Creating and sensing asymmetric lipid distributions throughout the cell

Eukaryotic cells have a compartmentalized architecture based on membranes composed of lipids and proteins. These lipids are very diverse but mostly belong to three classes: phospholipids, sphingolipids and sterols. In addition to serving as building-blocks — they organize into bilayers to constitute the cellular membranes — lipids play functional roles through interactions with specialized proteins. All are precisely distributed between the plasma membrane (PM), which is the limiting membrane of the cell, and the membranes that delimit its internal compartments i.e. organelles like the endoplasmic reticulum (ER) or Golgi apparatus [1,2]. A general hallmark of eukaryotic cells is the asymmetric distribution of different lipids along the secretory pathway, i.e. between a region corresponding to the ER and cis-Golgi and a region comprising the trans-Golgi, endosomes and the PM (Figure 1). This defines two membrane territories with different lipid composition and contrasting features, in terms of both fluidity, thickness and surface charge [1,2]. As a consequence, these territories can harbour specific molecular mechanisms that ensure local signalling pathways, protein exchange between organelles via particular vesicular trafficking routes, and the organization of the cortical cytoskeleton. What are the nature and specific roles of the lipids that are asymmetrically distributed?

Known examples are sterols and sphingolipids, which represent, respectively, ∼12–20% and 7.5–10% of membrane lipids in eukaryotic cells [3,4]. Sterol accounts for less than 5% of lipids in the ER but 10% in the trans-Golgi membrane and up to 40% in the PM [5–8]. Similar concentration gradients of sphingolipids have been observed [5,8]. This co-distribution of sterol and sphingolipids is critical for the cells, notably because, as they associate preferentially, these lipids increase the rigidity and thickness of membranes in which they are abundant [9]. This allows the PM to form an impermeable barrier between the cell interior and the external milieu but also to have a precise inner organization [10].

Another lipid is phosphatidylserine (PS), a negatively charged phospholipid species that accounts for 2–10% of cellular lipids [3,4,11]. In the ER membrane, its proportion among lipids is 3–7%, and is multiplied by 4–5 in the PM [6,12,13]. PS mostly concentrates in the inner layer of this membrane where it contributes negative charges and recruits cytosolic proteins with signalling functions via electrostatic interactions [14,15]. The last examples are phosphoinositides, a group of seven phospholipid species that represent ∼1–2% of lipids and derive from phosphatidylinositol (PI). They are absent from the ER and are mostly found in the trans-Golgi and endosomal membrane, and the PM, where they play pivotal roles in vesicular trafficking, cytoskeleton organization or signalling pathways [16,17].

Considerable efforts have been made to explain the origins of these asymmetries in lipid distribution. One goal is to understand how lipid species are enriched in the trans-Golgi or PM while being partially or entirely created in the ER. Another aim is to define how these asymmetries are maintained despite the bulk exchange of lipids between organelles by vesicular trafficking, but also fast lipid enzymatic conversion and uptake.

Current data suggest that lipid asymmetry within cells is at least partially created by small networks composed of lipid-synthesizing enzymes, located in distinct cellular membranes, and lipid transfer proteins (LTPs) that flow lipids between these membranes. LTPs belong to diverse families, but all have a domain with a cavity to encapsulate one or several lipids [18,19]. This allows these proteins to extract lipids from an organelle membrane and shield these hydrophobic molecules from the aqueous medium (i.e. cytosol) for delivery to a second membrane. Remarkably, these networks are often organized around contact sites where the membranes of two distinct organelles are less than 30 nm apart [20], which is an extremely short distance at the cellular scale.

A known network is at the core of sphingolipid asymmetry in human cells. It involves enzymes that synthesize a lipid called ceramide in the ER and an enzyme that converts ceramide into sphingomyelin in the Golgi membrane. Ceramide travels from the ER to the Golgi via CERT, an LTP present in contact sites between these organelles [21]. CERT could move ceramide bidirectionally between the two membranes, but the metabolic trapping of ceramide, i.e. its conversion into sphingomyelin creates the conditions for a one-way transfer. Indeed, there is always more ceramide in the ER than in the Golgi membrane and, following the law of mass action, CERT picks up ceramide where is it abundant to release it where it is scarce (Figure 2a).

Other networks that generate asymmetry involve LTPs with a lipid exchange activity (ORPs in humans, Osh proteins in yeast). They have different molecular configurations, but all comprise a domain called ORD (OSBP-related domain) with a lipid-binding pocket [22]. Biochemical studies have revealed that Osh4, a representative member of this protein family, consisting only of an ORD, could host either a molecule of ergosterol [23] (i.e. main yeast sterol) or PI4P [24]. This enables Osh4 to exchange these ligands between two membranes [24,25]. Like other phosphoinositides, the intracellular distribution of PI4P is programmed by the organelle-specific localization of enzymes involved in its synthesis and degradation [16,17]. In the trans-Golgi membrane and the PM, PI 4-kinases add a phosphate group to PI to form PI4P, whereas in the ER the Sac1 protein hydrolyzes the bulk of PI4P into PI. Thus PI4P is mostly present in the trans-Golgi and PM, where it serves as a molecular signpost, but another consequence is that PI4P concentration gradients are created between these compartments and the ER. Data obtained with Osh4 have suggested a scenario in which this PI4P asymmetry is used by this protein to generate a sterol asymmetry [24].

In this model, Osh4 extracts a sterol molecule from the ER membrane in which the lipid is synthesized. Then it exchanges sterol for PI4P at the trans-Golgi, carries PI4P to the ER and takes another molecule of sterol. The maintenance of a PI4P gradient by the Golgi PI 4-kinase and Sac1 allows for multiple exchange cycles and the build-up of sterol in the Golgi membrane (Figure 2b). In vitro assays have supported this idea, first by showing that Osh4 transfers sterol and PI4P along opposite routes at maximal speed when these lipid ligands are in two distinct membranes; this is because Osh4 preferentially unloads a ligand in a membrane in which the second ligand is present. Perhaps more importantly, this has revealed that Osh4 could generate a sterol gradient between these membranes by dissipating a PI4P gradient [25]. Subsequent cellular studies on OSBP, the founding member of the ORP/Osh family, have endorsed this model, showing that OSBP plays a substantial role in building asymmetric sterol distribution by sterol/PI4P exchange [26,27]. Observations that the absence of Sac1 results in the accumulation of PI4P in the ER have further supported this model [28].

Startlingly, a similar exchange process contributes to PS asymmetry in cells. Like sterol, PS originates from the ER [15] and concentrates in the PM. How this was accomplished was hardly known [13] until the finding that Osh6 and Osh7 were PS/PI4P exchangers [29,30]. In yeast cells, these proteins harness the asymmetry in PI4P at the ER/PM interface to supply the PM with PS (Figure 2c). In human cells, ORP5 and ORP8 perform the same function [31]. The fact that PI4P requires energy to be synthesized (i.e. ATP) and is consumed in exchange cycles has led us to consider PI4P as a universal ‘fuel' used by ORP/Osh proteins to create diverse lipid asymmetries throughout the eukaryotic cells [32].

Thus, the combination of lipid transfer with lipid conversion events allows establishing one-way fluxes of lipids between membranes and create lipid asymmetries. Besides this, the features of organelle membranes likely play significant roles. The ER membrane is fluid and mostly organized into tubules, and is, therefore, highly curved [33]. These characteristics might augment the propensity of lipids to leave the ER membrane and be extracted by LTPs, as suggested in vitro for sterol and phospholipids [25,34,35]. Also, interestingly, some LTPs specifically localize on highly curved areas of the ER membrane [36,37]. In contrast, in the trans-Golgi membrane and the PM, the privileged association of sterol with sphingolipid presumably acts as a thermodynamic trap that prevents any sterol return to the ER [38]. Corroborating this idea, kinetic assays have shown that the delivery of sterol by sterol/PI4P exchange is enhanced in sphingolipid-rich membranes [25]. It is not clear whether sterol, which can associate laterally with PS in the PM [39,40], promotes the enrichment of this membrane with PS supplied by an ORP/Osh-mediated exchange [41].

Many LTPs involved in lipid asymmetries are in contact sites for reasons that remain enigmatic. Possibly, the narrow gap between organelles allows for optimal transfer speed, but this idea is disputed and has rarely been addressed experimentally [42]. More likely, this LTP presence at contact sites guarantees local coupling between processes that make/modify lipids and those that transfer them, and, obviously, a high degree of accuracy, as lipids cannot be delivered by mistake to another organelle. Quite certainly, this serves to establish mechanisms that regulate lipid fluxes. The major one is a negative feedback loop encoded in the structure of many LTPs, that controls how long these proteins stay in contact sites and transfer lipids. First evidence has been provided by studies of OSBP in ER–Golgi contacts.

OSBP is a sterol/PI4P exchanger more complex than Osh4 that can connect the ER with the trans-Golgi membrane in addition to transferring lipids (Figure 3a); it comprises a short amino-acid motif to bind with the ER-resident protein VAPs [43] and a structural domain called PH (Pleckstrin Homology) to dock onto the trans-Golgi by recognizing PI4P [44]. It is surprising that OSBP depends on PI4P to associate with the Golgi membrane while having the ability to extract and transfer this lipid. This is in fact critical to implement a negative feedback loop that controls lipid transfer at ER–Golgi contacts [26]. As long as PI4P is available in the Golgi membrane, OSBP forms ER–Golgi contacts and exchanges PI4P for sterol, leading to the accumulation of cholesterol in the Golgi membrane. Once PI4P is exhausted, due to its transfer to the ER and hydrolysis by Sac1, OSBP dissociates from the Golgi membrane, exits contact sites and becomes inactive.

Moreover, via its exchange activity, OSBP controls the transfer activity of CERT, which also depends on PI4P to populate ER–Golgi contacts and be operative [45]. Consequently, the ceramide and cholesterol flows are synchronized, allowing for a strict control of the sterol/sphingolipid balance in the Golgi apparatus. Additional processes tightly regulate these flows. Notably, the accumulation in the Golgi membrane of sterol and DAG, a by-product of sphingomyelin synthesis, modulates both the synthesis of PI4P and the activities of CERT and OSBP [27,45–51]. OSBP must be spatially close to PI 4-kinases; otherwise lipid exchange is decoupled from PI4P synthesis, which results in uncontrolled PI4P waves along the Golgi apparatus and aberrant dynamics in ER–Golgi contacts formation [27]. Also, Sac1 associates with OSBP in complex with VAP [52]. Thus, in ER–Golgi contacts, OSBP and CERT are confined together with enzymes of the PI4P metabolism and auxiliary proteins to determine the lipid composition of the Golgi apparatus. This is critical for the formation of transport vesicles that carry proteins from the Golgi to the PM [52] and has an impact on the inner organization of the PM [53].

Finally, as suggested by these data, a reason why LTPs are secluded in contact sites is probably to combine their activities with those of local protein partners, and thereby perform specialized cellular functions. Recent studies have confirmed this notion, showing that diverse ORP/Osh proteins exchange lipids at contact sites to support specific vesicular trafficking routes. In yeast, Osh2 functions as a sterol/PI4P exchanger in ER–PM contacts, where it associates with Myo5, a protein required for actin polymerization and scission of endocytic vesicles (Figure 3b) [54]. Myo5 also interacts with ER-resident proteins involved in ergosterol biosynthesis [55]. Presumably, all these proteins co-operate to channel ergosterol to the PM and create sterol-rich membrane domains where actin polymerization is facilitated to promote endocytosis [55]. In contrast, it is likely that Osh1, which occupies ER–Golgi contacts, performs sterol/PI4P exchange to regulate post-Golgi vesicular trafficking [56]. In human cells, ORP10 exchanges PS for PI4P in ER-endosome contacts to deliver PS to the endosomal membrane (Figure 3c). This promotes the recruitment of the EHD1 protein whose membrane scission activity triggers the formation of tubulovesicular carriers that recycle proteins from the endosomes to the Golgi [57].

If today we are aware of the existence of protein networks that ensure one-way lipid transfers, it remains difficult to quantify how much they contribute to the impressive asymmetries in lipid distribution observed in the cell. First, although many approaches have been developed to map cellular lipid routes, for instance by using fluorescent lipids [58], lipid labelling procedures [58] or genetically encoded lipid sensors [58,59], these methods are not entirely suitable for precisely quantifying lipid fluxes. Second, in vitro assays provide key insights into the function of LTPs, but they are not realistic enough to define the exact speed at which these proteins carry lipids in the complex environment of the cell [60].

Attempts have been made to estimate how much lipid is transferred by given LTPs in cells considering their abundance and the speed at which they transfer their lipid ligands in vitro [25,60]. Calculations have also been made in specific cases in which the cellular abundance and contribution of LTPs to the synthesis of particular lipid species via the transport of their precursor were known [60]. Regarding ORP/Osh activity, insights came from the analysis of cellular PI4P levels. This lipid is present in minute amounts in the cell (in yeast, only 0.2% of lipids [61]) compared with sterol and PS, which seems incompatible with ORP/Osh-mediated exchange processes. Yet, in yeast lacking Sac1 or all Osh proteins, PI4P levels can be multiplied by up to ∼20 [62,63]. This suggests that substantial quantities of PI4P are constantly produced and consumed by networks involving Sac1 and ORP/Osh proteins to move lipids. In parallel, Osh6 and Osh7 were found to be responsible for 30% of PS delivery to the PM [29]. Later, investigations revealed that OSBP uses half of PI4P synthesized in a human cell to mediate 30–60% of ER-to-Golgi sterol transfer [27]. Thus ORP/Osh proteins seem able to transfer significant amount of lipids for creating lipid asymmetries.

However, in different cases, the raison d’être of ORP/Osh proteins is perhaps not to use the PI4P metabolism to deliver massive quantities of lipids to cellular membranes. Osh4 served as a great prototypical case to uncover PI4P-driven lipid exchange, but paradoxically there is no direct evidence besides a few clues that it delivers sterol in the Golgi in yeast cells [5,64]. Instead, it seems more certain that its role is to lower PI4P levels in post-Golgi secretory vesicles for functional purposes [65,66]. Also, OSBP was found to localize at ER-endosome contacts to control protein recycling not by delivering sterol but by adjusting the level of a small pool of endosomal PI4P [67]. Possibly, these proteins perform limited sterol/PI4P exchange to regulate the PI4P level of given membranes or control their sterol/PI4P balance. Another idea is that a LTP like Osh4 acts as a negative regulator of PI4P signalling by sequestering this lipid, and that a high concentration of sterol lifts its inhibitory action [68].

Moreover, studies of ORP5 and ORP8 in humans have shown that besides supplying the PM with PS, they also regulate the level of the main signalling lipid present in this membrane, which is PIP₂. These proteins are ER-anchored [69,70] and associate with the PM if the latter contains threshold PI4P and PIP₂ amounts, to join ER–PM contacts [31,71,72]. The PM is then enriched in PS but its content in PIP₂ decreases [71,72]. This is because PI4P, which is used as fuel for PS/PI4P exchange also serves as substrate for PIP₂ synthesis (it is unlikely that PS and PIP₂ are directly exchanged [41,72]). When PI4P and PIP₂ levels become too low, the proteins exit contact sites, PS transfer stops, and proper phosphoinositide levels are restored. So, ORP5 and ORP8 seem to constitute the molecular basis of a complex mechanism whose function is more to finely adjust the lipid composition of the PM, and thereby the signalling capacity of the cells [72], than to massively import PS in the PM (Figure 4). Side mechanisms are thought to regulate PS and PI4P flows. For instance, PS inhibits its own synthesis in the ER and this strongly influences PS/PI4P exchange [73]. Sac1 can be activated by PS [74], suggesting that an elevation of PS in the ER could generate a steeper PI4P gradient to sustain lipid exchange. PS might also promote the conversion of PI4P into PIP₂ by PI4P 5-kinase in the PM [75]. It is increasingly appreciated that dysregulation of this complex protein network can impact the recruitment of oncogenic proteins (e.g. KRAS) at the PM, and be associated with cancers [76].

Some networks involving LTPs are clearly tailored not to create lipid asymmetries but to sense a change in asymmetry and subsequently rectify the cellular level of particular lipids. One of these is critical to guarantee the signalling competence of the cell. In many signalling pathways, the activation of cell surface receptors triggers the hydrolysis of PIP₂ by phospholipase, which generates DAG and IP₃ as second messengers. PIP₂ levels must subsequently be restored so that the cells can respond once again to stimulation. This requires several enzymatic steps that constitute the PI cycle [77]. First, DAG is converted into phosphatidic acid (PA) in the PM. Next, PA is converted into PI in the ER membrane. Finally, in the PM, PI undergoes two successive phosphorylations leading to PIP₂ re-synthesis (Figure 5a). As hypothesized as early as 1975, the PI cycle can function only if PA and PI are exchanged between the ER and the PM to guarantee lipid flows between the spatially distant enzymes of this cycle [78]. As described below, long-term efforts have led to the identification of LTPs — Nir2 and Nir3 in humans — as the missing pieces of the puzzle,

These proteins are attached to the ER [79] and have a domain to alternatively capture PA and PI [80]. In addition, they possess a PA-binding domain and likely a DAG-binding domain to associate with the PM [81,82]. Recent data suggest the following scenario. The generation of DAG and PA triggers the translocation of Nir proteins to ER–PM contacts [82]. The accumulation of PA in the PM creates an asymmetry in PA at ER–PM contacts. Because PI is absent from the PM and already abundant in the ER [83,84], this likely sets the conditions for immediate PA/PI exchange as PA and PI are concentrated in distinct membranes. Next, PA is converted into PI in the ER, which sustains the exchange process until a full clearance of PA [82] and replenishment of the PIP₂ stock. Finally, as PA is cleared from the PM, Nir proteins disengage from contact sites. Interestingly, compared with Nir2, Nir3 has a weaker PA-transfer capacity but a higher capacity to detect small amounts of PA in the PM [85]. Consequently, Nir3 adjusts the PIP₂ level in resting cells, whereas Nir2 is only mobilized during acute receptor activation.

A second example was provided by the characterization of a newly discovered family of LTPs (Asters in humans, Lam/Ltc in yeast). These are ER-anchored proteins that have a PH-like domain called GRAM to associate with a second organelle and a sterol-transfer module [86,87]. Studies have shown that these proteins occupy contact sites between the ER and the mitochondrion, vacuole or PM to potentially serve as sterol transporters [86,87]. Remarkably, the GRAM domain of Asters (also called GRAMD1) has two binding sites for recognizing cholesterol and PS in a synergistic manner [88–90]. Consequently, in human cells, Asters can specifically join ER–PM contacts by sensing in the PM, in which PS is present, the transient expansion of an accessible pool of cholesterol resulting from the uptake of sterol by endocytosis of low-density or high-density lipoproteins [88,90–92]. Next, Asters transfer cholesterol down its concentration gradient to the ER; cholesterol is subsequently esterified to be stored in lipid droplets [91,92] (Figure 5b). Additionally, the sterol level increase in the ER down-regulates via the SREBP2 pathways the transcription of gene coding for the synthesis of cholesterol and its uptake [88,89,91,92]. Remarkably, Asters use their transfer activity to act as messengers. They inform the regulatory mechanisms that control the biosynthesis and uptake of sterol, present in the ER, of the overabundance of cholesterol in the PM, preventing the toxic accumulation of this lipid in the cell. This has important implications, since defects in this mechanism can result in a lack of steroid hormone synthesis in model animals [91] and might be linked to intellectual disability in humans [89].

LTPs assisted by lipid-synthesizing enzymes vectorially transfer lipids between organelles. It is critical to better quantify whether this serves to create lipid asymmetries throughout the cell and/or to fine-tune the lipid content of organelles. More work is also required to better define which lipid ligands are recognized by different LTPs and/or whether these LTPs function as lipid exchangers or mere lipid transporters. In some cases, the mechanism by which an LTP accumulates lipids in organelles remains particularly elusive [93]. In parallel, significant efforts are being undertaken to dissect networks present at ER-mitochondrion contacts and inside the mitochondria that contribute to the generation of key cellular lipids, among which specific mitochondrial ones [94]. The recent discovery of diverse LTPs at the mitochondrial surface raises exciting questions [95,96]. Finally, some evidence suggests that vesicular trafficking also contributes to creating lipid asymmetry via lipid sorting mechanisms whose characteristics remain to be defined [5,97,98]. In the future, a better understanding of how asymmetries in lipid distribution are created should provide key outcomes in basic cell biology and answer health-related questions.

• LTPs assisted by lipid-synthesizing enzymes can move lipids in a directional manner between two cellular membranes.

• Lipid transfers mostly occur in contact sites where membranes are extremely close for several functional purposes.

• These transfer processes contribute to creating asymmetric lipid distributions in the cell and/or fine-tuning the lipid composition of cellular membranes.

• Some LTPs play a key role in mechanisms that rectify the cellular level of particular lipids following changes in lipid asymmetry.

The author declares that there are no competing interests associated with this manuscript.

This work was supported by the Agence Nationale de la Recherche (ANR-20-CE13-0030-02 grant) and the CNRS.

G.D. conceptualized and wrote the manuscript.

I am grateful to Ms. Y. Van Der Does for her careful corrections and proofreading of the manuscript.

CERT

ceramide transfer protein

DAG

diacylglycerol

ER

endoplasmic reticulum

LTP

lipid transfer protein

ORD

OSBP-related domain

ORP

oxysterol-binding protein related-proteins

Osh

oxysterol-binding homology

PA

phosphatidic acid

PH

Pleckstrin Homology

PI

phosphatidylinositol

PI4P

phosphatidylinositol 4-phosphate

PIP₂

phosphatidylinositol 4,5-bisphosphate

PM

plasma membrane