How bilayer properties influence membrane protein folding

Abstract

The question of how proteins manage to organize into a unique three‐dimensional structure has been a major field of study since the first protein structures were determined. For membrane proteins, the question is made more complex because, unlike water‐soluble proteins, the solvent is not homogenous or even unique. Each cell and organelle has a distinct lipid composition that can change in response to environmental stimuli. Thus, the study of membrane protein folding requires not only understanding how the unfolded chain navigates its way to the folded state, but also how changes in bilayer properties can affect that search. Here we review what we know so far about the impact of lipid composition on bilayer physical properties and how those properties can affect folding. A better understanding of the lipid bilayer and its effects on membrane protein folding is not only important for a theoretical understanding of the folding process, but can also have a practical impact on our ability to work with and design membrane proteins.

Abbreviations

CHAPSO

3‐([3‐Cholamidopropyl]dimethylammonio)‐2‐hydroxy‐1‐propanesulfonate

DGDG

digalactosyldiglyceride

DHPC

l‐α‐1,2‐dihexanoyl‐sn‐glycero‐3‐phosphocholine

DMPC

l‐α‐1,2‐dimyristoyl‐sn‐glycero‐3‐phophocholine

DMPE

1,2‐dimyristoyl‐sn‐glycero‐3‐phosphoethanolamine

DMPG

1,2‐dimyristoyl‐sn‐glycero‐3‐phophoglycerol

DOPA

1,2‐dioleoyl‐sn‐glycero‐3‐phosphate

DOPC

l‐α‐1,2‐dioleoyl‐sn‐glycero‐3‐phophocholine

DOPE

l‐α‐1,2‐dioleoyl‐sn‐glycero‐3‐phosphoethanolamine

DOPG

l‐α‐1,2‐dioleoyl‐sn‐glycero‐3‐phosphoglycerol

DOPS

l‐α‐1,2‐dioleoyl‐sn‐glycero‐3‐phospho‐L‐serine

DOTAP

1,2‐dioleoyl‐3‐trimethylammonium‐propane

DPoPC

L‐α‐1,2‐dipalmitoleoylphosphatidylcholine

DPoPE

l‐α‐1,2‐ dipalmitoleoylphosphatidylethanolamine

GlcGlcDAG

diglucosyl diacylglycerol

lysoOPC

1‐oleoyl‐2‐dioleoyl‐sn‐glycero‐3‐phosphocholine

lysoPPC

1‐palmitoyl‐2‐hydroxy‐sn‐glycero‐3‐phosphocholine

MGDG

monogalactosyldiglyceride

MGlcDAG

monoglucosyldiacylglycerol

POPC

1‐palmitoyl‐2‐oleoyl‐glycero‐3‐phosphocholine

POPG

1‐palmitoyl‐2‐oleoyl‐glycero‐3‐phosphoglycerol

POPS

1‐palmitoyl‐2‐oleoyl‐sn‐glycero‐3‐phospho‐L‐serine

1. INTRODUCTION

The lipid bilayer is a complex system comprised of many different lipids. The specific composition and relative abundance of each type of lipid depends on the particular organism and organelle. For example, Escheria coli membranes are primarily comprised of phosphatidylethanolamine (PE), phosphatidylglycerol (PG), and cardiolipin (CL), while mammalian cytoplasmic membranes primarily consist of phosphatidylcholine (PC), phosphatidylethanolamine (PE), phosphatidylserine (PS), sphingomyelin, and cholesterol ¹ , ² (Table 1, Figure 1). The acyl chains of each lipid can range in length from 12–26 carbons, and can be variably saturated. ³ They can also be asymmetrically attached to the glycerol backbone (Figure 1). As a result, bacterial membranes can contain thousands of different kinds of lipids, while eukaryotic membranes can have orders of magnitude more. ⁴ , ⁵ Adding to this complexity is the fact that the bilayer lipid composition is dynamic. Bacteria can adjust the composition of their membranes in response to changes in pH, pressure, or temperature, ⁶ , ⁷ while acyl chain composition in humans can be affected by diet. ⁸ , ⁹

TABLE 1.

Lipid composition in various organisms and organelles ¹ , ⁶³ , ¹¹² , ¹¹³ , ¹¹⁴

	PE	PG	PC	PS	PI	CL	Ch	SM	MGlcDAG	DGlcDAG	DPG	BmP
Acholeplasma laidlawii a		10–20							20–43	31–56	9‐17 b
Escheria coli	70–75	20–25				5–10
Proteus mirabilis	80	10				5
Pseudomonas aeruginosa	60	21				11
Bacillus polymyxa	60	3				8
Bacillus subtilis	12	70				4
Streptococcus pneumoniae		50				50
Plasma membrane	11		23	8			34	17
Golgi	21		36	6	12		18	6
ER	20		54		11		8
Mitochondria	31		37		6	22
Endosomes/lysosomes	11		30		7		30	15				7

Note: All values listed are percentages of the total lipid content.

Abbreviations: BmP, bis(monoacylglycero)phosphate; Ch, cholesterol; CL, cardiolipin; DGDG, diglucosyldiglyceride, DPG, diphosphatidylglycerol; MGDG, monoglucosyldiglyceride; PC, phosphatidylcholine; PE, phosphatidylethanolamine; PG, phosphatidylglycerol; PI, phosphatidylinositol; PS, phosphatidylserine; SM: sphingomyelin.

^a The values listed are the range in which most tested strains fell.

^b Several strains only had trace amounts.

FIGURE 1.

The structures, properties, and bilayer‐forming tendencies of common membrane lipids. Lipids can have neutral (green), zwitterionic (blue), anionic (orange), or cationic (grey) head groups, and acyl chains of varying length (denoted R1 and R2). The glycerol backbone is colored red. Reprinted from Biochim Biophys Acta 1818(4), Dowhan W and Bogdanov M, Molecular genetic and biochemical approaches for defining lipid‐dependent membrane protein folding, 1,097–1,107, Copyright (2012), with permission from Elsevier ¹¹³

Early studies with bacteria suggested that bilayer properties serve a physiologically functional and vital role. In response to environmental cues, E. coli and Acholeplasma laidlawii were found to adjust the ratio of bilayer and nonbilayer lipids to keep the membrane just below the point of phase instability. ⁶ , ¹⁰ , ¹¹ , ¹² In response to a change in acyl chain saturation, A. laidlawii altered its lipid head group composition to maintain a stable membrane curvature. ¹³ When individual lipids were removed, E. coli was shown to alter the lipid composition of its membrane to compensate. ¹⁴ These experiments suggested that certain global properties of the bilayer need to be maintained for proper physiological functions.

Cells may control lipid composition partly because of its effect on membrane protein structure and folding. A particularly striking example is illustrated by the M. tuberculosis large‐conductance mechanosensitive channel (MscL), which opens upon membrane deformation induced by osmotic stress. Extensive distance measurements by electron paramagnetic resonance methods reveal a remarkable conformational change from the closed state when MscL is solubilized in asymmetric lipid bilayers instead of detergent ¹⁵ , ¹⁶ , ¹⁷ (Figure 2). Changes in lipid composition have also been shown to result in other extraordinary conformational changes, like flipping entire domains of various permeases within the membrane (see Figure 2 and below). ³ Furthermore, defects in membrane protein folding and insertion were seen in mutant E. coli that could not synthesize certain lipids. ¹⁸ In some cases, lipids with similar properties could be substituted to assure proper protein folding and function. ¹⁹ , ²⁰ , ²¹ Clearly, an understanding of membrane protein structure and folding must include lipid composition.

FIGURE 2.

Examples of dramatic structural changes upon alterations in lipid composition. (a) Structural changes in MscL in response to changes in lipid composition. When MscL is reconstituted into 18:1 dioleoylphosphatidylcholine, the closed state (left) is observed. When lysophosphatidylcholine is added to one side of the bilayer the open conformation (right) is stabilized. ¹⁶ , ¹⁷ Adapted from Curr Opin Struct Biol 13(4), Perozo E and Rees DC, Structure and mechanism in prokaryotic mechanosensitive channels, 432–42, Copyright (2003), with permission from Elsevier. ¹⁷ (b) Topological flipping of LacY, PheP, and GabP in the presence and absence of PE. ⁴⁷ , ⁴⁸ , ⁵⁴ , ⁵⁷ The approximate locations of positively charged residues are shown with red dots, and the locations of negatively charged residues are shown with blue dots. Figure adapted from Refs. 48, 53, 54, 55, 56, 57, 113

Given the complex and diverse composition of lipid bilayers, how can we even begin to elucidate the relationship between lipid properties and membrane protein folding? While lipids can bind specifically to membrane proteins, stabilizing their structure in a way unique to the individual protein, ²² , ²³ , ²⁴ , ²⁵ , ²⁶ , ²⁷ the global physical and chemical properties of a membrane can also play a critical role. Indeed, it has been shown that electrostatic interactions can modulate protein topology, while mechanical stresses can influence membrane protein insertion, folding, or activity. It is these general properties that will be the focus of this review. Although it is well known that bilayer properties can have a profound effect on β‐barrel membrane protein folding and stability, ²⁸ , ²⁹ we will concentrate on helix‐bundle membrane proteins.

2. LIPID AND BILAYER PROPERTIES

The electrostatic and mechanical properties ¹ of membranes are determined by the lipid headgroup size and charge, as well as acyl chain length and saturation. The headgroup can be either charged (cationic or anionic) or uncharged (neutral or zwitterionic), and can vary in size (Figure 1). Acyl chains have varying lengths and can be saturated or unsaturated. Saturated chains can pack more densely and reduce fluidity in the bilayer. Lipids with saturated acyl chains or those with an overall cylindrical shape tend to form planar bilayers. Lipids with unsaturated chains or with conical shapes tend to curve toward water. Lipids with single chains or with inverted conical shapes tend to curve away from water (Figure 3).

FIGURE 3.

Phase‐forming tendencies of lipids. Lipids with a cylindrical shape tend to form planar bilayer structures. Examples include phosphatidylcholine and phosphatidylserine. Lipids with an inverted cone shape tend to curve away from water and form micelles. Examples include lysophosphatidylcholine and other lysophospholipids. Lipids with a conical shape tend to curve toward water and form the inverted hexagonal phase. Examples include phosphatidylethanolamine and phosphatidic acid. Reprinted from FEBS Lett 593(17), Zhukovsky MA, Filograna A, Luini A, Corda D, Valente C, Phophatidic acid in membrane arrangements, 2,428–2,451, Copyright (2019), with permission from Wiley ¹¹⁵

All of these properties affect the pressure profile throughout the lipid bilayer ³⁰ (Figure 4). Electrostatic charges, steric hindrance, or hydration effects modulate the pressure in the headgroup area, often leading to positive pressure. Collisions among acyl chains lead to positive pressure in the chain area. The pressure at the polar/apolar interface is typically negative due to expansion to exclude water from the hydrocarbon core. ³¹ When nonbilayer hexagonal phase lipids are added to bilayer lipids, the pressure in the chain region becomes more positive as a result of more chain collisions. Because the net lateral pressure in the bilayer must remain zero to prevent the membrane from stretching, compressing, or distorting, the increase in chain pressure is balanced by a pressure decrease in the headgroup region or at the polar/apolar interface ³² , ³³ , ³⁴ (Figure 4). As shown in Figure 5, acyl chain lateral pressure may help shape membrane proteins by favoring conformations that are thinner in the middle of the bilayer. ²⁸ , ³⁵

FIGURE 4.

A typical lateral pressure profile through a bilayer. Adding nonbilayer lipids to bilayer lipids increases the pressure in the chain region while decreasing the pressure in the head group region. Adding lipids with longer acyl chains can also increase the lateral chain pressure, while lysophospholipids can decrease the pressure. Figure adapted from Ref. 34

FIGURE 5.

How bilayer pressure can shape membrane proteins or cause the bilayer to adapt to them. (a) Nonbilayer lipids can make an hourglass shape more stable. (b) The drive to shield hydrophobic side chains can either alter the protein conformation, the bilayer conformation, or both. Lipid shapes can facilitate or hinder bilayer adjustments

Acyl chain length and saturation can also affect bilayer thickness. This can lead to instances where the thickness is much greater or smaller than the length of the transmembrane hydrophobic domain (Figure 5). To minimize any mismatch, lipids may respond by altering their packing or degree of acyl chain ordering. ³⁶ , ³⁷ , ³⁸ , ³⁹ Protein helices may respond by tilting, bending, or stretching. ³⁸ , ³⁹ , ⁴⁰ , ⁴¹ Nonbilayer cone‐shaped lipids can make it easier to locally distort the bilayer outward to shield a thicker hydrophobic protein domain, or make it more difficult to accommodate a thinner one. ¹²

But which properties of the bilayer are most critical for membrane protein folding and how do they affect it? Properties like membrane curvature, bilayer rigidity, transmembrane lateral pressure, bilayer thickness, elastic curvature stress, and even charge are interrelated and determined by the lipid composition. Although the ways in which they influence membrane protein structure or folding seem to be disparate and numerous, they can all likely be ascribed to electrostatic and mechanical forces.

3. LIPID ELECTROSTATIC EFFECTS ON MEMBRANE PROTEIN FOLDING

3.1. Lipid charge and membrane protein topology

Pioneering work on E. coli lactose permease (LacY) suggested that bilayer mechanical properties affect membrane protein folding, ⁴² , ⁴³ and also yielded the first clues that protein‐lipid electrostatic interactions could play a role. ⁴³ , ⁴⁴ LacY activity was found to be consistently optimal in E. coli lipids or synthetic mixtures that approximated the lipid composition of E. coli membranes. ⁴² , ⁴⁴ , ⁴⁵ Bilayer‐forming PE lipids with saturated acyl chains supported proper folding, ⁴⁶ but nonbilayer‐forming PE with unsaturated acyl chains did not. ⁴³ However, if nonbilayer‐forming PE lipids were supplemented with another lamellar or bilayer‐forming phospholipid like PG, LacY renaturation could be observed. ⁴³ While this evidence suggested that proper LacY folding depends on lipid mechanical and phase properties, other experiments indicated that electrostatic interactions could play a critical role. In particular, negatively charged PS, a lipid that also contains a free amine in the headgroup, could substitute for PE. ⁴³ , ⁴⁴ Moreover, increasing methylation of the PE headgroup amine led to progressively decreasing LacY activity ⁴⁴ or renaturation. ⁴³ Indeed, PC, whose amino group is completely methylated, did not significantly support either LacY activity or renaturation. Together, this suggested that the ionizable nature or hydrogen‐bonding ability of the free primary amine might be critical for proper membrane protein folding.

These early studies led the Dowhan group on a surprising and fascinating journey. To elucidate the mechanism through which PE influences LacY folding, it was critical to understand the nature of the folding defect. Bogdanov et al. showed that PE can radically alter the topological organization of LacY. ⁴⁷ By labeling individual cysteines in the extracellular loops between transmembrane domains, they showed that the six N‐terminal transmembrane helices were inverted in PE‐deficient membranes (Figure 2). Remarkably, this inversion appeared to be reversible ² as it could be corrected by inducing PE expression to normal levels. Even more surprising, when LacY was reconstituted into liposomes, similar results were observed, regardless of whether the protein had originally been expressed in a PE‐expressing or PE‐deficient strain. ⁴⁹ Moreover, the quantity of PE in the liposomes directly correlated with the amount of topologically correct LacY. When the PE content was low, almost all reconstituted LacY displayed an inverted topology. But as the PE content increased, the amount of topologically correct LacY progressively increased, reaching normal levels at PE compositions above 50%. ⁵⁰ Remarkably, these extensive topological rearrangements occurred within minutes. ⁵¹ These results firmly demonstrate that lipid composition in general, and PE specifically, can determine whether the N‐terminal half of LacY is inverted. Moreover, the rearrangements can occur on a physiologically relevant timescale.

The role of lipid‐protein electrostatic interactions in LacY inversion became clearer following the observation that LacY's cytoplasmic loops between TMs I‐VI had many negatively charged residues. ⁴⁸ Replacing a single negative residue (Asp, Glu) with its neutral amide (Asn, Gln) in any of the three N‐terminal cytoplasmic loops was sufficient to prevent inversion of the N‐terminal helices in PE‐deficient cells. Removing both a positive and a negative charge from one loop, thereby keeping the net charge the same, resulted in inversion in PE‐deficient cells. Replacing positively charged residues in each loop with negatively charged residues induced inversion even in PE‐expressing cells. These experiments demonstrated that the charge across the cytoplasmic surface of LacY can determine its topology, and is modulated by lipid interactions. A sufficiently high negative charge across the cytoplasmic surface results in topological inversion. The presence of PE diminishes this effect, perhaps through pKa shifts that alter the charge state of ionizable side chains. ⁵²

Charge‐modulated topological inversion has been observed in other E. coli proteins, indicating that it is a more generalized phenomenon. Like LacY, phenylalanine permease (PheP) contains many acidic residues in its N‐terminal cytoplasmic loops. ⁵³ In the presence of PE, the effects of these charges are dampened and the loops favor a cytoplasmic orientation. In the absence of PE, TMIII acts as a hinge to invert TMI and TMII ⁵⁴ (Figure 2). When the net charge of the N‐terminus and second extracellular loop is engineered to be more positive, this inversion does not occur. ⁵⁵ The γ‐aminobutyric acid permease (GabP) has a similar charge distribution in its extracellular loops. ⁵⁶ Like PheP, in the absence of PE, the GabP TMIII acts as a topological hinge ⁵⁷ (Figure 2). Sucrose permease (CscB) also has a significant number of acidic residues in its cytoplasmic loops. Because of the higher hydrophobicity of its putative hinge domain, the N‐terminal TM helices do not invert when PE is absent. However, when an additional salt bridge is added to this region, the N‐terminal helices can invert when the cytoplasmic loops become sufficiently negative. ⁵⁵ The amount of topologically correct CscB is directly proportional to the amount of PE in the membrane, nearing approximately 100% at near‐physiological levels. ⁵⁸

Intriguingly, even charges from posttranslational modifications can affect protein topology. When phosphorylation sites were introduced into LacY, inversion occurred upon addition of kinases. ⁵⁹ However, because cholesterol and sphingomyelin also affect LacY topology, and because these components affect the electrostatic and mechanical properties of the bilayer, direct correlations between topology and charge were not drawn. ⁵⁹ Nevertheless, this work raises the possibility of signal transduction events flipping membrane protein topologies, although natural systems like this have, to our knowledge, not yet been discovered.

While several studies suggest that direct interaction between charged residues and lipid headgroups controls membrane protein function and topology, many other studies show that the regulatory nature of these electrostatic interactions is broader and more complex. When LacY was expressed in cells where PE was replaced with PC, it was able to assume a near wild‐type topological organization capable of uphill transport. ⁶⁰ However, this was only observed with a mixture of 70% PC, 27% CL, and 3% PG, where PC had at least one saturated alkyl chain or a Δ9 trans‐fatty acid. When reconstituted into liposomes with dioleoyl‐PC, native topology was observed, but uphill transport was not. ⁴⁴ , ⁴⁵ , ⁴⁹ When LacY was expressed in cells where PE was replaced with monoglucosyldiacylglycerol (MGlcDAG), normal topology and uphill transport occurred. ¹⁹ , ²¹ Yet diglucosyl diacylglycerol (GlcGlcDAG) only restored topology, but not active transport. ²⁰ The ability of PE, PC, and MGlcDAG to be used interchangeably to support native or near‐native topology and function of LacY suggests that the ethanolamine headgroup of PE is not a specific requirement. PC is a bilayer lipid whose head group cannot act as a hydrogen bond donor, while PE and GlcDAG are nonbilayer lipids that can potentially form hydrogen bonds. PE and PC are zwitterionic, while MGlcDAG is nonionic. This diversity suggests that other more global properties play a role in the interactions between lipids and membrane proteins.

Generally, it appears that there must be sufficient neutral or zwitterionic lipids to favor correct topology. It is possible that these neutral lipids effectively dilute the charged lipids and reduce the membrane surface charge density that the protein encounters. The high quantities of zwitterionic PE and neutral MGlcDAG in bacterial membranes, ¹⁴ , ¹⁹ , ²¹ , ⁶¹ , ⁶² , ⁶³ and the need for up to 70% of both in engineered systems ¹⁹ , ⁵⁰ , ⁵⁸ , ⁶⁴ , ⁶⁵ suggests that these net neutral lipids indeed play a crucial role in dampening the negative surface charge of anionic lipids. ²⁰ , ⁴⁸ Consistent with this idea, increasing the anionic lipid content increases the fraction of inverted LacY ⁶⁴ and can induce the translocation of Pf3 coat protein. ⁶⁶ Moreover, single molecule force spectroscopy experiments showed that LacY adopts its native conformation when reconstituted into PE, but is equally likely to assume an aberrant fold when reconstituted into PG. ⁶⁷ Altogether, studies overwhelmingly suggest that anionic lipids and acidic residues can drive changes in topology while net neutral lipids suppress this effect, perhaps by lowering repulsion between anionic lipids and the protein, or by altering protein charge by modulating the pKa of ionizable side chains. ⁶⁷

3.2. Lipid charge and in vitro reconstitution

Lipid charge is an important consideration in membrane protein reconstitution experiments as it influences membrane protein insertion and refolding yields. For example, MscL insertion and diacylglycerol kinase (DGK) folding yields were optimal in pure negatively charged DOPG liposomes. ⁶⁸ , ⁶⁹ Similarly, GlpG refolding rates improved as the negatively charged DMPG concentration in DMPC/CHAPSO bicelles increased, reaching an approximately sevenfold increase at a DMPG fraction of 30% compared to 0% DMPG. ⁷⁰

Membrane charge can also play a practical role in defining the orientation of proteins in artificial liposomes. The N‐terminus of proteorhodopsin has many negatively charged residues, and the C terminus has many positively charged residues. Insertion in positively charged DOPC‐DOTAP liposomes favors an outward‐facing N‐terminal orientation, while insertion into negatively charged POPC‐POPG liposomes favors an outward‐facing C‐terminal orientation. ⁷¹ In zwitterionic vesicles, neither orientation is favored, and pH, but not alkyl chain saturation, can be used to control proteorhodopsin orientation. Moreover, bacteriorhodopsin, which has a similar structure and approximately 24% sequence homology but does not have asymmetrically charged termini, does not show preferential topological organization. ⁷¹ Taken together, these results suggest that pR orientation is determined by its charged termini and the liposome lipid headgroups.

3.3. Lipid charge and membrane protein structure and stability

Negatively charged lipids can alter the structure and stability of the folded state. For example, dimer formation of E. coli MraY requires an anionic lipid. ⁷² PG is contained within the light‐harvesting complex of photosystem II, and may help stabilize or facilitate trimerization. ⁷³ Increasing the DOPG fraction in DOPC/DOPG vesicles increases the unfolding free energy of LeuT. ⁷⁴ KcsA tetramer stability also increases with negative charge, with DOPE/DOPG liposomes optimally stabilizing the KcsA tetramer. ⁷⁵ Replacing DOPG with DOPC decreases KcsA's melting temperature, while replacement with the negatively charged DOPA has no effect. In contrast, adding negatively charged PG, PS, or cardiolipin to PC bilayers destabilized the glycophorin A transmembrane helix dimer. ⁷⁶ This was attributed to interactions with positively charged residues that could possibly alter the tilt angle of the transmembrane helices. Indeed, when the charged residues at the end of the TM helix were replaced with neutral polar ones, negatively charged lipids did not affect dimer stability. Finally, ATR‐FTIR experiments showed that anionic lipids could alter the helical tilt of EmrE with respect to the bilayer ⁷⁷ : increasing the POPG fraction in POPG/POPE or POPG/POPC vesicles decreased the tilt angle, but no changes were observed in the net neutral POPE/POPC vesicles. ⁷⁷ These results clearly demonstrate that lipid charge can help define protein conformation.

3.4. π‐Interactions between membrane proteins and lipids

A prominent characteristic of membrane proteins is the so‐called “aromatic belt,” where aromatic side chains preferentially occupy the interfacial region of membrane bilayers. ⁷⁸ , ⁷⁹ , ⁸⁰ , ⁸¹ , ⁸² Aromatic residues may be preferred in this region because of favorable electrostatic interactions with the polar lipid head groups. These π‐interactions have been directly observed between PE or PC and aromatic residues in cytochrome c oxidase ²⁴ and in the human phosphatidyl transfer protein. ⁸³ Moreover, experiments with the aspartate transporter from Pyrococcus horikoshii (Glt_Ph) suggest that a cation‐π interaction between its Y33 residue and lipid headgroups can regulate aspartate transport and proper transmembrane orientation. ⁸⁴ These interactions can be quite energetically significant, as aromatic side chains in the interfacial region of the β‐barrel protein OmpA were found to contribute up to 2.6 kcal/mol to the unfolding free energy. ⁸⁵

3.5. The effects of pKa modulation by direct protein‐lipid hydrogen bonding

The bilayer environment can alter side chain pKa's through general solvent and electrostatic effects. ⁸⁶ , ⁸⁷ , ⁸⁸ Membrane lipids can also interact with proteins directly via hydrogen bonds, ²⁷ thereby modulating residue pKas. For example, when HorA was reconstituted into DOPE vesicles, it could hydrolyze ATP and transport the substrate Hoechst 33342. Reconstitution in DOPC abolished its transport activity, and decreased the mean angle of its transmembrane helices. This change in structure and activity was attributed to PE being a stronger hydrogen donor than PC. In support of this hypothesis, interactions between negatively charged residues and the PE head group have been observed in crystal structures. ⁸⁹ , ⁹⁰ Similarly, when LmpR was reconstituted in PC instead of PE, it suffered impaired activity. ⁹¹ Increasing methylation of PE, which progressively removes hydrogen bonding potential, stabilized the outward‐open conformation in a pH‐dependent manner, suggesting hydrogen bond interactions. ⁹² A single D68C mutation mimicked the effects of reconstituting wild‐type LmpR in PC, indicating that hydrogen bonding between D68 and PE could be critical for activity. ⁹¹ Further, when LmrP was reconstituted in liposomes containing PE, the mean carboxylate residue pKa shifted to 6.5. When PE was absent, the mean pKa returned to 4.6. ⁵² , ⁹²

4. BILAYER MECHANICAL PROPERTIES AND MEMBRANE PROTEIN FOLDING

General principles governing the effects of bilayer stresses on membrane protein folding have begun to emerge. Lower pressure in the head group region or low curvature stress can favor engagement with the membrane surface. Higher pressure in the center of the bilayer or a high curvature stress can impede insertion through the bilayer, but can stabilize the structure of the folded protein. Increasing the stress can increase the folding rate by facilitating helix packing. But, if the stress or pressure is too high, the folding rate can be slowed. Nonbilayer lipids can also influence the ease with which the bilayer can deform to accommodate hydrophobic mismatch (Figure 5). The importance of these various influences will depend on the protein and the specific circumstances. In the following sections, we provide examples of how mechanical properties influence insertion and folding in a variety of experimental systems.

4.1. Bilayer mechanics and membrane protein insertion

Numerous studies support the notion that high hydrocarbon chain lateral pressure can hinder uncatalyzed protein insertion through the bilayer. One of the first studies to directly demonstrate this link attempted to refold bR into various PC/PE vesicles. ⁹³ The regeneration yield of bR decreased as the PE content, and hence hydrocarbon chain pressure, increased. Moreover, at a fixed PE mole fraction of 0.16, the regeneration yield decreased as the alkyl chain length, and hence chain pressure or bilayer thickness, increased. Subsequent mechanistic work revealed that bilayer stress influenced insertion, rather than folding after insertion. ⁹⁴

Studies with other proteins mirrored the results observed with bR. The activation energy of insertion for a designed transmembrane helix was observed to increase as the fraction of DOPE in DOPC/DOPE vesicles increased. ⁹⁵ EmrE reconstitution into DOPC/DOPE liposomes likewise decreased as the DOPE fraction increased, though the observed decrease was found to depend on the reconstitution method. ⁹⁶ LacY reconstitution into DOPC/DOPE liposomes followed the same trend. ⁶⁴ When LacY was refolded into DOPE/DOPC liposomes, increasing the DOPE concentration decreased the yield of folded protein. ⁶⁴ Indeed, optimal reconstitution occurred at high DOPG fractions, which would likely reduce the bilayer bending rigidity ⁹⁷ while adding negative charge. Finally, DsbB preferentially inserted into DMPC liposomes, and had lower insertion rates in liposomes with high PE content. ⁹⁸

Although most studies suggest that high lipid chain pressure can impede protein insertion, it is likely to be protein‐dependent. GlpG had the highest insertion rates in 1:1 PG:PE bilayers, and the lowest insertion rates in pure DMPC liposomes, ⁹⁸ the exact opposite of DsbB. In single molecule folding experiments, GlpG refolding rates increased from approximately 15% to approximately 60% when the bilayer curvature was increased by reducing vesicle size, thereby presumably reducing the barrier to insertion into the head group region. ⁷⁰ GalP also preferred higher lateral pressures: optimal GalP refolding in DOPE/DOPC vesicles was observed as the DOPE fraction was increased up to 60%. ⁹⁹ Similarly, incorporation of KcsA into DOPE/DOPC vesicles increased as the fraction of PE increased. The authors postulated that the smaller head size of PE facilitated insertion, presumably at the stage of interaction with the membrane surface. ⁷⁵ Thermostabilized turkey β1AR also seemed to prefer an environment with high lateral chain pressure: solubilization in PC or PS lipids increased as the acyl chain length or degree of unsaturation increased, ¹⁰⁰ with DOPC (18:1/18:1 chains) being preferred over POPC (16:0/18:1) or DMPC (14:0/14:0), and DOPS (18:1/18:1) over POPS (16:0/18:1). Moreover, its activity increased as the acyl chain length of PC, PG, or PS lipids increased. However, the effects of bilayer thickness cannot be excluded, and the charged DMPG, POPG, and DOPG lipids solubilized β1AR nearly as well as DOPC.

Refolding experiments with diacylglycerol kinase (DGK) suggest that curvature elastic stress also affects the ability to insert into membranes. Adding increasing fractions of DMPC or lysoOPC to DOPC decreased the final yield of refolded protein by slowing the rate of formation of protein‐vesicle complexes. Both DMPC and lysoOPC decrease the stored curvature elastic stress, which lowers lateral chain pressure and increases pressure in the head group region. This result is thus consistent with the idea that higher pressure in the head group region can hinder the ability of proteins to engage the membrane surface. Interestingly, DOPE had little effect on the folding rate, and only slightly increased the final yield. DOPG had the biggest effect on both the folding rate and yield of functionally refolded protein. Charge was unlikely to be the primary factor since increasing the ionic strength of the buffers or substituting another anionic lipid had little effect on DGK folding rates or yields. Rather, the effect of PG was attributed to headgroup interactions. ⁶⁹

4.2. Effects of lipids on membrane protein folding rates

While relatively few studies have investigated the effects of lipid composition on the kinetics of membrane protein folding, those that have support the notion that membrane pressure can affect folding rates in addition to the insertion process. When EmrE was refolded or reconstituted into DOPC/DOPE and DOPG/DOPE vesicles, the rate of folding increased as the percentage of PE increased. However, the amount of functional protein recovered decreased. This is consistent with the hypothesis that an increase in lateral pressure can inhibit insertion but facilitate packing once transmembrane helices are inserted. ⁷⁷ A series of experiments with bR likewise demonstrated that membrane rigidity can control folding rates. When bR was folded into DMPC/DHPC micelles, the rate constants for a rate‐limiting folding intermediate decreased as the DMPC concentration (and thus lateral pressure) was increased. ³¹ Experiments in DPoPC bilayers, which are more stressed than DMPC/DHPC bilayers, suggested that bilayer lateral pressure directly affects the rate of formation and relative population of two intermediate bR folding states. ¹⁰¹ When the curvature stress of DPoPC vesicles was increased by adding DPoPE, the rate constants for noncovalent retinal binding and final folding to bR slowed. ⁹⁴

4.3. Lipid bilayer properties and membrane protein activity

Early studies suggested that lipid phase behavior was critical for proper folding and activity of membrane proteins. Both monogalactosyldiglyceride (MGDG) and PE lipids tend to form hexagonal II structures, and increasing the fraction of either lipid in DOPC vesicles increased Ca²⁺ uptake by Ca²⁺ ‐ATPase. Methylation of DOPE or glycosylation of MGDG, which decreases their ability to form the hexagonal II phase and increases their tendency to form bilayers, decreased ATP‐dependent Ca2+ uptake. Activity in the bilayer lipids PS, PI, DOPC, or DOPG was minimal. ¹⁰² Similarly, leucine transport by the Branched‐Chain Amino Acid Transport System of S. cremoris increased as the PE fraction in PE/PC membranes increased, and as the glycolipid (GL) content in GL/PC membranes increased. Higher transport was observed with MGDG than digalactosyldiglyceride (DGDG), and leucine transport decreased as PE was increasingly methylated. ⁶² These results validated those with the Ca²⁺ ‐ATPase, indicating that nonbilayer lipids impart properties to the membrane critical for optimal protein activity. Because high activity for the Branched‐Chain Transporter was also observed in POPC/POPS vesicles, and because POPS can increase the bending rigidity of POPC bilayers, ⁹⁷ the effect of these nonbilayer lipids on membrane protein activity is likely mechanical in nature.

Several studies support the notion that variations in protein activity observed upon changes in lipid composition may be due to structural alterations. ³⁵ When alamethicin was reconstituted into DOPC/DOPE vesicles, the probability of higher conductance levels increased as the DOPE mole fraction increased. The increase in probability was proportional to an increase in the curvature stress. Because high conductance states are associated with multimeric channel aggregates, this suggests that the increased curvature stress stabilized alamethicin in a multimeric form. ¹⁰³ Further, when DGK was reconstituted into PC liposomes, optimal activity was observed with 18 carbon acyl chains, which minimize the degree of hydrophobic mismatch. Longer or shorter chains resulted in lower activities. ¹⁰⁴ Similar results were observed for the Ca²⁺‐ATPase, ¹⁰⁵ and for MelB. ¹⁰⁶ Tilting of the transmembrane helices or altered helix packing likely explains the effect of bilayer thickness on protein activity. ¹⁰⁴ , ¹⁰⁵

4.4. Bilayer mechanics and membrane protein stability

High bilayer curvature stress and high lateral pressure have been shown to increase membrane protein stability. Bacteriorhodopsin was more stable to irreversible thermal denaturation in mixed DMPC/DOPC vesicles than in DMPC vesicles, ⁹⁴ presumably due to the lower curvature stress of DMPC. The unfolding free energy of LeuT was found to increase as the PE content in PC liposomes increased. ⁷⁴ Increasing the DOPE fraction in DOPE/DOPC vesicles increased the lifetime of alamethicin channels and the denaturation temperature for LacY. ⁶⁴ , ¹⁰³ Adding PE to PC vesicles stabilized the GpATM dimer, while adding lysoPC (and thus lowering chain lateral pressure) destabilized it. ⁷⁶ KcsA tetramer formation in DOPC/DOPE vesicles increased as the PE fraction increased to approximately 40%. ⁷⁵ Furthermore, a higher fraction of the denaturant trifluoroethanol (TFE) was required to induce tetramer dissociation in vesicles containing PE compared to those without PE. In this case, PE may have stabilized the protein directly, or countered the bilayer destabilization and pressure‐reducing effects of TFE.

Lateral pressure can also influence thermodynamic stability, as suggested by studies with the peptide alamethicin. ¹² Alamethicin insertion is reversible, allowing free energy measurements between the water‐solubilized and membrane‐inserted states. The free energy of alamethicin insertion into DOPC/DOPE vesicles increased with increasing DOPE. Moreover, the free energy decreased with each methyl group added to the DOPE headgroup, or as the alkyl chains were increasingly saturated, thereby reducing curvature stress. ¹² These results suggest that the inserted state is destabilized relative to the common water‐solubilized reference state, and is directly correlated with bilayer mechanics. Alamethicin has a relatively short hydrophobic segment, implying that the bilayer must thin to accommodate the inserted form. It was therefore suggested that nonbilayer lipids destabilize the inserted state by making it harder to locally distort the bilayer (see Figure 5).

5. CONCLUSION

The lipid bilayer is a complex system that interacts with membrane proteins and affects their structure and function. Indeed, the bilayer's electrostatic and mechanical properties can have diverse effects on membrane protein insertion, folding, stability, and activity. Although most studies focused on the effects of electrostatic or mechanical interactions in isolation, a growing number demonstrate that both are critical. ²⁰ , ⁶⁴ , ⁷⁴ , ⁷⁵ , ¹⁰⁰ , ¹⁰⁷ , ¹⁰⁸ , ¹⁰⁹

Lipid mixtures may be necessary to balance the disparate effects of electromechanical properties on membrane proteins. The bilayer must contain enough bilayer lipids to maintain structural integrity, but enough nonbilayer lipids to allow for essential destabilizing events like budding, mitosis, and fusion. ¹⁰ , ³⁰ The membrane must be relaxed enough to allow for insertion, but stressed enough for optimal folding and stability. ¹² , ⁹⁴ , ¹⁰³ There must be enough neutral lipids to dilute negative surface charge to promote correct topology, but enough charged lipids to stabilize tertiary structures and enhance activity. It is no surprise then, that mixtures of lipids are generally more favorable for folding than vesicles made from pure lipids. ⁴⁵ , ⁶⁴ , ⁶⁷ , ⁷⁵ , ¹⁰⁷ The composition cells choose likely facilitates all critical functions without necessarily optimizing them ⁶⁴ (Figure 6). Many studies found that the optimal lipid content for in vitro folding studies was near physiological values, ⁵⁰ , ⁶² , ⁶⁵ , ⁶⁷ , ⁷⁵ , ¹⁰⁷ , ¹⁰⁹ , ¹¹⁰ , ¹¹¹ which could perhaps be expected since the protein and bilayer composition must have evolved together.

FIGURE 6.

Lipid mixtures are necessary to simultaneously ensure proper folding, stability, insertion, and activity. LacY reconstitution, refolding, topology, and activity were assayed in different compositions of DOPE, DOPG, and DOPC, and the optimal concentrations for each individual parameter, as well as for all of them together, are shown. (a) Legend showing the relative fractions of DOPE, DOPG, and DOPC on each gridline. The corners labeled DOPE, DOPG, and DOPC represent samples with 100% of each lipid, respectively. The edges represent bilayer mixtures. For example, the edge between DOPE and DOPC represents samples made of those two lipids in varying percentages, where the orange numbers along the edge state the DOPE fraction. Each point or corner in the center of the large triangle represents samples with a mixture of all three lipids, with the numbers along the edge describing the exact composition at each point. (b) The reconstitution of LacY is optimal at high DOPG fractions, and impeded at low DOPE fractions. (c) The refolding of LacY is inhibited at high DOPE and DOPC fractions. (d) High DOPG fractions impede correct topological organization of LacY. (e) LacY transport activity is optimal at high DOPE fractions. (f) The lipid fractions at which LacY reconstitution, refolding, topology, and activity are all favorable, even if they are not all optimal. Only a subset of lipid compositions can support both proper protein folding and function. In (b)‐(e), each colored dot represents the results of an experiment at the indicated lipid composition. The color of the dot shows how well each lipid composition supports proper LacY structure and activity, with green dots representing optimal function or folding and red dots illustrating when the protein is most compromised. Adapted from Sci Rep 7(1):13056, Findlay HE and Booth PJ, The folding, stability, and function of lactose permease differ in their dependence on bilayer lipid composition, Copyright (2017) ⁶⁴

Although some general principles governing lipid‐protein interactions have begun to emerge, most studies on lipid composition and membrane protein folding have been anecdotal and hard to generalize. Indeed, lipids can have disparate and even opposite effects on different proteins. This may reflect differences in the energetic barriers to folding. For example, the differential effects of lateral packing pressure on membrane protein insertion may reflect the relative importance of initial engagement of the protein with the bilayer interface versus insertion through the bilayer. To better understand the interplay between membrane proteins and bilayer properties, systematic studies varying both lipid composition and protein sequence will likely be necessary. It may also be necessary to characterize the mechanics and phase behavior of additional lipid mixtures. This information can accelerate future membrane protein studies, which may aid in the understanding of misfolding diseases and the development of drugs.

CONFLICT OF INTEREST

K.C. declares no conflict of interest. J.U.B. has founded a company involved in the production of natural chemicals that bears no direct relation to this work.

AUTHOR CONTRIBUTIONS

Karolina Corin wrote the initial draft and Karolina Corin and James U. Bowie worked together on revisions.

Corin K, Bowie JU. How bilayer properties influence membrane protein folding. Protein Science. 2020;29:2348–2362. 10.1002/pro.3973