Seeing the Membrane from Both Sides Now: Lipid Asymmetry and Its Strange Consequences

Abstract

Almost all biomembranes are constructed as lipid bilayers and, in almost all of these, the two opposing monolayers (leaflets) have distinct lipid compositions. This lipid asymmetry arises through the concerted action of a suite of energy-dependent enzymes that maintain living bilayers in a far-from-equilibrium steady-state. Recent discoveries reveal that lipid compositional asymmetry imparts biophysical asymmetries and that this dualistic organization may have major consequences for cellular physiology. Importantly, while transbilayer asymmetry appears to be an essential, near-ubiquitous characteristic of biological membranes, it has been challenging to reproduce in reconstituted or synthetic systems. Although recent methodological developments have overcome some critical challenges, it remains difficult to extrapolate results from available models to biological systems. Concurrently, there are few experimental approaches for targeted, controlled manipulation of lipid asymmetry in living cells. Thus, the biophysical and functional consequences of membrane asymmetry remain almost wholly unexplored. This perspective summarizes the current state of knowledge and highlights emerging themes that are beginning to make inroads into the fundamental question of why life tends toward asymmetry in its bilayers.

LIVING MEMBRANE ASYMMETRY: PAST AND PRESENT

One of the foundational facts of biology is that cells are enveloped and internally subdivided by fluid lipid bilayers. Shortly after the widespread acceptance of the elegantly named “fluid mosaic” model, cracks began to show in the tile, revealing that living membrane organization is much more intricate and variable than a homogenous two-dimensional patchwork. From lipidome complexity (Shevchenko and Simons 2010; Harayama and Riezman 2018) to lateral domains (Sezgin et al. 2017; Levental et al. 2020), to partitioning (Kusumi et al. 2012; Fujimoto and Parmryd 2016) and shaping (Bonifacino and Lippincott-Schwartz 2003; Simunovic et al. 2019) of membranes via proteins, cell membranes are textured and structured at nearly all length scales and involved in nearly all physiological functions. This perspective focuses on lipid asymmetry (i.e., distinct compositions of the two bilayer leaflets), a facet of membrane complexity that was initially described more than 50 years ago (Bretscher 1972; Verkleij et al. 1973), yet whose form and function remain mysterious to this day. Lipid asymmetry was originally discovered by experiments using enzymes or chemicals that could not penetrate the bilayer and thus could only modify external leaflet lipids (Bretscher 1972; Verkleij et al. 1973). These studies revealed that the outer leaflet in mammalian erythrocytes was composed nearly exclusively of relatively unreactive choline-headgroup lipids (phosphatidylcholine [PC] and sphingomyelin [SM]), whereas charged and amino-headgroup lipids (phosphatidylserine [PS], phosphatidylinositol [PI], and phosphatidylethanolamine [PE]) were confined to the inner, cytoplasmic plasma membrane (PM) leaflet (Fig. 1; Devaux 1991). These chemical/enzymatic approaches have been supported and complemented by immunolabeling of freeze-fractured cell membranes (Murate et al. 2015), together suggesting that all mammalian PMs, almost all eukaryote PMs (with some exceptions [Shiomi et al. 2021]), and likely most cell membranes have some degree of lipid asymmetry (Kobayashi and Menon 2018; Doktorova et al. 2020).

Figure 1.

Organization of plasma membrane components. (A) Exoplasmic leaflets in mammalian plasma membranes are composed of phospholipids with relatively saturated acyl chains and uncharged lipid headgroups. This leads to more tightly packed and less mobile leaflets than their cytosolic counterparts, which are rich in anionic, unsaturated lipids. (B) Divergent results for cholesterol interleaflet distribution makes it challenging to piece together the precise organization of membrane components.

Since the 1980s, much of the effort in this field has focused on identifying and characterizing the protein machinery that establishes, maintains, and releases lipid asymmetry. We point the curious reader to a comprehensive recent review on these protein regulators (Sakuragi and Nagata 2023). Briefly, lipids are driven against their equilibrium (symmetric) distribution by ATP-dependent transporters known as “flippases” (which force lipids into the inner leaflet) and “floppases” (for outward translocation). There are likely many such enzymes with specificity for various headgroups. Asymmetry can then be released by relatively nonselective (Malvezzi et al. 2013) “scramblases,” which—despite the name—are not enzymes but rather lipid channels that form a hydrophilic passage within the hydrophobic bilayer core through which lipid headgroups can quickly traverse and equilibrate between leaflets (Kobayashi and Menon 2018). Particularly notable among these is the TMEM16F scramblase (Suzuki et al. 2010; Malvezzi et al. 2013), which is gated by calcium and thus potentially implicates lipid scrambling among the many facets of calcium-associated signaling.

BOTH SIDES NOW: BIOPHYSICAL AND FUNCTIONAL ASYMMETRY

While the foundational facts of lipid asymmetry have been known for decades, the physiological purpose of highly asymmetric lipid distributions remains unknown. This may be a surprising claim, as there have been extensive reports and reviews on functional roles of lipid asymmetry (Op den Kamp 1979; Fadeel and Xue 2009; Bevers and Williamson 2016; Kobayashi and Menon 2018; Doktorova et al. 2020). However, in almost all such cases, the cellular function in question is mediated by exposure of PS onto the outer leaflet (Bevers and Williamson 2016), that is, the loss of lipid asymmetry. PS exposure is a hallmark of apoptosis, required for coagulation, and involved in diverse functions including cell–cell fusion (Verma et al. 2018; Whitlock et al. 2018), viral infection (Husby et al. 2022), and synaptic pruning (Bevers and Williamson 2016). However, the notion that asymmetry exists to be released has a whiff of teleology. We argue that the significant expenditure of metabolic resources required to pump nearly all PM lipids far away from their equilibrium distribution suggests that asymmetry per se is likely functionalized by cells.

Recently, a detailed analysis extended the previously established chemical asymmetry of hydrophilic headgroups to physical asymmetries associated with lipid acyl chains (Fig. 1A; Lorent et al. 2020). Specifically, the acyl chains of lipids comprising the outer PM leaflet were found to be largely saturated (i.e., containing few double bonds), whereas inner PM leaflets were mostly polyunsaturated, with four or more double bonds per lipid. Acyl chain saturation determines how tightly lipids can pack, which in turn determines how fast they can move (diffusion), how easily they deform (stiffness), and how readily stuff can slip between them (permeation). Thus, since the two PM leaflets have different degrees of saturation, they may also be expected to have different physical properties. This hypothesis was confirmed by several independent measurements that converged on the conclusion that inner leaflets are less tightly packed and more diffusive than outer leaflets (Fig. 1A; Gupta et al. 2020; Lorent et al. 2020). Intriguingly, this transbilayer lipid-packing asymmetry correlates with asymmetric structures of protein transmembrane domains (TMDs) (Sharpe et al. 2010; Lorent et al. 2020), which are on average bulkier in the loosely packed inner leaflet and slimmer in the tightly packed outer leaflet. Because TMD structures can be inferred directly from sequences, their asymmetries can be predicted for any organism for which genomic information is available. This approach was used to make the striking prediction that structural asymmetry may be a feature common to most eukaryote PMs (Lorent et al. 2020).

LIKE A ROLLING STONE: CHOLESTEROL ASYMMETRY

The transbilayer distributions of membrane phospholipids have been extensively investigated and comprise one of the core dogmas of membrane biology, contrasting sharply with the lack of consensus about the asymmetry of the mammalian PM's most abundant component (∼40 mol% [Symons et al. 2021]): cholesterol (Fig. 1B; Steck and Lange 2018). This knowledge gap persists because sterols are chemically distinct from phospholipids in ways relevant for measuring their asymmetry. Specifically, the charged moieties of phospholipids prevent their rapid movement between leaflets (through the hydrophobic core), which is often called flip-flop and proceeds on the timescale of hours (Marquardt et al. 2017; Doktorova et al. 2019). Conversely, most membrane sterols have small uncharged “headgroups” (hydroxyl for cholesterol) and are flat and rigid, allowing them to easily and rapidly traverse the space between leaflets (Steck et al. 2002; Bennett et al. 2009; Bruckner et al. 2009; Steck and Lange 2012). In fact, this movement is so fast that there is little consensus about just how fast it is; most experiments are too slow to measure it accurately and estimates range from the microseconds to minutes timescale (Lange et al. 1981; Leventis and Silvius 2001; Steck et al. 2002; Bruckner et al. 2009; Steck and Lange 2012). This rapid equilibration between leaflets presents a unique difficulty in trying to measure cholesterol's residence in a single leaflet, since each cholesterol molecule samples both leaflets frequently and thus labeling with an external reagent (e.g., cholesterol oxidase) simply leads to all cholesterol being rapidly labeled (Lange et al. 1981).

Despite this daunting conundrum, many attempts have been made to measure cholesterol's asymmetric distribution in living PMs. Perhaps unsurprisingly, these attempts have yielded wildly divergent conclusions. As early as the 1970s, a report claimed up to three-fold enrichment in the outer leaflet (Fisher 1976), later contradicted by claims of the same degree of cholesterol enrichment, but in the cytoplasmic leaflet (Schroeder et al. 1991). Subsequent reports continued the discrepant trend, with convincing evidence for cytoplasmic (Igbavboa et al. 1996; Mondal et al. 2009; Solanko et al. 2018), exoplasmic (Bittman and Rottem 1976; Hale and Schroeder 1982; Ingólfsson et al. 2014), or neither (Lange and Slayton 1982) leaflet being cholesterol enriched, and some attempts to justify the differences based on cell types (Hale and Schroeder 1982). This trend persists to recent years, with two high-profile papers again conflictingly claiming that cholesterol is either highly enriched in the inner PM leaflet (Courtney et al. 2018b) or almost exclusively (>90%) in the outer leaflet (Liu et al. 2017). The latter report used an elegant assay, relying on sensor proteins engineered to bind membranes with exquisite sensitivity to their cholesterol content. These probes were used to probe both PM outer leaflets (via external labeling by purified probe) and inner leaflets (by expressing in the cytoplasm) (Liu et al. 2017; Buwaneka et al. 2021). The same approach was also used to implicate possible regulators of cholesterol asymmetry, identifying the transporter ABCA1 and the Hedgehog-morphogen receptor Patched (Zhang et al. 2018). While this clever design appears straightforward, several possible contaminating effects were identified that cast doubt on the core conclusions (Courtney et al. 2018a; Steck and Lange 2018), although some of these were addressed in follow-up experiments (Buwaneka et al. 2021). Equally concerning to technical artifacts was the conceptual puzzle of how a molecule that rapidly flips between leaflets could possibly be so highly enriched in one of them. Our recent experiments (Doktorova et al. 2023) and clever theory and simulations (Varma and Deserno 2022) from the Deserno group provide several possible answers: (1) the two bilayer leaflets are not absolutely constrained to have the same number of phospholipids—if one leaflet has more than the other, cholesterol will naturally “flip to fill the gaps”; and (2) cholesterol prefers to interact with lipids that reside on outer leaflets (sphingolipids, saturated acyl chains) rather than inner leaflet lipids (often highly unsaturated). These factors (and others) combine to determine cholesterol steady-state distribution in cells. Ultimately, while several lines of evidence lead us to predict that cholesterol is likely enriched on the PM outer leaflet, the exact asymmetry of cholesterol in any given context remains a matter of debate.

Ch-ch-CHANGES: INTERLEAFLET COUPLING OF MEMBRANE PROPERTIES

Biophysical studies of synthetic membranes have played a key role in the evolution of PM models. The vast majority of such studies have relied on symmetric bilayers, mostly because of the historic unavailability of asymmetric model systems. Although methods for preparing and characterizing model asymmetric bilayers are seeing rapid development (Cheng and London 2011; Lin and London 2014; Doktorova et al. 2018; Enoki and Feigenson 2019), they still present technical challenges (Doktorova et al. 2018). Thus, there remain substantial gaps in our understanding of the biophysical consequences of lipid asymmetry. A key factor determining such consequences is interleaflet coupling, which (at the most basic level) reflects the extent to which properties of different monolayers are influenced by their mutual interactions in a bilayer. Put another way, interleaflet coupling is the physical manifestation of the bilayer's desire to achieve a minimum free energy by exploring available degrees of freedom within the individual leaflets. These can include various types of lipid deformations (e.g., stretching, splay, tilt, protrusion, etc.) as well as degrees of freedom associated with lipid mixing within and between leaflets.

A useful experimental strategy for investigating interleaflet coupling is to compare a particular property measured in an asymmetric bilayer to that measured in a symmetric bilayer of identical or similar overall composition, with the signal preferably isolated to a single leaflet. In this way, various groups have investigated how asymmetry influences lipid order (Cheng and London 2011), packing (Heberle et al. 2016; Frewein et al. 2022), and lateral diffusion (Chiantia et al. 2011; Chiantia and London 2012), membrane bending rigidity (Elani et al. 2015; Lu et al. 2016; Rickeard et al. 2020) and compressibility (Lu et al. 2016), and the line tension (Enoki et al. 2021) and thermal stability (Cheng et al. 2009) of ordered domains. Perhaps unsurprisingly, the picture that emerges is complex and does not yield a simple recipe for predicting the behavior of asymmetric membranes. Three case studies exemplify the challenge of elucidating general rules for interleaflet coupling: (1) London and coworkers observed strong coupling of lipid diffusion in asymmetric vesicles with brain SM in the outer leaflet and mixed chain PC in the inner leaflet, evidenced by reduced lateral diffusion in both leaflets (Chiantia and London 2012). Curiously, the change in diffusion was not accompanied by increased order in the inner leaflet, suggesting that different properties can couple to different extents within the same bilayer. (2) The Ces group found anomalous stiffening of asymmetric giant vesicles: the bending rigidity of an asymmetric bilayer of DOPC (di18:1 PC) and POPC (16:0,18:1 PC) leaflets was more than twofold greater than that of either symmetric bilayer (Elani et al. 2015). (3) Eicher et al. discovered remarkable differences in coupling depending on the sidedness of the asymmetry. Using asymmetric liposomes in which the inner leaflet was either POPC or POPE and the outer leaflet a mixture of the two, strongly coupled leaflets (evidenced by a single melting transition and similar lipid packing in both leaflets) were observed when the inner leaflet was POPE, but an apparent lack of coupling (i.e., two independent melting transitions) was found when the inner leaflet was POPC (Eicher et al. 2018). Moreover, neither asymmetric bilayer showed evidence for coupling at higher temperatures, where the membrane was a fully melted fluid. If any robust conclusion can be drawn from these and other studies, it is that the extent of interleaflet coupling of membrane structural and material properties depends not only on lipid composition, but also other factors (e.g., temperature, curvature, relative packing).

PHASED AND CONFUSED: COUPLING OF PHASE BEHAVIOR

On timescales where lipid translational degrees of freedom are relevant (i.e., >microseconds for fluid phases), interleaflet coupling can also influence the spatial organization of lipids—and potentially induce phase changes—in either or both leaflets (Fig. 2). Because of the relevance of asymmetry and domain coupling to PM rafts, many experimental and theoretical studies of asymmetry have focused on a scenario involving leaflets that would exhibit disparate phase behaviors in symmetric bilayers. Such phase asymmetries could hypothetically be related to the steady-state arrangement of mammalian PMs (Lorent et al. 2020), whose cytoplasmic leaflet is composed largely of unsaturated glycerolipids (which would not be expected to form ordered phases) and whose exoplasmic leaflet composition—rich in SM, lower melting phosphocholine lipids, and cholesterol—would seem to fall squarely within a two-phase coexistence regime (Fig. 2; Veatch and Keller 2003; Heberle et al. 2010). Experimental studies of this sort have produced outcomes ranging from no apparent coupling of phase behavior, to induced lipid segregation in the model cytoplasmic leaflet, to suppressed lipid segregation in the model exoplasmic leaflet (Crane et al. 2005; Kiessling et al. 2006; Garg et al. 2007; Collins and Keller 2008; Wan et al. 2008; Lin and London 2015; Heberle et al. 2016; Wang and London 2018; Enoki and Feigenson 2019, 2022; Enoki et al. 2021). While the sheer variety of membrane geometries (i.e., supported and unsupported planar bilayers vs. vesicles of various sizes), leaflet compositions, and experimental techniques used in these studies makes it difficult to tease out systematic trends and organizing principles, it is clear that the full spectrum of coupling scenarios are possible in biomimetic membranes. These observations make an urgent call for studies of interleaflet coupling in which the lipid compositions of both leaflets are systematically varied. A further exciting direction would be evaluating the effects of bilayer spanning components (i.e., proteins), which would likely enhance coupling.

Figure 2.

A thermodynamic model for interleaflet coupling of phase behavior. In mammalian plasma membranes (PMs), the outer leaflet contains a mixture of saturated and unsaturated lipids, while the inner leaflet contains polyunsaturated lipids (the transbilayer distribution of cholesterol is uncertain, but likely enriched in the outer leaflet). Symmetric models of outer leaflet mixtures have a strong tendency to form coexisting ordered and disordered domains, while inner leaflet models tend to form a uniform disordered phase. The magnitude of interleaflet coupling is determined by a balance between line tension that favors phase separation in the outer leaflet, and interleaflet surface tension that opposes it. If line tension dominates (i.e., weak coupling), lipid organization within each leaflet reflects that of cognate symmetric bilayers. If surface tension dominates, both leaflets are uniformly mixed. The lipid raft hypothesis may represent an intermediate between these extremes, with clustering or phase separation of outer leaflet lipids potentially inducing segregation of inner leaflet lipids.

A key parameter determining interleaflet coupling of domains is the surface tension at the midplane interface of a mismatched or “anti-registered” domain (e.g., one with an L_o-like composition in one leaflet and an L_d-like composition in the other) (Fig. 2). This surface tension, often called mismatch free energy γ (units of k_BT/nm²), is defined as the free energy cost per unit area to create an asymmetric domain. For mismatched ordered/disordered liquid phases, γ has been estimated from heuristic arguments (Collins and Keller 2008), calculated from molecular mean-field theory (Putzel et al. 2011), and measured in both experiment (Blosser et al. 2015) and simulation (Risselada and Marrink 2008), culminating in a disparate set of values that span nearly two orders of magnitude. Although the underlying molecular origins of mismatch free energy are unclear, a reasonable assumption is that γ scales with the difference in acyl chain order of the opposing phases, such that large transbilayer differences in outer and inner leaflet order are energetically disfavored. In bilayers exhibiting asymmetric compositions and propensities to form domains (like the mammalian PM), the free energy minimum therefore reflects a compromise between line tension λ, which promotes domain formation in the outer leaflet, and midplane surface tension (γ) that opposes it. Large values of λ would thus favor mismatched raft domains in a sea of registered disordered phase, with γ serving to tune the outer/inner order difference and with it the partitioning of membrane components between raft and nonraft environments (Fig. 2).

These ideas can be quantitatively explored with thermodynamic models, resulting in phase diagrams that predict the number and types of coexisting phases (i.e., registered and antiregistered) as a function of leaflet composition and the strength of in-plane versus midplane lipid interactions, represented, respectively, by λ and γ (Allender and Schick 2006; Wagner et al. 2007; Putzel and Schick 2008; Williamson and Olmsted 2015, 2018). Conceptually, these phase diagrams broadly recapitulate the more common framing of interleaflet coupling in terms of leaflet dominance: does a leaflet with a tendency to phase separate induce domains in an opposing leaflet that would otherwise mix uniformly? Or does the uniform leaflet instead impose its behavior and abolish phase separation in the opposing leaflet? Viewed through the lens of mean-field theory, the former corresponds to weak interleaflet coupling, in which γ is large enough to induce the lateral partitioning of inner leaflet components (perhaps including signaling molecules such as acylated proteins or phosphorylated lipids), but not so strongly unfavorable that it overwhelms outer leaflet line tension and destroys the raft entirely (i.e., inner leaflet dominance). This behavior has been observed in PM-mimetic asymmetric giant unilamellar vesicles (GUVs) that showed induced partitioning of a fluorescent probe in a leaflet where phase separation would not be expected (Lin and London 2015; Enoki and Feigenson 2019). An advantage of the mean-field framework is its emphasis on coupling strength as a continuous variable that can be controlled by the cell, which could result in a continuum of inner and outer leaflet lipid partitioning behaviors between the two extremes of leaflet dominance (Fig. 2).

For example, a resting cell PM with sufficiently large γ might reside in a region of the asymmetric phase space in which both leaflets are uniformly mixed, although with the outer leaflet poised to phase separate owing to a large λ. Any change in composition that would tend to make the inner leaflet more ordered or the outer leaflet more disordered could potentially induce raft formation by moving the system into a phase coexistence region. Lipid scrambling (for example, through the activation of calcium-dependent scramblases [Suzuki et al. 2010; Malvezzi et al. 2013]) could have that effect, while flippases and floppases would act to restore a larger order difference and abolish rafts. Molecules that localize to the midplane (i.e., between leaflets) may have their own unique effects by lowering the surface tension of the mismatched phase and expanding the phase coexistence region. Crossing a phase boundary can induce not only mismatched and disordered phases, but also phases in which both leaflets of the domain are relatively ordered, raising the possibility of multiple types of raft phases. Transmembrane proteins could have distinct preferences for these phases, which may in turn have unique functional consequences for the proteins. It is not difficult to imagine a scenario in which a protein is nonfunctional in the resting asymmetric phase but is functionalized upon partitioning into a biophysically distinct registered-disordered phase.

UNDER PRESSURE: FROM MODEL MEMBRANES TO CELLS

The chemical and functional complexity, and inherent collectivity, of biological membranes complicates investigation of specific mechanisms underlying their involvement in cellular phenomena. It is unsurprising, therefore, that the main historical approach to understanding membrane behavior has been to strip down their functional roles and simplify their underlying chemistry—in other words, to create and characterize model membranes. Until recently, two factors dominated the conceptual models underlying membrane biophysics: the structures of the constituent lipids and their in-plane interactions (Marsh 2013). Clearly demonstrated in symmetric bilayers, these two factors were analogously extended to asymmetric ones, where mixtures modeling the lipids of exoplasmic and cytoplasmic leaflets could be studied in isolation or combined to create synthetic asymmetric liposomes (St. Clair et al. 2017; Scott et al. 2021). In such asymmetric membranes, simple mixtures modeling the cellular exoplasmic leaflet affected, and were affected by, opposing leaflets composed of some disordered lipids but not others, and this applied to some properties (e.g., lipid diffusion) but not others (e.g., lipid order) (see above).

Importantly, recent discoveries highlight a key characteristic inherent to compositionally asymmetric bilayers that exists in parallel to the chemical composition of lipid membranes: resting stresses that accumulate in asymmetric leaflets and modulate their properties (Doktorova and Weinstein 2018; Hossein and Deserno 2020). Such leaflet tension was shown in simulations of simple (single-component) bilayers with mismatched numbers of lipids in the two leaflets: the overpopulated leaflet becomes compressed while the underpopulated one becomes tensed (Fig. 3A; Esteban-Martín et al. 2009; Park et al. 2015). Importantly, the stresses are equal and opposite in the two leaflets, thus the bilayer has no net tension. Similar behavior can occur in compositionally asymmetric bilayers, but with an unpredictable dependence on the relative lipid abundances in the two sides (Doktorova and Weinstein 2018). For example, both leaflets could have the same number of phospholipids, but the lipids of one leaflet may pack more tightly than the other (Fig. 3A). This phenomenon, now formally defined as differential stress (Hossein and Deserno 2020), refers to coexisting, perfectly balanced stresses within leaflets resulting from their equal-and-opposite suboptimal lipid packing densities in a bilayer. Analogous to the structural and dynamical effects of net bilayer tension, the properties of compositionally identical bilayers can vary dramatically depending on the amount of differential stress (tension or compression) in their leaflets.

Figure 3.

Regulation of differential stress. (A) Differential stress can arise in asymmetric membranes when one leaflet has a different preferred surface area than another. This situation can arise when one leaflet has more lipids than the other (left) and/or in leaflets with equal lipid numbers but very different packing preferences (right). (B) Cells can regulate the differential stress in their membranes by tuning the relative lipid compositions and total abundances in the two leaflets.

While initially observed and characterized in computational models, differential stress is likely to also occur in synthetic and biological membranes. After all, for a compositionally asymmetric membrane, the truly tension-free bilayer is very much a special case: the preferred surface areas of all lipids in a given leaflet must perfectly balance those of a compositionally different one (Doktorova and Weinstein 2018). Such a fine-tuned balance seems unlikely to arise spontaneously; thus, many experimentally prepared asymmetric membranes may harbor differential stresses. For example, a combination of simulations and theory explained an initially puzzling experimental result—the anomalous stiffening of asymmetric giant vesicles (Elani et al. 2015; Hossein and Deserno 2020)—by invoking the presence of differential stress. Similar asymmetry-induced membrane stiffening was also reported in neutron spin-echo experiments of extruded vesicles prepared via cyclodextrin-mediated lipid exchange (Rickeard et al. 2020). Both observations suggest that, even in the absence of external energy input (i.e., enzymes), lipid-only asymmetric bilayers can be trapped in relatively stable differentially stressed states.

There is also evidence to suggest that biological membranes may carry significant differential stresses. The complexity of live cell lipidomes and the intricate, dynamic, and interwoven biosynthetic networks that produce them would seem to preclude perfectly balanced leaflet areas. And indeed, experimental measurements of the relative packing densities of the two PM leaflets in mammalian cells show large differences between exoplasmic and cytoplasmic monolayers (Gupta et al. 2020; Lorent et al. 2020). Moreover, the mammalian PM contains a high concentration of cholesterol. While cholesterol may be expected to buffer tension via its fast transbilayer redistribution (Miettinen and Lipowsky 2019), preferential interactions with saturated lipids can dominate over those tendencies and drive its accumulation in a particular leaflet, thereby generating differential stresses (Doktorova and Levental 2022; Varma and Deserno 2022). The exoplasmic leaflet in mammalian PMs is much more saturated than the cytoplasmic one (Lorent et al. 2020), which likely serves to recruit cholesterol (Steck and Lange 2018; Buwaneka et al. 2021) and produce a relative compression therein. Thus, the enormous configuration space available by varying relative lipid abundances and compositions between leaflets may provide cells with yet-unexplored flexibility in regulating differential stresses and the molecular processes affected by them (Fig. 3B).

Considering these converging lines of evidence, we argue that deep characterization of differential stress in asymmetric membranes represents a critical need in membrane biology and biophysics. While currently experimentally inaccessible, the presence of differential stress can be inferred from observables including changes in membrane morphology, bending rigidity, packing density, and phase transitions (Foley et al. 2023). Combining simulations with wet-lab approaches can help reveal the molecular details, magnitude, and effects of differential stress. It should also be emphasized that the differential stress in an asymmetric membrane cannot be approximated from the lipid compositions of its leaflets alone (Scott et al. 2021; Foley et al. 2023). Other essential determinants are the mixing behavior of the lipids, lipid abundance imbalances between leaflets, and the yet-unknown effects of interleaflet coupling. Thus, at present, it is not possible to predict or directly measure—and very difficult to even infer—the differential stress in any experimental system.

Ultimately, robust understanding of differential stress is essential for designing, executing, and interpreting manipulations of lipid composition and distribution in cells. Quantifying the actual magnitude of the differential stress in a living cell PM is not yet achievable, and it likely varies within and between cells. However, other related questions may be more generalizable and within reach: How can cells alter differential stress? How is leaflet tension coupled to lipid composition and interleaflet interactions? How do the properties of the membrane leaflets change when differential stress is varied, and how can those changes affect biology at the PM? Addressing these questions will begin to reveal the versatile functional roles of actively maintained PM asymmetry in cells.

STILL HAVEN'T FOUND WHAT WE'RE LOOKING FOR: OUTLOOK AND PERSPECTIVE

While many questions remain about the structure and function of cell membranes, one fact is clear: the asymmetric distribution of their lipids cannot be ignored. An asymmetric membrane has many more degrees of freedom, providing access to a wider range of biophysical properties. In that respect, both the steady-state organization of the membrane and any deviations from it can be integrated into cellular function. There exists a large variety of lipids in the PM, and actively maintaining the asymmetry of some may drive the asymmetry of others (Girard and Bereau 2023). A clear example is cholesterol, which can redistribute in response to changes in phospholipid composition and abundance (Varma and Deserno 2022). Further, lipid asymmetry can be released (scrambled), sometimes completely (Segawa and Nagata 2015) (as in apoptosis) but sometimes also in more localized and transient ways (Doktorova et al. 2020). Notably, this release of asymmetry may involve intermediate asymmetric states that have functional consequences (Hammill et al. 1999). Characterizing the compositional states accessible by cells in different contexts and understanding the biophysical principles governing the emerging behavior of asymmetric membranes are critical next steps for deciphering the enigma of PM lipid organization.