Improved pendant-bearing glucose-neopentyl glycols (P-GNGs) for membrane protein stability

Abstract

Membrane proteins are biologically and pharmaceutically significant, and determining their 3D structures requires a membrane-mimetic system to maintain protein stability. Detergent micelles are widely used as membrane mimetics; however, their dynamic structures often lead to the denaturation and aggregation of encapsulated membrane proteins. To address the limitations of classical detergents in stabilizing membrane proteins, we previously reported a class of glucose-neopentyl glycols (GNGs) and their pendant-bearing versions (P-GNGs), several of which proved more effective than DDM in stabilizing membrane proteins. In this study, we synthesized additional GNG derivatives by varying the lengths of the pendant (P-GNGs), and by introducing hemi-fluorinated pendants to the GNG scaffold (fluorinated pendant-bearing GNGs, or FP-GNGs). The synthetic flexibility of the GNG chemical architecture allowed us to create a diverse range of derivatives, essential for effective optimizing of detergent properties effectively. When tested with two model membrane proteins (a transporter and a G-protein coupled receptor (GPCR)), most of the new (F)P-GNGs demonstrated superior stabilization of these membrane proteins compared to DDM, original GNG (OGNG)), and a previously developed P-GNG (i.e., GNG-3,14). Notably, several P-GNGs synthesized in this study were as effective as or even better than lauryl maltose neopentyl glycol (LMNG) in stabilizing a human GPCR, beta2 adrenergic receptor (β2AR). Enhanced protein stability was particularly observed for the P-GNGs with a butyl (C4) or pentyl (C5) pendant, indicating that these pendant sizes are optimal for membrane protein stability. The volumes of these pendants appear to minimize the empty spaces in the micelle interiors, thereby enhancing detergent-detergent interactions in micelles complexed with the membrane proteins. Additionally, we identified one FP-GNG that was more efficient at extracting the transporter and more effective at stabilizing the GPCR than DDM. Thus, the current study demonstrates that both chain length and number of fluorine atoms in the pendants of the P-GNGs were crucial determinants for membrane protein stability. We not only developed a few (F)P-GNGs that are significantly more effective than maltoside detergents (LMNG/DDM) for protein extraction and stability, but we also provided an effective strategy for detergent design through the utilization of partially fluorinated pendants of varying length.

Membrane proteins residing within biological membranes perform critical functions, including material transport, signal transduction and cell-cell communications.¹ In humans, malfunctions of these proteins are implicated in various diseases such as asthma, cancer, Alzheimer’s, and Parkinson’s, underscoring their additional importance in drug development.² Over 50% of current drugs target this class of bio-macromolecules, making detailed understanding of their high-resolution 3D structures essential for structure-based drug design.^3,4 Despite their significant roles in biological research and drug development, progress in elucidating membrane protein structures remains slow, even with recent technical advances. As of October 30, 2024, only around 2,000 out of a total 210,000 known protein structures represent unique membrane proteins.⁵ The challenges in determining membrane protein structures stem from their intrinsic amphiphilic nature. Unlike soluble proteins, membrane proteins have a hydrophobic region that is embedded in membranes, flanked by two hydrophilic regions that project into aqueous environment on either side of the membrane. When the hydrophobic regions are exposed to the aqueous environment, membrane proteins tend to denature and/or aggregate, complicating structural characterization.^6–8 Therefore, in-vitro studies of membrane proteins require membrane mimetics capable of encapsulating the hydrophobic protein surface, thus maintaining membrane protein stability in aqueous conditions.^9,10 The first crystal structure of membrane protein was obtained for Rhodopseudomonas viridis photosynthetic assembly using the detergent lauryldimethylamine-N-oxide (LDAO) in 1985.¹¹ Since then, micelles formed by conventional detergents, such as n-octyl-β-D-glucoside (OG) and n-dodecyl-β-D-maltoside (DDM)—small amphiphilic molecules—have been widely used for membrane protein structure determination.^12,13 Similar to phospholipids in cell membranes, these conventional detergents possess both hydrophilic and hydrophobic moieties, enabling them to form self-assemblies in water. However, due to their conical shapes, they form globular micelles rather than planar bilayers.^14,15 Detergent micelles can associate with the hydrophobic surfaces of membrane proteins in aqueous environments, resulting in protein-detergent complexes (PDCs) that serve as starting materials for in-vitro studies of membrane protein structures and functions.¹²

Despite the availability of more than 100 conventional detergents with a single head and tail group, they often fail to stabilize membrane proteins, particularly G protein-coupled receptors (GPCRs) and large membrane protein complexes.¹⁶ The highly dynamic nature of micelles formed by conventional detergents compared to lipid bilayers leads to the instability of encapsulated membrane proteins.¹⁷ Over the past two decades, various membrane mimetic systems have been developed to address the limitations of conventional detergents.^18,19 Notable examples include bicelles,^20,21 nanodiscs (NDs) derived from membrane scaffold protein (MSP),^22,23 styrene-maleic acid (SMA) copolymer,^24,25 or saposin A,²⁶ and peptide-based detergents such as lipopeptide detergents (LPDs)²⁷ and beta-peptides (BPs).²⁸ Those alternatives that closely mimic lipid bilayers have proven highly effective in maintaining the native structures of membrane proteins. However, they are typically inefficient at extracting proteins from the membranes and often show poor compatibility with protein crystallization.²⁹ In contrast, detergent micelles have the advantages of being applicable across all stages of in-vitro membrane proteins studies from protein extraction and solubilization to purification and structure determination.³⁰ Additionally, detergent micelles are compatible with various methods for protein structure determination, including X-ray crystallography, cryo-electron microscopy (cryo-EM), and nuclear magnetic resonance (NMR) spectroscopy. Therefore, developing new detergents with enhanced efficacy for protein stabilization is crucial to advancing membrane protein research.³¹ Significant efforts have been made over the past two decades to develop novel detergents, leading to the creation of various amphiphilic molecules, including facial amphiphiles (FAs),^32,33 oligoglycerol detergents (OGDs),³⁴ neopentyl glycol-based amphiphiles [e.g., maltose neopentyl glycols (MNGs),^35,36 and glucose neopentyl glycols (GNGs)^37,38], rigid hydrophobic group-bearing amphiphiles [e.g., chobimalt³⁹ and glyco-diosgenin (GDN)⁴⁰], tandem amphiphiles [tandem malonate-derived glucoside (TMG),^42,43 tandem facial amphiphiles (TFAs),⁴⁴], triazine-based detergents [1,3,5-triazine-based di-maltosides (TEMs)⁴⁶ and melamine-based glucosides (MGs)⁴⁷], and hybrid detergents⁴⁸. These novel detergents have shown superiority to the gold standard DDM in terms of membrane protein stability. In particular, LMNG has been used in over 300 structural studies of membrane proteins, including more than 150 GPCRs, since its introduction in 2010, highlighting the significant impact of novel detergents in membrane protein structure determination.⁴⁹ Fluorinated surfactants (FSs) are an additionally interesting detergents that can enhance membrane protein stabilization.⁵⁰ However, this class of detergents has generally been shown to be ineffective at extracting membrane protein from membranes. This limitation has led to the development of partially fluorinated or hemi-fluorinated surfactants (HFSs) that feature both a hydrocarbon chain and a fluorinated chain in the hydrophobic region.^51–53 Due to the presence of a hydrocarbon chain, these surfactants exhibit stronger interactions with membrane protein surfaces compared to FSs, resulting in improved membrane protein extraction and solubilization efficiency.⁵⁴

The first class of GNG amphiphiles we reported previously contains a branched diglucoside head group and two identical alkyl chains.³⁷ Although these GNGs, glucoside versions of LMNG, were generally not as effective as DDM in stabilizing membrane proteins, one representative (OGNG) has facilitated structure determinations of 10 membrane proteins since its introduction in 2012.^55–58 We recently reported a second generation of GNGs featuring two different alkyl chains — one long main chain and a small pendant (methyl, ethyl, or propyl).⁵⁹ These pendant-bearing GNGs (P-GNGs) demonstrated superior performance compared to DDM in stabilizing various membrane proteins, with the best results observed for propyl pendant-bearing GNGs (e.g., GNG-3,14). In the current study, we prepared additional P-GNG derivatives by introducing longer pendants (butyl, pentyl, or hexyl) into the detergent scaffold. This design was based on our hypothesis that increasing the pendant length of P-GNGs would enhance the hydrophobic density in the micelle interiors, resulting in micelle stability and protein stability. We also synthesized a new set of GNGs by incorporating a partially fluorinated alkyl chain as a pendant into the P-GNG scaffold, designated fluorinated pendant-bearing GNGs (FP-GNGs). We conceived that the presence of a partially fluorinated pendant in this GNG scaffold could mimic the favorable features of HFSs. When tested with two model membrane proteins [a transporter and a GPCR], several (F)P-GNGs outperformed DDM, the original OGNG, and previously developed GNG-3,14 in stabilizing these proteins. Notably, some butyl or pentyl pendant-bearing GNGs prepared in this study were even superior or comparable to LMNG in stabilizing the GPCR. The enhanced detergent-detergent interactions achieved through pendant elongation likely contribute to the significantly improved efficacy of these P-GNGs for protein stabilization.

Results

Rationale for new detergent design

Original GNG (OGNG) tends to form smaller PDCs compared to DDM, likely due to the presence of a bulky, branched diglucoside head group.^37,60 This property endows OGNG with high compatibility for membrane protein crystallization, as evidenced by the fact that all 10 membrane protein structures determined using this detergent were obtained via the in surfo crystallization method.^49,55,56 However, due to its limited efficacy in stabilizing membrane proteins, OGNG has primarily been useful in the structural studies of relatively robust membrane proteins, such as channels and transporters, rather than aggregation-prone GPCRs. The limited protein stabilization efficacy of OGNG is believed to originate, at least in part, from its shorter hydrophobic length compared to DDM and LMNG (C8 vs. C12), the two most widely used detergents for membrane protein structural studies. For this reason, we strategically lengthened the hydrophobic region of GNGs in our previous study by introducing two different alkyl chains: one long main chain (C11/C12/C13/C14) and one short pendant (methyl [C1]/ethyl [C2]/propyl [C3]).⁵⁹ When tested with membrane proteins, these P-GNGs, which have sufficient hydrophobic lengths, generally outperformed DDM in terms of protein stability. The efficacy of the P-GNGs for protein stabilization tended to improve with increasing pendant length, with the best performance detected for the C3 pendant-bearing GNGs (GNG-3,13 and GNG-3,14). We hypothesized that the propyl pendant effectively fills empty spaces in the micelle interiors, thereby increasing the alkyl chain density of micelle interiors, resulting in enhancement in both micelle stability and protein stability.⁶¹ However, it is not possible to determine a priori which chain — propyl (C3) or a longer alkyl chain — will provide the best fit in these micellar spaces. Thus, in the current study, we prepared new P-GNGs having longer alkyl pendants (butyl [C4], pentyl [C5] and hexyl [C6]) than the propyl chain (C3). The main chains of the new P-GNGs varied from C12 to C15 based on our previous findings, which indicates that membrane protein stability increased with longer main chain lengths from C11 to C14. By varying both the main (C12 to C15) and pendant chain lengths (C4 to C6), we synthesized nearly 20 new P-GNGs. The structures of these new detergents, along with those of previously developed P-GNGs, are depicted in Figure S1, according to their main and pendant chain lengths. The newly prepared P-GNGs were designated based on lengths of their main and pendant chains, as exemplified by GNG-4,15, which has a C4 (butyl) pendant and a C15 main chain. Additionally, we prepared a second set of P-GNGs with partially fluorinated pendants (FP-GNGs) to investigate the impact of this fluorine-bearing pendant on membrane protein stability within the P-GNG architecture. This set of detergents shares structural similarities with hemi-fluorinated surfactants (HFSs),⁵³ but the FP-GNGs are among a few examples that incorporate a partially fluorinated chain as a pendant, rather than as a main chain segment.^51,62

HFSs are known to be effective at maintaining the subtle integrity of membrane protein complexes due to the reduced tendency of hemi-fluorinated chains to dissolve lipophilic cofactors embedded in protein interiors or lipids specifically bound onto membrane protein surfaces, compared to their hydrocarbon counterparts.^63,64 We expect that our FP-GNG design may mimic such favourable effects of the HFSs on protein stability. Additionally, a fluorinated chain tends to form strong interactions with other fluorinated chains in the hydrophobic environments. Therefore, positioning fluorinated chains at the hydrophilic-hydrophobic interfaces could enhance detergent-detergent interactions, which is likely to be beneficial for protein stability. The three FP-GNGs prepared in this study contain the same number of total carbons (C21) in the hydrophobic region, allowing us to derive instructive detergent design principles from their comparative evaluation with membrane proteins. Of note, these detergents were designated to indicate the lengths of both the fluorinated and hydrogenated components within the pendant chain, as well as the main chain length. For instance, GNG-F1,2,15 contains a trifluoromethyl ethyl (CF₃-CH₂-CH₂-) pendant and a C15 main chain.

The hydrophilic-hydrophobic balance (HLB) is one main property influencing membrane protein stability.^65,66 It has been reported that detergents with a certain range of HLB (11 to 14) are favourable for membrane protein solubilisation, stabilization and crystallization, as exemplified by those of OG (12.3), DDM (13.4) and LMNG (13.6).^66,67 When calculated according to Griffin’s method,⁶⁸ most of the new P-GNGs yielded HLB values ranging from 10.5 to 11.5, lower than those of conventional detergents (OG and DDM) (Table S1). For the FP-GNGs, we calculated their HLB values using Davies’ method, as Griffin’s method is inapplicable to fluorinated detergents.⁶⁹ As summarized in Table S2, the FP-GNGs have lower HLB values (9.2 or 9.6) compared to the P-GNGs, partly due to the differences in the calculation methods. The low HLB values indicate that the new (F)P-GNGs are substantially more hydrophobic than DDM and LMNG. As we will discuss shortly, a couple of P-GNGs such as GNG-5,14, GNG-4,15 and GNG-5,15 have HLB values of 10.5/10.7 yet were comparable to or more effective than DDM in stabilizing membrane proteins (vide infra). Therefore, the established range of optimal HLB values (11 to 14) for membrane protein studies may need to be revised to include these HLB values. The high hydrophobicity of the new P-GNGs compared to DDM and LMNG is likely favourably associated with membrane protein stability through the promotion of effective hydrophobic interactions between detergent molecules in PDCs.

Detergent synthesis and physical characterization

The new P-GNGs were prepared with slight modifications to a previously reported protocol.⁵⁹ The FP-GNGs were synthesized using partially fluorinated alkyl chains instead of hydrocarbon chains. All P-GNGs in CD₃OD exhibited a single doublet peak at 4.31/4.32 ppm in their individual NMR spectra, with a vicinal coupling constant of J = 8.0 Hz (Figures S2), confirming β-stereochemistry of the two glycosidic bonds present in these GNGs. We then measured the water-solubility of the new detergents at room temperature, as summarized in Table 1. The (F)P-GNGs with relatively short pendants (C1 to C4) exhibited more than 5.0 wt% water solubility. However, as the pendant length increased to C5/C6, the water-solubility of the detergents tended to decrease to less than 5.0 wt%. The three FP-GNGs demonstrated good water-solubility (5.0/10 wt%). We did not evaluate the poorly water-soluble P-GNGs (less than 5.0 wt%) for membrane protein stability as these GNGs are unlikely to be useful for membrane protein studies.

Table 1.

Molecular weights (MWs), critical micelle concentrations (CMCs), and water solubility of new (F)P-GNGs along with control detergent (DDM, OGNG, and GNG-3,14), and hydrodynamic radii (R_h) (mean ± S.D., n = 5) of their micelles.

Detergent	M.W.a	CMC (mM)	CMC (wt%)	Rh (nm)b	Solubility (wt%)
GNG-0,12	568.7	~0.20	~0.011	3.1 ± 0.1	~10
GNG-4,12	624.8	~0.025	~0.0016	3.2 ± 0.1	~10
GNG-5,12	638.8	~0.025	~0.0016	26 ± 1.0	~10
GNG-6,12	652.9	~0.02	~0.0013	34 ± 6.7	~5.0
GNG-0,13	582.7	~0.08	~0.0047	3.3 ± 0.0	~10
GNG-4,13	638.8	~0.025	~0.0016	3.2 ± 0.1	~10
GNG-5,13	652.9	~0.015	~0.0010	24 ± 0.6	~5.0
GNG-6,13	666.9	~0.01	~0.0007	44 ± 7.1	~1.0
GNG-0,14	596.8	~0.03	~0.0018	3.3 ± 0.0	~1.0
GNG-4,14	652.9	~0.025	~0.0016	3.4 ± 0.0	~10
GNG-5,14	666.9	~0.02	~0.0013	27 ± 1.5	~5.0
GNG-6,14	680.9	ND	ND	ND	insoluble
GNG-0,15	610.8	~0.02	~0.0012	3.7 ± 0.1	~1.0
GNG-1,15	624.8	~0.018	~0.0011	3.7 ± 0.1	~10
GNG-2,15	638.8	~0.015	~0.0010	3.8 ± 0.2	~10
GNG-3,15	652.9	~0.015	~0.0010	3.7 ± 0.1	~10
GNG-4,15	666.9	~0.01	~0.0007	3.7 ± 0.1	~10
GNG-5,15	680.9	~0.01	~0.0007	73 ± 6.0	~5.0
GNG-6,15	694.9	ND	ND	ND	insoluble
GNG-F2,2,14	714.8	~0.035	~0.003	9.6 ± 0.6	~5.0
GNG-F1,3,14	706.8	~0.02	~0.041	3.3 ± 0.1	~10
GNG-F1,2,15	706.8	~0.02	~0.041	3.8 ± 0.2	~10
GNG-3,14c	638.8	~0.025	~0.0015	3.2 ± 0.0	~10
DDM	510.6	0.17	~0.0087	3.4 ± 0.1	~10
OGNG c	568.7	1.0	~0.058	4.4 ± 0.3	~10
LMNG	1005.2	0.01	~0.001	10.6 ± 0.2	~10

^aMolecular weight of detergents.

^bHydrodynamic radius of micelles was determined at 1.0 wt% by dynamic light scattering.

^cReported in the literature.⁵⁹ ND = Not-determined.

The (F)P-GNG assemblies were then further characterized by their critical micelle concentrations (CMC) and self-assembly sizes in water at room temperature, determined using the hydrophobic dye [diphenylhexatriene (DPH)] and dynamic light scattering (DLS), respectively.⁷⁰ The CMCs of (F)P-GNGs were determined by measuring changes in the fluorescence intensity of DPH as a function of detergent concentration (Figure S3). As summarized in Table 1, the new P-GNGs exhibited a narrow range of CMCs (0.01 to 0.025 mM) despite substantial variations in the lengths of their main and pendant chains. These CMCs are comparable to those of LMNG (0.01 mM) and the previously reported GNG-3,14 (0.025 mM), but significantly lower than those of DDM (0.17 mM) and OGNG (1.0 mM). The CMCs of the P-GNGs were lower than those of the monopod counterparts (GNG-0,12, GNG-0,13, GNG-0,14, and GNG-0,15) and tended to decrease with increasing pendant length. This trend is generally consistent with that observed for pendant-bearing DDM analogs in a previous study.⁷¹ The addition of the pendant chain in the hydrophobic region increases the detergent hydrophobicity, which accounts for this observation. Additionally, due to the direct effect of detergent hydrophobicity on the HLB value, we found a correlation between the CMC and HLB values of the P-GNGs (Table 2 and Table S1). The CMCs of the FP-GNGs were higher than or comparable to those of their hydrocarbon counterparts depending on the number of fluorine atoms in the pendant region. The pentafluoroethyl (CF₃CF₂)-ended pendant-bearing GNG (GNG-F2,2,14) exhibited a CMC of 0.035 mM, higher than that of GNG-4,14 (0.025 mM), a trend observed from HFSs.^54,72 The FP-GNGs with the trifluoromethyl (CF₃)-terminated pendant (GNG-F1,3,14 and GNG-F1,2,15) showed CMCs of 0.025 and 0.020 mM, respectively, comparable to their hydrocarbon counterparts.

Next, we measured micelle sizes formed by the new detergents at 1.0 wt%, estimated by the hydrodynamic radii (R_h) of the micelles. Micelles formed by the new P-GNGs tended to increase in size with the main chain length. In contrast, the pendant chain length appeared to have little effect on micelle size unless it increased beyond a C4 pendant (Figure 2a). For example, the micelle sizes formed by the C15 main chain P-GNGs remained relatively unchanged (3.7 to 3.8 nm) as the pendant chain length extended from C0 (no pendant) to C1 (methyl) to C2 (ethyl) to C3 (propyl) to C4 (butyl). A similar trend was observed for previously described P-GNGs with the C1/C2/C3 pendant.⁵⁹ When the pendant length of the P-GNGs was increased to C5 (pentyl) or C6 (hexyl), however, we noted a dramatic increase in micelle sizes (Figure 2a). For example, the C5 pendant-bearing GNGs (GNG-5,13, GNG-5,14 and GNG-5,15) exhibited R_h values of 24, 27 and 73 nm, respectively, significantly larger than their C4 pendant-bearing counterparts. A similar trend was observed for pendant-bearing maltoside detergents;⁷¹ the micelle sizes formed by these DDM analogs showed little variation when a small pendant (e.g., methyl) was appended to the main chain. However, the introduction of larger pendants (e.g., ethyl or propyl) into the main chain substantially increased their micelle sizes. This interesting trend can be explained by the relative size/volume of the pendants compared to the empty spaces in the micelle interiors formed by the corresponding monopod GNGs (Figure 1c; m = −3). In other words, the micelle sizes remained relatively constant for the GNGs with the C1/C2/C3/C4 pendants, as the empty spaces in the micelle interiors formed by the monopod GNGs were large enough to accommodate these relatively small pendant chains. However, these empty spaces appear limited to accommodate larger pendants (C5/C6). The size/volume mismatch between the empty spaces in the micelle interiors and the larger pendants (C5/C6) would induce micelle restructuring to relieve steric hindrance exerted by the large pendants, resulting in micelle expansions. Based on this interpretation, the DDM analogs with ethyl/propyl pendant form large micelles due to the limited available spaces in their micelle interiors. As for the P-GNGs, the butyl (C4) pendant appears optimal for effectively filling the empty spaces within the micelle interiors, thereby maximizing alkyl chain density in this region, while maintaining a small micelle size. For the FP-GNGs, GNG-F1,3,14 and GNG-F1,2,15 formed small micelles with R_h values of 3.3 and 3.8 nm, respectively, comparable to their hydrocarbon counterparts (GNG-4,14 at 3.4 nm and GNG-3,15 at 3.7 nm). In contrast, GNG-F2,2,14 formed substantially larger micelles (9.6 nm) than its hydrocarbon counterpart (GNG-4,14 at 3.4 nm). These results indicate that the CF₃-ended butyl and propyl pendants are small enough to fit into the empty spaces of the micelle interior, while the CF₃CF₂-ended butyl pendant is too large for this fitting.

Figure 2.

(a) Variations in hydrodynamic radius (R_h) of micelles formed by the P-GNGs at 1.0 wt% at room temperature with increasing pendant chain length. Micelle sizes formed by GNG-6,14 and GNG-6,15 were not obtained due to their limited water-solubility (< 1.0 wt%). (b,c) R_h of micelles formed by GNG-4,15, GNG-5,14, GNG-F1,2,15, DDM and LMNG over a range of detergent concentrations (0.5 to 2.0 wt%) and temperatures (15 to 65 °C).

Figure 1.

Chemical structures of octyl glucose neopentyl glycol (OGNG) (a), the previously developed GNG-3,14 (b) and newly prepared (F)P-GNGs (c,d). The newly prepared P-GNGs include monopod GNGs with no pendant (m = −3) in (c). The new GNGs were designed either by extending the pendant chain length of the previous GNGs (P-GNGs) or by introducing hemifluorinated alkyl pendants into the GNG scaffold (FP-GNGs). The main chain length varied from C12 to C15, while the pendant chain length ranges from C4 to C6 in the P-GNGs. The FP-GNGs all contain a total of 21 carbons in the hydrophobic region and feature either a trifluoromethyl (CF₃)-ended propyl (C3) or butyl pendant (C4), or a pentafluoroethyl (CF₃CF₂)-ended butyl pendant.

The micelle sizes formed by the representative (F)P-GNGs (GNG-4,15, GNG-5,14 and GNG-F1,2,15) were further investigated using transition emission microscopy (TEM) and the results were generally consistent with those obtained from DLS (Figure S4). When we investigate the size distribution of micelle populations, micelles formed by most of the (F)P-GNGs showed unimodal distributions (Figures S5 and S6). The presence of large aggregates or the appearance of multiple peaks in the intensity-weight DLS charts is likely due to the high sensitivity of light scattering to large aggregates. The sizes of micelles formed by the (F)P-GNGs were further examined by varying the detergent concentration and solution temperature, using the selected detergents (GNG-4,15, GNG-5,14, and GNG-F1,2,15), along with DDM and LMNG as references. As expected, the micelle size formed by DDM showed little variation with changes in detergent concentration from 0.5 to 2.0 wt% or solution temperature from 15 to 65 °C (Figure 2b,c). Similar behavior was observed for the small micelle-forming GNGs (GNG-4,15 and GNG-F1,2,15). Interestingly, micelle sizes formed by GNG-4,15 tended to slightly increase with increasing solution temperature. In contrast, the micelle size formed by LMNG tended to increase with increasing detergent concentration and decrease with increasing solution temperature, consistent with previous results.^47,49 Similar trends were observed for the large micelle-forming detergent, GNG-5,14. These results suggest that a strong dependency of micelle size on detergent concentration and/or solution temperature is a typical characteristic of detergents that forms large micelles.

Detergent evaluation with model membrane proteins

We began the detergent evaluation with the prokaryotic transporter melibiose permease from Salmonella typhimurium (MelB_St).^73,74 MelB_St, overexpressed in E. coli membranes, was extracted using 1.5 wt% of DDM, GNG-3,14, and the individual GNGs for 90 minutes at 0 °C. The membrane extracts were then incubated at elevated temperatures (45, 55, or 65 °C) for an additional 90 minutes. The amounts of solubilized MelB_St were measured using Western blot and density quantification following ultracentrifugation and presented as percentages of the total MelB_St present in the untreated membranes. The amounts of soluble MelB_St extracted at low temperature (0 °C) and elevated temperatures (45, 55, or 65 °C) reflect membrane extraction efficiency and protein stabilization efficacy of the new detergents, respectively. Thus, this temperature-ramp protocol effectively estimates a detergent’s ability to both extract membrane protein and maintain it in a stable state, two crucial properties for membrane protein studies.

At 0 °C, DDM yielded a quantitative amount of soluble MelB_St, consistent with its extensive use for membrane protein extraction. resulted in approximately 75% soluble MelB_St. All new P-GNGs including GNG-3,14, except for GNG-0,12, were efficient at extracting MelB_St, indicating that the P-GNG architecture is highly compatible with membrane protein extraction.³⁷ All the P-GNGs, including the previous GNG-3,14, yielded amounts of soluble MelB_St comparable to DDM (~100%), except for GNG-4,14 (~70%) and the monopod GNG (GNG-0,12 and GNG-0,13) (less than 20%). When the incubation temperature further increased to 55 °C, the amounts of soluble MelB_St varied significantly, primarily depending on pendant chain length. At this high temperature, DDM and the monopod GNGs failed to stabilize the transporter, whereas previous GNG-3,14 yielded around 50% of soluble MelB_St. Notably, most P-GNGs with C13/C14/C15 main chains and C4/C5 pendant chains were generally more effective than DDM and GNG-3,14 at stabilizing the transporter, as exemplified by GNG-4,13, GNG-5,13, GNG-5,14, and GNG-4,15. The use of the C6 pendant-bearing GNG (i.e., GNG-6,12) or C12 main chain P-GNGs (e.g., GNG-4,12) resulted in the limited amounts of soluble MelB_St (10 to 30%). Importantly, the optimal length of the pendant chain for MelB_St stability was nearly the same (C4 or C5) irrespective of the main chain length: the C4 pendant length was optimal for the P-GNGs with the C12, C13, or C15 main chains (e.g., GNG-4,13 and GNG-4,15), while the C5 pendant length was most suitable for the P-GNGs with the C14 main chain. The significant dependency of detergent efficacy on pendant length for protein stabilization highlights the importance of alkyl chain density at the hydrophilic-lipophilic interfaces of detergent micelles in maintaining membrane protein stability.

The FP-GNGs exhibited a wide range of behaviors depending on the specific structures of the pendants. GNG-F2,2,14 yielded 60% soluble MelB_St following protein extraction at 0 °C, but an increase in the incubation temperature above 45 °C completely diminished MelB_St solubility. In contrast, when GNG-F1,3,14 and GNG-F1,2,15 were used at 0 °C, these CF₃-ended pendant-bearing GNGs proved to be as efficient as DDM in extracting the transporter from the membranes. This is notable, as many hemifluorinated surfactants (HFSs) or partially fluorinated detergents have been shown to be less effective than their hydrocarbon counterparts for membrane protein extraction.⁵² When the incubation temperature was increased to 55 °C, GNG-F1,2,15 fully retained solubility of the transporter, demonstrating that this FP-GNG outperforms the other (F)P-GNGs in terms of extraction and stabilization of MelB_St. These results indicate that the presence of a small number of fluorine atoms (-CF₃) can significantly impact on detergent behaviours for both protein extraction and stabilization.

The promising properties of the new detergents with MelB_St led us to evaluate their abilities to stabilize a more challenging membrane protein, β₂AR.⁷⁵ This human G protein-coupled receptor (GPCR) was extracted using DDM and purified in 0.01% LMNG. The LMNG-purified β₂AR was then diluted into buffer solutions containing the individual detergents (DDM, GNG-3,14 and (F)P-GNGs) at a final concentration of 0.1 wt%. This dilution method would lead to mixed micelles comprising (F)P-GNG (major) and LMNG (minor). Consequently, the receptor stability would be influenced by the residual LMNG as well as (F)P-GNG in the samples, likely determined by the combined/averaged effect of the two detergents on protein stability. Therefore, the detergent efficacy order for protein stabilization remains unchanged whether tested with a single detergent or a combination of two detergent components.⁷⁶ Of note, we included LMNG as a control, as this novel detergent is known to be particularly effective for GPCR stability. Receptor stability in each detergent was assessed by monitoring the binding of the radiolabeled antagonist ([³H]-dihydroalprenolol (DHA)) at regular intervals over an 8-day incubation period at room temperature.⁷⁷ Consistent with previous results, DDM and OGNG were ineffective at retaining GPCR ligand-binding capability.^78,79 As expected, LMNG and the previously described GNG-3,14 allowed the receptor to bind DHA effectively at day 0, but this activity was gradually lost over time. The new P-GNGs exhibited variable receptor stability based on the lengths of their pendant and main chains. For the P-GNG series with identical pendant groups, receptor stabilization peaked with the main chain length of C14/C15. Among the C4 pendant-bearing GNGs, for instance, GNG-4,14 and GNG-4,15 were the most effective, followed by GNG-4,13 and GNG-4,12. More importantly, detergent efficacy for receptor stabilization improved with pendant chain length increasing from C1 to C5 within the same main chain series. For instance, among the C14 main chain GNGs, GNG-5,14 was the most effective, followed by GNG-4,14 and GNG-3,14. A similar pattern was found with the C15 main chain GNG. Across all main chain length (C12/C13/C14/C15), the optimal pendant length for the P-GNGs for receptor stability were C4 or C5. The combinations of C14/C15 main chain with the C4/C5 pendant were the best for receptor stability, as exemplified by GNG-5,14, GNG-4,15, GNG-5,13, and GNG-4,14. These detergents showed comparable or superior receptor stability compared to LMNG, a significant achievement given the prevalent use of LMNG in GPCR structural studies.⁴⁷ When examining the FP-GNGs under the same conditions, we observed that the CF₃-ended pendant-bearing GNGs (GNG-F1,3,14 and GNG-F1,2,15) demonstrated significantly improved efficacy compared to DDM and GNG-F2,2,14. Intriguingly, GNG-F1,2,15 excelled at stabilizing the receptor, performing significantly better than both DDM and its hydrocarbon counterpart (GNG-3,15). This result is noteworthy given the scarcity of HFSs that support GPCR stability.⁸⁰

Discussion

We initiated our research with an aim of enhancing detergent efficacy for protein stabilization by either extending the pendant length of previous P-GNGs or incorporating partially fluorinated alkyl groups into the pendant region. The pendants of the newly prepared GNGs have varying chain lengths (P-GNGs) or differing numbers of fluorine atoms (FP-GNGs). The synthetic simplicity (only four synthetic steps) and flexibility for structural modification allowed us to efficiently prepare a large number of P-GNGs with diverse structural variations, enabling the optimization of detergent properties for membrane protein studies. In the evaluation with MelB_St and β₂AR, we observed the generally favorable effect of the new P-GNGs on protein stability as the main chain length increased from C12/C13 to C14 or C15. This is likely due to the improved compatibility of these longer main chain lengths (C14/C15) with the hydrophobic dimensions of the membrane proteins. Notably, detergent efficacy for protein stabilization tended to improve significantly as the pendant chain length gradually increased from C0 to C4, with a clear, consistent trend observed for both MelB_St and β₂AR. In contrast, despite variations in pendant chain length, micelle sizes formed by the P-GNGs were similar, ranging from 3.1 to 3.7 nm. These results suggest that the pendant chain length, rather than micelle size, plays a dominant role in determining membrane protein stability for this scaffold. When the pendant chain length increased further from C4 to C5 or C6, detergent efficacy showed little improvement (C5) or decreased (C6). Together, the results from both the current study and our previous reports demonstrate that the optimal pendant length for effective membrane protein stabilization by the P-GNGs lies between C4 and C5.

A question arises when comparing the effects of pendant chain length on micelle size versus protein stability. Micelle size formed by the P-GNGs increased abruptly when the pendant chain length increased from C4 to C5, suggesting that the C4 pendant has the best fit into the available space in the micelle interiors formed by the monopod GNGs. Thus, it is reasonable to expect that the C4 pendant-bearing P-GNGs are most suitable for stabilizing membrane proteins compared to other pendant variants. This is supported by the best protein stabilization performance observed for the P-GNGs with the C4 pendant, as demonstrated by GNG-4,13 (MelB_St), GNG-4,15 (MelB_St), and GNG-4,15 (β₂AR). However, in some cases, the P-GNGs with the C5 pendant exhibited overall superior stability for membrane proteins, as seen with GNG-5,13 (β₂AR) and GNG-5,14 (MelB_St and β₂AR). We propose a couple of possible explanation for this unexpected result. First, since the membrane protein stability assays were conducted in buffered solutions rather than pure water, the micelle sizes of the P-GNGs in the buffered solutions rather than in pure water are more relevant to the discussion. We measured the micelle sizes of the P-GNGs in HEPES buffer (20 mM, pH 7.4) and observed a similar trend in micelle size variation with pendant length, as seen in pure water. Therefore, the variation in optimal pendant length between dense micelle formation (C4) and enhanced protein stability (C4/C5) is unlikely to stem from difference in the experimental medium (pure water vs. buffered solution). Second, the micellar architecture in PDCs may differ substantially from that of pure detergent micelles. Upon association with membrane protein surfaces, detergent molecules in micelles rearrange to find optimal positions and conformations. This micelle remodeling could generate empty spaces in the micelle interiors that differ in size from those in pure micelles. In other words, the C5 pendant might fit better into the interiors of micelles associated with membrane proteins, even though the C4 pendant was optimal for pure GNG micelles. Thus, information from pure detergent micelles may only partially reflect the detergent properties relevant to membrane proteins.

Building on the unique behaviors of HFSs, which differ from fully fluorinated or hydrocarbon-based detergents, we prepared three FP-GNGs with different hemifluorinated pendant groups. The presence of a partially fluorinated chain in the detergent hydrophobic region can enhance protein stability by improving hydrophobic/fluorophobic interactions between detergent tail groups (detergent-detergent interactions) in aqueous solutions. Generally, the CF₃CF₂-ended pendant-bearing GNG (GNG-F2,2,14) was inferior to its hydrocarbon counterpart (GNG-4,14) in stabilizing MelB_St and β₂AR. This suggests that the pendant chain of GNG-F2,2,14 may be too bulky to fit into the empty spaces within the micelle interiors or too lipophobic to promote strong detergent-detergent interactions. In contrast, the CF₃-ended pendant-bearing GNG-F1,3,14 and GNG-F1,2,15, generally favored membrane protein stabilization, with GNG-F1,2,15 showing the best performance. Notably, GNG-F1,2,15 outperformed DDM at extracting MelB_St from the membranes and was highly effective at stabilizing β₂AR (a GPCR). The favorable behaviors of the FP-GNG for protein extraction and GPCR stabilization have rarely been observed with other HFSs, with a small exception.⁸¹ Consequently, this FP-GNG can be effectively used in all stages of membrane protein manipulation - extraction, purification, and structural study. These favourable properties appear to stem from its optimal structural features, including the number (3) and position (at the hydrophilic-lipophilic interface) of fluorine atoms, in addition to the main chain (C15) and pendant chain lengths (C3) in the GNG architecture.

Conclusion

Three sets of GNG detergents— monopod GNGs, P-GNGs and FP-GNGs—were prepared through structural variations in the pendant chain. Our comparison of their abilities to stabilize membrane protein demonstrated that a sufficiently large pendant size in this scaffold is critical for membrane protein stability, with the best performance observed for the GNGs bearing the C4 or C5 pendant. Additionally, the FP-GNG study indicates that the sizes of fluorinated carbons (i.e., the number of fluorine atoms) in the pendant region significantly affect the detergent’s ability to stabilize membrane proteins. We identified several P-GNGs (GNG-5,13, GNG-5,14, and GNG-4,15) and one F-GNG (GNG-F1,2,15) as new detergent tools that are remarkably effective for membrane protein stability. Notably, GNG-5,14 outperformed LMNG in stabilizing β₂AR, while GNG-F1,2,15 was superior to DDM for the extraction and stabilization of MelB_St. The markedly enhanced efficacy of these two glucoside detergents is attributed to the incorporation of multiple favorable factors, such as hydrophilic-lipophilic balance (HLB), pendant size and main chain length, into a compact detergent structure. Therefore, the current study not only suggests detergent tools for membrane protein research but also provides guidelines for detergent design aimed at enhancing membrane protein stability. This contribution will advance membrane protein research and facilitate the development of novel detergents for structural study.

Figure 3.

Thermo-solubility of MelB_St solubilized in the new (F)P-GNGs. DDM and GNG-3,14 were used as controls. Western blotting (top panel): the amounts of soluble MelB_St after protein extraction from E. coli membranes using the individual detergents at 0 °C and further incubation of the resulting membrane extracts at elevated temperatures (45, 55, or 65 °C). Histogram (bottom panel): the amounts of soluble MelB_St in the individual detergents are expressed as percentages of total MelB_St present in the untreated membranes (referred to as ‘Control’). Error bars represent n = 3, SEM.

Figure 4.

Time-course stability of β₂AR solubilized in the (F)P-GNGs at 0.1 wt%. DDM, OGNG, LMNG, and GNG-3,14 were used as controls. Receptor stability are evaluted by monitoring its ability to bind the radiolabeled antagonist ([³H]-dihydroalprenolol (DHA)) during an 8-day incubation at room temperature. Error bars represent n = 3, SEM.