regulated.In fact, MCS plasticity which is under the control of post-translational modifications like phosphorylation or ubiquitination may have well been the decisive feature that made cellular pathways which are split across different compartments more successful than less complex versions with the same biochemical output.
Tuning organelle function through inter-organelle crosstalk at MCS is a powerful mode of adaptation which is universally applicable.Re-arranging organelle contacts during starvation stress and mounting a tailored response to changing nutrient conditions is only one example where regulated remodeling of MCS are important.Similar scenarios will also apply during the functional implementation of transcriptional instruction from differentiation programs.While the organelles adopt new functions as cells differentiate from less specialized precursors, they in turn define through their collective remodeling novel cell states, for instance when cells transition from one differentiation stage to the next.
Specialization of cells, commitment to their new identity, and adaptation of metabolism to nutrient stress is only possible through concerted remodeling of organelle networks.Regulated exchange of metabolites, lipids, and signaling molecules at MCSs is at the core of the machinery that mediates underlying metabolic adaptation and controls cellular fate during differentiation.

1276FEBS
Letters 598 (2024) 1274-1291 ª 2024 The Authors.FEBS Letters published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.

1278FEBS
Letters 598 (2024) 1274-1291 ª 2024 The Authors.FEBS Letters published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.