The evolution of lipids from solvents to substrates

Abstract

While the role of water in soluble protein structure and function is well-established, the analogous role of lipids as a solvent for membrane proteins is less understood. Bacterial membranes exhibit extraordinary lipid diversity, with Escherichia coli synthesizing over 1,800 distinct glycerophospholipids. This lipid diversity gives rise to bulk membrane properties and specific lipid-lipid and lipid-protein interactions that directly impact α-IMP folding, assembly, and function. In this review, we use the same thermodynamic framework for understanding the solvation of soluble proteins to examine bacterial α-helical integral membrane protein (α-IMP) interactions with chemically diverse lipid environments. We propose that preferential solvent interactions were essential evolutionary drivers enabling lipids to evolve as protein cofactors and substrates, with lipid chemical diversity creating unique evolutionary pressures distinct from aqueous systems.

1. INTRODUCTION

The role of water in soluble protein structure, stability, and function is well established through decades of research that has revealed how aqueous environments shape protein evolution and biochemistry. In contrast, the role of lipids as a solvent for membrane protein structure, stability, and function remains poorly understood, despite membrane proteins comprising approximately 30% of all proteins and serving as targets for over 50% of pharmaceutical drugs. This knowledge gap represents a fundamental limitation in understanding membrane protein biology, biochemistry, and cellular processes.

1.1. Unifying framework

For soluble proteins, water serves as the solvent through three distinct mechanisms: bulk solvent properties involving water-water interactions, ordered solvation shells around polar protein surfaces, and specific ordered water molecules essential for protein fold and function (Figure 1, top). Membrane proteins operate under analogous principles, but their lipid solvent environment is considerably more complex. Lipids assemble into membranes with diverse pairwise interactions between different lipid species and can form microdomains containing single lipids or specific lipid mixtures. While this framework provides a solid foundation, membrane protein systems present unique complexities absent in aqueous environments. The unfolded state of membrane proteins remains poorly characterized and has only recently begun to be systematically investigated. While the number of solved membrane protein structures continues to grow, the determinants of their stability are less understood compared to soluble proteins. The identity and energetics of lipid solvation shells around membrane proteins are actively debated, and the molecular interactions driving lipid–protein associations remain poorly defined. Like water molecules, lipids can function as essential cofactors and substrates. We propose that preferential solvent interactions were essential drivers that enabled both water and lipids to evolve as protein cofactors and substrates.

Figure 1-. Comparison of molecular interactions in aqueous and lipid environments.

The same thermodynamic principles that determine water and soluble protein interactions (top row) also apply to lipid and membrane proteins (bottom row). Solvent-solvent interactions need to be disrupted to solubilize the protein. Interactions between the protein and the solvent dictate the distribution of hydrophobic (yellow) and hydrophilic (blue) amino acids. The resulting protein surface interacts with solvent creating solvation shells of ordered solvent molecules. In some cases, the solvent is necessary for the protein to carry out its function. We propose that the evolution of solvent as a cofactor or substrate is dependent on the preferential solvation of the protein. Shown here are structures of trypsin in complex with N-amidinopiperidine (N-AmPip) and structured water near critical residue Asp 189 (PDB – 5MO2) (top row), ubiquinone (UQ8) bound to the QH site of cytochrome bo₃ oxidase stabilized by two phosphatidylglycerol (PG) molecules (PDB – 7CUB) (bottom row).

1.2. Scope and Thesis of Review

This review systematically examines how lipid properties impact bacterial α-helical integral membrane proteins (α-IMPs) through three interconnected mechanisms that parallel the established water-protein framework. The review establishes lipids as a complex solvent with vast chemical diversity (Figure 2) that contributes to lipid-lipid interactions that determine the membrane environments. The discussion then explores how this diversity manifests as macroscale biophysical properties, including hydrophobic thickness, membrane fluidity, and electrostatic potentials that serve as bulk solvent characteristics affecting α-IMPs. The review next investigates preferential lipid solvation, where specific lipids become enriched around individual membrane proteins, creating local solvent environments analogous to hydration shells around soluble proteins. Direct lipid-protein interactions are examined, where lipids function as essential cofactors and substrates, directly participating in membrane protein structure and catalysis, much like water molecules in soluble protein active sites. Finally, these findings are synthesized to propose that preferential solvent interactions were essential evolutionary drivers that enabled water and lipids to evolve as protein cofactors and substrates, with the chemical diversity of lipid solvents creating unique evolutionary pressures that shaped membrane protein structure and function.

Figure 2-. Bacterial membrane lipid diversity.

The structure of three types of membrane lipids is shown. (a) Glycerophospholipids (GLs) have a variety of hydrophilic headgroups, including PhosphatidylEthanolamine (PEtn) for PE, PhosphatidylGlycerol (PGro) for PG, PhosphatidylInositol (PIns) for PI, phosphatidylserine (PS) for PS, diPhosphatidylGlycerol (PPtdGro) for CL, and hydroxyl group (OH) for DAG. GLs are characterized by an ester linkage between the hydrophobic tails and the distinct headgroup. The fatty acid chains of these GLs may vary in length, degree of unsaturation, cyclization, as well as branching. (b) Although ornithine lipids (OLs) are similar to GLs in that they have a hydrophilic headgroup as well as two hydrophilic tails, their chains are connected by amide and ester linkages instead of the usual ester/ether linkage. (c) Hopanoids (HOPs) have hydrophobic tails composed of several isoprenoids and can have a variety of headgroups.

2. Lipids as a Solvent

Soluble proteins fold and maintain stability because of water's bulk properties and interactions with polar and charged functional groups on the protein's surface (Figure 1, top). These interactions create layers of more ordered water molecules called “solvation shells”. In addition to the bulk properties and solvation, water molecules can be essential for active site coordination and catalysis. The solvent contribution to soluble protein folding and stability is a balance between the entropic and enthalpic interactions broken by the disruption of water molecules with themselves and those formed between water molecules and the protein (Figure 1, top). The same thermodynamic balances apply to lipid–lipid and lipid–membrane protein interactions (Figure 1, bottom). Specifically, van der Waals dipole–dipole (including hydrogen bonding), ion-dipole, and ion–ion interactions occur between lipids and α-IMPs. The lipid membrane protein interactions can be further categorized by lipid head group and lipid alkyl chain interactions with interfacial and hydrophobic residues of the membrane protein, respectively. A complexity of lipid-lipid and lipid-membrane protein interactions that is different from the water-soluble protein comparison is the chemical diversity of lipid molecules.

2.1. Lipid diversity

Phospholipids are abundant, comprising around 95% of all lipids found within the inner bacterial membrane in E. coli (83). Despite having three abundant lipids (PE, PG, and CL), E.coli can synthesize other lipids which vary depending on the environmental niche and stressors (56). According to the E. Coli Metabolome Database (ECMDB), 1862 distinct glycerophospholipids have been identified (40). These additional lipids are phospholipids such as phosphatidylcholine (PC), phosphatidylinositol (PI), phosphatidylserine (PS), phosphatidylethanolamine lipids (PELs, chemically modified PEs), as well as other types of lipids, including hopanoids (HOPs), glycolipids (GLs), ornithine lipids (OLs), diacylglycerol (DAG), diacylglyceryl-N,N,N-trimethylhomoserine (DGTS), sporulenes, carotenoids, and sterols (3) (Figure 2). This lipid chemical diversity is generated using different synthetic pathways, with some lipids serving as the starting precursor to others (e.g., if a bacterium has PE, then there will be some percentage of PS).

2.1.1. Lipid alkyl chain diversity

The alkyl chains of lipids are synthesized with a set of highly conserved enzymes. These chains are synthesized via the type II fatty acid synthesis (FASII) pathway (27), but also may involve additional modifying enzymes, such as fatty acid biosynthesis protein H (FabH) for branched chain fatty acids or FabA for unsaturated fatty acids (UFAs) (22, 34). The acetyl-CoA carboxylase enzyme complex and FabD transfer acetyl-CoA to acyl carrier protein (ACP), producing malonyl-ACP (26). Elongation of the alkyl chain occurs through cycles of condensation of malonyl-ACP with acetyl-CoA, as well as necessary steps of reduction and dehydration by the family of Fab enzymes (116). The elongation terminates with acyl-ACP in the cycle. Acyl ACP is further elongated with Fab enzymes or processed by acyl-ACP thioesterase, producing fatty acids (77). The fatty acid is transferred to CoA to form acyl-CoA; then the acyl chain is transferred to glycerol-3-phosphate (G3P) to form phosphatidic acid (77). The phosphatidic acids are converted to cytidine diphosphate-diacylglycerol (CDP-DAG) by CDP-DAG synthase I. The CDP-DAG is a substrate in multiple pathways that transfer the alkyl chains to lipid headgroups and precursors (104).

Polyunsaturated fatty acids (PUFAs) can be produced during synthesis through the FabA-FabB pathway or by desaturases and are attached to phospholipids (25). The PUFA attachment occurs through an ester linkage, replacing other saturated or UFAs along the glycerol backbone (49). Long-chain polyunsaturated fatty acids (LC-PUFAs) have chain lengths longer than 20 carbons and have been identified in bacterial membranes (121). In marine bacteria, LC-PUFAs are theorized to aid in adaptation within aquatic environments (121) and may increase bacterial membrane stability and fluidity (101). In E. coli, eicosapentaenoic acid, a common LC-PUFA, maintains colony growth when exposed to oxidizing agents (74). The roles of many PUFAs and LC-PUFAs in membrane biophysics and bacterial function are still unknown.

2.1.2. Lipid headgroup diversity

Many bacteria contain the synthetic pathways for PS, PE, and PG. PS is a negatively charged lipid and is synthesized by phosphatidylserine synthase (PssA) using CDP-DAG and serine as substrates (68). PE is synthesized from PS, which is decarboxylated by the enzyme PS decarboxylase (PSD) (21). Then, an ethanolamine group is attached to the phosphate headgroup, creating the zwitterionic PE lipid and increasing the lipid amphiphilicity (17). PG has a glycerol attached to the phosphate group. In the first step, PG phosphate (PGP) is formed via phosphatidylglycerol phosphate synthase (PgsA) using the precursor CDP-DAG with G3P (104). PGP is dephosphorylated in the presence of phosphatidylglycerophosphatase (Pgp) A/B/C in the final step of the synthesis of PG (104). Additional structural diversity exists from different linkages between the head group and the alkyl chains. In bacteria, ester linkage between the glycerol-1-phosphate backbone and the alkyl chains is the most abundant, while in archaea, the most abundant is an ether linkage (16).

Additional lipids that may be found in abundance in certain bacteria are not present in many other bacterial species. For instance, CL is the third most abundant lipid in E.coli, but is not found in Campylobacter jejuni and many others (104). CL is synthesized by CL synthase using two phospholipid molecules, creating a lipid with two phosphate headgroups and four fatty acid chains (29, 104). PC, another relatively rare phospholipid, is the second most abundant lipid after PE in Legionella pneumophila (45). PC is important for the virulence of particular pathogenic bacterial species. In L. pneumophila, PC increases binding to macrophages and reduces secretion of virulence factors (24, 55). In addition to modulating virulence, certain bacteria have evolved unique lipid structures and associated synthetic pathways to adapt to environmental niches.

Much of the lipid diversity in bacterial inner membranes arises from many lipids that are not abundant. In certain bacteria, GLs, DGTS, DAGs, and OLs were all found to replace phospholipids to decrease the use of phosphorus as a response to a phosphorus-depleted environment (36, 106) and/or homeoviscous adaptation (70). Although OLs and PELs contain a positive amino group, OLs have a negative carboxyl group instead of a phosphate group (71) (Figure 2). While OLs are rare within bacterial membranes, they contribute to greater order within the bilayer (71). HOPs are structurally similar to sterols (Figure 2), which are commonly found in eukaryotes, and HOPs are thought to be the prokaryotic equivalent (6, 35). HOPs are considered isoprenoids, specifically polycyclic triterpenoids consisting of five rings (23). The replacement of other lipids with HOPs in bacterial species may increase membrane fluidity in the presence of increased temperature (52). The function of sterols in membrane ordering and the formation of lipid rafts has been suggested for HOPs in bacteria (90, 91).

Many lipids have been thought to be exclusively found in eukaryotes, but are now being identified in bacteria. The identification of lipids and lipid synthetic pathways is still emerging. For instance, ceramides have been recently discovered in gram-positive species; however, the synthetic pathway is different from that found in eukaryotes (107). Bioinformatics, number of sequences, and phylogenetic analysis coupled with biochemistry experiments continue to identify new bacterial lipids with unique functions.

2.2. Lipid–lipid interactions

The complexity of a biological bacterial membrane is a significant challenge to understanding the impacts of pairwise lipid interactions on the native bulk properties and preferential solvation of α-IMPs. However, the combination of molecular simulations, which can interrogate complex membranes (8), and biophysical measurements of liposomes with controlled mixtures of synthetic lipids has allowed for identifying and quantifying lipid-lipid interactions (32) and correlating them with bulk measurements. The geometry of an amphiphile determines the self-assembly structures. In the case of detergents and lipids, conical structures (often quantified by the ratio of the volume or surface area of the head group to that of the tail) form micelles, and cylindrical structures, such as many phospholipids, form bilayers (112). The strength of association of the lipids in the bilayer determines the phase behavior of the lipid bilayer, with the dominating attractive force being van der Wal interactions between the alkyl chains (112). Dipole-dipole (including hydrogen bonding) (11, 72, 80, 103, 122), ion-dipole (76), and ion-ion interactions (attraction and repulsion) between head group atoms (directly or indirectly through bridging waters and ions) also occur despite a level of hydration and a diffuse electric double-layer from localized ions (18) (Figure 1, bottom). Insights into the energetics of lipid-lipid interactions of identical lipids can be calculated using computational approaches to estimate the free energy of desorption, which is the energy of transferring a lipid from water to the center of the bilayer. For dipalmitoylphosphatidylcholine (DPPC), the free energy of desorption was 19 kcal/mol (114), and, in a different study, the free energy of desorption was calculated to be 9, 12, 19, 20, and 22 kcal/mol for 1,2-dilauroyl-sn-glycero-3-phosphocholine (DLPC), dimyristoylphosphatidylcholine (DMPC), DPPC, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC), and 1,2-dioleoyl-sn-glycero-3-phosphocholine (96). Experimentally, the Gibbs free energy of interaction between unlike lipids in a membrane has been estimated for binary mixtures and varies between −0.300 and +0.300 kJ (28). Although the identity and strength of a few lipid-lipid interactions are known, the identity and energetics of the interactions of most biological lipids in a complex mixture are poorly understood.

3.1. Macroscale biophysical properties and their impact on α-IMPs

These chemically diverse lipids come together to exhibit macroscale or bulk properties essential for bacterial membrane function. While much is understood about the macroscale biophysical properties of lipids and bilayers, the effect of the molecular diversity of lipids on such properties in vivo is not understood. Membrane biophysical properties, such as hydrophobic thickness, lipid fluidity, and electrostatics, may impact cellular responses by regulating α-IMPs and the subsequent signalling pathways.

3.2. Hydrophobic thickness

Bacteria have different lipid membranes and peptidoglycan layers that create a semi-permeable barrier between the inside of the cell and the extracellular environment. In all cases, the membrane immediately adjacent to the cytoplasm is the inner or cytoplasmic membrane. All integral membrane proteins in the inner membrane adopt an α-helical architecture. α-IMPs have transmembrane domains (TMDs) that are solvated by amphipathic lipids. Modulating this membrane thickness can impact α-IMP folding, topology, assembly, and function, with different effects on the molecular details depending on the number of proteins and lipids involved (2, 30, 47, 67, 79, 89).

When the hydrophobic dimensions of α-IMPs differ from the average thickness of the surrounding lipid bilayer, energetically unfavorable interfaces arise from the exposure of hydrophobic surfaces to water or hydrophilic surfaces to lipid tails, and this phenomenon is referred to as “hydrophobic mismatch”. In cases where there isn’t a match between the bulk hydrophobic dimension of the bilayer, a local deformation due to preferential lipid enrichment allows for the matching of the hydrophobic surface of the protein (perilipidome) to occur (59, 89). Bilayers formed by a single type of lipid are limited in the ability to modulate hydrophobic thickness around proteins, which leads to deformation of the bilayer and/or tilting of α-helics of membrane proteins to adapt to hydrophobic mismatch. Many in vitro studies use well-defined bilayers composed of one or two types of lipids and observe hydrophobic mismatch as a membrane defect. When the transmembrane α-helical peptide gramicidin is reconstituted in DLPC and DMPC bilayers, gramicidin thickens the DLPC bilayer by ≈ 1.3 Å and thins the DMPC bilayer by ≈ 2.6 Å on a bulk scale (43). While changes in membrane thickness can lead to the stabilization of the transmembrane peptides, in the case of amphipathic peptides, such as alamethicin and melittin, hydrophobic mismatch facilitates the formation of pores (20). Peptides reconstituted into solid-supported multilamellar 1,2-diphytanoyl-sn-glycero-3-phosphocholine (DPhPC) and DOPC bilayers thin the membrane to compensate for hydrophobic mismatch (20, 46, 51, 119). When α-IMPs such as K⁺ channel of streptomyces A (KcsA), mechanosensitive channel of large conductance (MscL), and twin-arginine translocation (Tat) transporter are reconstituted in bilayers containing PC lipids with varying acyl chain lengths, the hydrophobic mismatch between the bilayer and the proteins lead to conformation changes in the proteins (helix tilt or bending) (Figure 3a) (81, 108, 120).

Figure 3-. Examples of biophysical interactions between lipids and α-IMPs in bacteria.

(a) E. coli MscL (EcMscL) undergoes structural rearrangements in the N-terminal domain between the closed (Chai-1 generated model (113)) and open (AlphaFold 3 generated model (1)) states (3, 4). The open state has a thinning of the hydrophobic region of MscL resulting in membrane thinning. Methods such as EPR (Electron Paramagnetic Resonance) spectroscopy, single-channel recording (patch clamp electrophysiology), and molecular dynamics (MD) simulations indicate the N-terminus amphipathic helix of MscL is a major mechanosensitive domain which couples membrane tension and hydrophobic mismatch to the opening of the channel (3, 5). (b) Methods such as single-molecule photobleaching and MD simulations have revealed that CLC-ec1 Cl-/H+ antiporters exist mostly as dimers (PDB – 2FEE) in a POPE/POPG bilayer due to the free energy cost of solvating the dissociated monomers by the lipids. Titration of short-chain lipids (12:0/12:0 PE and PG) leads to the shift in equilibrium towards the monomeric form because of the preferential solvation of the dimer interface by the short chain lipids (19).

3.3. Membrane fluidity and lipid packing

Bacterial membranes exhibit phase coexistence, a phenomenon in which different lipid phases, such as liquid-ordered (Lo) and liquid-disordered (Ld), exist simultaneously in the lipid bilayer (14, 62, 69, 73, 100). This phenomenon arises because of the liquid-liquid demixing observed in biological membranes due to their extensive diversity in lipid composition (44, 110). The demixed regions are called microdomains, which have different viscosities and lipid packing than the bulk. Microdomains can partition α-IMPs and/or modulate their function. Experimental and molecular simulations have begun to detect and characterize the existence of such phase co-existence and microdomains in bacterial membranes.

Super-resolution microscopy studies report a heterogeneous population of α-IMP diffusion states in Bacillus subtilis, attributed to localization to membrane microdomains in addition to localization at the poles of the bacteria (63). Examples include proton-coupled osmosensor and transporter, ProP, and the mechanosensitive channel of small conductance (MscS), which have been reported to localize towards the poles of bacteria rich in curvature-inducing lipid, CL (87, 88). Under osmotic stress conditions, there is increased localization of ProP towards the poles due to an increase in CL concentrations in the membrane, leading to enhanced uptake of osmoprotectants like proline and glycine betaine (88). The same has been observed in the case of MscS channels, where under osmotic stress, the MscS channels get selectively recruited into the poles of bacteria, which are rich in CL-enriched microdomains (85). Planar patch-clamp studies have established that CL directly modulates the function of MscS channels in the microdomains by lowering the membrane tension threshold required for the activation of the channels (86). Interestingly, in addition to lipid composition modulating fluidity, membrane proteins may modify the local lipid environment and modulate membrane fluidity. Some α-IMPs, such as rhomboid intramembrane proteases, share a unique structural architecture called the rhomboid fold, which has a loop that protrudes into the lipid bilayer, leading to distortion of the surrounding lipid bilayer (57). The resulting membrane defects are hypothesized to increase the surrounding membrane fluidity and rhomboid mobility in the membrane, resulting in a greater probability of collisions with transmembrane substrates (57). In vitro studies have also shown that the UFAs of the membrane phospholipids enhance SecA and SecYEG-mediated protein translocation. UFAs lead to weaker packing of lipids in the membrane and increase the overall fluidity of the membrane, promoting SecA binding to the lipid membrane. This increased SecA binding leads to elevated levels of protein translocation (50).

While many studies have been conducted on the adaptive changes in lipid membrane fluidity in response to temperature changes, we are just beginning to understand the effects of other environmental stress conditions on membrane fluidity and their implications for membrane protein function. Thus, further work is needed to understand better how cells respond to such stress conditions and changes in membrane fluidity by controlling the activity of α-IMPs.

3.4. Electrostatic properties

Membrane electrostatic potentials are manifested from the asymmetric organization of water, ions, lipids, and integral membrane proteins. Several electrostatic potentials are associated with the membrane, including the transmembrane, dipole, and surface potential. The transmembrane potential is an electrical potential that results from the difference in concentration of ions between the intracellular (cytoplasm) and extracellular sides of the lipid membrane. Most actively growing bacteria have a transmembrane potential of ≈ −100 to −170 mV, depending on the external pH (33, 61). The membrane dipole potential occurs mainly due to the alignment of polar bonds in phospholipids and water dipoles in the membrane-water interface (78, 97). Based on the structural properties of lipids, the dipole potential can vary in magnitude from ~ Δ100 to Δ400 mV. The membrane surface potential arises from charged lipid headgroups and the ions near the membrane surface. These membrane electrostatic properties impact α-IMP insertion, assembly, folding, and function. This review will focus on the surface and dipole potentials, which are directly affected by lipid identity and composition.

Based on environmental cues, bacteria can optimize the lipid composition of their inner membrane to modulate these potentials, which can modulate α-IMP structure and function. Bulk changes in the concentration of negatively charged lipids, such as PG and CL, and zwitterionic lipids, such as PE, can lead to changes in the surface electrostatic potential. The selective enrichment of PE lipids in the inner leaflet and PG and CL in the outer leaflet (9) results in changes in the surface electrostatic potential of the inner and outer membrane leaflets and impacts membrane protein insertion (118). If this membrane surface potential asymmetry is disturbed, it can affect the insertion and the topology of α-IMPs. For instance, the native topology of E. coli lactose permease (LacY) has N- and C-termini facing the cytoplasm. Depletion of PE lipid leads to inversion of the N-terminal helix bundle motif relative to the C-terminal helix bundle (10, 99, 117), decreasing the overall sugar transport across the membrane and, thus, the available metabolic carbon source. Bacteria have also evolved mechanisms to modify the overall membrane surface charge by changing the bulk lipid composition of the membrane to defend against antimicrobial peptides (AMPs). Bacterial pathogens such as Mycobacterium tuberculosis, Staphylococcus aureus, and Clostridium perfringens fine-tune their membrane surface charge in response to AMPs by increasing the expression of multiple peptide resistance factor (MprF) proteins (37, 42, 102, 105). These MprF proteins play a central role in the aminoacylation of PG lipids through the enzymatic conversion of the anionic lipid PG to either zwitterionic alanyl-PG or cationic lysyl-PG (37, 42, 102, 105). The resulting change in membrane charge is suggested to reduce AMP-membrane binding and hinder downstream AMP activity.

4.1. Preferential lipid solvation of α-IMP

The use of the word annulus is controversial due to the term used in a series of publications in the 1970s that reported the detection of ordered lipids in a narrow time regime (48). Other spectroscopic evidence did not indicate that lipids at the membrane protein interface were more ordered (15, 75, 98). These experiments were performed with single lipid reconstituted membranes and not biologically relevant mixtures of lipids. A recent publication continues the debate and refutes the existence of an annulus and the use of such a term (39). To avoid being associated with these early experiments and to appease these naysayers, other terms such as perilipidome and nano-environment have been introduced. Here, we will use “preferential solvent” (most recently identified in membrane proteins experimentally in (19)) to further dissociate from the controversies of the word annulus (Figure 1, bottom). Preferential solvent includes the concept of lipid solvation, which is not controversial, in the context of complex lipid composition, where all lipids are not equal in their interactions with each other or the membrane protein.

4.2. Role of lipid diversity in preferential solvation of α-IMPs

Advances in computational and experimental methods are needed to detect and characterize the preferential solvation of membrane proteins. α-IMPs must be maintained in a bilayer environment that differs dramatically between species and changes within a bacterium based on the cell cycle and the external environment. Thus, some membrane proteins may be agnostic to their lipid environment by evolving to bulk properties ubiquitous in most bacterial membranes (95). Other α-IMPs may adapt to the diverse lipid environment by evolving interactions with specific lipids. Both adaptations may occur.

Several membrane proteins have asymmetric hydrophobic thickness that impacts the organization of lipids around the protein. There are many instances in which the hydrophobic thickness has caused membrane thinning in the context of a single lipid reconstituted bilayer. In the case of synthetic bilayers composed of one type of lipid, the membrane thickness may not match the hydrophobic thickness of the protein, and the interfacial lipids deform to adjust the membrane thickness near the protein. However, in a membrane with diverse lipids, the membrane mismatch can be accommodated by recruiting lipids with chain lengths that match the protein thickness. Thus, appropriate membrane thickness is achieved without lipid packing deformations.

4.3. Evidence of preferential solvation of α-IMPs

The preferential solvation to match protein thickness was demonstrated experimentally and computationally with CLC-ec1 Cl−/H+ antiporter. In 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine (POPE)/1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoglycerol (POPG), the antiporter forms predominantly dimers (Figure 3b). However, the dimerization equilibrium is shifted towards monomer in the presence of short-chain lipids (Figure 3b). The dimer interface has a hydrophobic thickness shorter than the bilayer formed by POPE and POPG. Addition of a short-chain lipid to the POPE/POPG bilayer disrupts the dimer by preferentially solvating the thinner hydrophobic region of the antiporter with the short-chain lipid (19). Concentrations as low as 1% of a short-chain lipid effectively shifted the equilibrium to monomer. At concentrations less than 1%, a linear dependency of ΔΔG on the logarithm of the co-solvent activity was observed, indicating a thermodynamic linkage model of preferential solvation (66, 84, 111, 115).

Like water molecules in soluble protein crystallography, the electron density of lipids and lipid-like molecules is often observed in membrane protein crystal structures. These molecules frequently have high B-factors and are not entirely resolved. Interestingly, in some cases, the interactions captured are with headgroups, and others are with the alkyl chain, highlighting that the preferential solvation layer is not just due to ion-ion and dipole-dipole interactions between the membrane protein and lipid head group. The alkyl chain packing suggests that specific unsaturation patterns could exclude certain lipids from interacting at these solvent interaction sites.

Mass spectrometry of intact membrane proteins in different lipid environments has identified lipid preferences for the bacterial aquaporin, AqpZ, and ammonium transport protein, AmtB. For example, AmtB has selective binding preferences with high selectivity for PG lipids, some selective sites for PC, and non-selective binding for PE (53). AqpZ has different preferences, favoring both PC and PG over PE, but preferring PC most of all (53). Coupled with activity assays and crystal structures, specific lipids were identified that modulate protein structure and function, with CL being the most impactful for AqpZ, and PG for AmtB (58).

5.1. Lipid-protein interactions

In addition to bulk properties and preferential solvation impacting assembly, structure, and function, lipids can specifically bind to α-IMPs as cofactors and substrates (Figure 1, bottom). Lipid substrates and cofactors can be identified from structures of membrane proteins bound to lipids, coupled with biochemical functional assays and molecular dynamics simulations. Functional assays alone do not differentiate between bulk properties, preferential solvent, or direct binding of substrates and cofactors, emphasizing the need for high-resolution structures.

5.2. Lipid cofactors

Several known systems have lipids as cofactors. One such complex found in the thylakoid membrane of cyanobacteria is photosystem II (PSII). PSII uses light to drive the synthesis of plastoquinone from water. PSII is a large homodimeric protein with each monomer consisting of 10 different subunits and several cofactors. The X-ray crystal structure of PSII from Thermosynechococcus elongatus reveals 25 integral lipid cofactors bound to the dimeric structure of the protein. These lipids were identified as 11 monogalactosyl-diacylglycerol (MGDG), 7 digalactosyl-diacylglycerol (DGDG), 5 sulfoquinovosyl-diacylglycerol (SQDG) and 2 PG molecules bound to each monomer, which reflects most of the lipids found in the thylakoid membrane of cyanobacteria (41). Studies using lipid-synthesis gene mutants have shown that depletion of these thylakoid lipids causes defects in PSII function (33, 92-94)The combined structural and functional studies suggest these lipids are required for the complex's structural integrity.

Another complex dependent on lipid cofactors is the multidrug efflux transporter, AcrB, a homotrimer that uses the transmembrane proton gradient to pump drugs out of bacteria. High-resolution cryo-EM studies of AcrB reconstituted in styrene maleic anhydride (SMA) polymer nanodiscs have revealed native lipid molecules trapped inside the core of the trimer with no pathway for exchange of those lipids with the surrounding or bulk lipids. The native lipid patch is suggested to support and regulate the opening of the AcrB pore during the efflux of drug molecules (82, 109). In another instance, lipids help stabilize substrate binding. The bacterial respiratory cytochrome bo₃ oxidase has two inner leaflet PE molecules inserted in a groove formed by TM0, TM1, and TM2 close to the binding site for ubiquinone-8 (60). The hydrophobic tails of the PG lipids may stabilize the binding of ubiquinone to the QH site of cytochrome bo₃ oxidase (Figure 2).

In the case of the SecYEG protein translocation machinery and KcsA, a bacterial potassium channel, specific lipids impact structure and function by modulating oligomeric and conformational states. CL stabilizes the dimer formation of SecYEG that provides a stable high-affinity surface for the attachment of SecA ATPase, which further drives protein translocation by ATP hydrolysis (38). In addition, increasing PG lipids in the membrane leads to a significant increase in the open channel probability of the bacterial K⁺ ion channel, KcsA, in a cooperative manner, with lipids affecting the conformational states of the protein (65). Direct binding as a cofactor was supported by observing specific binding sites for anionic lipids using crystal structures. Although not structurally identified, the addition of phosphoinositides increases the probability of the KcsA open state (54).

In some proteins, there is evidence that lipids regulate conformational changes associated with function, but their role as a cofactor is less defined. For example, E. coli intramembrane rhomboid protease GlpG has bound phospholipids near TMH 2 and 5, which form a lateral gate for the substrate’s access to the enzyme's active site (7). Atomistic simulation studies indicate that these lipids could affect the docking of the transmembrane substrate by competing for the active site of GlpG (12, 13). However, more studies are needed to investigate the identity, energetics, and specificity of lipid cofactor interactions with α-IMPs in lipid membranes that better replicate native membranes.

5.3. Lipid substrates

Bacterial lipoproteins are essential for a wide range of functions in bacteria. Lipoprotein biosynthesis in bacteria starts with the insertion of signal peptides of pre-prolipoproteins, after which they undergo diacylglyceryl modification by PG:proLipoprotein diacylGlyceryl Transferase (Lgt) to form diacylglyceryl-prolipoproteins (64). Lgt specifically has binding sites for two PG molecules which serve as diacylglyceryl donors. While negatively charged phosphatidic acid (PA) and PS lipids can also act as diacylglyceryl donors they are less efficient than PG (64). In X-ray crystal structures, one PG molecule and another diacylglyceryl group of another PG molecule were observed in the central cavity of Lgt. In the first binding site, the negatively charged phosphate headgroup of PG interacts with the N-terminal end of TMH7, potentially through charge-dipole interactions (64). This interaction may be crucial for the recognition of PG lipids by Lgt.

Similar to Lgt, membrane protein MprF (an aminoacyl transferase and scramblase (42)) uses PG lipids as substrates for making aminoacyl PG lipids (Ala-PG and Lys-PG). After forming aminoacyl lipids, MprF also translocates these lipids to the outer leaflet. An increase in aminoacyl PG lipids reduces binding of cationic AMPs, which provides resistance to antimicrobial agents (31). The cryo-EM structure of MprF from Pseudomonas aeruginosa (PaMprF), supported by MD simulation studies, revealed three lipid binding sites in transmembrane domain (TMD) 1, at the TMD1/2 interface, and the cytoplasmic end of TMH8 (42).

The recognition of specific lipids as substrates of α-IMP in a heterogeneous lipid population suggests a selection of lipids from the lipid solvation layer of α-IMP enzymes. The substrate must be selected from a competing population of lipids through interactions with the membrane protein.

6. Discussion and outlook

Similar to water as a solvent for soluble proteins, lipids solvate membrane proteins (Figure 1). The same thermodynamic principles apply to both systems. The non-covalent interactions between solvent molecules dictate the bulk properties of the solvent. Solvation of proteins occurs through non-covalent interactions between the solvent and protein, which orders the solvent molecules (e.g. hydration shells for soluble proteins) compared to the bulk solvent. Both water and lipids can be cofactors and substrates (ligands) and are directly involved in protein function. We propose that these different solvent and protein interactions (i.e., bulk, preferential, and ligand) follow an evolutionary path in which the thermodynamics of the bulk properties of the solvent is a selective pressure on the evolution of protein structure, with the selection of amino acids to interact with and exclude solvent. Similarly, the solvation shells could provide opportunities to evolve solvent as cofactors or substrates. The structural diversity of lipids is distinctly different from that of water. For membrane proteins to evolve selective substrates and cofactors, this review suggests that some proteins might have a preferential solvation of specific lipids that could provide a selective pressure to evolve lipids as substrates and cofactors. The emergence of bioinformatics and artificial intelligence, coupled with the ability to study proteins in biological or complex membranes experimentally and computationally, provides the opportunity to test this hypothesis.