Toward Rational Design of Protein Detergent Complexes: Determinants of Mixed Micelles That Are Critical for the In Vitro Stabilization of a G-Protein Coupled Receptor

Abstract

Although reconstitution of membrane proteins within protein detergent complexes is often used to enable their structural or biophysical characterization, it is unclear how one should rationally choose the appropriate micellar environment to preserve native protein folding. Here, we investigated model mixed micelles consisting of a nonionic glucosylated alkane surfactant from the maltoside and thiomaltoside families, bile salt surfactant, and the steryl derivative cholesteryl hemisuccinate. We correlated several key attributes of these micelles with the in vitro ligand-binding activity of hA₂aR in these systems. Through small-angle neutron scattering and radioligand-binding analysis, we found several key aspects of mixed micellar systems that preserve the activity of hA₂aR, including a critical amount of cholesteryl hemisuccinate per micelle, and an optimal hydrophobic thickness of the micelle that is analogous to the thickness of native mammalian bilayers. These features are closely linked to the headgroup chemistry of the surfactant and the hydrocarbon chain length, which influence both the morphology and composition of resulting micelles. This study should serve as a general guide for selecting the appropriate mixed surfactant systems to stabilize membrane proteins for biophysical analysis.

Introduction

Integral membrane proteins constitute one-third of the human proteome and are targeted by half of all marketed pharmaceuticals (1). Rational drug design strategies often rely on determining the high-resolution structure of soluble proteins; however, similar strategies for membrane proteins are limited by our ability to reconstitute these proteins in a membrane-mimetic environment with native stability and ligand-binding activity. Surfactant micelles have long been used as membrane-mimetic systems to enable membrane protein integrity in vitro via solubilization in protein detergent complexes (PDCs) (2). PDCs are particularly attractive vehicles for evaluating membrane protein folding (3,4) and protein-protein interactions (5,6), and can facilitate structure determination through crystallization and solution NMR (7–9). However, strategies to reconstitute membrane proteins in PDCs are often met with complications arising from changes incurred to the protein's native folded state within micelles (10) or depletion of critical native interactions with membrane-resident constituents (11). This is especially true for complex eukaryotic membrane proteins, whose function is often modulated by lipids within bilayers (11).

Despite the widespread use of surfactant micelles, no detailed rules have been established that can direct the choice of surfactants and formulation of micelles to stabilize the structure of a membrane protein of interest within PDCs. Hundreds of surfactants are commercially available, each of which has a unique set of properties and can be used individually or combined in seemingly infinite combinations to create mixed micelles. Synthesis of new surfactants that are superior for biophysical studies of membrane proteins is an area of active research (12–15), yet critical information to better inform the design of these amphiphiles is still lacking.

A number of anecdotal studies of membrane protein stability in various micellar systems were recently reviewed (16,17), and their findings were reduced to rules of thumb regarding selection of surfactants for the study of membrane proteins. Specifically, it is believed that membrane protein stability generally increases with increasing hydrophobic tail size and decreases with increasing headgroup size and charge, although how these rules are manifested in PDCs is unclear. These heuristics implicitly assume that molecular chemistry is the most, if not the only, important variable in selecting a micellar system to retain the in vitro stability of membrane proteins. This typically leads to strategies in which pure micelles of a particularly mild surfactant are used as a first attempt (e.g., dodecyl maltoside), and the micellar system is fine-tuned through the use of other detergents or additives until a suitable system is found for the protein and biophysical method of interest (16,18). Such trial-and-error methodology is time-consuming and resource-intensive, especially considering the low yields that are typically achieved for expression and purification of most membrane proteins, and does not lead to information that is readily translatable across different systems.

The primary limitation of this approach is that it ignores the effects of variables other than molecular chemistry on PDC formation, even though the micellar structure of a majority of commonly used surfactants (e.g., glucosides, maltosides, and ethoxylated alkanes) are known to depend strongly on solution conditions (e.g., concentration and temperature) (19–23). An alternative approach is to consider the effects of molecule selection and solution conditions on the micellar structure of PDCs, and ultimately on membrane protein folding. Indeed, it is often speculated that the prototypical micellar system for successful in vitro protein stabilization would be one that closely mimics the native bilayer (5,24). However, few authors have attempted to directly test this hypothesis, owing to a lack of detailed characterization of micelle morphology and composition for the systems of interest. Thus, the optimal structure of different micellar aggregates for these applications remains largely unknown, even though the geometry and composition of such micelles are likely to impact membrane protein folding. This is particularly the case for mixed surfactant micelles (25–27) and surfactant/lipid micelles (10,28–31), which have been shown to be useful for promoting the in vitro stability of a growing number of membrane proteins, and for studying the thermodynamics of membrane protein folding/unfolding (32).

In previous work, we showed that the composition and morphology of mixed micelles containing the nonionic alkyl glucoside surfactant dodecyl-β-D-maltoside, the cholesterol derivative cholesteryl hemisuccinate (CHS), and the bile salt surfactant 3-(3-cholamidopropyl)-dimethylammoniopropane sulfate (CHAPS) were critically linked to the conformational stability of the human adenosine A₂a receptor, a seven-transmembrane domain G-protein coupled receptor (GPCR) (31,33,34). Specifically, we found that mixed micelle compositions that preserved the optimal ligand-binding activity of hA₂aR bore a striking similarity to the geometry and cholesterol content of mammalian membranes (33). The addition of cholesterol derivatives to PDCs also induced changes in the curvature and size of the micelles that one would not predict by considering molecular architecture alone.

In this work, we sought to test the generality of these results and achieve a better understanding of the critical parameters that govern the stability and structure of PDCs by extending our study to different micellar systems with divergent properties. After an initial screen of commonly used surfactants (see Fig. S1 in the Supporting Material), we focused specifically on micelles based on nonionic alkyl maltosides (C_nβG₂) and thiomaltosides (C_nβSG₂) in the presence of the additives CHS and CHAPS. Specifically, we studied hA₂aR activity (defined here as specific binding to an A₂aR-specific agonist) as a function of CHS/CHAPS content within PDCs, and correlated successful reconstitution of the receptor with aspects of the micelle environment that have been observed to promote receptor activity. Simultaneously, we used small-angle neutron scattering (SANS) to determine the molecular composition and aggregate structure of these micelles. This allowed us to identify characteristic traits of surfactant micelles that can be linked to hA₂aR's conformational stability. These studies will enable more rational selection of surfactants to facilitate structural studies of membrane proteins.

Materials and Methods

Micelle solution preparation

Surfactants and lipid analogs were purchased from Anatrace (Maumee, OH) at the highest purity available. Pure component and mixed micelles were prepared from concentrated aqueous surfactant stock solutions in 50 mM sodium phosphate buffer at a pH of 7.0. Stock solutions of CHS were made in CHAPS such that the final stock solution was 6% CHAPS/1.2% CHS (w/w). To enhance contrast for the SANS experiments, the aforementioned stock solutions were also prepared in buffer containing D₂O (Cambridge Isotope Laboratories, Andover, MA).

To facilitate the comparison of different surfactant micelles as a function of CHAPS and/or CHAPS/CHS addition, solutions were prepared as a function of the lipid-analog content δ (33), defined as follows:

$$\delta =\frac{\text{mass of CHAPS}+\text{mass of CHS}}{\text{total mass of solute}}.$$ (1)

Entire surfactant series were tested as a function of δ, reflecting the addition of CHAPS or CHAPS/CHS to the micellar solutions. The composition of the micelles can similarly be defined using the mole fraction, x_CHAPS,CHS, of CHAPS and CHS as

$$x_{CHAPS,CHS}=\frac{\text{moles of CHAPS}+\text{moles of CHS}}{\text{total moles of solute}}.$$ (2)

We note that, due to the similar molecular weights of the constituent molecules, the values of x_CHAPS,CHS for the micellar systems studied here are nearly identical to the corresponding values of δ.

hA₂aR-His₁₀ purification, reconstitution, and activity measurement

Recombinant human adenosine hA₂a receptor (hA₂aR) with a C-terminal decahistidine tag was overexpressed in yeast as described previously (31,33–35). hA₂aR-His₁₀ was purified from yeast membranes after cell lysis, solubilization, and immobilized metal affinity chromatography (31,33,34). Receptors were solubilized in 2% dodecyl-β-D-maltoside (C₁₂βG₂) containing 1% CHAPS and 0.2% CHS while bound to nickel resin particles, which was reduced to 0.1% C₁₂βG₂/0.1% CHAPS/0.02% CHS during purification. To facilitate surfactant exchange for hA₂aR-His₁₀, resin-bound receptors were aliquoted and then centrifuged at 3000 × g to sediment the resin, and the supernatant was removed with a gel-loading pipette tip. Fresh surfactant (of varied type and concentration) at a volume 100-fold greater than that of the resin was added to the resin-bound receptors, which were incubated in an end-over-end mixer for 10–15 min. The samples were again centrifuged and incubated with fresh surfactant, and this process was repeated four to five times to ensure sufficient surfactant exchange. We measured the ligand-binding activity for resin-bound hA₂aR-His₁₀ by competitive radioligand binding experiments using the tritiated agonist ³[H]-CGS-21,680 (PerkinElmer, Waltham, MA) and unlabeled competitor N⁶-cyclohexyladenosine (CHA; Sigma, St. Louis, MO) as described previously (33,36).

SANS measurements

SANS experiments were performed at the National Institute of Standards and Technology Center for Neutron Research in Gaithersburg, MD. Measurements were made on both the NG3 and NG7 30 m SANS instruments. Samples were loaded into standard 2 mm path length titanium cells with quartz windows, and placed in a temperature-controlled sample environment (10CB) held at 25°C. Two-dimensional sample scattering spectra were obtained at instrument detector distances of 1.35 m, 4.5 m, and 13.1 m. We reduced the measured spectra according to standard protocols using the NCNR Igor Pro software package (37), which yielded the absolute scattered intensity, I(q), where q is the momentum transfer vector.

SANS measurements were performed on mixed micelles comprised of C_nβG₂/CHAPS/CHS and C_nβSG₂/CHAPS/CHS at various values of the surfactant fraction δ and total solute concentration of 1 wt% in D₂O buffer to enhance contrast. All SANS data were analyzed according to a previously described method (33) to determine micelle morphology and composition (refer to Supporting Material).

results

Activity of hA₂aR-His₁₀ in nonionic mixed micelle systems

A number of surfactants have proven useful for the solubilization and characterization of membrane proteins; however, the interplay between surfactant type and folding of membrane proteins is not well understood (2). In previous studies, a mixed micelle system consisting of 0.1% C₁₂βG₂, 0.1% CHAPS, and 0.02% CHS was identified, which maintained the active conformation of a decahistidine-tagged human adenosine A₂a receptor (hA₂aR-His₁₀) in vitro (31). We aimed to explore critical parameters within the micelle environment and evaluate different surfactant systems for their ability to retain the activity of hA₂aR-His_10.

To investigate the effect that different surfactant micelles have on the structural integrity of hA₂aR-His₁₀, we swapped various surfactants (Fig. S1) for C₁₂βG₂, and performed competitive radioligand binding to monitor receptor activity. Because reconstitution of hA₂aR in 0.1% C₁₂βG₂ or 0.1% C₁₂βG₂/0.1% CHAPS alone failed to preserve ligand-binding activity for hA₂aR-His₁₀ (31,33), all of the surfactant systems discussed below contained 0.1% CHAPS and 0.02% CHS. In all cases (other than CHAPS), the concentration was ∼16 times above the pure component critical micelle concentration (cmc). Surfactants were chosen for this activity screen because of their widespread use in membrane protein studies (2,7,38), and varied in both headgroup properties and hydrophobic tail length.

Members of the maltoside and related thiomaltoside surfactant series were analyzed, because activity was previously observed for hA₂aR in C₁₂βG₂ mixed micelles (31). As shown in Fig. 1, activity of hA₂aR-His₁₀ was preserved in mixed micelles including C₁₂βG₂, tridecyl-maltoside (C₁₃βG₂), and tetradecyl-maltoside (C₁₄βG₂; Fig. 1 A), signified by B_max values ∼5000–6000 cpm, whereas no appreciable activity was observed in decyl-maltoside (C₁₀βG₂) or undecyl-maltoside (C₁₁βG₂). A similar trend was seen in the related thiomaltoside series: activity of hA₂aR-His₁₀ was observed in dodecyl-thiomaltoside (C₁₂βSG₂), yet no activity was seen in decyl-thiomaltoside (C₁₀βSG₂; Fig. 1 B). Some degree of ligand-binding activity was achieved for hA₂aR-His₁₀ reconstituted in mixed micelles including undecyl-thiomaltoside (C₁₁βSG₂; Fig. 1 B), although no appreciable binding was seen in C₁₁βG₂ (Fig. 1 A), which has an identical hydrocarbon tail length.

Figure 1.

Competitive ligand-binding activity assays characterize the activity of hA₂aR-His₁₀ after surfactant exchange from C₁₂βG₂/CHAPS/CHS into the indicated surfactant. All samples contain 0.1% CHAPS and 0.02% CHS. Members of the (A) maltoside series (• C₁₀βG₂, □ C₁₁βG₂, ♦ C₁₂βG₂, Δ C₁₃βG₂, ▿ C₁₄βG₂), (B) thiomaltoside series (• C₁₀βSG₂, □ C₁₁βSG₂, and ▴ C₁₂βSG₂), and (C) other surfactants (including • octyl-glucoside, ▪ nonyl-glucoside, ♦ C8E4, ▴ C12E9, ▾ LDAO, ▵ fos-choline 12, ▴ cymal-5, ○ triton X-100, and × triton X-305) were analyzed. The depicted fits to a one-site competitive ligand-binding model are indicated by lines, and experimental data are represented by points. Error bars represent the standard deviation from the average of three independent points.

hA₂aR-His₁₀ activity in other representative surfactant micelles

Given that protein activity was identified for mixed micelles consisting of maltosides or thiomaltosides in the presence of CHAPS/CHS, we explored other surfactant types with divergent properties for their ability to stabilize the active conformation of hA₂aR in vitro. These surfactants included other sugar-based surfactants (glucosides, Cymal-5), zwitterionic surfactants (fos-choline 13), polyoxyethylene surfactants (C₈E₄, C₁₂E₈, Triton X-100, Triton X-305), and lauryldimethylamine-oxide (LDAO); (Fig. S1). Fig. 1 C shows activity results obtained from competitive radioligand binding experiments performed on hA₂aR reconstituted in these mixed micelle systems with 0.1% CHAPS and 0.02% CHS. In contrast to the receptor activity generally observed in maltoside and thiomaltoside-based surfactant mixed micelles, other surfactant systems failed to stabilize the active conformation of purified receptors under these conditions (Fig. 1 C). Moderate ligand binding was identified only in mixed micelles containing C₁₂E₉, an ethoxylated alkane surfactant with a 12-hydrocarbon tail; other surfactant systems largely displayed nonspecific binding.

Morphology and composition of nonionic mixed micelles

To better understand the relationship between surfactant micelle architecture and receptor activity observed in different mixed micelle systems, we characterized empty micelles by SANS. Because the micelles used in this work exist on a nanometer scale, SANS is an ideal method for probing the structure and composition of these single-component and mixed micelles for membrane protein study (39). Furthermore, because SANS is an absolute measurement, this technique also allows for the calculation of aggregation numbers within micelles, in contrast to alternate approaches based on light or x-ray scattering (40).

Mixed micelles consisting of members of the maltoside and thiomaltoside surfactant series, with varied amounts of CHAPS/CHS, were systematically analyzed via SANS to determine size and shape information. Given the critical necessity of sterols for hA₂aR activity (31), we particularly sought to understand how micelle structure changed upon addition of various amounts of CHS. Because other surfactant types generally were not observed to stabilize the active state of hA₂aR-His₁₀ through radioligand-binding experiments (Fig. 1 C), we performed SANS only on mixed maltoside and thiomaltoside micelles.

Surfactant monomers have a wide range of properties, resulting in different micelle morphologies encountered in solution above the critical micelle concentration. The aggregate morphology and composition of micelles are sensitive to both the overall concentration and relative composition of solute species (5,41). To systematically study this variation among different mixed micelle systems, we characterized micelles with varying values of δ, acting as a measure of CHAPS and/or CHS concentration in solution (refer to Materials and Methods). For SANS experiments, the overall concentration of species in solution was fixed at 1 wt % to allow for adequate comparison between different surfactants. We characterized δ-values of 0, 0.1, 0.2, 0.55, and 0.85 (Table 1 and Table S1) corresponding to increasing amounts of CHAPS/CHS that span the range tested in radioligand-binding studies. To isolate the specific role of CHS in micelle shape change, we also analyzed samples at identical δ-values with CHAPS in the absence of CHS (Table S2 and Table S3).

Table 1.

Relevant parameters of indicated maltoside-based micelles and mixed micelles at given δ, determined based on an ellipsoidal model, except for those for which a cylindrical model was used (italicized)

CnβG2 n	δ	xCHAPS,CHS	a (Å)	b (Å)	Rg (Å), model	Rg (Å), Guinier	NS	NCHAPS	NCHS
n = 10	0	0.000	26.3 ± 0.1	15.2 ± 0.1	18.0 ± 0.3	17.4 ± 0.1	68 ± 1	–	–
	0.1	0.096	26.0 ± 0.1	15.8 ± 0.1	17.9 ± 0.5	17.3 ± 0.2	62 ± 1	6 ± 1	2 ± 1
	0.2	0.193	25.2 ± 0.1	15.4 ± 0.1	17.4 ± 0.4	17.1 ± 0.3	48 ± 1	11 ± 1	4 ± 1
	0.55	0.539	24.2 ± 0.1	13.2 ± 0.1	16.4 ± 0.3	16.2 ± 0.2	26 ± 1	21 ± 1	7 ± 1
	0.85	0.844	24.0 ± 0.1	12.3 ± 0.1	15.8 ± 0.4	15.5 ± 0.2	8 ± 1	23 ± 1	6 ± 1
n = 11	0	0.000	29.3 ± 0.1	16.5 ± 0.1	19.9 ± 0.3	19.4 ± 0.2	89 ± 1	–	–
	0.1	0.099	28.3 ± 0.1	17.1 ± 0.1	19.4 ± 0.4	18.7 ± 0.2	73 ± 1	9 ± 1	2 ± 1
	0.2	0.198	27.4 ± 0.1	17.4 ± 0.1	19.2 ± 0.3	18.3 ± 0.2	58 ± 1	15 ± 1	5 ± 1
	0.55	0.546	25.6 ± 0.1	16.3 ± 0.1	17.9 ± 0.3	16.3 ± 0.2	31 ± 1	24 ± 1	10 ± 1
	0.85	0.848	24.6 ± 0.1	13.3 ± 0.1	16.7 ± 0.4	15.2 ± 0.2	10 ± 1	26 ± 1	7 ± 1
n = 12	0	0.000	32.4 ± 0.1	17.5 ± 0.1	21.9 ± 0.2	22.3 ± 0.2	121 ± 1	–	–
	0.1	0.101	30.6 ± 0.1	18.3 ± 0.1	21.0 ± 0.4	21.1 ± 0.2	96 ± 1	7 ± 1	3 ± 1
	0.2	0.202	29.4 ± 0.1	18.6 ± 0.1	20.3 ± 0.3	19.8 ± 0.3	82 ± 1	12 ± 1	5 ± 1
	0.55	0.553	26.5 ± 0.1	17.2 ± 0.1	18.4 ± 0.3	17.4 ± 0.2	36 ± 1	27 ± 1	12 ± 1
	0.85	0.852	24.8 ± 0.1	13.9 ± 0.1	16.8 ± 0.3	15.4 ± 0.2	10 ± 1	29 ± 1	9 ± 1
n = 13	0	0.000	47.0 ± 0.3	18.6 ± 0.1	25.6 ± 0.8	24.5 ± 0.5	194 ± 1	–	–
	0.1	0.104	33.3 ± 0.1	19.1 ± 0.1	22.7 ± 0.4	23.3 ± 0.2	108 ± 1	8 ± 1	3 ± 1
	0.2	0.206	30.4 ± 0.1	19.3 ± 0.1	20.5 ± 0.3	21.6 ± 0.3	75 ± 1	14 ± 1	7 ± 1
	0.55	0.560	27.0 ± 0.1	17.2 ± 0.1	18.8 ± 0.4	17.8 ± 0.2	30 ± 1	29 ± 1	18 ± 1
	0.85	0.855	24.9 ± 0.1	13.6 ± 0.1	16.9 ± 0.3	15.9 ± 0.2	7 ± 1	27 ± 1	12 ± 1

δ, mass fraction of surfactant in solution; x, mole fraction of CHAPS, CHS in solution; a, minor axis; b, major axis; R_g (model), model-predicted radius of gyration; R_g (Guinier), radius of gyration from Guinier analysis; N_s, aggregation number of surfactant indicated; N_CHAPS, aggregation number of CHAPS; N_CHS, aggregation number of CHS.

After reduction of the SANS data, we fit the data to a model to determine the approximate morphology and relevant axial dimensions for each surfactant mixture studied. Representative SANS spectra obtained for C₁₂βG₂ and C₁₂βSG₂ as a function of δ are shown in Fig. S2. These curves display the absolute scattered intensity (I) of neutrons from the sample as a function of the inverse length scale q. To extract morphological information from the spectra, we performed a model-independent Guinier analysis, yielding the radius of gyration (R_g) of micelles in solution, as shown for C₁₂βG₂ and C₁₂βSG₂ in Fig. S2 (insets).

Table 1 and Table S1 list experimentally determined values of R_g as a function of δ for mixed micelles with CHAPS and CHS added to maltosides and thiomaltosides. At low δ, R_g increases in accordance with the hydrocarbon chain length for surfactant micelles analyzed within each surfactant series, where C₁₃βG₂ and C₁₂βSG₂ have the greatest calculated values for R_g among their respective series (Table 1 and Table S1). Upon addition of CHAPS and CHS to solution (increasing δ), R_g generally decreases toward a value of ∼17 Å, in similarity to the size of pure CHAPS micelles (42). Note that the reduction in R_g is significantly more pronounced in the absence of CHS (Table 1, Table S1, Table S2, and Table S3).

Subsequently, various morphological models were applied to describe the scattering data. These morphological models predicted a specific value of R_g, which was compared with the model-independent value of R_g calculated from the Guinier analysis. This comparison was the primary metric used to evaluate the applicability of the chosen model fit to the data (Table 1 and Table S1). As such, an oblate ellipsoid model was ultimately selected to describe the morphology of most compositions, and it is in quantitative agreement with previous SANS studies of pure-component C₁₂βG₂ micelles (43), thereby validating the model used. However, in several cases (italicized in Table 1 and Table S1), this model did not fit the data well, and we instead used a cylindrical model that more accurately described the scattering data for these particular systems (Table 1 and Table S1). These results indicate an ellipsoid-to-rod transition for micelles with increasing chain length of maltoside and thiomaltoside surfactants for low values of x_CHAPS/CHS. This can be understood in terms of the increase in the packing parameter of the surfactant with increasing alkyl chain length (44), leading to a reduction of aggregate curvature toward elongated structures. Of interest, this ellipsoid-to-rod transition is more pronounced for the thiomaltoside surfactants, for which C₁₂βSG₂ micelles are highly elongated. This observation is likely due to the significantly greater hydrophobic character of the thiol linkage of C₁₂βSG₂ compared with the ester linkage of C₁₂βSG₂ (45), which has been shown to promote elongation of micelles in other sugar-based surfactants (46).

The resulting morphological parameters a and b for each micelle and mixed micelle system, including both ellipsoidal and cylindrical micelles, are summarized in Table 1 and Table S1. Upon an increase in the amount of CHAPS/CHS in solution, a proportional decrease in the major axial dimension (a) is observed, consistent with the overall reduction in R_g (Table 1, Table S1, and Fig. S3). Thus, the addition of CHAPS and CHS results in an increase in micellar curvature. In cases where cylindrical micelles are obtained at δ = 0, this increase in curvature serves to counteract the ellipsoid-to-rod transition, and results in micelles that retain an ellipsoidal morphology. In contrast, the minor axis (b) systematically exhibits nonmonotonic behavior with increasing δ (Fig. 2) added to the micelles for both maltoside and thiomaltoside-based micelles.

Figure 2.

Micellar minor axis (A) and number of CHS monomers per micelle (B) as a function of δ for mixed n-alkyl-β-D-maltoside/CHAPS/CHS micelles for the indicated alkyl chain lengths (n). Boxed region in the upper panel indicates the radial thickness of a typical mammalian membrane (∼16 Å), ± 0.5 Å deviation.

The aggregation numbers for surfactant monomers (N_S), CHAPS (N_CHAPS), and CHS (N_CHS) obtained from model fitting are also listed in Table 1 and Table S1. As expected, N_s systematically decreases and N_CHAPS systematically increases with increasing x_CHAPS/CHS due to the replacement of maltoside and thiomaltoside surfactant with CHAPS in solution. However, in all cases, N_CHS exhibits nonmonotonic behavior, increasing from approximately one to three CHS monomers at low δ to a maximum of seven to 18 CHS monomers per micelle, with the maximum in N_CHS occurring between δ = 0.55–0.85 for all systems studied. This observation is consistent with previous measurements on mixed C₁₂βSG₂/CHAPS/CHS micelles (33), and indicates that maximum solubilization of CHS within maltoside and thiomaltoside-based mixed micelles does not simply correspond to the largest amount of CHS in solution. Given how critical CHS addition is for hA₂aR activity, this result has significant implications for the conformational stability of GPCRs within mixed micelle PDCs (33).

hA₂aR-His₁₀ activity as a function of CHS content within PDCs

Previous experiments have shown that addition of CHS to hA₂aR-His₁₀ PDCs is critical for ligand-binding activity (28,31,33). To investigate whether the lack of hA₂aR-His₁₀ activity observed in certain surfactant systems could be a direct result of micelle morphology or CHS content, we analyzed mixed micelles at fixed and varied δ-values for their ability to stabilize the active state of hA₂aR-His₁₀.

PDCs made up of different chain length maltosides at the same δ-value were analyzed via point competition radioligand binding experiments to correlate hydrocarbon tail length with protein activity. The C₁₂βG₂/CHAPS/CHS mixed micelle system that was successfully used to retain hA₂aR activity had δ = 0.55, and generally contained 12 CHS molecules per micelle (Table 1), so this value of δ was chosen to compare activity of hA₂aR in mixed micelles containing structurally related maltosides. As shown in Fig. 3 A, some degree of ligand-binding activity was observed for all samples, irrespective of chain length, at δ = 0.55. Significantly greater activity for hA₂aR was achieved in C₁₁βG₂ mixed micelles compared with previous results (Fig. 1 A), where δ was 0.2. The relative receptor activity in this case was generally proportional to the length of the maltoside's hydrocarbon tail up to a chain length of 12 (Fig. 3 A). The highest degree of activity was achieved in mixed micelles containing C₁₂βG₂ and C₁₃βG₂ at δ = 0.55 (Fig. 3 A), consistent with the elevated protein activity observed for these surfactants in previous experiments.

Figure 3.

(A) Specific activity measurements for hA₂aR-His₁₀ reconstituted in different chain length maltoside surfactants, with CHAPS/CHS of hydrophobic chain length n, at a δ ratio of 0.55 (n = 10 designates C₁₀βG₂; n = 11 designates C₁₁βG₂; n = 12 designates C₁₂βG₂; n = 13 designates C₁₃βG₂). (B) Summary of specific radioligand binding for hA₂aR-His₁₀ reconstituted in mixed micelles over various δ ratios for C₁₁βG₂ (n = 11) and C₁₂βG₂ (n = 12). Inset: Specific activity for hA₂aR-His₁₀ determined via radioligand binding versus aggregation number of CHS (N_CHS) in both C₁₁βG₂ (light gray) and C₁₂βG₂ (dark gray) mixed micelles. Error bars represent the standard deviation from the average of three independent samples.

To evaluate hA₂aR-His₁₀ activity as a function of CHS content, we analyzed mixed micelles containing C₁₁βG₂ and C₁₂βG₂ over a range of δ-values. These micelles were chosen for comparison because although they are made up of virtually identical surfactant monomers that differ by only one hydrocarbon tail length, they displayed dramatic differences in their ability to stabilize activity of hA₂aR (Fig. 1). Consistent with previous results (33), no protein activity was seen in the single-component micelles lacking CHS (δ = 0; Fig. 3 B). Negligible protein activity was observed for hA₂aR reconstituted in C₁₁βG₂ micelles at low (0) or high (0.85) δ-values. Comparatively, a similar trend for mixed micelles containing C₁₂βG₂ was observed, although appreciable activity was seen at δ ratios of 0.1 and 0.2. For both systems, a ratio of 0.55 corresponded to the greatest degree of hA₂aR-His₁₀ activity as measured through radioligand binding (Fig. 3 B). However, compared with C₁₁βG₂ micelles, C₁₂βG₂ micelles were superior at promoting the native ligand-binding conformation of hA₂aR-His₁₀ at all δ-values.

It was previously demonstrated that hA₂aR activity in maltoside-based mixed micelles was directly correlated with the number of CHS monomers, N_CHS, contained within the corresponding empty micelles (33). Fig. 3 C shows a similar correlation obtained from the activity data in mixed C₁₁βG₂ and C₁₂βG₂ micelles using the values of N_CHS determined by SANS. As was previously observed for C₁₂βG₂ mixed micelles (33), the results indicate that hA₂aR activity is directly proportional to CHS content within micelles (Fig. 3 B) for both the maltoside chain lengths studied. However, it should be noted that the critical number of CHS monomers required for retention of significant activity for C₁₁βG₂ micelles is at least twice that required for C₁₂βG₂, consistent with the increased level of activity observed with increasing alkyl chain length.

Correlating hA₂aR activity with CHS content and surfactant morphology

With receptor activity obtained for a range of different mixed micelles, and the morphology and composition of those micelles known, we sought to correlate micellar properties with those that promote hA₂aR activity in PDCs. Specifically, because earlier experiments indicated that hA₂aR activity is dependent on hydrocarbon tail length (Fig. 1) but can change as a function of x_CHAPS/CHS (Fig. 3), we sought to identify optimal parameters that would promote receptor activity. Fig. 4, top, shows contour plots that illustrate how hA₂aR radioligand-binding activity correlated with the choice of surfactant vis-à-vis the maltoside hydrocarbon tail length (n) and relative CHAPS/CHS content in solution (x_CHAPS/CHS). Such a representation reveals a clear optimum in corresponding hA₂aR activity between a hydrocarbon tail length of 12–13 and δ ∼ 0.6, where ∼40% of the mixed micelles consists of maltoside or thiomaltoside surfactant and 60% consists of CHAPS/CHS (Fig. 4, top).

Figure 4.

Contour plots show the impact of micellar composition and structure on hA₂aR-His₁₀ activity reconstituted in C_nβG₂/CHAPS/CHS micelles. The plots compare hA₂aR-His₁₀ ligand-binding activity, where shades of gray indicate radioactive counts per minute (cpm) for (Top) δ versus hydrocarbon chain length n and (Bottom) aggregation number of CHS in micelles (N_CHS) versus minor axis of micelles (b).

In a previous study (33), we observed that although hA₂aR activity in mixed C₁₂βG₂ micelles varied systematically with the extensive variable δ, this dependence was fundamentally related to intensive properties of the micellar system, including both the average number of CHS monomers per micelle (N_CHS) and the minor axis of the ellipsoidal micelles (b), along which hA₂aR was believed to be oriented within PDCs. To extend this analysis to incorporate other surfactants, in Fig. 4, bottom, we show how N_CHS and b are correlated with hA₂aR activity for the C_nβG₂/CHAPS/CHS micelle systems considered in this study (with n = 10–13). From this analysis, we observe that hA₂aR activity increases monotonically with increasing N_CHS and is highest for > 12 CHS monomers per mixed micelle, whereas activity exhibits a mild nonmonotonic dependence on b, with optimal retention of activity observed in the range of b ∼ 15–18 Å (Fig. 4, bottom), or a total micelle thickness of 30–36 Å.

Discussion and Conclusions

The role played by specific surfactants and lipids in biophysical studies and high-resolution structure determination of membrane proteins is quite poorly understood, even though these entities form critical hydrophobic contacts with transmembrane domains that are necessary for structural stabilization in polar solvents (7). PDCs remain the most robust membrane-mimetic platform to facilitate structural studies of GPCRs via techniques based on AUC, calorimetry, NMR, and crystallization. Some GPCRs (e.g., rhodopsin and β₁-adrenergic receptor) have been crystallized in short chain length, sugar-based surfactants (eight to nine hydrocarbons) (7,47–49). By contrast, inactive forms of both human β₂-adrenergic and adenosine A₂a receptors have been crystallized in lipidic cubic phases, largely due to difficulties associated with their instability in mixed micelles (50,51). More recently, crystal structures for several GPCRs in an active conformation have been released (52–55), yet difficulties associated with crystallization have necessitated extensive modification of these receptors from their wild-type form. As such, our primary motivation for this study was to better understand how surfactant micelles stabilize the active conformation of wild-type membrane proteins, enabling rational selection of micellar PDC systems that maintain membrane protein conformational stability for study using several biophysical approaches, including high-resolution structure determination.

As with many mammalian membrane proteins, hA₂aR requires the presence of a lipid-like cholesterol analog to maintain conformational stability in vitro (28,31). In agreement with previous results (31,33), we found that the ligand-binding activity of hA₂aR was preserved only upon addition of the cholesterol analog CHS within micellar PDCs. In previous work, we identified that this effect was due in part to specific receptor-sterol interactions that require a minimum aggregation number of CHS within the PDC structure (33). This suggests that micellar PDC systems must be selected so as to allow for adequate solubilization of sterols within micelles, afforded here by the addition of the bile salt surfactant CHAPS to the surfactant micelles of interest. Furthermore, because sterols provide conformational rigidity in native membranes (11), CHS may also provide critical changes in fluidity of the resultant PDCs to promote the stability of hA₂aR; however, further studies of the dynamics of PDCs are required to determine whether this is so.

We found that the surfactant architecture, in terms of both the hydrophobic tail group and the hydrophilic headgroup, significantly affected the conformational stability of the receptor. For example, measurable hA₂a ligand-binding activity was preserved only for surfactants, including the alkyl maltosides, thiomaltosides, and ethoxylates, whose tail group consisted of an n-alkyl chain with length n ≥ 11. This is in agreement with the rule of thumb discussed above, according to which increased chain length better promotes protein stability in many systems (16,17). However, we note several results from this study that would not be predicted by, or in some cases refute, this general guideline. For example, we found that surfactants with cyclic tail group moieties, such as Cymal-5 and the Triton surfactants, failed to retain the ligand-binding activity of hA₂aR. Because both of these surfactants have hydrocarbon tails with n > 10, this effect suggests that the hydrocarbon architecture also plays a significant role in maintaining activity. Furthermore, hA₂aR exhibits a nonmonotonic dependence of ligand-binding activity on chain length and minor axis b for the C_nβG₂ surfactant series (Fig. 3 A). Overall, because PDC morphology is typically thought to adopt the conformation of a surfactant-sterol belt equatorial to the transmembrane domains of the membrane protein (5), these results indicate that the hydrophobic moiety must be long enough to ensure that the PDC aggregate structure is sufficient to span the width of the hydrophobic transmembrane domains of the receptor. However, longer hydrophobic moieties lead to a reduction of conformational stability, possibly due to forced interactions between the soluble loops of the protein and the hydrophobic micellar core.

We also found that the hydrophilic headgroup of the surfactant influenced the receptor activity. For example, of the micellar systems composed of n-alkyl surfactants having n = 12, we found a clear trend in ligand-binding activity, where C₁₂E₉ < C₁₂βSG₂ < C₁₂βG₂. This difference in behavior between the alkyl ethoxylate and maltoside-based micellar systems suggests that the specific chemistry of the surfactant group plays a significant role in maintaining an active conformation, possibly by influencing the conformational stability of the extracellular ligand-binding domain of the protein. For example, the sugar surfactants (C_nβG_m and C_nβSG_m) exhibit greater hydrophilic character than the ethoxylated surfactants due to increased hydrogen bonding of the headgroup (46), and also retained greater activity of hA₂aR. It is important to note, however, that surfactants with headgroups structurally related to the maltosides, such as the single-sugar glucoside, two-sugar Cymal series, and two-sugar maltoside surfactant families, produced quite different results in terms of receptor activity (Fig. 1). This difference between the two classes of sugar-based surfactants is in agreement with previous studies that revealed increased thermal stability of a truncated form of the hA₂aR in various maltoside-based micelles compared with those made from glucosides (51). Furthermore, the significant difference in ligand-binding activity between the maltoside and thiomaltoside surfactants suggests that the hydrophilicity of the headgroup was also important for determining the optimal PDC formulation. This stems from an increase in relative hydrophobicity of the thiol versus ether linkage (45),which is also reflected in the respective cmc values for both types of surfactants.

In the context of previous results (33), several chemical and structural features of surfactants can be postulated to promote the activity of hA₂aR, including increased solubilization of CHS, the self-assembly of optimal micellar and PDC geometries, and specific interactions of micellar constituents with the receptor itself. In our detailed studies of the morphology of C₁₂βG₂/CHAPS/CHS and C₁₂βSG₂/CHAPS/CHS micelles presented here, we found that the ability of the surfactant to solubilize CHS, as quantified by the aggregation number N_CHS within the mixed micelles (Table 1 and Table S1), generally increased with increasing n-alkyl tail length from n = 10 to n = 13. Correspondingly, the measured activity of hA₂aR significantly increased over this same range (Fig. 3 A). This resulted in a systematic increase in protein activity with increasing N_CHS that was nearly independent of micelle morphology (Fig. 4, bottom). The fact that there appeared to be a minimum value of N_CHS required to retain measurable activity suggests that specific sterol-receptor interactions must be present within the PDC environment, in agreement with previous reports of cholesterol-binding domains identified for hA₂aR (56). However, the observation that ligand-binding activity continued to increase for N_CHS > 10 suggests that CHS also provided nonspecific stability, possibly by providing a more rigid conformation of the PDC much like cholesterol does in native membranes (11,57).

By contrast, we found no significant difference between the ability of maltoside and thiomaltoside to solubilize CHS, which suggests that the difference in observed ligand-binding activity between the two surfactant types was solely due to differences in micelle morphology. In light of previous results (33), we first sought to determine any correlation between hA₂aR activity and the overall size of the micelles (independently of shape) when they were reconstituted in these systems. Some studies have inferred that overall micelle size directly influences membrane protein activity (17,58). However, in the case of hA₂aR, the R_g of the micelles did not correlate exclusively with protein activity. For example, C₁₂βG₂/CHAPS/CHS mixed micelles with δ = 0.55 as well as C₁₀βG₂/CHAPS/CHS mixed micelles with δ = 0.1 and 0.2 all have approximately identical R_g-values (Table 1). However, although hA₂aR activity was observed for the C₁₂βG₂ mixed system, it was not detected for C₁₀βG₂ mixed micelles with similar R_g (Fig. 1 A). Furthermore, hA₂aR activity showed no clear correlation with relative size, as the R_g-values for mixed micelle systems that retained activity ranged from 15.4 to 19.8 Å (Table 1, Table S1, and Fig. 3). Because we did not identify a clearly optimal R_g-value for hA₂aR, we explored other characteristics.

Generally, we found that mixed micelles that retained hA₂aR activity shared some obvious similarities, specifically in the minor axial dimension of the micelles (b). This parameter can be likened to the thickness of native mammalian bilayers, where GPCRs normally reside. Axial dimensions for previously used mixed micelles of C₁₂βG₂ with CHAPS and CHS at a δ ratio of 0.55 that maintained hA₂aR activity were determined to have b = 17.2 Å (33) (Table 1), such that the micelles were similar in thickness to the native membranes. By extending our analysis to other members of the maltoside surfactant series, and the related thiomaltoside series, we found that similar morphological characteristics were correlated with hA₂aR activity among different micellar systems (Figs. 2 and 4, bottom). Specifically, micellar systems that exhibited optimal activity had minor axes in the range of b = 15–18 Å (Fig. 4, bottom).

To summarize, we found several trends that are unexplained by the conventional rules of thumb for selecting surfactants to stabilize hA₂aR in vitro. These include the observations that 1), ligand-binding activity of hA₂aR depends nonmonotonically on alkyl chain length for the maltoside surfactant series; 2), the same set of molecules (for both C₁₁βG₂/CHAPS/CHS and C₁₂βG₂/CHAPS/CHS) can produce order-of-magnitude changes in conformational stability depending on the composition of the solution, even when micelles are in great excess; and 3), the maltoside surfactants retain superior stability compared with the more-hydrophobic thiomaltoside surfactants.

These trends can be easily rationalized in terms of their underlying dependence on aggregate structure and composition. Specifically, we found that hA₂aR activity was correlated with several key aspects of micellar architecture, including CHS content in the micelles, the relative hydrophobic thickness of micelles along the presumed orientation of the protein, and (to a lesser extent) surfactant headgroup chemistry, which together are necessary for activity of the receptor in micelles. In this study, hA₂aR activity was observed only in micelles with five or more CHS monomers, and activity increased proportionally with CHS content. A minor axial dimension of these micelles that closely mimics native mammalian bilayers appears to be optimal for retention of hA₂aR activity. The generalization of these findings to a number of sugar-based, nonionic surfactants represents a significant advance toward building a more rational approach for designing micellar systems for in vitro membrane protein stabilization. This approach may be further applied to the study of other membrane proteins, including GPCRs, to better understand and explore how mixed micelles support native protein structure.