Quaternary arrangements of membrane proteins: an aquaporin case

It has been nearly 25 years since the hypothesis was put forward that incorporating most integral (or intrinsic) polytopic α-helical membrane proteins (spanning membrane and classified in Types-I to Types-VI) in cellular membranes can be considered a two-stage process. During stage I, hydrophobic regions of membrane proteins fold into α-helices crossing a lipid bilayer to give rise to functional proteins during stage II [1,2]. Since then, a large body of evidence has accumulated to specify the roles of structural properties of membrane proteins and lipid bilayers, where membrane proteins fold and form functional assemblies. Biophysical studies such as cryogenic-electron microscopy (cryo-EM) combined with 3D reconstruction were instrumental in defining the interactions between transmembrane protein α-helices and lipid bilayers [3]. These efforts contributed to formulating the unifying models for α-helical protein biogenesis, which emphasize the contributions of lipids, termed finely-tuned molecular machines [4–7] in membrane protein folding and oligomerization. For example, in human aquaporin 1 [AQP1; Protein Data Bank (PDB) accession 1H6I], a polar tripartite motif at transmembrane α-helices 2 and 5 was a key to directing its α-helical packing, topogenesis, monomer folding and tetramerization. This mechanism carved by evolutionary divergence could have general importance [8].

This mini-review is not exhaustive and focuses on the last three years of progress with earlier concepts contextually mentioned.

The approaches describing biochemical and biophysical properties of α-helical membrane proteins have significantly expanded since the formulation of the two-stage hypothesis [1,2]. For example, more recent methods to monitor protein binding in membranes include switchSENSE® technology [9], while cell imaging, cryo-EM, and in situ cryogenic-electron tomography (cryo-ET) visualize macromolecules in situ [10]. However, cryo-ET must improve interpretability or assignment identity with diminish low-resolution and noise problems via template-free or template-based setups [11]. An example of cryo-ET achievement is a near-complete atomic model of a human endoplasmic reticulum (ER)-bound translocon [12]. Other techniques of studying membrane proteins include synthetic chemistry combined with nanodevices [13] and model liposomes linked with microfluidics sizing-based assays, circular dichroism, fluorescence correlation spectroscopy, thermodynamic quantifications and mechanistic approaches to protein embedding in lipid bilayers [14]. These methods could be combined with high-precision single-molecule total internal reflection fluorescence microscopy to evaluate, amongst other properties, distances and dynamics between membrane oligomers in bilayers of live cells [15,16]. Additional approaches include pulsed interleaved excitation fluorescence cross-correlation spectroscopy, which examines the oligomerization of membrane proteins in live cells and measures single-cell-based protein expression levels and diffusion coefficients [17], and single-molecule-based tracking exploring the dynamics of protein movements [18]. These methods could be supplemented by techniques that evaluate scattering length density profiles of bilayers via neutron and X-ray diffraction to determine the location and orientation of molecules between asymmetric membrane leaflets [19].

Theoretical and computational approaches also contributed to descriptions of membrane proteins and nuanced mechanisms of bilayer asymmetry that govern structural and dynamic coupling between compositionally distinct bilayer leaflets [20]. The definitions of force-field parameters [21] allowed describing properties of homogeneous bilayers in bacterial, mammalian and cancer cells (area, volume, and thickness properties) [22] and mechanical properties of solid-ordered bilayers (area per lipid, density vesicle profiles, bending rigidity coefficient, area compressibility properties) via molecular dynamics (MD) simulations [23]. The latest developments include the coarse-grained Martini 3 OliGo̅mers method, developed for large-scale simulations, to realistically describe higher-order biomolecules while considering folding and unfolding events [24]. The Martini 3 OliGo̅mers method was applied to the assembly/disassembly of tetrameric bovine AQP1 (PDB accession 1J4N), revealing that it formed an exceptionally stable tetrameric structure. It also suggested that stability was due to a mismatch between the AQP1 hydrophobic regions and the bilayer thickness [24].

To assist these methods, structural repositories provide a wealth of information: specifically, the PDB [25] with >227 000 entries and nearly 1.1 million computed models (as of November 15, 2024), combined with the Cambridge Structural Database, enlisting >1.25 million small-molecule structures [26]. Other valuable resources are 2024 Nobel Prize-awarded AlphaFold [27,28] and RoseTTAFold [29,30] combined with AlphaFill [31], AlphaFold-Multimer [32], and Cosmic² [33], which integrate structural analyses with machine learning.

Some integral polytopic α-helical membrane proteins form ordered oligomeric quaternary assemblies, which underlie their function with profound physiological consequences [34,35]. Although a membrane protein monomer could sometimes function, oligomeric structures offer advantages. Often, in lipid bilayers, the occurrence of mono- and oligomeric membrane proteins could take place simultaneously, as shown with AQPZ (PDB accession 1RC2) using high-speed atomic force microscopy (HS-AFM) and kinetic membrane elastic theory, which quantify membrane-mediated interactions at high resolution and define elastic properties in the bilayers [36].

The two major features that underscore the integration and formation of α-helical membrane protein quaternary assemblies are properties of: (i), membrane proteins, such as interaction patterns and topology, which confine and underly the orientation of protomers to facilitate assembly [37]; and (ii), a lipid bilayer-forming solvent, which potentiates folding of membrane proteins in a native environment [38,39]. Factors that affect membrane transport protein integration and oligomerization can be subdivided into those of 3D structural properties, the presence of GXXXG and leucine heptad sequence motifs, properties of loops facing extracellular and intracellular environments, N- and C-terminal termini modifications, lipid release potential, structural stability, quaternary contacts, cooperativity between protomers, transport activity, and hydrophobic matching within a lipid bilayer, while those of bilayer properties include lateral pressure, lipid composition and chemistry, and the thickness of a bilayer core [35,36,39–42]. Other factors that play roles in these processes are post-translational modifications such as e.g. phosphorylation and lipidations (discussed below in detail), and changes through interactions with ‘proteomimic’. The latter term involves interactions with soluble or membrane proteins, peptides, peptide mimetics and small molecule ligands, which could enhance, modify or block protein function; this could be exploited in drug and herbicide design [39]. In other transporters, such as solute carrier six family members, a single-molecule fluorescence microscopy revealed that they form a range of monomeric to multimeric assemblies with complex stoichiometries, where all transmembrane α-helices contributed to oligomer interfaces, except a bundle domain [42].

Membrane proteins interact with annular lipids localized in their immediate vicinity, which restricts membrane proteins’ mobility, with an overall entropy increase, and even so much more after membrane proteins form higher-order quaternary structures when the lipid-membrane-protein surface areas are progressively lowered [35]. In the latter case, only obligate oligomeric membrane proteins are functional. Conversely, facultative membrane proteins may operate as monomers or oligomers — such is the case of AQPs [35,36,41], although this distinction is not always clear-cut.

Other functions that underscore membrane protein integration and organization in bilayers are the roles of lipid-binding modules (e.g. 150 amino acid residue discoidin-type C2 or 30-residue long basic-rich domains occurring in yeasts, mammals and plants) and protein lipidation, which targets proteins to intracellular membranes allowing them to operate spatially at defined cellular locations [43]. Specifically, in Saccharomyces cerevisiae plasma membranes monitoring with fluorescent probes-tagged lipid-binding domains in real-time visualized lipid species and how they associate with cellular membranes. This approach helped to understand the signalling roles of protein complexes in plasma membranes after being targeted to plasma membranes. Similarly, co- and post-translational prenylation (at C-termini of Cys residues), myristoylation (at N-termini of Gly residues), palmitoylation (additions to Cys residues along protein chains), and O- and N-acylation (Ser and Lys residues) could increase membrane attachments through a higher hydrophobicity [43]. These adjustments target co- and post-translationally modified membrane proteins to their locations and combined with engineered membrane proteins, it is possible to clarify their exact function in plasma membranes.

Integral polytopic α-helical membrane proteins are incorporated into ER membranes after they are synthesized by ribosomes, integrated into ribosome-associated ER membranes and trafficked via the Golgi system to their target membrane locations through elaborate trafficking machinery [19,42,44–46]. This machinery in eukaryotic cells includes SEC24C or SEC24D protein transport proteins, a Coat Protein Complex I (COPI) and a heptameric Coat Protein Complex II (COPII), which traffics membrane proteins in cells through COPII-vesicular machinery [45]. It is also significant that plasma membrane curvature regulates lipid homeostasis and contacts with ER membranes, which stabilize the inserted membrane protein oligomers [47]. This could affect their post-translational modifications, sorting, and trafficking [36]. During insertion and trafficking, the synthesized membrane proteins must be correctly folded, including the packing of their C-terminal α-helices facing ER membranes. It has been suggested that protein components of SEC24 might sense the folding of transporters or recognize their conformational equilibrium rather than participate in oligomerization [42,44]. Although, these events may not be neatly separated and could coincide [48]. To clarify these aspects, new revolutionary approaches, such as single-molecule atomic force microscopy (AFM) and single-molecule-forced unfolding using magnetic tweezers, could unravel forces [hydrogen bonds (H-bonds), hydrophobic and electrostatic interactions, van der Waals forces] as the source of entropy for folding and oligomerization of membrane proteins [48].

Several principles of AQPs homo- or heteromeric quaternary structure formation were described, and exact determinants differ in each case. Dynamic imaging resolved the association/dissociation processes of AQP monomers [24,36], providing evidence that the mismatch between proteins’ hydrophobic regions and membrane thickness could drive membrane protein oligomerization. In AQPs, at the structural levels, several other factors could favour oligomerization, such as the chemical nature of lipids and lipidic environments where AQPs reside, lipid release during hydrophobic mismatch, quaternary contacts of monomers, structural and proteolytic stability of monomers, which could lead to forming lipid-binding cavities. It would be crucial to reveal how these cavities could affect AQP oligomerization and gating and if their formation is linked to changes in the distributions of secondary structural elements. These factors could modulate AQP function and cooperativity amongst monomers and yield new activity with or without beneficial features [35].

AQPs could be monomerized by even a single mutation, as shown in E. coli glycerol facilitator GlpF (PDB accession 1LDA), where polar glutamate was replaced by the apolar Ala residue. In GlpF, four Gln residues co-ordinated a single Mg²⁺ ion in the pore, while a neighbouring Mg²⁺ ion was co-ordinated by four Trp residues; it is unknown if either coordination has functional consequences. This single glutamate to alanine mutation weakened the stability and ability of the GlpF oligomer to form a tetramer, meaning that tetramerization stabilized its quaternary structure. Monomerized versus tetramerized forms of GlpF, where the tetramer oligomerized through dimerization, exhibited differences in permeation, and the monomer failed to facilitate polyalcohol flux across the E. coli cytoplasmic membranes [49].

However, further, studies of bacterial and human Major Intrinsic Protein families, permeating water and other solutes, indicated that AQPs could not oligomerise upon a single (as detailed above) or several mutations. Here, the exact structural descriptors, such as loop E in human GlpF (PDB accession 1FX8) were responsible for homo-tetramerization [50,51]. Meanwhile, in human AQP2 (PDB accession 4NEF), glutamate 258, residing on a C-terminal cytoplasmic loop, was found to be critical, which upon expression in Xenopus oocytes of wild-type or targeted mutants, formed monomeric (mutated into cysteine) or homo-tetrameric (mutated into lysine and wild-type) structures [50]. But, upon co-expression of AQP2 versions, the AQP2-lysine mutant, but not the AQP2-cysteine mutant hetero-tetramerized with wild-type; these tetramers localized to respective ER and the Golgi complex. This inhibition of a correct routing of the hetero-tetrameric AQP2-lysine mutant from the Golgi complex to the plasma membrane exemplified that an impaired routing rather than an impaired function of a mutant was the cause of the nephrogenic diabetes insipidus disease in humans [52].

Several cases of homo- or hetero-oligomerization of AQP quaternary structures (which could be evolutionarily conserved) from Arabidopsis, rice, maize, tobacco, grape and spinach were described upon expression in plant cells or Xenopus oocytes. The focus has been on Plasma Membrane Intrinsic Proteins (PIP) isoforms (e.g. PIP1 and PIP2 families) [53–57], although the question remained open if homo- or hetero-tetramers had altered selectivity under normal or adverse environmental settings [58].

For example, hetero-tetramerization of OsPIP1;3 with OsPIP2;2 or OsPIP2;4, when expressed in root cells indicated increased water permeability, while that with OsPIP2;3 did not have the same effect [59]. This effect was assigned to the trafficking of hetero-tetramers to a plasma membrane. In Beta vulgaris PIP AQPs, only the BvPIP2;2 isoform formed functional hetero-tetramers with BvPIP1;1 (important for regulation and fine-tuning of water transport), while BvPIP2;1 failed to homo-tetramerize with either BvPIP isoform upon expression in Xenopus oocytes [60]. Notably, when analysed by confocal fluorescent microscopy, BvPIP2;1 could not translocate fluorescently labelled BvPIP1;1 from the intracellular to a plasma membrane environment when co-expressed, contrary to BvPIP2;2. Based on the MD simulations and mutational studies, it was concluded that the failure of BvPIP2;1 to hetero-tetramerize was in sequence variations of extracellular loop A. In BvPIP2;1, this loop A folded differently and failed to facilitate inter-chain hydrophobic contacts with other BvPIP isoforms, which hampered tetramer formation [60]. It would be of interest to compare the selectivity of BvPIP homo- and hetero-tetramers and if they permeate other molecules besides water. It was also shown that the Beta vulgaris PIP2-PIP1 hetero-tetramers, upon expression in Xenopus oocytes, assembled in various stoichiometries (3:1, 1:3, 2:2) depending on cell expression levels [61]. Given that in BvPIP2;1, loop A did not form inter-chain associations with other BvPIP isoforms, and in spinach PIP2;1 loop D, controlled gating via pore occlusion [62], it would be vital to examine how loop movements outside the hydrophobic cores of AQPs affect gating and if this is linked to oligomerization.

Another study of plant homo- and hetero-tetramerization included maize PIP1 and PIP2 AQP isoforms, exhibiting different water-channel activities upon expression in Xenopus oocytes [63]. When maize PIP1;2 and PIP2;5 were co-expressed, they were assembled as homo- and hetero-dimers. Structural assessments and analyses of mutant versions AQPs assigned predominantly positively charged, neutral and hydrophobic determinants for α-helical cores of PIP1;2 (lysine, glutamine, Trp and Phe residues) and PIP2;5 (Leu, Phe, Trp and Tyr residues). These side-chains via hydrophobic, aromatic–aromatic, aromatic–sulfur, cation–pi, and H-bonds in the α-helical regions of AQPs were critical for a hetero-oligomeric quaternary assembly. Using microscopy, the fluorescently labelled wild-type and mutant PIP1;2 AQPs co-localized in ER membranes, while after co-expression with PIP2;5, the PIP1;2-PIP2;5 hetero-oligomer localized in a plasma membrane, suggesting strong complementary interactions between the two isoforms.

To analyse how homo-tetrameric bacterial E. coli AQPZ (PDB accession 2ABM) [64] and human AQP5 (PDB accession 3D9S) [65], both solved by X-ray crystallography, form contacts and interact with lipids in bilayers of known composition, we conducted MD simulations [22] of tetramers embedded in symmetric bilayers mimicking bacterial or mammalian membranes.

Multiple sequence alignment of AQPZ and AQP5 (Figure 1A) and their monomeric structures (chains A) (Figure 1B) indicated that these AQPs have similar dispositions of α-helices 1–6, and re-entrant α-helices 1–2, and NPA1-NPA2 motifs (cpk sticks); the latter two are involved in water and other solute permeation. However, the hydrophobic properties of α-helices 1–6, representing Ala, Gly, Val, Ile, Leu, Phe and Met residues (Figure 1B; in green), differed between the two AQPs, pre-empting that their lipid binding properties might also vary.

Analyses of monomeric interactions of homo-tetrameric E. coli AQPZ and human AQP5, using PDBePISA [67] indicated that these contacts were mediated through 30 and 42 residues, respectively (Figures 2A and 3A, left panels). In both AQPs, an equal number of hydrophobic residues (18), were exposed to the neighbouring protomers in A-chains, nine of which were Phe, Tyr and Trp side-chains. From these residues, Phe side-chains dominated with five and six in bacterial and human AQPs, respectively (Figures 2A and 3A, right panels).

The tetrameric E. coli AQPZ was embedded in a symmetric bilayer mimicking a bacterial membrane [22] [POPE 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine):POPG (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoglycerol) = 2:1 ratio]. Human tetrameric AQP5 was embedded in a symmetric bilayer mimicking a mammalian membrane [22] [CHOL (cholesterol):POPE:POPC (1-palmitoyl-2-oleoylphosphatidylcholine):POPS (1-hexadecanoyl-2-(9Z-octadecenoyl)-sn-glycero-3-phosphoserine):PSM (N-palmitoyl-d-sphingomyelin) = 3.1:1.9:1.6:1.5:1 ratio]. In both cases, we used the Chemistry at Harvard Macromolecular Mechanics Graphical-User-Interface (CHARMM-GUI) for embedding [68,69]. Tetrameric complexes were minimized in YASARA [70] with knowledge-based Yasara2 forcefield (bond distances, planarity of peptide bonds, bond angles, Coulomb terms, dihedral angles, van der Waals forces) combined with the particle-mesh-Ewald energy function for long-range electrostatics at a cut-off 8.0 Å to obtain smoothed electrostatic potentials [71]. Incorrect covalent geometry and conformational stress were removed by short steepest descent minimization (time 5000 fs, 1 ft time steps, 298 K) followed by simulated annealing with 1 fs time step, atom velocities scaled down by 0.9 every 10th step, until convergence at 710 steps with energy improvement of >0.05 kJ/mol per atom during 200 steps. The absence of plant tetrameric AQP co-ordinates and membrane force-field parameters disallowed similar analyses of plant AQPs. Structural images were generated in PyMOL Molecular Graphics System v3.0.7.3 (Schrődinger LLC).

After embedding bacterial and human AQPs in lipid bilayers, on average around 18 and 17 respective lipid-exposed residues were identified in each protomer (Figures 2B and 3B, left panels). Given the dissimilar compositions of bacterial and human membrane bilayers [22], these interactions reflected differences in anchoring AQPs. Tetrameric E. coli AQPZ made close contacts with up to 10 POPE (zwitterionic) and POPG (anionic) lipid molecules through the hydrophobic tail regions at the periphery of each homo-tetramer (Figure 2B). Interactions between E. coli AQPZ and lipids were also mediated via electrostatic forces, at the separations of 2.6–3.7 Å, e.g. between a positively charged Lys residue and a phosphate group in the hydrophobic head region of POPG. These interactions were also formed via H-bonds between the Arg, Glu, Gln, Tyr and Thr residues and the oxygen atoms of POPG and POPE, containing unsaturated palmitoyl and oleoyl (POPE and POPG; both C16.1/C18:1) fatty acids attached to respective phosphatidylglycerol or phosphatidylethanolamine backbones (Figure 2B). An example of how AQPZ (chains A and B) interacted with POPE and POPG lipid molecules illustrates that Trp, Phe and Thr residues formed contacts with polar head regions, while hydrophobic tails of lipids aligned with hydrophobic surfaces of α-helices (Figure 2B; right panel).

We also revealed molecular interactions with lipids in human AQP5 embedded in the lipid bilayer composed of five types of lipids (Figure 3B). Here, all five lipids: CHOL (neutral), POPE (zwitterionic), POPC (zwitterionic), POPS (anionic), and PSM (anionic) containing unsaturated palmitoyl and oleoyl (POPE; C16.1/C18:1), palmitoyl and oleoyl (POPC; C16.0/C18:1), palmitoyl and oleoyl (POPS; C16.0/C18:1 and phospho-l-serine) and palmitoyl and oleoyl (PSM; C16:0/18:1) fatty acids, made contacts with human AQP5. These lipids participated in aliphatic, aromatic, electrostatic and H-bond interactions with AQP5, with the latter two types of interactions represented by 20 short contacts at the separations between 2.5 and 3.7 Å (Figure 3B). Amongst these contacts, Arg and Lys residues interacted with phosphate headgroups of POPE, POPC, POPS and PSM, and OH groups of CHOL, and their hydrophobic tails were wedged between α-helices. In the latter case, the Tyr106 side-chain formed an H-bond with the CHOL head group at the separation of 2.7 Å, while its hydrocarbon tail aligned with one of the α-helices of AQP5 (Figure 3B; left and right panels).

Our computational analyses of bacterial AQPZ [64] embedded in lipid bilayers mimicking bacterial membranes [22] relate to the experimental findings of E. coli aquaglyceroporin GlpF with low permeation activity in liposomes made of native E. coli lipids. This finding was, however, surprising, as GlpF should work optimally in its natural membrane environment [72]. In this experimental setup, GlpF permeation was strongly affected by negatively charged lipids, irrespective of the chemical nature of lipid headgroups, and was insensitive to a lateral bilayer pressure. These data imply that GlpF could be modulated by a negative charge density.

Furthermore, E. coli AQPZ interaction energies were quantified as a function of a bilayer thickness [36], when embedded in a bilayer of [1,2-dioleoyl-sn-glycero-3-phosphoethanolamine (DOPE), 1,2-dioleoyl-sn-glycero-3-phospho-l-serine (DOPS) and 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC) at the ratio 8:1:1], using HS-AFM and kinetic membrane elastic theory. This work revealed that oligomerization was more favourable in lipids matching the protein hydrophobic thickness, whilst interactions with multiple neighbours during protein array formation were favourable in bilayers with a great mismatch. Meanwhile, the archetypal E. coli AQPZ interactions with negatively charged phospholipids were conserved between positively charged residues and stabilized its quaternary tetrameric assembly [41,73].

Finally, the computational analyses of dimyristoyl-glycero-phosphocholine (DMPC) bilayer properties with AQP0 (PDB accession 3M9I) indicated that the AQP0 surfaces could induce specific fluid- and gel-phase prone areas [74] and that these regions could influence its mobility. These AQP0-lipid sorting interactions were compatible with a square array oligomerization of tetrameric lens-specific AQP0 (PDB accession 2B6O) encircled by DMPC lipids [75]. This type of oligomerization was recently confirmed for the same AQP0 forming square arrays in lens membranes enriched in sphingomyelin and cholesterol [76]. The latter study of AQP0 in a closed pore conformation solved by cryo-EM in double-layered two-dimensional crystals showed up to seven DMPC molecules with each monomer, forming contacts via Tyr and Trp side-chains.

• Importance in the field: Structural, biophysical, biochemical and in vivo studies emphasized the fundamental importance of mechanistic knowledge for our understanding of quaternary oligomeric arrangements and thermodynamics of membrane transport proteins assemblies and AQPs to understanding their modus-operandi.

• Summary of current thinking: A large body of evidence has cemented the roles of structural properties of membrane proteins and lipid bilayers in forming functional membrane proteins. Although this work describes mechanisms using current approaches, the view is open on how future methods could stimulate progress, including interpretability constraints, in emerging single-molecule-based and cryo-ET methods. These improvements will become critical to unravelling the source of entropy of membrane proteins and pathways underlying their integration, folding, oligomerization and quaternary structure formation with binding partners.

• Future directions: Future exceedingly interdisciplinary approaches will form the basis for creating transport proteins and AQPs with optimized features to inspire bioengineering and develop a sustainable and healthy society. We need to resolve and develop: (i), How are homo- and hetero-oligomeric AQP assemblies formed at atomic levels; (ii), Do post-translational modifications contribute/regulate these processes and what is their importance in forming functional proteins in any given biological context; (iii), What are the structural determinants that play roles in the selectivity function of AQP tetramers under normal and a variety of environmental conditions; (iv), For predictions of conformational dynamics of plant AQP tetramers in bilayers, the force-field parameters of plant lipids need to be defined.

Data contained in the article are available within the submitted material and Figures 1–3.

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

This work was supported by funding from the Australian Research Council (DP120100900) to M.H.

Open access for this article was enabled by the participation of University of Adelaide in an all-inclusive Read & Publish agreement with Portland Press and the Biochemical Society under a transformative agreement with CAUL.

M.H. wrote, edited and finalized the text of the article, conducted experiments and generated graphics.

M.H. thanks the University of Adelaide and the Waite Research Institute (Australia) for support. Contributions, suggestions and critical reading from Professor Stephen D. Tyerman (University of Adelaide) are valued. Dr Elmar Krieger is acknowledged for advice on molecular dynamics simulations, and with Dr Kornel Ozvoldik, thanked for implementing force-field parameters of lipid molecules in YASARA (1993–2024 YASARA Biosciences GmbH, Vienna, Austria).

3D

three-dimensional

AFM

atomic force microscopy

AQP

aquaporin

CHARMM-GUI

Chemistry at Harvard Macromolecular Mechanics Graphical-User-Interface

CHOL

cholesterol

COPI

Coat Protein Complex I

COPII

Coat Protein Complex II

cryo-EM

cryogenic-electron microscopy

cryo-ET

cryogenic-electron tomography

DMPC

dimyristoyl-glycero-phosphocholine

DOPC

(1,2-dioleoyl-sn-glycero-3-phosphocholine (1-palmitoyl-2-oleoylphosphatidylcholine)

DOPE

1,2-dioleoyl-sn-glycero-3-phosphoethanolamine

DOPS

1,2-dioleoyl-sn-glycero-3-phospho-l-serine

ER

endoplasmic reticulum

GlpF

glycerol facilitator

HS-AFM

high-speed atomic force microscopy

PIP

plasma membrane intrinsic protein

POPE

1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine

POPG

1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoglycerol

POPS

1-hexadecanoyl-2-(9Z-octadecenoyl)-sn-glycero-3-phosphoserine

PSM

(N-palmitoyl-d-sphingomyelin)

PDB

Protein Data Bank.