THE ROLE OF THE MEMBRANE IN TRANSPORTER FOLDING AND ACTIVITY

Abstract

The synthesis, folding, and function of membrane transport proteins are critical factors for defining cellular physiology. Since the stability of these proteins evolved amidst the lipid bilayer, it is no surprise that we are finding that many of these membrane proteins demonstrate coupling of their structure or activity in some way to the membrane. More and more transporter structures are being determined with some information about the surrounding membrane, and computational modeling is providing further molecular details about these solvation structures. Thus, the field is moving towards identifying which molecular mechanisms - lipid interactions, membrane perturbations, differential solvation, and bulk membrane effects, are involved in linking membrane energetics to transporter stability and function. In this review, we present an overview of these mechanisms and the growing evidence that the lipid bilayer is a major determinant of the fold, form, and function of membrane transport proteins in membranes.

GRAPHICAL ABSTRACT

INTRODUCTION

By definition, a cell must be distinctly separated from the external world by a chemical and physical barrier. Biology adopted an elegant solution here, taking advantage of the self-assembly properties of phospholipids and other lipoidal molecules in water to form a lipid bilayer. This structure provides a stable, low-dielectric layer that confers an electrostatic barrier to ions and larger charged or polar molecules [1]. However, the presence of the lipid bilayer introduces a new problem. How does the cell now move key molecules like nutrients and salts that are critical for its survival across this physical barrier? This responsibility falls on the membrane-embedded proteins including passive ion channels and uniporters, secondary active transporters, and active ATP-driven pumps. While all of these types of transport proteins work together for different purposes, it is the secondary active transporters and ATP-driven pumps that act as the batteries of the cell, fighting the uphill battle of establishing electrochemical gradients that become dissipated by ion channels and uniporters.

This balance between channels, tranporters, and pumps largely influences the physiological function of each cell. If we peer into the human body, we can find many examples of how transporters come together in a single membrane to enable extraordinary physiological feats. A prime example of transport fortitude is demonstrated in our kidneys (Figure 1A), in which the nephrons dispose of nearly all contents in our plasma including glucose, salts, and toxic waste products, only to then recover with exquisite precision proper concentrations of water, salts, and nutrients to maintain homeostasis [2]. Similarly astonishing is the function of the blood-brain barrier (Figure 1B), a practically impenetrable guard that protects our precious central nervous system, arising due to the significant presence of ATP coupled exporters like P-glycoprotein that continuously expell unwanted substances that passively permeate the membrane [3]. Finally, transport balance can connect our human physiology with the external world, or in the following example, the internal world. Consider our gut microbiome that consists of trillions of bacterial cells, including E.coli (Figure 1C). These organisms take up amino acids, ferment carbohydrates, and produce essential vitamins but also participate in a dynamic cross-talk with the host impacting the development and function of our immune system [4–6]. The proper physiological workings of every single cell within the human body rely on a precise balance of many different transport proteins working together in the membrane in spatial and temporal harmony.

Figure 1. Physiological membranes are complex molecular environments for membrane transporters.

Transporters must assemble and function in complex reaction environments to contribute to proper physiological cell function. In each of these systems, representative proteins have been built in model membranes using CHARMM-GUI [144]illustrating that cellular membranes are multi-component systems with many lipid species and the potential for higher-order assemblies of membrane proteins. (A) The kidney. Transport proteins are involved in the finely tuned reabsorption process of water, salts, and nutrients. (B) The blood-brain barrier. Exporters like P-glycoprotein expel unwanted small molecules that passively crossed the blood-brain barrier back into the bloodstream. (C) Gut bacteria. Bacterial transporters enable the uptake of nutrients and key substances and permit cross-talk with the host of the microbiome. Systems are visualized using the software VMD [145].

When this balance is disrupted, the membrane activity can be pushed towards pathophysiological states. This can occur via changes in the number of active transport molecules in the membrane, which depends on synthesis, folding, and trafficking. In addition, it can occur through modifications in functional activity, altering properties such as gating or unitary transport rates. Deviations in any of these factors can lead to changes in the permeation properties of the membrane that may escalate to cellular malfunction. Take for example the Cystic Fibrosis Transmembrane Regulator (CFTR), a chloride ion channel in the ABC transporter family for which more than 1700 mutations have been identified. Each one has been found to contribute to dysfunction at one of the stages of synthesis, folding, trafficking, or function, ultimately resulting in the disease of Cystic Fibrosis [7]. It is a frustrating situation that so many independent mutations lead to a loss of protein function and by so many distinct mechanisms. Yet, studies indicate that even the wild-type CFTR faces challenges to fold faithfully within cells [8], making one wonder if the robustness of this pathology highlights a problem that is intrinsic to membrane proteins. Perhaps CFTR is an unfortunate example of a protein that gets tested at every step of the complex pathway that a membrane protein must take before becoming a functionally folded transporter in the membrane.

Why is a membrane protein more challenging to make than a soluble protein? Of course, the main difference is that its folding and functional activity all occur within the lipid bilayer, a structurally constrained sheet of hydrocarbons about 30 Å thick. Upon synthesis, a membrane protein must first overcome the hurdle of partitioning a polar and charged polypeptide into this hydrophobic environment, then solve the problem of assembling into an ensemble of folded structures that can stably reside within the greasy bilayer core. Functional activity requires that multiple stable conformations are permitted to provide changes in exposure of protein surfaces to the intracellular and extracellular spaces, enabling direct or alternate access across the membrane. Thus, membrane proteins must pass many trials before becoming folded and functional, and the common factor for each step is that they must be achieved in the hydrophobic solvent of the lipid bilayer. Compounding this problem, membranes possess the added complexity of changes in local structure and chemical composition. At this time, we understand relatively little about what defines the stability of a membrane protein in the membrane, and whether the folded state offers a robust solution, or is inherently tenuous as may be the case for CFTR. Thus, to build a fundamental understanding of the factors impacting membrane transport physiology, we must develop a better understanding of the role the membrane plays in defining protein folding and function.

In this review, we will provide an overview of the emerging work in the field that identifies how the molecular properties of membranes influence the structure and activity of transport proteins. We first give a brief description of the physical and chemical environment that a protein encounters while embedded within the hydrophobic lipid bilayer. Next, we present evidence from the field showing how topology, folding, oligomerization, and function of membrane transporters are affected by molecular contributions from the membrane. Finally, we end with a brief perspective on key questions that require further investigation within the field. While we hope that this review provides an overview of the current research, we acknowledge that we cannot be comprehensive here, as this body of work is broad and continuing to expand. We thus refer the reader to other excellent reviews on this topic for further examination [9–18].

HOW THE MEMBRANE INFLUENCES MEMBRANE PROTEINS

From studies of soluble protein stability and ligand binding, we know that solvation forces play a key role in defining the free energy of these reactions [19]. Here, there is only one relevant type of solvent– water, a small, polar molecule that forms an isotropic liquid at typical physiological temperatures. Many of the defining properties of water arise from its ability to form hydrogen bonds with itself and the other species in the system - ions, co-solvents, and proteins. This feature is also at the root of the hydrophobic effect, a major driving force for protein folding. Since the unfolded ensemble exposes non-polar residues to the solvent, the number of hydrogen-bonded configurations available for water is typically limited in the unfolded state resulting in an entropic penalty. The hydrophobic effect, therefore, offers a generalizable driving force for protein assembly that does not consider specific interactions of the protein with water, but rather the perturbation that the protein imposes on the solvent as a whole [20–22]. In addition, the stability of soluble proteins is often dependent on co-solvents, salts, and chaotropic denaturants, which interact with the protein surface by preferential solvation [23–26]. Finally, consider that solvation forces play a defining role in determining the free energies of ligand binding. A notable example from the membrane protein field is the mechanism of selectivity in potassium channels, which achieve an exquisite selection of K⁺ over Na⁺ despite Na⁺ having a smaller ionic radius [27,28]. The free energy difference that defines the selectivity in the channel arises in large part due to the difference in the hydration free energies of K⁺ and Na⁺ in bulk solution [29]. Therefore, from protein folding to ligand binding, solvation forces play a pivotal role in defining the equilibria of protein reactions.

By extension, it is logical to hypothesize that the membrane solvent must play a similar role in defining the equilibrium reactions of membrane proteins. However, it is critical to acknowledge that the membrane provides a solvent that is completely different from water being comprised predominantly of amphipathic phospholipids. The lipids self-assemble into a stable bilayer structure where they align with a particular average orientation within each leaflet. The core of the lipid bilayer is comprised of hydrophobic acyl-chains that form substantial van der Waals interactions with neighboring lipids and proteins, establishing a highly viscous fluid that at certain temperatures, still allows its components to freely diffuse within the membrane plane. Maintaining the fluidity and other global bilayer properties is critical for cellular physiology and achieved by ‘homeoviscous adaptation’ [30] where cells enzymatically control the synthesis and modification of lipids. This highlights another key difference between the reactions of soluble proteins and membrane proteins. Soluble proteins have evolved with respect to a single molecular solvent species – water. On the other hand, there appear to be infinite combinations of possibilities for the chemical composition of biological membranes [31]. With this, the role of lipids as solvent and the importance of chemical diversity in membranes remain key questions in understanding the equilibrium behavior of membrane proteins.

To understand the degree to which membrane solvation defines membrane protein behavior, consider a thermodynamic cycle that describes the free energy difference of a membrane protein in two states (Figure 2A). In this example, we consider the conformational change of lactose permease LacY between outside-facing (OF) and inward-facing (IF) conformations. The free energy of the conformational change, ΔG_{OF⟶IF,memb.}, is depicted on the right leg of the cycle and reflects each protein state in a lipid bilayer where the lipid molecules can alter their configurations and dynamics to optimally solvate each protein structure. To isolate the contributions coming from that component of the solvation, we can define a reference state as depicted on the left leg of the cycle. This reflects the minimally solvated reaction, where the protein is simply inserted into the membrane but the conformation and dynamics of the lipids surrounding the transporter are maintained as in the bulk membrane. The total free energy difference of the conformational change in the membrane can therefore be decomposed into the differential work required for the membrane to solvate each protein structure relative to the bulk reference (top and bottom legs of the cycle), ΔG_{OF,bulk⟶memb.} —ΔG_{IF,bulk⟶memb.}, and the change in free energy between the two protein states in an unperturbed membrane, ΔG_{OF⟶IF,memb.}:

$$\Delta G_{OF\to IF,memb.}=\left(\Delta G_{OF,bulk\to memb.}-\Delta G_{IF,bulk\to memb.}\right)+\Delta G_{OF\to IF,memb.}$$

Figure 2. From two-dimensional solvent to three-dimensional material - the reactive world of the lipid bilayer.

(A) A thermodynamic cycle of the conformational change of LacY from outward-facing (OF) to inward-facing (IF), conformations (IF - PDB ID: 1PV6 [146], OF - PDB ID: 6C9W [147]. The right leg depicts the actual reaction in a realistic membrane that can rearrange to solvate the conformational change. The free energy of this reaction can be decomposed into the free energy change of a reference reaction where the protein undergoes the OF-IF conformational change without any change in the membrane, left leg of the cycle, and the difference in the solvation free energies of the OF and IF conformations with respect to a bulk-like membrane, top and bottom legs of the cycle. (B) A map of regions of differential membrane linkage with membrane proteins. Lipids are classified into four categories according to the radial distance R from the protein’s solvent accessible surface boundary, defined as R = 0 Å. Region 1: lipids within the protein. Cofactors are embedded within the protein’s surface (R < 0 Å), and are essential for the activity of the protein. Region 2: lipids within the first solvation shell. These annular lipids (0 < R < 10 Å) can exhibit solvent characteristics that can be distinguished from the bulk or may exhibit ligand-like behavior with titratable and reversible binding. Region 3: long-range membrane perturbations. Protein-induced membrane deformations like bending and thinning can extend far beyond the first shell of lipids. Region 4: the bulk membrane. Sufficiently far away from a protein, the lipids assume conformations and behavior that is intrinsic to the membrane itself, serving as a thermodynamic reference state. (C) A cartoon representation of different types of membrane effects that impact the reactions of membrane proteins. From left to right - lipid domain formation, which may sequester lipids and impact the energetics of the bulk membrane; changes in hydrophobic thickness due to matching around protein surfaces; long-range curvature strain due to membrane bending; co-factors and site-specific lipids as ligands, for example lipid substrates that are transported or that bind reversibly and in a titratable manner; and differential solvation where certain lipids are sequestered around proteins as preferential solvents, stabilizing that state.

Thus, this cycle highlights the change in the solvation free energies associated with the lipids reassembling and rearranging around each protein state, which links directly to the conformational equilibrium.

From this perspective, it becomes apparent that the membrane has the potential to modulate the reactions of membrane proteins through solvation or other forms of molecular linkage. To consider the different physical mechanisms, let us first define the spatial regions where the membrane or lipids contact the protein (Figure 2B). Since proteins come in various shapes and sizes, it is appropriate to define the protein as the volume within the solvent accessible surface. With this, the different regions pertaining to different molecular mechanisms can be defined in terms of the radial distance, R, from this boundary surface, defined as R = 0 Å. Lipids that are within the protein will therefore be defined as having a negative radial distance, i.e. R < 0 Å, whereas lipids, that solvate the protein are defined as R > 0 Å. In the following section, we examine the different types of molecular linkage that can occur in each of these defined regions (Figure 2C).

Region 1: lipids within the protein

First, let us consider the lipids that are observed to be embedded within the protein. As described above, if we define the solvent accessible boundary as R = 0 Å, then an embedded lipid is defined as having a radial distance of R < 0 Å. A lipid that co-localizes with a membrane protein in this manner may be considered a cofactor if it is deemed essential for the activity of the protein. The term cofactor is often more broadly used to describe a prosthetic group that generally remains bound to the protein. In membrane proteins, a classic example of a cofactor is the chromophore retinal, a lipoidal carotenoid that is covalently associated with bacteriorhodopsin and rhodopsin homologues. This molecule is intrinsic to the structure, function, and fold of these proteins, and is not dissociable [32]. However, cofactors do not need to be covalently linked to the protein but can be non-exchangeable within a protein structure. This type of arrangement is found in the multidrug efflux pump AcrB (Figure 3A) as many lipids are found within the trimer core, but there is no visible access pathway to indicate equilibrium exchange with the surrounding membrane [33]. Another example is the two-dimensional crystalline lattice of the water channel Aqp0 that is a component of the lens of the eye (Figure 3B). These lipids are arranged on the outside of each tetramer and so they may not participate as cofactors for the individual protein but may be cofactors for the larger assembly and formation of the crystalline lattice [34]. In addition to acting as cofactors, embedded lipids can also be substrates for transport as in MsbA, an ATP-binding cassette transporter that transports lipid A and lipopolysaccharide (LPS) from the cytoplasmic to the periplasmic leaflet of the inner membrane [35]. Cryo-electron microscopy (Cryo-EM) structures show that LPS binds deep inside MsbA at the height of the periplasmic leaflet, establishing extensive hydrophilic and hydrophobic interactions [36]. In cases like this, substrate lipids may be considered as ligands because they are dissociable, and we will describe the details of this distinction in the next section.

Figure 3. Evidence of membrane-protein coupling in transport proteins.

(A) Lipids (yellow) trapped within the trimeric assembly of AcrB act as cofactors, PDB ID: 6BAJ [33]. (B) Lipids in 2D crystals of Aqp0 appear to be essential for the formation of the crystalline lattice and are therefore cofactors of the larger assembly, adapted from [148]. (C) X-ray crystallography reveals cholesterol resolved on the surface of the DAT transporter, adapted from [43]. (D) Coarse-grained molecular dynamics simulations of Aqp1 in a mixed lipid bilayer showing differential enrichment of polyunsaturated lipids around the protein, adapted from [54]. (E) Time-average analysis of the instantaneous 3D conformations of lipid molecules residing at different positions across the membrane plane around the CLC-ec1 monomer, adapted from [55]. The helices of the dimerization interface are shown in yellow. (F) Cryo-EM study of Piezo in liposomes shows membrane perturbations around the protein, adapted from [59]. (G) Cryo-EM structures of TMEM16A in nanodiscs showing membrane bending and thinning, postulated to be important for the mechanism of lipid scrambling, adapted from [60].

Region 2: lipids in the first solvation shell

In bilayers, lipids have an average cross-sectional molecular area in the range of 0.5–0.8 nm² [37]. Therefore, we can generally define the first solvation shell as the lipids that come within 10 Å of the protein’s solvent accessible surface, i.e. 0 < R < 10 Å. These lipids are more commonly classified as annular lipids, as they reside outside of the protein and are in direct contact with its surface [38]. In recent years, structural studies of membrane proteins by x-ray crystallography or Cryo-EM, have revealed many examples where lipids have been resolved on the protein surface [33,39–42]. In many cases, structural lipids are co-purified with the protein, implying a potential interaction in the original membrane and a slow dissociation rate off the protein. One of these is the dopamine transporter, DAT, for which crystal structures reveal an associated cholesterol molecule [43] (Figure 3C). The question remains whether these structurally resolved lipids identify ligand binding sites that are relevant in membranes or whether the conditions of the experiment drive the otherwise dynamic lipid solvent to these observable configurations. High-resolution structures of soluble proteins often show crystallographic water molecules, but these are generally interpreted as solvent molecules that become ordered in the crystal and are not specifically associated with the protein under physiological conditions. For membrane proteins, factors such as the micelle environment, non-physiological temperatures, crystal packing, or nanodisc compartmentalization may all drive the observation of ordered lipids in structural studies. In these cases, such an observation may highlight positions within the solvation shell where there is a higher probability of observing a lipid, but not necessarily a site of higher affinity in the context of a fluid lipid bilayer.

To test this, further studies are needed to better understand how these lipids interact with the protein while in the dynamic environment of the lipid bilayer. In biochemistry, the term ligand refers to the reversible, specific, and dose-dependent binding of a substance to a protein to form a complex [44,45]. These ligands may include peptides, small molecule drugs, hormones, metabolites, and in the membrane - lipids. Thus, lipids that are visualized in membrane protein structures may be ligands if they exhibit the concentration-dependent and reversible association that is characteristic of ligand binding. Carrying out a titration of a lipid in a membrane environment is often associated with technical challenges and so these types of experiments are not commonly conducted. However, recent work indicates that titration studies can yield interesting analyses and are worth further examination. For example, one study monitored the interaction of lipids with the ammonia channel AmtB from E. coli in detergent micelles using native mass spectrometry. While these experiments were conducted outside of the bilayer environment, they demonstrate the ability to titrate and monitor lipid association with a membrane protein in a concentration‐dependent manner [46]. Alternatively, functional activity can offer a rigorous report of an interaction and may be reasonable to measure in ion channels. One such study monitored the activity of reconstituted human K_IR K⁺ channels as a function of % PIP₂ in the lipid bilayer [47]. It was possible to fit these data with a Hill equation to provide information about the affinity and cooperativity [48], though the mechanistic interpretation of these parameters relating to channel function remains unclear. Still, these titration results demonstrate that lipids can behave as ligands within a dynamic equilibrium membrane.

Another possibility is that annular lipids do not exhibit site-specific binding behavior in the context of the lipid bilayer, but demonstrate non-specific solvent characteristics that are distinctly different from those lipids that are in the bulk. Indeed, electron spin resonance (EPR) studies of spin-labeled lipid dynamics around the muscle sarcoplasmic reticulum Ca²⁺-ATPase showed that annular lipids exhibited restricted rotational mobility compared to lipids in the bulk membrane [49–51]. However, there was only a two-fold decrease in exchange rates [10], implying that annular lipids do not exhibit increased residence times indicative of tight ligand binding. Still, the changes in mobility do describe non-specific associations, and multi-site binding models across the total number of annular sites can be applied to quantify differential lipid affinities around membrane proteins. These have been reported for several spin-labeled lipids interacting with different channels and transporters, with ΔΔG_assoc./k_BT values ranging from 0 to −2 compared to the reference lipid PC [14]. Similarly, computational calculations of lipid binding free energies show a general affinity of lipids to annular sites, particularly with negatively charged lipids [52]. Still, it remains unclear whether the observation of an association of a lipid with a membrane protein necessarily implies an action on the protein’s stability or activity. This relationship requires further investigation of the equilibrium of channel and transporter dynamics linked to changes in lipid composition.

One way in which non-specific lipid association may modulate transporter equilibrium is via the mechanism of preferential solvation. Recent coarse-grained molecular dynamics simulations show that in mixed composition bilayers, the local lipid densities around a membrane protein exhibit enrichment and depletion of specific lipids such that each protein appears to have its own ‘lipid fingerprint’ (Figure 3D) [13,53–56]. This means that for every state of a membrane protein, and every bulk membrane composition, there is a thermodynamically favored local solvent distribution. If the different states of a protein exhibit differential enrichment or depletion of a cosolvent lipid, then the equilibrium between these states is linked to the activity of these lipids by the mechanism of preferential solvation [14]. Recently, it was shown that this effect impacts the dimerization equilibrium of the CLC-ec1 chloride/protein antiporter in membranes, which introduces a thinned membrane defect in the monomeric state that is preferentially solvated by short-chain, di-lauryl lipids (Figure 3E) [55]. Experimentally, this resulted in a titratable destabilization of dimerization up to ≈ 4.2 k_BT, linked to the amount of short-chain lipids in the membrane. Quantification of differential lipid solvation has also been made for the active and inactive states of G-protein coupled A_2A adenosine receptor using all-atom molecular dynamics simulations [57]. While both states preferentially recruit unsaturated di-oleoyl PC (DOPC) over saturated di-palmitoyl PC, the active state does so in larger proportions corresponding to a ≈ 0.5 k_BT shift towards the active state linked to DOPC, expected to change the active vs. inactive ratio by 60%. Therefore, annular lipids can have a significant impact on protein equilibrium reactions even when the interactions are weak and non-specific.

Region 3: long-range membrane perturbations

Oftentimes, membrane perturbations extend far beyond the first lipid shell. For example, cryo-EM studies of the Piezo mechanosensitive channel in nanodiscs and liposomes (Figure 3F) demonstrate that the channel deforms the membrane into an extended dome shape [58,59]. In addition, pronounced membrane bending has been observed in cryo-EM structures of lipid scramblases of the TMEM16 family in nanodiscs (Figure 3G) [60,61], with functional and computational studies indicating that the formation of this membrane defect may be key to the mechanism of lipid scrambling [62]. Even when the membrane is not resolved in a structural study, physical modeling by molecular dynamics simulations offers a refined molecular picture of the membrane structure [63]. Unfortunately, since lipids are highly viscous, there is considerable computational cost associated with all-atom simulations of membranes, and this impedes the proper sampling of lipid dynamics or bilayer changes. Thus, coarse-grained molecular dynamics is an appropriate approach here as the simulations can be run efficiently to yield converged sampling of the dynamics and long-range structural changes of the membrane around a membrane protein [55,64,65].

When long-range membrane deformations occur, it is usually in connection to a region of hydrophobic mismatch of the membrane protein and the surrounding lipid bilayer. This means that there is a difference between the hydrophobic span of a protein and the hydrophobic thickness of the lipid bilayer [66]. In these situations, the system will rearrange to minimize this mismatch in one of two ways. In the first option, the protein may alter its orientation relative to the membrane plane, or change in conformation to accommodate the hydrophobic environment. In the second option, the membrane changes its shape to accommodate the protein, resulting in local or long-range membrane perturbations. Since membrane transport proteins are often large and anchored in the membrane in a particular orientation, it appears that protein-induced membrane deformations are preferred and have been observed on numerous occasions by micelle or bilayer deformations in cryo-EM studies (Figure 3F, G). For transport proteins that undergo large conformational changes, it is conceivable that each state in the transport cycle may alter the shape of the membrane in a unique way (Figure 2A), thus connecting the functional mechanism to the energetics associated with rearranging the membrane solvent.

An important question that follows is whether the cost of deforming the membrane is of significant quantity to drive the equilibria of transport reactions. Past experimental studies demonstrate that this can be the case. A classic example is the dimerization reaction of Gramicidin A, where the equilibrium association of two monomers in a head-to-head configuration forms a conducting channel across the lipid bilayer [67]. The functional dimer is shorter than a typical biological membrane comprised of C16 or C18 acyl-chains and therefore introduces a hydrophobic mismatch in the assembled state [67]. By examining the monomer-dimer equilibria in lipid bilayers with varying membrane thickness, it was found that dimerization is stabilized as the membrane becomes thinner [68–70]. The cost of the membrane deformation free energy can be estimated using a continuum model that considers the lipid bilayer as an elastic material. Here, the classical Helfrich-Canham functional is used to calculate the cost of deforming the membrane as a function of macroscopically defined bending and curvature moduli [71]. With this, the energetic consequence of the mismatch can be quantified and linked to the overall free energy of dimerization in a matched membrane.

However, through structural and molecular dynamics studies, we are learning how membranes around membrane proteins exhibit molecular complexity, with non-uniform defects and compositional diversity. Continuum elasticity models have been developed to account for asymmetric defects [72], however, variations in local lipid distributions require a molecular model of the membrane. Recently, a molecular dynamics approach was developed to calculate the potential of mean force of a membrane perturbation according to a set of collective variables that define the density of the membrane on a grid [73]. With this, conventional methods of calculating the free energy, such as umbrella sampling or metadynamics can be employed to quantify the cost of a membrane deformation based on the molecular fluctuations of the system with respect to a predefined density distribution. Examining sinusoidal deformations, it was found that the energies calculated using this method converged with the Helfrich-Canham model for larger membranes > 100 Å. However, at shorter length scales, the continuum model appears to overestimate the energetic cost compared to the molecular calculations. This suggests that continuum models overestimate the local stiffness of the membrane, leading to a quantitative breakdown as we move closer to the protein. While there is still further work to identify whether the values obtained from the molecular calculations reflect an accurate cost, this methodology opens up a pathway to studying the impact of changes in local lipid distributions and an ability to dissect the molecular contributions that define the energetics of lipid solvation. Still, all of this comes at a cost since molecular dynamics simulations of lipid bilayers require extensive simulation times. Recent studies demonstrate that adequate sampling of competing lipid distributions requires more than 15 μs to converge [74]. This makes sense considering the high viscosity of lipids within the bilayer, but indicates that when we view simulations of lipids, we are watching a dynamic molecular system that moves in excruciatingly slow motion; hundreds of nanoseconds of simulation time is just not enough to determine whether a lipid has resided near the protein longer than the average residence time. The use of coarse-grained lipid force-fields like MARTINI [75] alleviates this sampling problem and thus offers a best-practices choice when modeling membrane deformations or lipid distribution profiles around proteins.

Region 4: the bulk membrane

As illustrated in the thermodynamic cycle discussed previously (Figure 2A), the bulk membrane plays an essential role in defining the reference state to which the membrane deformation is compared. It is important to consider that this reference is affected when the global membrane composition is changed, either by changes in membrane thickness, or other macroscopic properties such as changes in elasticity, bending, tilt, and compression moduli [76]. In fact, the concept of hydrophobic matching exhibited in the Gramicidin dimerization studies exemplifies the idea that one can change the reaction equilibrium by changing the bulk membrane properties. In this situation, the global membrane thickness was changed to match the dimer state, thus minimizing the penalty associated with solvating that state and shifting the reaction equilibrium towards dimers.

In physiological membranes that contain multiple components, there is also another way in which the bulk membrane may impact the solvation free energy. As mentioned previously, local differences in lipid composition are often observed around membrane proteins when simulated in mixed membranes. If the states in the reaction are differentially solvated by a particular cosolvent, then by the mechanism of preferential solvation, the equilibrium will be linked to the activity of that cosolvent in the membrane [14]. In this case, the bulk membrane can be a key determinant in defining the global cosolvent activity. For instance, there has been long-standing evidence suggesting that cellular membranes are laterally heterogeneous, forming distinct, ordered lipid domains alongside less organized and more fluid regions [77]. Consider a lipid that interacts with a protein as a preferential cosolvent or a ligand, in a membrane where there is a significant ordering of lipids in the bulk. The formation of a “raft-like” domain could sequester this particular lipid, stabilizing it in the bulk, shifting equilibrium away from the protein. This type of phenomenon has been suggested for PIP₂, a key signaling lipid, that has been found to be comprised of polyunsaturated fatty acyl chains that localize in the liquid-disordered phases in the membrane. Meanwhile, PIP₃, the phosphorylation product of PIP₂, is primarily comprised of saturated or monounsaturated lipid acyl chains and localizes in cholesterol-rich ordered domains [78,79]. Therefore, this acyl chain-based localization of PIPs in the plasma membrane may be important for signaling processes like ion channel activation [80,81] defined by the bulk sequestration in addition to interactions with membrane proteins.

MEMBRANE INFLUENCES ON TRANSPORTER ASSEMBLY AND FUNCTION

In considering the reactions of membrane proteins, let us first discuss the primary reaction that brings these proteins into existence - biogenesis. Early studies demonstrated that proteolytic fragments of bacteriorhodopsin spontaneously reassemble into a folded structure when in the same membrane, leading to the development of the basic two-stage model of membrane protein folding [82]. This model involves: (1) the partitioning of transmembrane α-helices into the membrane environment and (2) the assembly of these helices into the folded structure. However, with the identification of the holo-translocon complex as an essential component for the biological synthesis of most membrane proteins, this minimal model has been expanded to a fuller picture [18]. For most transport proteins, the membrane is encountered soon after the initiation of translation and all aspects of assembly occur at the membrane interface or inside the hydrophobic core (Figure 4). In this regard, the lipid solvent is likely to play a key role in membrane protein synthesis and folding, analogous to how the energetics of water drive the folding of soluble proteins. In this next section, we review the findings that show how the membrane influences topology, folding, oligomerization, and conformational equilibrium of membrane proteins, thus defining the ultimate transport activity in cellular membranes.

Figure 4. Synthesis, topology formation, folding and conformational equilibrium of transport proteins in membranes.

(A) Membrane protein translation and helix formation. Co-translational folding results in the formation of α-helices that lowers the energetic cost of partitioning a peptide chain into the hydrophobic core of the membrane. (B) Partitioning of helices into the bilayer. In the cell, the partitioning of non-polar helices (blue) into the membrane is usually facilitated by the translocon (orange) and additional insertases. (C) Subunit folding. These helices then rearrange and assemble into a thermodynamically favorable folded structure in the membrane. (D) Conformational changes. Additional reactions pertain to the folded transporter, including oligomerization and conformational exchange that are central to the functional transport mechanism. The example shown here is xylose transporter XylE, in inward-facing conformation PDB ID: 4JA4 [149], and outward-facing conformation PDB ID: 4GBY [150].

Membrane dependent influences of transporter topology

For α-helical transporters, synthesis begins with translation by the ribosome, usually in a complex with a membrane-embedded holo-translocon complex that facilitates the formation of non-polar helices in the lipid bilayer. The mechanism of this co-translational partitioning is complex, but with the rate of peptide translation by the ribosome, ≈ 20 amino acids per second in E. coli or one transmembrane helix per second, there is sufficient time for a helix to thermodynamically sample the membrane environment and find its free energy minimum state [18]. It remains unclear what route these helices take to get into the membrane, whether this process involves the passage of the protein through the translocon directly, or whether the holo-complex acts as an environmental chaperone providing an external pathway where the helix can interact with the membrane interface to facilitate partitioning into the core [18]. In any case, helical partitioning in cellular membranes appears to be a thermodynamically driven process. The amino acid hydrophobicity scale measured from translocon studies [83] correlates well with hydrophobicity scales derived from amino acid partitioning into organic solvents or equilibrium folding of the OmpLA beta barrel protein into lipid bilayers, even if the magnitudes of these free energies differ due to changes in the solvent environment [84–86]. Therefore, since partitioning is thermodynamically driven, a helix that is sufficiently non-polar will become embedded within the membrane core, while a polar helix may translocate across or be retained on the cytoplasmic side.

The options of translocation, membrane partitioning, and cytoplasmic retention define the topological fold of a membrane protein. With more and more structures being determined, it is apparent that the topologies of transporters can be remarkably complex. Oftentimes, these proteins possess reentrant helices and polar segments that are buried within the protein fold and participate in forming an aqueous site or pathway for charged or polar substrates. In addition, many secondary active transporters possess an inverted topology repeat architecture [87], which means that each subunit can be divided into two similar folds that are connected in opposite orientation with respect to the membrane. This may intrinsically encode the potential for alternating conformational access within the protein’s fold. A variation on this is the dual topology transport proteins like the fluoride channel Fluc and transporters of the small multidrug resistance family, EmrE and Gdx-Clo for which structural data show antiparallel topology where a single polypeptide chain inserts into the membrane in opposite orientations with equal probabilities [88,89]. Given all of the topological decisions that must be made during the co-translational synthesis of membrane transporter, it seems remarkable that these proteins fold faithfully in the biological environment.

With this, it seems necessary that such a process is thermodynamically driven, as the physical situation of the cell can then guide these topological choices. Membrane proteins generally follow a “positive-inside rule”, identified based on the observation that over 80% of cytoplasmic loops carry a net positive charge, and these segments partition near the cytoplasmic inner leaflet where anionic lipids are found [90]. Since this mechanism is directly linked to the charge state of the membrane, it argues that topology should be sensitive to lipid composition. Indeed, the topological orientation of leader peptidase, Lep, or signal peptidase I, is dependent on negatively charged phospholipids [91]. In addition, studies show that genetic modulation of E. coli that cannot synthesize phosphatidyl-ethanolamine (PE), leads to a topological misfolding of lactose permease, LacY [92]. Since these changes do not rely strictly on the charge distribution of the protein, a ‘charge balance rule’ has also been proposed to incorporate lipid-protein interactions during topology formation [93].

In contrast to co-translationally determined topology, there is growing evidence that topology can dynamically change after the initial partitioning step. In vitro studies of reconstituted LacY show that the topological inversion can even occur when PE is removed from vesicles [94], suggesting that topology reflects a thermodynamic sampling of states within the membrane environment. Another example is the multidrug exporter EmrE that appears to exist in the membrane in a dynamic equilibrium between two topological states where one helix flips in and out of the membrane with dimerization shifting the equilibrium to the fully assembled state [95]. Similarly, maturation of Aqp1 in the endoplasmic reticulum has been shown to involve a topological reorientation of three internal transmembrane segments and two peptide loops [96]. Therefore, it appears that for some transporter proteins, topology is not determined at the point of partitioning, but is rather a dynamic equilibrium within the lipid bilayer where the free energy may depend on the membrane environment.

Membrane dependent influences of transporter folding

In the second stage of transporter folding, the membrane-embedded helices assemble into a folded tertiary structure (Figure 4C). This has been shown to be a thermodynamically driven process based on early studies of bacteriorhodopsin, where proteolytic fragments were reconstituted into different vesicles and fusion of these membranes led to a functionally folded protein capable of conjugating retinal [97]. Similarly, lactose transport activity can be obtained by co-expression of LacY fragments [98]. However, in these cases, quantification of the assembly free energy of α-helical membrane proteins has been considerably challenging to measure in the membrane phase. While the study of soluble protein folding has been made accessible by the use of denaturants, it does not seem possible to fully chemically denature an α-helical membrane protein without impacting the lipid bilayer in a significant way. For instance, detergents can be used to destabilize the protein assembly but may affect the integrity of the membrane thus altering the reference state and confounding thermodynamic interpretation. However, recent studies have taken alternate approaches that pave the way towards reversible folding. Single-molecule force microscopy measurements have been carried out studying the unfolding of the monomeric CLC-ec1 transporter, showing that the protein disassembles in two independent steps corresponding to the inverted topology repeat within the protein [99]. In addition, the steric trapping method using biotin-streptavidin binding equilibrium to bias the folding equilibrium has enabled measurement of folding free energies of multi-helix membrane proteins like the rhomboid protease GlpG [100]. Most of these studies are not in membranes but are in detergent micelles or bicelles that include a representation of a lipid bilayer environment. However, it has been possible to study the dimerization equilibrium of the single transmembrane helix Glycophorin-A in lipid bilayers using the steric trapping approach, paving the way towards studying the underlying physical forces for multi-helix folding in membranes [100,101]. These studies also reveal how these reactions are sensitive to lipid composition, showing that dimerization depends on phosphatidyl-choline (PC), phosphatidyl-glycerol (PG), and PE headgroup changes, linking the reaction to both electrostatic and elasticity effects [102].

While chemical denaturation of membrane proteins within intact lipid bilayers may be fraught with challenges, recent studies have demonstrated that re-folding of membrane transport proteins into membranes from denatured states is indeed possible. Titrations of urea and detergents have been found to successfully unfold and re-fold membrane transporters like EmrE, CLC-ec1, and Glt_Ph into their proper oligomeric assemblies in lipid bilayers [103–106]. These re-folded proteins regain transport function, which is the most rigorous metric for the achievement of a proper biological fold. As membrane proteins have long been assumed to unfold irreversibly, these results are surprising, demonstrating that equilibrium folding studies of membrane proteins may be more accessible than what was previously thought. While free energies of folding can be measured in this manner, it is important to remember that the use of detergents perturbs the solvent structure of the lipid bilayer, thus disconnecting the reaction from an appropriate reference state. However, a recent investigation of the reversible folding of the bacterial leucine transporter LeuT with a urea titration required only a small amount of octyl-β-glucoside detergent proposed to maintain the bilayer structure [107]. With this, the effect of lipid composition could also be examined, identifying that an increase in chain lateral pressure and a negative headgroup charge both increased LeuT stability. Finally, using an in vitro cell-free expression system that avoids detergents, synthesis, spontaneous insertion, and functional folding of members of the Major Facilitator Superfamily, LacY and XylE have recently been demonstrated [108]. In line with the LeuT folding studies, increased lateral chain pressure through the addition of di-oleoyl PE (DOPE), and negatively charged headgroups by the addition of di-oleoyl PG (DOPG) were found to increase insertion yields.

Membrane dependent oligomerization of transporters

An extension of the second stage of transporter folding is the tertiary assembly of subunits into oligomers. Approximately 70% of membrane proteins in the Protein Data Bank of Transmembrane Proteins (PDBTM) form oligomers [87,109], with some of these assembled as apparently obligatory complexes where the tertiary arrangement forms a combined transport pathway (Figure 5A). This is the case for many ion channels such as K⁺ channels [110] and the fluoride channel Fluc [88], as well as transporters such as SemiSWEET [111], members of the small multidrug resistance family like Gdx-Clo [112], and the G subfamily of ABC transporters [113]. Alternatively, numerous transporters adopt an oligomeric assembly that exhibits a parallel pathways architecture, where each subunit contains its own distinct and seemingly independent transport pathway (Figure 5B). For some of these, the oligomeric structure is required for function, for example, the trimeric aspartate transporter Glt_Ph and similar homologues require the trimerization scaffold to support the large, elevator-like conformational changes of the transport domains [114].

Figure 5. Oligomerization of membrane transport proteins.

Based on structures in the PDBTM, approximately 70% of membrane proteins adopt oligomeric assemblies. (A) Oligomers where the transport pathway is along the oligomerization interface. Dimers - Gdx-Clo (PDB ID 6wk5 [89]), Fluc-ec2 (PDB ID 5a43 [88]), semiSWEET (PDB ID: 4QND [151]), ABCG2 (PDB ID: 6HZM [152]); Tetramer - Kv1.2 (PDB ID: 2A79 [153]); Pentamer - nAchR (PDB ID: 2BG9 [154]). (B) Oligomers that exhibit parallel pathways architecture, where the transport pathway is contained within each subunit. Dimers - CLC-ec1 (PDB ID: 1OTS [155]), AdiC (PDB IDL 3NCY [156]); UraA (PDB ID: 5XLS [157]); Trimers - BetP (PDB ID: 4DOJ [158]); Tetramers - Aqp0 (PDB ID: 2B6P [148]); Pentamers - FocA (PDB ID 3KCU [159]). (C) Oligomeric assemblies of transport proteins participate in equilibrium association behavior in membranes. Shown here is the example of CLC-ec1 dimerization for which a major part of the driving force for dimerization comes from the change in membrane energetics in the dissociated vs. associated states [55].

However, there are many other examples where the oligomeric assembly is not the required state for function. For example, CLC-ec1 is a Cl^-/H⁺ antiporter from the CLC superfamily of ion channels and transporters. The ion channels were first identified as homodimeric in structure due to their characteristic “double-barreled shotgun” single-channel recordings exhibiting two independent permeation pathways [115]. More recently, it was discovered that this dimer assembly could be shifted to a monomeric form by substituting tryptophans at the dimerization interfaces, resulting in a dissociated subunit that maintains stoichiometric Cl^-/H⁺ transport activity [116]. Similarly, the water channel AqpZ has been shifted to a monomeric form by modulation of the hydrophobic interface, while maintaining functional water permeation [117]. Alternatively, there are cases where the role for oligomerization has been identified as part of a regulatory mechanism of the protein’s function. For example, while BetP can transport betaine in a monomeric form, only the trimeric assembly is sensitive to osmotic shock [118]. Altogether, it appears that in many cases, oligomerization is not a requirement for transport function, but may be important for the regulatory physiology of these membrane proteins.

Implicit in these oligomerization reactions is the exposure of buried protein interfaces to the surrounding membrane upon dissociation of the complex. Thus these reactions may be inherently sensitive to changes in the solvation free energies of each state [14]. Modulating oligomeric assemblies by changes in lipid composition or regulatory molecules could have physiological value. Yet, for this to be the case, oligomeric transporters would have to participate in dynamic equilibrium association reactions. Growing evidence indicates that this is the case, even for large membrane transporter complexes. Following the equilibrium measurements of transmembrane helix oligomerization in detergent micelles [119–122], it was discovered that some purified transporters exhibited dynamic oligomerization in detergent as well. For example, the fluoride selective CLC^F transporter exhibits dimerization that depends on the protein to detergent molar ratio [123]. Oligomeric transporter assemblies also appear to be sensitive to lipids, with titrations of lipids into micelles at small amounts leading to changes in the protein stoichiometry. For example, the eukaryotic UapA transporter dissociates into monomers upon removal of co-purified lipids - PC, PE, and phosphatidyl-inositol (PI), while the addition of PE or PI leads to re-formation of dimers [124]. Similarly, mass spectrometry studies of the bacterial sugar transporter SemiSWEET identified a solution-phase monomer-dimer equilibrium that is dependent on cardiolipin [125]. In these cases, lipids could bind specifically and stabilize the oligomeric assembly, or integrate into the micelles and allosterically modulate the protein-micelle assembly by changing the structure of the protein, micelle, or both. While the isolation of the molecular mechanism may be complicated by the fact that the addition of lipids often changes the micelle shape or size on its own, these studies demonstrate that lipids are linked to the higher-order assembly of membrane proteins while solvated by detergent.

Oligomeric membrane-embedded transporters also exhibit dynamic association within lipid bilayers. A study of EmrE dimerization in liposomes showed path-independent monomer-dimer equilibrium behavior [126]. Further, it has been established that the CLC-ec1 Cl^-/H⁺ antiporter demonstrates equilibrium dimerization behavior within lipid bilayers (Figure 5C), with full binding isotherms measured using single-molecule photobleaching microscopy [127,128]. The reaction is dynamic and reversible, and depends on the protein to lipid mole fraction, exhibiting a strong affinity for the wild-type dimer, ≈ −11 kcal/mole in the 2:1 POPE/POPG lipid composition relative to the 1 subunit/lipid standard state. A recent study revealed that a major part of the driving force for dimerization comes from the change in membrane energetics in the dissociated vs. associated states [129]. Upon dissociation, the dimerization interface is exposed to the membrane and introduces a thinned, non-bilayer defect where lipids are significantly tilted and interleaflet chain contacts are increased (Figure 3E). With this, it was hypothesized that the C16:0/C18:1 acyl chains in the PO lipid bilayer act as poor solvents for the dimerization interface and add an energetic penalty to the monomeric state via the solvent alone. If this is the case, then the addition of smaller solvent molecules that are more structurally suited to the thinned perturbation would be expected to stabilize the monomeric state. Indeed, the addition of the short-chain, C12:0 lipids shifted dimerization equilibrium towards monomeric populations, with the co-solvent linkage corresponding to a mechanism of preferential solvation. Note, the effects of these co-solvent lipids occur at low proportions within the membrane in a range where structural bilayer changes or protein functional changes are not observed, supporting a molecular mechanism of modulating solvation energetics as has been previously proposed by others [51,82].

Membrane dependent effects on the transport cycle

If the equilibrium of transporter oligomerization depends on the energetics of lipid solvation, then other reactions that involve a change in protein exposure to the membrane should be affected as well. Secondary active transporters must provide alternate access across the membrane, usually via conformational changes that differentially expose a central binding site to either side of the membrane, while transiting through an occluded intermediate [130]. These conformational changes have been described as rocker-switch, rocking-bundle, and elevator mechanisms - all implying a change of exposure of protein interfaces to the aqueous solutions and possibly the membrane [131]. As depicted in the thermodynamic cycle in Figure 3A, the net transport depends on the relative energetics of these conformational states that include the membrane solvation free energy, the kinetic barriers between these states, as well as the coupling of this energy landscape to driving ions, substrates, and the membrane potential [65].

Advances in cryo-EM technologies have introduced an era where the molecular structures for many states are becoming accessible [132]. However, while more and more studies reveal multiple conformations of a single transporter, we are often missing a key part of the picture that defines the free energy balance - the membrane solvation structure. Recent work using nanodiscs or liposomes is starting to reveal these structures, showing how the membrane shape changes for different conformational states. For example studies of different conformations of Glt_Ph in nanodiscs reveal buckling of the surrounding solvation structure around the protomers of the trimer in the inward-facing state (Figure 6A) [114]. Parallel computational modeling studies of the different conformational states of Glt_Ph not only showed similar membrane deformations but also aimed to calculate the cost associated with these long-ranged yet mutually independent membrane deformations (Figure 6B, C) [65]. By calculating the potential of mean force based on a two-dimensional grid of the membrane [73], the free-energy cost of the perturbation was estimated to be 6–7 kcal/mol per protomer, indicating a potential 20 kcal/mole penalty of the inward-facing state due to membrane deformations alone. Single-molecule FRET experiments indicate that the intrinsic free-energy difference between the outward and inward-facing states of the transporter is approximately zero [133–135], therefore, the inward-facing state must balance the cost somehow. This can happen in various ways but likely involves optimization of the protein and lipid interactions in that state. More generally, when membrane deformations around protein structures are observed, the question becomes how the associated energetic costs are compensated. This compensation can happen in various ways and is probably distinct for each transporter, but likely involves optimization of the protein-lipid interactions.

Figure 6. Membrane coupling to conformational changes in the transport cycle of Glt_Ph.

(A) Cryo-EM structures of Glt_Ph in nanodiscs reveals membrane deformations dependent on the conformational state of the protein [114]. (B) Coarse-grained molecular dynamics simulations of the outward and inward conformations of Glt_Ph showing the density map for the lipid bilayer alkyl chains within 10 Å of the protein surface (yellow) [65]. (C) Top-view membrane deflection map around the trimeric Glt_Ph for each of the states quantified by calculating the mean value of the perpendicular Z coordinate of the bilayer across the X-Y plane. Positive values reflect upward bending while negative values inward bending corresponding to the deflections in (B) [65].

If optimization of protein-lipid interactions can account for energetic costs associated with membrane deformations, then the cell could regulate these interactions by varying the lipid composition. Along these lines, recent studies indicate that lipid composition does modulate the distribution of transporter conformations. Combining molecular dynamics simulations with hydrogen-deuterium exchange mass spectrometry experiments, it was identified that the conformational equilibrium between outward and inward-facing states of XylE and LacY is modulated by PE lipids that interact with certain residues on the protein [136]. Similar results were obtained for the proton-powered multidrug transporter LmrP using systematic double electron-electron resonance (DEER) in nanodiscs [137]. Further, the activity of the large amino acid transporter Lat1 is dependent on cholesterol [138] and upon depletion, the transport kinetics decrease while the substrate affinity remains constant. The authors propose that cholesterol binds in a specific manner stabilizing the transporter, similar to the proposed mechanisms of cholesterol stabilization of the outward conformations of hDAT and hSERT [139–141]. A final example that is particularly interesting is the proposed mechanism of the energy coupling factor (ECF) family of ABC transporters that involve the in-membrane dynamic assembly of a lone S-component substrate-binding domain to a hydrophobic ECF module to enable transport [142]. These transporters not only involve multi-subunit assembly within membranes but also exhibit a mechanism where the S-component changes orientation within the bilayer to enable binding to the scaffold domain. Recent computational studies indicate that the ECF module introduces a perturbation in the surrounding membrane, facilitating a membrane-assisted toppling of the S-component that enables binding [143]. In addition, functional experiments show that the transport rates are dependent on lipid composition, supporting the participation of the membrane in the underlying mechanism. Overall, a growing number of studies are suggesting that the lipid bilayer is a critical component in defining the conformational equilibrium of membrane transporters.

PERSPECTIVES

Recent research has provided an abundance of data showing that the membrane is a key factor in defining the synthesis, folding, and functional activity of transport proteins. New structures of transporters in different conformational states reveal associated lipids and large-scale membrane deformations that seem to tie into the molecular mechanisms of these proteins. Yet, research must push forward to dissect the molecular linkage of lipids and the membrane to the reactions of membrane proteins in the context of the lipid bilayer. Only then can we learn how transport proteins evolve to harness the physical consequences of lipids acting as site-specific ligands, preferential solvents, or impactors of bulk bilayer effects. With further research in this area, we may understand why there is so much diversity in the chemical composition of membranes across biology, why so many regulatory molecules are lipoidal in nature, and the mechanisms of general anesthetics. While this review does not discuss trafficking or degradation pathways, these are also critical factors in determining physiological transport activity and are likely to be impacted or regulated by the lipid bilayer. Finally, the coupling observed between membrane proteins and the membrane indicates that there are inherent driving forces for these proteins to self-assemble into higher-order organizations that may offer another level of complexity to physiological function. Most importantly, all of these questions will require an additional understanding of the natural variability of physiological membrane conditions to connect these physical relationships to the biological context. All in all, the membrane presents a strange new universe for the reactions of membrane transport proteins, one that is very different than the sea of inert lipid molecules that was proposed long ago. A productive understanding of membrane transporters, and their potential for pharmacological modulation, will thus depend strongly on our ability to dissect the molecular relationship between the membrane and membrane transport proteins.

HIGHLIGHTS.

• Molecular and physical properties of membranes influence the activity of transporters

• Lipids act via site-specific binding, preferential solvation and bulk effects

• Membrane coupling to transporter synthesis, folding, oligomerization and conformational equilibrium