==== Front
bioRxiv
BIORXIV
bioRxiv
Cold Spring Harbor Laboratory

37398072
10.1101/2023.05.30.542905
preprint
1
Article
Lipid Bilayer Strengthens the Cooperative Network of a Membrane-Integral Enzyme
Muhammednazaar Shaima 1
Yao Jiaqi 1
Guo Ruiqiong 1
Rhee May S. 1
Kim Kelly H. 2
Kang Seung-gu 3
Hong Heedeok 12
1 Department of Chemistry, Michigan State University, East Lansing, MI 48824, USA
2 Department of Biochemistry & Molecular Biology, Michigan State University, East Lansing, MI 48824, USA
3 Computational Biology Center, IBM Thomas J. Watson Research Center, Yorktown Heights, NY 10598, USA
Author contributions. S.M., J.Y., R.G., S-g.K., and H.H. designed research. S.M., J.Y., R.G., M.S.R., S-g.K., and H.H. performed research. S-g.K. and R.G. provided new analytical tools. S.M., J.Y., M.S.R., S-g.K., and H.H. analyzed data. S.M., J.Y., R.G., S-g.K., and H.H. wrote the paper.

Correspondence and requests for materials should be addressed to H.H. (honghd@msu.edu) and S-g.K. (sgkang@us.ibm.com).
31 5 2023
2023.05.30.542905https://creativecommons.org/licenses/by/4.0/ This work is licensed under a Creative Commons Attribution 4.0 International License, which allows reusers to distribute, remix, adapt, and build upon the material in any medium or format, so long as attribution is given to the creator. The license allows for commercial use.
nihpp-2023.05.30.542905.pdf
Lipid bilayer provides a two-dimensional hydrophobic solvent milieu for membrane proteins in cells. Although the native bilayer is widely recognized as an optimal environment for folding and function of membrane proteins, the underlying physical basis remains elusive. Here, employing the intramembrane protease GlpG of Escherichia coli as a model, we elucidate how the bilayer stabilizes a membrane protein and engages the protein’s residue interaction network compared to the nonnative hydrophobic medium, micelles. We find that the bilayer enhances GlpG stability by promoting residue burial in the protein interior compared to micelles. Strikingly, while the cooperative residue interactions cluster into multiple distinct regions in micelles, the whole packed regions of the protein act as a single cooperative unit in the bilayer. Molecular dynamics (MD) simulation indicates that lipids less efficiently solvate GlpG than detergents. Thus, the bilayerinduced enhancement of stability and cooperativity likely stems from the dominant intraprotein interactions outcompeting the weak lipid solvation. Our findings reveal a foundational mechanism in the folding, function, and quality control of membrane proteins. The enhanced cooperativity benefits function facilitating propagation of local structural perturbation across the membrane. However, the same phenomenon can render the proteins’ conformational integrity vulnerable to missense mutations causing conformational diseases1,2.
==== Body
pmcThe solvent plays a pivotal role in the folding and function of proteins3,4. For water-soluble proteins, the hydrophobic effect (i.e., the unfavorable ordering of water molecules around nonpolar solutes) provides a crucial driving force for folding by inducing collapse of nonpolar residues in the protein interior, leading to expulsion of solvating water to the bulk aqueous medium5. Involving the collective formation and disruption of water hydrogen (H)-bond networks, the solvent further mediates cooperativity in the folding and allosteric protein–ligand interactions4,6–9.

In contrast, membrane proteins fold and function in a lipid bilayer. The folding of helical membrane proteins can be described using the two-stage model10: In Stage I, transmembrane (TM) helices are established across the bilayer driven by the hydrophobic effect that induces the burial of nonpolar segments of the polypeptide chain in the bilayer, further benefitted by the favorable formation of backbone H-bonds therein11,12. In Stage II, the helices associate into a compact native structure. While the hydrophobic effect is not likely strong within the bilayer due to the lack of water, various molecular forces are known to drive this stage including interhelical van der Waals (vdW) packing and polar interactions13–16, the backbone and side-chain entropy17,18, and the bilayer forces (e.g., the lipid-packing pressure and lipid deformation by the hydrophobic mismatch between the bilayer and protein)19–21.

Nonetheless, the precise role of lipid solvation in the folding energy landscape and cooperativity of membrane proteins is poorly understood. The “lipophobic effect” (the lipid-mediated protein association in the membrane) has been proposed as an analogous phenomenon to the hydrophobic effect in water21–24, but such effect remains untested for the folding of bona fide membrane proteins. Cooperativity, which links the behavior of distant sites25, underlies the function of membrane proteins (e.g., receptors and transporters) enabling propagation of physical or chemical stimuli from one side of the protein to the other across the bilayer. It is unknown if lipids play a role in mediating cooperativity for membrane proteins as water does for globular proteins.

Using GlpG, a member of the universally conserved rhomboid protease family, we hypothesize that, if the lipid bilayer simply serves as an inert hydrophobic solvent for membrane proteins, the information pertaining to stability and cooperativity is solely encoded in the amino acid sequence. Consequently, the characteristics of a hydrophobic medium would have no impact on the strengths of individual residue interactions or their overall interaction network. To test this hypothesis, we conducted a comparative analysis on the stability, residue interaction network, and solvation dynamics of GlpG in two distinct hydrophobic media: bicelles (a lipid bilayer edge-stabilized by detergents) composed of 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC) and 3-[(3-cholamidopropyl) dimethylammonio]-1-propanesulfonate (CHAPS), which serves as a native-like bilayer medium, and dodecylmaltoside (DDM) micelles, which represents a nonnative hydrophobic medium (Fig. 1a). These specific amphiphilic assemblies are widely used in structural, folding, and functional studies of membrane proteins26,27.

Bilayer increases GlpG stability.

Quantification of the thermodynamic stability of a membrane protein (ΔGoN-D, the free energy difference between the native and denatured states) is a daunting task due to the inherent difficulty of achieving the folding reversibility in a lipid bilayer28. Here, we overcame the challenge by employing the steric trapping strategy, which capitalizes on the coupling of the spontaneous denaturation of a doubly biotinylated protein to the simultaneous binding of two bulky monovalent streptavidin (mSA) molecules (Fig. 1b)29–31. This strategy successfully recapitulates Stage II of membrane protein folding (see below).

Cryo-electron microscopy of bicelles indicates the formation of uniform discoidal bilayers with a diameter of ~90 Å (Extended Data Fig. 1), large enough to accommodate both native and sterically denatured GlpG32. To site-specifically biotinylate GlpG, we employed BtnPyr, the thiol-reactive biotin derivative with pyrene (Fig. 1b and Extended Data Fig. 2)30. Upon conjugation to GlpG, pyrene fluorescence reports the binding of quencher-labeled mSA30. We generated the double-biotin variants, 95N172M‒BtnPyr2 and 172M267C‒BtnPyr2 (N, M, and C: the residue positions of engineered cysteine, the N-terminal, Middle, and C-terminal helices, respectively) to measure the stability at the N- and C-terminal halves of GlpG (i.e., N- and C-subdomains), respectively (Fig. 1c). The binding isotherms between GlpG and a weaker biotin-affinity mutant, mSA-E51S (Extended Data Figs. 3–5) displayed a tight, unhindered first binding followed by an optimally attenuated second binding (Figs. 1b and 1d). The inactivation phase of proteolytic activity of GlpG, which was used as a folding indicator, agreed with the second mSA binding phase regardless of whether GlpG was native or sterically denatured prior to incorporation into bicelles (Extended Data Figs. 6–8). Sterically denatured GlpG became highly susceptible to proteolysis by Proteinase K, indicating the increased conformational flexibility and water accessibility relative to native GlpG (Extended Data Fig. 9). This result establishes the folding reversibility and coupling between the second mSA binding and denaturation, thereby validating the steric trapping scheme.

The second mSA binding phases were fitted to yield ΔGoN-D,bicelleN = −7.0 ± 0.2 kcal/mol for N-subdomain and ΔGoN-D,bicelleC = −6.7 ± 0.2 kcal/mol for C-subdomain in bicelles (Fig. 1e). For comparison, the two subdomains exhibit distinct folding properties in micelles: N-subdomain (ΔGoN-D,micelleN = −5.8 ± 0.2 kcal/mol), whose disruption leads to global denaturation, is more stable than C-subdomain (ΔGoN-D.micelleC = −4.7 ± 0.1 kcal/mol), which undergoes subglobal denaturation30. Thus, relative to micelles, bicelles stabilized N- and C-subdomains by −1.2 ± 0.3 kcal/mol and −2.0 ± 0.3 kcal/mol, respectively, and induced a near-uniform subdomain stability (|ΔGoN-D,bicelleN – ΔGoN-D,bicelleC | = 0.3 ± 0.3 kcal/mol). The stability that we determined directly under native conditions (ΔGoN-D = −12kBT) is larger than that obtained from extrapolation to zero force (–6.5kBT) in the molecular tweezer study, likely due to the difference in conformation of the denatured states (Extended Data Fig. 10)33–35.

Bilayer facilitates residue burial.

Next, we investigated whether the features of hydrophobic environment affect the contribution of individual residue interactions to GlpG stability. To this end, 37 residues with various degrees of burial in the protein interior were chosen for large-to-small mutation except for A253V, G261A, and A265V. All mutation-induced stability changes measured at both subdomains in micelles vs bicelles (ΔΔGoWT-Mut,micelle vs ΔΔGoWT-Mut,bicelle) (Fig. 2a-left; Extended Data Tables 1–2; Extended Data Figs. 11–12) displayed linear correlation with the slope close to 1 (m = 1.1 ± 0.1). This may indicate that individual residue interactions make a similar contribution to the stability in both environments. However, the mutational impacts displayed differential environmental sensitivity depending on the region where the stability was measured (Fig. 2a-right): While the mutations caused destabilization of N-subdomain to a similar extent in micelles and bicelles (m = 1.0 ± 0.1), they induced larger destabilization of C-subdomain in bicelles than in micelles (m = 1.3 ± 0.1) (p<0.005 from Chow’s test).

In-depth analysis of the mutational impacts based on the degree of residue burial of mutated residues provided insights into the origin of the environmental sensitivity. For the mutations of residues completely buried in the protein interior (Fig. 2b-left), the fitted slopes exceeded one (m = 1.2 ± 0.2 at N-subdomain and m = 1.7 ± 0.2 at C-subdomain). Thus, an impact of disrupting internal packing was amplified in bicelles, indicating that the bilayer induced more favorable burial of wild-type residues in the protein interior than micelles. This tendency was more pronounced at C-subdomain than at N-subdomain (m = 1.7 vs 1.2, p<0.05) regardless of the position of mutation. As the degree of residue exposure increased, the slope decreased to m = 0.5–0.8 (p<0.005).

The differential environmental sensitivity of mutational impacts observed at the two subdomains can be attributed to their distinct packing motifs (Extended Data Fig. 13). In C-subdomain, the stability primarily arises from the backbone–backbone contact between the conserved Gly-zipper motifs (Gly–x3–Gly–x3–Gly: Gly can be Ala or Ser, and x is any residue) in TM4 and TM6, respectively30,36–38. Due to the weakly polar nature of the backbone contact, the bilayer which has a more dehydrated hydrocarbon core than a micelle provides the enhanced stability (Fig. 1e)39,40. Thus, disruption of this backbone contact leads to the larger destabilization31 (m = 1.7). On the other hand, the stability of N-subdomain mainly relies on extensive vdW contacts involving large and small nonpolar residues30,38. The impact of disrupting the vdW contacts showed a modest sensitivity to the specific features of the hydrophobic environment (m = 1.2).

Bilayer enhances cooperativity.

Are the lipid effects on GlpG stability localized only to the specific region under investigation (i.e., the subdomain or the site of mutation) or do they globally impact the residue interaction network? To address this question, we employed our cooperativity profiling analysis allowing us to investigate whether a given residue engages in local or cooperative interactions with its surrounding30. This experimental approach measures the degree of spatial propagation of structural perturbation induced by a point mutation, as quantified by the differential effect of the mutation on the stability of the two subdomains (i.e., ΔΔΔG = ΔΔGoWT-MutN − ΔΔGoWT-MutC)30. We used four regular cut-off values, ΔΔΔG = −2RT, −RT, +RT and +2RT (R: gas constant and T: absolute temperature) to define the cooperativity profile of each residue30.

The cooperativity profiles mapped on the GlpG structure unveiled distinct types of residue interactions, classified as “cooperative” (a mutation similarly destabilizes the two subdomains), “localized” (a mutation preferentially destabilizes the subdomain bearing the mutation), and “overpropagated” (a mutation on one subdomain induces the larger destabilization of the other) (Fig. 3). In micelles, cooperative interactions clustered into multiple distinct regions, including the packing core near the bilayer center (Met100, Cys104, Leu174 and Thr178)30, the water-conduction channel (Ser201, Met249, His150 and Asn154) connected to the catalytic dyad (Ser201–His254)41, and the TM4/TM6 interface where many residues (Ala253, His254, Gly261, Ala265 and Asp268) engaged in overpropagated interactions30.

Strikingly, the cooperativity map exhibited a substantially different pattern in bicelles. Most of the localized and overpropagated interactions in micelles turned into cooperative interactions in bicelles. Resultantly, the entire packed regions of GlpG formed a single cooperative unit in bicelles. The narrower cutoff values (ΔΔΔG = −RT, −1/2RT, +1/2RT and +RT) yielded the cooperativity profiles resembling those in micelles with the regular cutoff values (Extended Data Fig. 14). Thus, the cooperativity profiles in micelles were partially preserved in bicelles.

This result provides compelling evidence that the bilayer tightly engages the residue interaction network compared to micelles, thereby facilitating propagation of structural perturbation throughout the protein. Thus, the lipid effects on the stability stem from the globally strengthened cooperative network. The proteolytic mechanism of GlpG involves the coordinated motions of multiple structural segments (TM4, TM6, L4 and L5) upon substrate binding42. Remarkably, bicelles elicited a five-fold increase in GlpG activity relative to micelles for both membrane-bound (Extended Data Figs 15–16) and water-soluble substrates43, which may be attributed to the augmented cooperativity in bicelles.

Bilayer weakly solvates protein.

Finally, to elucidate the molecular basis underlying the environmental dependence of GlpG stability and cooperativity, we carried out all-atom MD simulation of the GlpG–bilayer and GlpG–micelle complexes, as well as the micelles alone up to 2.3 μs. In the micelle simulation, we chose two aggregation numbers of DDM, NA,DDM = 120 (DDM120) and 150 (DDM150) within the experimental range (NA,DDM = 90–150)44 (Extended Data Fig. 17). Although a comprehensive analysis of protein stability requires the study of both the native and denatured states, modeling of the denatured state ensemble (DSE) in atomic details is yet challenging for membrane proteins. Thus, our simulation focused on the native state to infer the amphiphile effects on the stability.

Upon reaching the conformational equilibration of both protein and amphiphiles (Fig. 4a), we analyzed the solvation dynamics of amphiphiles by calculating the contact autocorrelation as a function of time for the protein–amphiphile and amphiphile–amphiphile interactions, respectively, and their residence times (τR’s) (Fig. 4b)45. The tR was defined as the time at which the amplitude of autocorrelation reached 1/e of its initial value. All contact autocorrelation decayed to <1%, indicating that the intermolecular interactions involving amphiphiles were largely equilibrated during simulation. Interestingly, amphiphiles exhibited longer residence times on GlpG (80–90 ns) than on themselves (20–40 ns).

Based on the equilibrated solvation dynamics and the residence times, we were able to quantify the solvation free energy (ΔGoSolv = −RT•ln[τR,protein-amphiphile/τR,amphiphile-amphiphile]) as a measure of an amphiphile’s affinity for the first solvation shell of GlpG relative to the likewise favorable interaction with another amphiphile in the bulk. Surprisingly, contrary to the expectation that a lipid molecule with double aliphatic tails would form stronger vdW contacts with the protein than a detergent with a single tail, lipids exhibited weaker solvation on GlpG than detergents (ΔGoSolv,Lip = −0.50 ± 0.02 kcal/mol vs ΔGoSolv,DDM120 = −0.85 ± 0.03 kcal/mol and ΔGoSolv,DDM150 = −0.61 ± 0.02 kcal/mol) (Fig. 4b-left). This weaker lipid solvation was attributed to the longer τR of lipid–lipid contacts and the comparable or shorter τR of protein–lipid contacts than the respective τR’s of detergents.

Notably, the increase in NA,DDM of a micelle from 120 to 150 led to the weakening of solvation possibly due to the increased detergent-mixing entropy in the larger micellar volume. Furthermore, the headgroup and tail regions of an amphiphile differently contributed to the solvation (Fig. 4b-middle and right). For lipids, the solvation was primarily driven by the headgroup whereas for detergents, by the tail.

MD simulation suggests that the bilayer acts as a poorer solvent than micelles mainly due to the strong lipid–lipid interactions, which facilitate dissociation of lipids from the protein. This leads to the intriguing scenario that the lipid-induced enhancement of stability and cooperativity arises from the dominant intraprotein interactions which outcompete the weak lipid solvation. Despite the small difference in ΔGoSolv between lipids and detergents (~0.2 kcal/mol), considering the number of amphiphile molecules in the first solvation shell (~40 for lipids and ~70 for detergents), the cumulative difference in ΔGoSolv piles up to ~30 kcal/mol. This difference can impact the folding energetics of GlpG, highlighted by the dramatic influence of lipid environment on the protein’s stability and cooperativity.

Conclusion and Outlook.

Our findings demonstrate a profound impact of lipid solvation on the internal organization of a membrane protein. Contrary to the prevalent view as an inert medium, the bilayer actively participates in folding by promoting compaction of polypeptide chains. The bilayer is also known to induce contraction of the DSE of GlpG allowing partial association of the TM helices32. Both the studies point to the general lipophobic effect as a key driving force for membrane protein folding bearing an analogy to the hydrophobic effect for water-soluble proteins. The magnitude of the solvation free energy predicted from our simulation (ΔGoSolv = ~–0.6 kcal/mol-lipid) is comparable to the thermal energy (~0.6 kcal/mol). Thus, protein interactions, either intra- or intermolecular, whose strengths surpass the thermal fluctuation can drive compaction of the polypeptide chains in the bilayer. This suggests that the solvation force as a minimal safety threshold for preventing nonspecific collapse of polypeptide chains may not be strong in the bilayer. Notably, the cellular folding and maturation is known to be inherently inefficient for membrane proteins, rendering the actions of protein quality control mechanisms necessary to alleviate misfolding stresses in all types of subcellular organelles46.

The lipid-induced enhancement of cooperativity is a double-edged sword in the function, folding, and quality control of membrane proteins. Many membrane proteins require conformational changes spanning the entire lengths of the proteins to transmit chemical or physical stimuli across the bilayer47–49. Our finding indicates that the bilayer serves as an adequate conductive medium of such stimuli, facilitating conformational changes in a cooperative manner. However, the enhanced cooperativity can render the conformational integrity of membrane proteins susceptible to disease-causing mutations (Fig. 2b-left). Most of disease-causing mutations on proteins are known to be detrimental to the stability rather than to disrupt active-site residues2,50,51. The mapping of disease-causing mutations on the structures of G-protein coupled receptors, ion channels and transporters displays a strong bias of finding such mutations in the TM regions over the extramembraneous regions1. In the TM regions, the propensity is even more pronounced for the residues buried in the protein interior than for the residues exposed to the lipid environment1. Collectively, our study reveals a foundational physical principle underlying various molecular events that occur in the cell membranes.

MATERIALS AND METHODS

Expression and purification of GlpG.

E. coli BL21(DE3)RP cells were transformed with pET21a vector encoding the transmembrane (TM) domain (residues 87–276) of GlpG1. The cells in Luria-Bertani (LB) broth were grown at 37°C until OD600nm reached 0.9. Protein expression was induced at 0.5 mM IPTG followed by overnight culture at 15°C. Harvested cells were resuspended in 50 mM 2-amino-2-(hydroxymethyl)propane-1,3-diol;dihydrochloride (Tris-HCl) buffer (pH 8.0, 5 mM ethylenediamine-tetraacetic acid (EDTA), 0.5 mM Tris-(2-carboxyethyl)phosphine (TCEP), 0.5 mM phenylmethylsulfonyl fluoride (PMSF)). After the removal of aggregates, the supernatant of cell lysates was centrifuged to isolate the total membrane fraction at 24,000 rpm for 2 h in the 45Ti rotor using ultracentrifuge (Beckman-Coulter). Membrane resuspension in 50 mM Tris-HCl buffer (pH 8.0, 200 mM NaCl, 1 mM TCEP, 0.25 mM PMSF) was solubilized by the addition of 0.7%-w/v n-dodecyl-β-D-maltoside (DDM). After the removal of aggregates, GlpG in supernatant was purified using nickel-nitrilotriacetic acid (Ni-NTA, Qiagen) affinity chromatography in 50 mM Tris-HCl buffer (pH 8.0, 200 mM NaCl, 0.1% DDM).

Biotinylation of GlpG.

50 μM of the double-cysteine variant of GlpG (P95C/G172C or G172C/V267C) was incubated with a 10-molar excess of TCEP for 1 h at 25°C. A 40-molar excess of BtnPyr-IA in dimethyl sulfoxide (DMSO) was added and the reaction proceeded overnight at 25°C. Excess free labels were removed by washing GlpG bound to Ni-NTA resin with 0.1%-w/v DDM and further by dialysis. Labeling efficiency was determined by measuring the absorbance of pyrene (ε346nm = 42,000 M−1cm−1) and protein concentration by 660 nm assay (Bio-Rad). An SDS-PAGE shift assay for isolating single-mSA bound, double-mSA bound, and no-mSA bound GlpG was carried by mixing 10 μl of 5 μM GlpG with 10 μl of the SDS sample buffer (30 min), adding 10 μl of 25 μM mSA-WT (30 min), and running SDS-PAGE on ice without sample boiling.

Preparation of monovalent streptavidin.

Detailed procedures are described in the previous literatures1–3. Streptavidin (active or inactive) encoded in pET21a vector was expressed in E. coli BL21(DE3)RP cells in inclusion body. To label mSA with a thiol-reactive dabcyl quencher (dabcyl-maleimide, AnaSpec), Tyr83 near the biotin-binding pocket in the active subunit was mutated to cysteine. Active subunit: wild-type streptavidin or its weaker biotin affinity variants (W79M, S45A, S27A, and E51S) with a C-terminal His6 tag; Inactive subunits: the triple mutant (N23A/S27D/S45A) without His6 tag2.

Expression and purification of GlpG substrate SN-LYTM2.

The construct containing the second TM segment of E. coli lactose permease (LYTM2)4 fused to the C-terminus of staphylococcus nuclease (SN) (SN–LYTM2) was encoded in pET30a vector1. SN–LYTM2 was expressed and purified using the protocol described previously1. In LYTM2, the residue at five residue-upstream of the scissile bond was mutated to cysteine for conjugating the thiol-reactive, environment-sensitive fluorophore, iodoacetyl-7-nitrobenz-2-oxa-1,3-diazol (IA–NBD amide, Setareh Biotech). The initial slope of NBD fluorescence change vs time represented proteolytic activity of GlpG. NBD fluorescence were monitored using a SpectraMax M5e plate reader (Molecular Devices) with λExcitation = 485 nm and λEmission = 535 nm.

Cryo-electron microscopy of bicelles.

3%-w/v 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC)/3-((3-cholamidopropyl)dimethylammonio)-1-propanesulfonate (CHAPS) bicelles ([DMPC]/[CHAPS], q = 1.5) was prepared without GlpG. Cryo-EM grids were frozen using a Vitrobot Mark IV (ThermoFisher). Briefly, 3.5 μL of each sample was applied to a glow-discharged Quantifoil Cu 1.2/1.3 holey carbon 200-mesh grid. The grid was blotted for 3.5 s prior to plunge freezing in liquid ethane. Cryo-EM images were recorded on a Talos Arctica (ThermoFisher) operated at 200 kV and equipped with a Falcon 3EC direct electron detector camera. Images were recorded in counting mode using EPU software at a nominal magnification of x92,000 (1.12 Å/pixel), with a defocus of −2.5 μm. Micrographs were collected as single-frame images with a total exposure time of 1.5 s and a total dose of 30 electrons/Å2. A total of 6,538 particles from 10 images were auto-picked and extracted into 192 x 192-pixel boxes. The particles were then subjected to 2D classification using cryoSPARC into 50 classes. The diameter of the bicelle in each class average was measured to estimate the size distribution of the bicelles.

Preparation of native and sterically denatured GlpG in micelles.

20 μM of the double-biotin variants of GlpG (95N172M–BtnPyr2 or 172M267C–BtnPyr2) was incubated with 2.4 molar excess of mSADAB-E51S in 20 mM N-2-hydroxyethylpiperazine-N-2-ethane sulfonic acid (HEPES) buffer (pH 7.5, 40 mM KCl, 0.5 mM DTT, 5 mM DDM) at 25°C. The extent of denaturation was monitored every 24 h using GlpG activity as a folding indicator. For 172M267C–BtnPyr2, maximum denaturation was reached in 24 h. For 95N172M–BtnPyr2, 8 mM SDS was added in the beginning to facilitate the denaturation and incubated for 5 h.

Fluorescence quenching assay to measure incorporation of GlpG into bicelles.

As a positive control for complete incorporation of GlpG in bicelles, DMPC lipids were mixed with dabcyl–1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine (POPE) (Avanti polar lipids) at the molar ratio of 99.5:0.5 in chloroform in a glass tube and dried under stream of nitrogen. After further dried in vacuum for 4 h, the lipid mixture was solubilized in 500 μL of 20 mM HEPES buffer (pH 7.5, 5%-w/v β-octylglucoside (Anatrace)) at the final lipid concentration of 7.5%-w/v. GlpG variant 95N172M–BtnPyr2 or 172M267C–BtnPyr2 in DDM was added to the resuspension and incubated on ice for 30 min. Biobeads (Bio-Rad) were added to remove the detergents in three steps (for each step, 0.2 g/mL of wet Biobeads for 6–12 h at 25°C). Resulting proteoliposomes were extruded using a 0.2 mm pore size membrane. The total phospholipid concentration was measured using an organic phosphate assay. Based on the lipid concentration, CHAPS was added to form bicelles of q = 1.5. The protein concentration was measured using a 660 nm assay. As a negative control for no incorporation, water-soluble mSA-E51S/Y83C labeled with N-(1-pyrene)maleimide (ThermoFisher) was used. To prepare an experimental sample, native or sterically denatured GlpG (95N172M–BtnPyr2 or 172M267C–BtnPyr2) in DDM was directly injected to the bicelles containing dabcyl-POPE (20 mM HEPES buffer, pH 7.5, 40 mM KCl). In the control and experimental samples, the final concentrations of pyrene labels, DDM, and bicelles were adjusted to 1 μM, 5 mM, and 3%-m/w, respectively. After the incubation of the mixtures at 25°C for 24 h, pyrene fluorescence was measured with λExcitation = 345 nm and λEmission = 390 nm. The degree of quenching, which was related to the degree of GlpG incorporation into bicelles, was determined by the equation, [FNegative control – FExperiment]/[FNegative control – FPositive control] (F: fluorescence intensity of pyrene).

Testing the folding reversibility of GlpG.

Native or sterically denatured GlpG in DDM micelles was directly injected into DMPC/CHAPS bicelles (q = 1.5, 3%-w/v) at various concentrations of mSADab-E51S in 20 mM HEPES buffer (pH 7.5, 40 mM KCl, 0.5 mM DTT) to initiate denaturation and refolding at 25°C, respectively (the final concentration of GlpG = 0.5 μM). To monitor mSA binding, pyrene fluorescence was measured with λExcitation = 345 nm and λEmission = 390 nm on a plate reader every 24 h until an equilibrium was reached (48–72 h). To test the coupling between GlpG denaturation and mSA binding, GlpG activity as a folding indicator was measured at a 20-times molar excess of the substrate incorporated in bicelles.

Proteinase K digestion of native and sterically denatured GlpG.

Native or sterically denatured GlpG (95N172M–BtnRG2 or 172M267C–BtnRG2) was directly injected into DMPC/CHAPS bicelles (3%-w/v, q = 1.5) in 20 mM HEPES buffer (pH 7.5, 40 mM KCl, 5 mM DDM) at the final concentrations of 5 μM and 25 μM for GlpG and mSA-WT, respectively. After incubating at 25°C for 24 h, Proteinase K (the final concentration of 3.4 mg/mL) was added to initiate digestion. The samples were withdrawn at each time point followed by the addition of 10 mM PMSF to quench proteolysis. 10 mM DTT was added and incubated for 1 h to dissociate bound mSA-WT from biotinylated GlpG by cleaving the disulfide bond. SDS-PAGE was run on ice.

Measuring the intrinsic biotin affinity of mSA variants in bicelles.

The mSA variant with a weaker biotin-binding affinity, mSADAB-W79M (FRET acceptor) was titrated to 100 nM of GlpG singly labeled with BtnPyr (FRET donor) at P95C, G172C, or V267C in 20 mM HEPES buffer (pH 7.5, 3% DMPC/CHAPS bicelles, 40 mM KCl, 0.5 mM DTT). Pyrene fluorescence was measured with λExcitation = 345 nm and λEmission = 390 nm. After 24 h, excess free biotin was added to the final concentration of 2 mM and incubated for another 24 h to dissociate bound mSA. The measured pyrene fluorescence serves as a background. Background-subtracted data were fitted to Eq.1 to obtain Kd,biotin of mSADAB-W79M in bicelles.1 Eq.1 F=A1⋅(PT+[mSA]+Kd,biotin)−(PT+[mSA]+Kd,biotin)2−4PT⋅[mSA]2PT+A2

, where F is the measured fluorescence intensity, PT is the total GlpG concentration, [mSA] is the total mSA concentration, Kd,biotin is the dissociation constant of mSADAB from biotin, A1 is the total net change in fluorescence and A2 is the fluorescence level without mSADAB.

The Kd,biotin of a stronger biotin-binding variants (W79M, S45A, or S27A) was measured by a FRET-based competition assay. 1 μM G172C–BtnPyr was pre-equilibrated with a 2- or 5-times molar excess of the dabcyl-labeled mSA variants for 3 h at 25°C, which represented a quenched state. Next, a weaker biotin-affinity mSA variant without the dabcyl label was titrated against the quenched state. Dequenching of pyrene fluorescence was measured with λExcitation = 345 nm and λEmission = 390 nm. Once equilibrium was reached (24–48 h), 2 mM biotin was added to dissociate bound mSA and further equilibrated for 7–24 h. The fluorescence data served as a background signal. Background-subtracted data were fitted to Eq. 2. Eq.2 F=A1⋅−[PT+[mSA]+KunlabelKlabel⋅(CT−PT)]+(PT+[mSA]+KunlabelKlabel⋅(CT−PT))2+4PT⋅[mSA]⋅KunlabelKlabel2PT⋅KunlabelKlabel+A2

, where Kunlabel is the dissociation constant of mSA without dabcyl from BtnPyr, Klabel is the dissociation constant for mSADAB from BtnPyr. Fitted values include A1, A2, and Kunlabel or Klabel. For mSADAB-E51S, 1.5 μM G172C–BtnPyr was first titrated with various concentrations of mSA-S27A without a label, and pyrene fluorescence was measured. This signal served as a background. Then, 2 μM of mSADAB-E51S was added, and quenching of pyrene fluorescence was measured. After reaching an equilibrium (48–72 h), background-subtracted data were fitted to Eq. 2.

Construction of binding isotherms to determine ΔGoN-D of GlpG.

1 μM of GlpG (95N172M–BtnPyr2 or 172M267C–BtnPyr2) was titrated with various concentrations of mSADAB in 20 mM HEPES buffer (pH 7.5, 3%-w/v bicelles or 5 mM DDM micelles, 40 mM KCl, 1 mM DTT). Depending on the stability of GlpG mutant, a series of mSA variants with a weaker biotin affinity (mSADAB-W79M, S45A, S27A, and E51S) were screened until an optimal second binding phase was obtained. The titrated samples were transferred to a 96 well plate, sealed with a polyolefin tape, and incubated at 25°C. After the equilibrium was reached, the binding was measured by quenching of pyrene fluorescence with λExcitation = 345 nm and λEmission = 390 nm on a plate reader. Data were averaged from three fluorescence readings. The second attenuated binding phase was fitted to Eq.4.

Fitting of the second binding phase to obtain DGoN-D of GlpG.

The attenuated second binding of mSA was fitted to the equation derived from the following scheme1: Eq.3 N⋅mSA⇄KDD⋅mSAwhereKD=[D⋅mSA][N⋅mSA]D⋅mSA+mSA⇄Kd,biotinD⋅2mSAwhereKd,biotin=[D⋅mSA][mSA][D⋅2mSA]

The fitting equation was: Eq.4 F=1[1+(Kd,biotin+Kd,biotinKD)⋅1[mSA]]⋅(F∞−Fo)+Fo

Eq.5 ΔGN−Do=−RT⋅ln(1KD)

, where F is the measured fluorescence intensity and Fo and F∞ are the fluorescence intensities from BtnPyr conjugated to GlpG at [mSA] = 0 and at [mSA] = ∞, respectively. [mSA] is the total mSA concentration, Kd,biotin is the unhindered biotin affinity of mSA, and KD is the equilibrium constant for denaturation of GlpG. After obtaining the fitted KD, DGoN-D was calculated.

Cooperativity profiling1.

Specific residue interaction is perturbed by a single point mutation in the background of the double-biotin variants 95N172M–BtnPyr2 (ΔGoN-DN) or 172M267C–BtnPyr2 (ΔGoN-DC), which is set as ‘WT’. Then, the stability change induced by the same point mutation was measured by steric trapping for each WT background (ΔΔGoN-D,WT-MutN = ΔGoN-D,WTN− ΔGoN-D,MutN or ΔΔGoN-D,WT-MutC = ΔGoN-D,WTC− ΔGoN-D,MutC). Then, the differential effect of the mutation on the stability of the two subdomains is quantified as follows: Eq.6 ΔΔΔG=[ΔGN−D,WToN−ΔGN−D,MutoN]−[ΔGN−D,WToC−ΔGN−D,MutoC]=ΔΔGN−D,WT−MutoN−ΔΔGN−D,WT−MutoC

We apply four cut-off values, ΔΔΔG = −2RT, −RT, RT and 2RT (i.e., five sets of the cooperativity profile; R: gas constant and T: absolute temperature) to resolve the degree of cooperativity of each residue interaction. For a given ΔΔΔG value, we assign the cooperativity profile as follows. +2RT<ΔΔΔG: highly localized in N-subdomain; +RT<ΔΔΔG≤+2RT: moderately localized in N-subdomain; −RT≤ΔΔΔG≤+RT: cooperative; −2RT≤ΔΔΔG≤–RT: moderately localized in C-subdomain; ΔΔΔG<–2RT: highly localized in C-subdomain.

Molecular dynamics simulation.

MD simulation setups were based on the crystal structure of E. coli GlpG (PDB code: 2IC8). The bicelle was approximated to a lipid bilayer composed of 315 DMPC molecules, which was constructed using the CHARMM-GUI membrane builder5. Two micellar systems were set up with 120 (DDM120) and 150 DDM (DDM150) molecules per micelle, modeled by symmetrically enclosing the TM domain of GlpG with DDM molecules. Each of the GlpG–bilayer and GlpG–micelle composite systems were immersed in the TIP3P water solvent6 followed by charge neutralization and ionization with 150 mM NaCl. Each system was composed of >90,000 atoms in a 115×115×89 Å3 box. Independently, we prepared the DDM120 and DDM150 micelles without GlpG as controls.

All inter- and intramolecular interactions were enumerated under the CHARMM36 force field7. The nonbonding van der Waals and short-range electrostatic interactions were treated with a typical cutoff distance of 12 Å, while the long-range electrostatic contributions were evaluated with the particle-mesh Ewald method8. All simulations were carried out using GROMACS software9 parallelized in the GPU-accelerated IBM Power8 machine. Each system was first subject to 10,000 steps of conjugate gradient energy minimization to remove any unfavorable atomic crash with lipids and GlpG, which were restrained to preserve their conformation and relative positions. The systems were pre-equilibrated along 6 scheduled steps as gradually removing the external restraints until no constraints. The simulations were proceeded with a 2-fs timestep in the semi-isotropic isobaric and isothermal (NPT) ensemble of 1 atm and 310 K, where the pressure and temperature were controlled by Parrinello-Rahman barostat10 and Nosé-Hoover thermostat11,12, respectively. The pressure was decoupled between the xy-plane and the z-axis, so the membrane normal fluctuated independently from the isotropic lateral motions (xy-plane). >2 μs-long simulation trajectories were generated for each system.

Assessment of the equilibration of protein and amphiphiles.

The equilibration of systems was evaluated for: (1) the conformation of protein, (2) the conformation of amphiphiles, and (3) the solvation dynamics of interfacial amphiphiles on protein. The equilibration of GlpG conformation was examined by calculating the RMSD’s of all heavy atoms referenced to the crystal structure13. Regarding the equilibration of amphiphile conformation, we assessed the time-autocorrelated RMSD (τ) as a function of the time lag τ by averaging over all lipid or detergent molecules in the bulk (i.e., not in contact with the protein) as follows: Eq.7 RMSD(τ)=1NL∑i=1NL<RMSDi(t,t+τ)>t

, where NL is the number of the lipid or detergent molecules, and <RMSDi (t, t + τ)>t is the heavy-atom RMSD between the i-th amphiphile’s conformations at the time t and t + τ, averaged over all available time t’s. The bulk lipid molecules were selected from the ones not in contact with the protein over the analysis period while the bulk detergent molecules from the control micelles without GlpG. Prior to the RMSD calculation, the amphiphiles under comparison at each t and t + τ were structurally aligned with each other by transrotating the heavy-atom conformations.

For the solvation dynamics of interfacial amphiphiles on protein, we analyzed how fast the lipid (or detergent) would dissociate from the protein by measuring the residence time (τR) from the autocorrelation function on time for the amphiphile heavy atom in contact (distance cutoff of 5 Å) with GlpG as follows:13–15 Eq.8 c(τ)=1Nc∑i=1Nc<ci(t,t+τ)>t

, where Nc is the number of contact events, and a single contact event is defined as a consecutive contact of an amphiphile with no non-contacting time gap longer than the amphiphile relaxation time measured above (i.e., 20 ns). The autocorrelation function at the time τ of the i-th contact event, <ci (t, t + τ)>t is defined by <qi(t)•qi(t + τ)/qi2(t)>t the normalized product of heavy atom contact numbers of an amphiphile in the i-th contact event, qi(t) and qi(t + τ) at two time moments (t and t+ τ), averaged over the time t.

Residence time of amphiphiles.

The τR was defined as the time when the amplitude of contact autocorrelation reached 1/e of the initial value without any assumption on the dissociation mechanisms based on Eq.8. Dissociation dynamics of an amphiphile from GlpG was assessed as a whole or parts (i.e., head group and tail). The tail of DMPC was defined as the atoms in two aliphatic chains (C22–C214 and C32–C314) while that of DDM included all carbons in the dodecyl chain (C1–C12), thus leaving the rest as the head group. As a control, we assessed self-dissociation of an amphiphile from another amphiphile, where the pairs were selected from the molecules in contact with each other in the bulk. DMPC molecules were selected from ones of no explicit protein contact in the bilayer, and DDM molecules from the micelles without GlpG. In each system, the residence time for self-dissociation was served as a reference for evaluating the relative preference toward protein.

Solvation free energy (ΔGoSolv) of amphiphiles on GlpG.

ΔGoSolv was derived from the following equations: Eq.9 P+L⋅L⇄koff,L⋅Lkoff,P⋅LP⋅L+L

Eq.10 ΔGSolvo=−RTlnKSolv=−RTln(koff,L⋅Lkoff,P⋅L)=−RT(τR,P⋅LτR,L⋅L)

P and L•L denote the protein and the amphiphile–amphiphile complex outside of the first solvation shell (i.e., in the bulk), respectively. P•L denotes the protein–amphiphile complex. koff,P•L and koff,L•L are the dissociation rates of an amphiphile from the protein–amphiphile and amphiphile–amphiphile complexes (i.e., the inverse of τR), respectively. A negative ΔGoSolv indicates that an amphiphile favorably makes a contact with protein (τR,P•L) overcoming the likewise favorable contacts between the amphiphiles in the bulk (τR,L•L).

Supplementary Material

Supplement 1

Acknowledgement.

This work was supported by the National Institute of Health grants R01GM118685 and R35GM144146 to H.H., and R35GM147522 to K.H.K. K.H.K. thanks for support from the Michigan State University RTSF Cryo-EM facility.

Fig. 1│ Establishment of the reversible folding of GlpG in bicelles.

(a) Solvation dynamics of amphiphiles on GlpG in the DMPC bilayer and DDM micelles studied by all-atom MD simulation. The illustration shows the tracking of the 40 lipid or detergent molecules (green) on GlpG (orange) in the first solvation shell at the simulation time, t = 0.2 μs. DDM120: the aggregation number (NA) = 120; DDM150: NA = 150.

(b) Steric trapping scheme. GlpG is labeled with the biotin tags at two specific residues, which are close in space in the native state but distant in the amino acid sequence. A first monovalent streptavidin (mSA, 52 kDa) binds unhindered to either biotin tag (ΔGoBind). Due to the steric hindrance with the first bound mSA. A second mSA binds only when the tertiary contacts between the biotinylated sites are denatured. The coupling between GlpG denaturation and mSA binding attenuates the apparent second binding (ΔGoBind + ΔGoN-D). ΔGoN-D is determined by fitting the second binding phase to Eqs. 3–5 (Online Methods).

(c) Crystal structure of GlpG (PDB: 3B45)37 annotated with the secondary structural elements, N (cyan)- and C (orange)-subdomains, and the positions of biotin pairs.

(d) Reversibility of the folding and mSA binding, and their coupling in bicelles. Native (“from folded”) or sterically denatured (“from denatured”) GlpG in micelles before transfer to bicelles at an increasing concentration of mSA-E51S (Kd,biotin = 52 ± 19 pM) or mSA-S27A (Kd,biotin = 3.2 ± 0.8 nM). The signals for mSA binding and denaturation were normalized to the amplitude of the second binding phase and the change in activity of GlpG, respectively. Dashed lines: unhindered mSA binding without coupling to denaturation.

(e) The effect of the bilayer environment on the stabilities of N- and C-subdomains of GlpG.

(d–e) Errors denote ± SEM (n = 3).

Fig. 2│ GlpG stability measured at N- and C-subdomains in micelles and bilayers.

(a) (Left) Comparison of the mutation-induced stability changes (ΔΔGoN-D,WT-Mut’s) in micelles vs bicelles. The plot includes all ΔΔGoN-D,WT-Mut values measured at N- and C-subdomains. (Right) The ΔΔGoN-D,WT-Mut values in micelles vs bicelles were separately plotted depending on the location of the biotin pair at which the stability was measured.

(b) Comparison of the ΔΔGoN-D,WT-Mut values in micelles vs bicelles depending on the location of the biotin pair and the degree of burial of the residues targeted for mutation (fASA: the fraction of solvent-accessible residue surface area).

(a–b) Errors denote ± SD from fitting. The statistical significance of the difference in correlation slope was evaluated using the pairwise Chow’s test (NS: p>0.05; *: p<0.05; **: p<0.005)

Fig. 3│ Cooperativity maps of GlpG in micelles and bicelles.

The cooperativity profile of each residue is color-coded: “cooperative” (green, |ΔΔΔG|≤RT = 0.6 kcal/mol), “moderately localized in N-subdomain” (tin, 2RT≥ΔΔΔG>RT), “localized in N-subdomain” (blue, ΔΔΔG>2RT), “moderately localized in C-subdomain” (orange, −RT>ΔΔΔG≥−2RT), and “localized in C-subdomain” (red, −2RT>ΔΔΔG).

Fig. 4│ All-atom MD simulation of GlpG WT in the lipid bilayer and micelles.

(a) The RMSD’s of the backbone heavy atoms (top). The RMSD (τ) of the lipid or detergent conformation in the bulk amphiphilic phase measured as a function of the time lag τ (middle). The residue RMSF’s of GlpG WT in each environment (bottom).

(b) The contact autocorrelation on time for measuring the residence time of any heavy atoms in the lipid or detergent molecules on GlpG or on themselves (left). The contact autocorrelation was also separately monitored for the heavy atoms in the headgroup (middle) or in the tail (right) of the amphiphiles. The solid line is the fitted result of each data to a triple-exponential decay function (Extended Data Table 3), implying heterogeneous modes of dissociation. Errors denote ± SD from fitting.

The authors declare no competing financial interests.
==== Refs
REFERENCES

1 Oberai A. , Joh N. H. , Pettit F. K. & Bowie J. U. Structural imperatives impose diverse evolutionary constraints on helical membrane proteins. Proc Natl Acad Sci U S A 106 , 17747–17750 (2009). 10.1073/pnas.0906390106 19815527
2 Kopito R. R. & Ron D. Conformational disease. Nat Cell Biol 2 , E207–209 (2000). 10.1038/35041139 11056553
3 Levy Y. & Onuchic J. N. Water mediation in protein folding and molecular recognition. Annu Rev Biophys Biomol Struct 35 , 389–415 (2006). 10.1146/annurev.biophys.35.040405.102134 16689642
4 Bellissent-Funel M. C. Water Determines the Structure and Dynamics of Proteins. Chem Rev 116 , 7673–7697 (2016). 10.1021/acs.chemrev.5b00664 27186992
5 Dill K. A. Dominant forces in protein folding. Biochemistry 29 , 7133–7155 (1990). 10.1021/bi00483a001 2207096
6 Heyden M. & Tobias D. J. Spatial dependence of protein-water collective hydrogen-bond dynamics. Phys Rev Lett 111 , 218101 (2013). 10.1103/PhysRevLett.111.218101 24313531
7 Chan H. S. , Zhang Z. , Wallin S. & Liu Z. Cooperativity, local-nonlocal coupling, and nonnative interactions: principles of protein folding from coarse-grained models. Annu Rev Phys Chem 62 , 301–326 (2011). 10.1146/annurev-physchem-032210-103405 21453060
8 Jiang L. Real-time monitoring of hydrophobic aggregation reveals a critical role of cooperativity in hydrophobic effect. Nat Commun 8 , 15639 (2017). 10.1038/ncomms15639 28561067
9 Salvay A. G. , Grigera J. R. & Colombo M. F. The role of hydration on the mechanism of allosteric regulation: in situ measurements of the oxygen-linked kinetics of water binding to hemoglobin. Biophys J 84 , 564–570 (2003). 10.1016/S0006-3495(03)74876-0 12524309
10 Popot J. L. & Engelman D. M. Membrane protein folding and oligomerization: the two-stage model. Biochemistry 29 , 4031–4037 (1990).1694455
11 Hessa T. Recognition of transmembrane helices by the endoplasmic reticulum translocon. Nature 433 , 377–381 (2005). 10.1038/nature03216 15674282
12 Cao Z. , Hutchison J. M. , Sanders C. R. & Bowie J. U. Backbone hydrogen bond strengths can vary widely in transmembrane helices. J Am Chem Soc 139 , 10742–10749 (2017). 10.1021/jacs.7b04819 28692798
13 Joh N. H. Modest stabilization by most hydrogen-bonded side-chain interactions in membrane proteins. Nature 453 , 1266–1270 (2008). 10.1038/nature06977 18500332
14 Joh N. H. , Oberai A. , Yang D. , Whitelegge J. P. & Bowie J. U. Similar energetic contributions of packing in the core of membrane and water-soluble proteins. J Am Chem Soc 131 , 10846–10847 (2009). 10.1021/ja904711k 19603754
15 Mravic M. Packing of apolar side chains enables accurate design of highly stable membrane proteins. Science 363 , 1418–1423 (2019). 10.1126/science.aav7541 30923216
16 Fleming K. G. & Engelman D. M. Specificity in transmembrane helix-helix interactions can define a hierarchy of stability for sequence variants. Proc Natl Acad Sci U S A 98 , 14340–14344 (2001). 10.1073/pnas.251367498 11724930
17 Corin K. & Bowie J. U. How physical forces drive the process of helical membrane protein folding. EMBO Rep 23 , e53025 (2022). 10.15252/embr.202153025 35133709
18 O’Brien E. S. Membrane Proteins Have Distinct Fast Internal Motion and Residual Conformational Entropy. Angew Chem Int Ed Engl 59 , 11108–11114 (2020). 10.1002/anie.202003527 32277554
19 Hong H. & Bowie J. U. Dramatic destabilization of transmembrane helix interactions by features of natural membrane environments. J Am Chem Soc 133 , 11389–11398 (2011). 10.1021/ja204524c 21682279
20 Andersen O. S. & Koeppe R. E. 2nd ,. Bilayer thickness and membrane protein function: an energetic perspective. Annu Rev Biophys Biomol Struct 36 , 107–130 (2007). 10.1146/annurev.biophys.36.040306.132643 17263662
21 Chadda R. Membrane transporter dimerization driven by differential lipid solvation energetics of dissociated and associated states. Elife 10 (2021). 10.7554/eLife.63288
22 Lague P. , Zuckermann M. J. & Roux B. Lipid-mediated interactions between intrinsic membrane proteins: a theoretical study based on integral equations. Biophys J 79 , 2867–2879 (2000). 10.1016/S0006-3495(00)76524-6 11106595
23 Jiang Y. Membrane-mediated protein interactions drive membrane protein organization. Nat Commun 13 , 7373 (2022). 10.1038/s41467-022-35202-8 36450733
24 Yano Y. , Kondo K. , Kitani R. , Yamamoto A. & Matsuzaki K. Cholesterol-induced lipophobic interaction between transmembrane helices using ensemble and single-molecule fluorescence resonance energy transfer. Biochemistry 54 , 1371–1379 (2015). 10.1021/bi501528e 25629582
25 Hilser V. J. , Dowdy D. , Oas T. G. & Freire E. The structural distribution of cooperative interactions in proteins: analysis of the native state ensemble. Proc Natl Acad Sci U S A 95 , 9903–9908 (1998). 10.1073/pnas.95.17.9903 9707573
26 Niebling S. Biophysical Screening Pipeline for Cryo-EM Grid Preparation of Membrane Proteins. Front Mol Biosci 9 , 882288 (2022). 10.3389/fmolb.2022.882288 35813810
27 Harris N. J. Methods to study folding of alpha-helical membrane proteins in lipids. Open Biol 12 , 220054 (2022). 10.1098/rsob.220054 35855589
28 Bowie J. U. Solving the membrane protein folding problem. Nature 438 , 581–589 (2005). 10.1038/nature04395 16319877
29 Chang Y. C. & Bowie J. U. Measuring membrane protein stability under native conditions. Proc Natl Acad Sci U S A 111 , 219–224 (2014). 10.1073/pnas.1318576111 24367094
30 Guo R. Steric trapping reveals a cooperativity network in the intramembrane protease GlpG. Nat Chem Biol 12 , 353–360 (2016). 10.1038/nchembio.2048 26999782
31 Hong H. , Blois T. M. , Cao Z. & Bowie J. U. Method to measure strong protein-protein interactions in lipid bilayers using a steric trap. Proc Natl Acad Sci U S A 107 , 19802–19807 (2010). 10.1073/pnas.1010348107 21041662
32 Gaffney K. A. Lipid bilayer induces contraction of the denatured state ensemble of a helical-bundle membrane protein. Proc Natl Acad Sci U S A 119 (2022). 10.1073/pnas.2109169119
33 Min D. , Jefferson R. E. , Bowie J. U. & Yoon T. Y. Mapping the energy landscape for second-stage folding of a single membrane protein. Nat Chem Biol 11 , 981–987 (2015). 10.1038/nchembio.1939 26479439
34 Lu W. , Schafer N. P. & Wolynes P. G. Energy landscape underlying spontaneous insertion and folding of an alpha-helical transmembrane protein into a bilayer. Nat Commun 9 , 4949 (2018). 10.1038/s41467-018-07320-9 30470737
35 Choi H. K. Watching helical membrane proteins fold reveals a common N-to-C-terminal folding pathway. Science 366 , 1150–1156 (2019). 10.1126/science.aaw8208 31780561
36 Kim S. Transmembrane glycine zippers: physiological and pathological roles in membrane proteins. Proc Natl Acad Sci U S A 102 , 14278–14283 (2005). 10.1073/pnas.0501234102 16179394
37 Wang Y. , Zhang Y. & Ha Y. Crystal structure of a rhomboid family intramembrane protease. Nature 444 , 179–180 (2006). 10.1038/nature05255 17051161
38 Baker R. P. & Urban S. Architectural and thermodynamic principles underlying intramembrane protease function. Nat Chem Biol 8 , 759–768 (2012). 10.1038/nchembio.1021 22797666
39 Park H. , Yoon J. & Seok C. Strength of Calpha-H...O=C hydrogen bonds in transmembrane proteins. J Phys Chem B 112 , 1041–1048 (2008). 10.1021/jp077285n 18154287
40 Zhou H. X. & Cross T. A. Influences of membrane mimetic environments on membrane protein structures. Annu Rev Biophys 42 , 361–392 (2013). 10.1146/annurev-biophys-083012-130326 23451886
41 Zhou Y. , Moin S. M. , Urban S. & Zhang Y. An internal water-retention site in the rhomboid intramembrane protease GlpG ensures catalytic efficiency. Structure 20 , 1255–1263 (2012). 10.1016/j.str.2012.04.022 22705210
42 Cho S. , Baker R. P. , Ji M. & Urban S. Ten catalytic snapshots of rhomboid intramembrane proteolysis from gate opening to peptide release. Nat Struct Mol Biol 26 , 910–918 (2019). 10.1038/s41594-019-0296-9 31570873
43 Arutyunova E. An internally quenched peptide as a new model substrate for rhomboid intramembrane proteases. Biol Chem 399 , 1389–1397 (2018). 10.1515/hsz-2018-0255 30044760
44 Lipfert J. , Columbus L. , Chu V. B. , Lesley S. A. & Doniach S. Size and shape of detergent micelles determined by small-angle X-ray scattering. J Phys Chem B 111 , 12427–12438 (2007). 10.1021/jp073016l 17924686
45 Brunne R. M. , Liepinsh E. , Otting G. , Wuthrich K. & van Gunsteren W. F. Hydration of proteins. A comparison of experimental residence times of water molecules solvating the bovine pancreatic trypsin inhibitor with theoretical model calculations. J Mol Biol 231 , 1040–1048 (1993). 10.1006/jmbi.1993.1350 7685828
46 Marinko J. T. Folding and Misfolding of Human Membrane Proteins in Health and Disease: From Single Molecules to Cellular Proteostasis. Chem Rev 119 , 5537–5606 (2019). 10.1021/acs.chemrev.8b00532 30608666
47 Drew D. & Boudker O. Shared Molecular Mechanisms of Membrane Transporters. Annu Rev Biochem 85 , 543–572 (2016). 10.1146/annurev-biochem-060815-014520 27023848
48 Matthews E. E. Thrombopoietin receptor activation: transmembrane helix dimerization, rotation, and allosteric modulation. FASEB J 25 , 2234–2244 (2011). 10.1096/fj.10-178673 21402716
49 Latorraca N. R. , Venkatakrishnan A. J. & Dror R. O. GPCR Dynamics: Structures in Motion. Chem Rev 117 , 139–155 (2017). 10.1021/acs.chemrev.6b00177 27622975
50 Yue P. , Li Z. & Moult J. Loss of protein structure stability as a major causative factor in monogenic disease. J Mol Biol 353 , 459–473 (2005). 10.1016/j.jmb.2005.08.020 16169011
51 Molinski S. V. Comprehensive mapping of cystic fibrosis mutations to CFTR protein identifies mutation clusters and molecular docking predicts corrector binding site. Proteins 86 , 833–843 (2018). 10.1002/prot.25496 29569753
1 Guo R. Steric trapping reveals a cooperativity network in the intramembrane protease GlpG. Nat Chem Biol 12 , 353–360 (2016). 10.1038/nchembio.2048 26999782
2 Howarth M. Monovalent, reduced-size quantum dots for imaging receptors on living cells. Nat Methods 5 , 397–399 (2008). 10.1038/nmeth.1206 18425138
3 Hong H. , Chang Y. C. & Bowie J. U. Measuring transmembrane helix interaction strengths in lipid bilayers using steric trapping. Methods Mol Biol 1063 , 37–56 (2013). 10.1007/978-1-62703-583-5_3 23975771
4 Akiyama Y. & Maegawa S. Sequence features of substrates required for cleavage by GlpG, an Escherichia coli rhomboid protease. Mol Microbiol 64 , 1028–1037 (2007). 10.1111/j.1365-2958.2007.05715.x 17501925
5 Jo S. , Kim T. , Iyer V. G. & Im W. CHARMM-GUI: a web-based graphical user interface for CHARMM. J Comput Chem 29 , 1859–1865 (2008). 10.1002/jcc.20945 18351591
6 Jorgensen W. L. , Chandrasekhar J. , Madura J. D. , Impey R. W. & Klein M. L. COMPARISON OF SIMPLE POTENTIAL FUNCTIONS FOR SIMULATING LIQUID WATER. Journal of Chemical Physics 79 , 926–935 (1983). 10.1063/1.445869
7 Best R. B. Optimization of the additive CHARMM all-atom protein force field targeting improved sampling of the backbone phi, psi and side-chain chi(1) and chi(2) dihedral angles. J Chem Theory Comput 8 , 3257–3273 (2012). 10.1021/ct300400x 23341755
8 Darden T. , York D. & Pedersen L. PARTICLE MESH EWALD - AN N.LOG(N) METHOD FOR EWALD SUMS IN LARGE SYSTEMS. Journal of Chemical Physics 98 , 10089–10092 (1993). 10.1063/1.464397
9 Pall S. Heterogeneous parallelization and acceleration of molecular dynamics simulations in GROMACS. J Chem Phys 153 , 134110 (2020). 10.1063/5.0018516 33032406
10 Parrinello M. & Rahman A. Polymorphic transitions in single crystals: A new molecular dynamics method. Journal of Applied Physics 52 , 7182–7190 (1981). 10.1063/1.328693
11 Nosé S. A molecular dynamics method for simulations in the canonical ensemble. Molecular Physics 52 , 255–268 (1984). 10.1080/00268978400101201
12 Hoover W. G. Canonical dynamics: Equilibrium phase-space distributions. Phys Rev A Gen Phys 31 , 1695–1697 (1985). 10.1103/physreva.31.1695 9895674
13 Gaffney K. A. Lipid bilayer induces contraction of the denatured state ensemble of a helical-bundle membrane protein. Proc Natl Acad Sci U S A 119 (2022). 10.1073/pnas.2109169119
14 Brunne R. M. , Liepinsh E. , Otting G. , Wuthrich K. & van Gunsteren W. F. Hydration of proteins. A comparison of experimental residence times of water molecules solvating the bovine pancreatic trypsin inhibitor with theoretical model calculations. J Mol Biol 231 , 1040–1048 (1993). 10.1006/jmbi.1993.1350 7685828
15 Song W. PyLipID: A Python Package for Analysis of Protein-Lipid Interactions from Molecular Dynamics Simulations. J Chem Theory Comput 18 , 1188–1201 (2022). 10.1021/acs.jctc.1c00708 35020380