Self-assembly of single integral membrane proteins into soluble nanoscale phospholipid bilayers


  • Timothy H. Bayburt,

    1. Department of Biochemistry, Urbana, Illinois 61801, USA
    2. The Beckman Institute, University of Illinois, Urbana, Illinois 61801, USA
    Search for more papers by this author
  • Stephen G. Sligar

    Corresponding author
    1. Department of Biochemistry, Urbana, Illinois 61801, USA
    2. Department of Chemistry, Urbana, Illinois 61801, USA
    3. The Beckman Institute, University of Illinois, Urbana, Illinois 61801, USA
    • Department of Biochemistry, University of Illinois, 116 Morrill Hall, 505 S. Goodwin, Urbana, IL 61801; fax: (217) 265-4073.
    Search for more papers by this author


One of the biggest challenges in pharmaceutical research is obtaining integral membrane proteins in a functional, solubilized, and monodisperse state that provides a native-like environment that maintains the spectrum of in vivo activities. Many of these integral membrane proteins are receptors, enzymes, or other macromolecular assemblies that are important drug targets. An example is the general class of proteins composed of seven-transmembrane segments (7-TM) as exemplified by the G-protein–coupled receptors. In this article, we describe a simple system for self-assembling bacteriorhodopsin, as a model protein containing 7-TM helices, with phospholipids to form a nanometer-scale soluble bilayer structure encircled by a 200 amino acid scaffold protein. The result is the single molecule incorporation of an integral membrane protein target into a soluble and monodisperse structure that allows the structural and functional tools of solution biochemistry to be applied.

Many of the exciting signal transduction and molecular machines of the living cell are membrane proteins. Indeed, these amphipathic receptors and enzymes represent a large class of important pharmaceutical targets. A major hurdle to understanding the structure and function of membrane proteins has been the unavailability of a simple and reproducible procedure for isolating these systems in a monodisperse form that maintains the phospholipid–protein native structure and functionality. In this article, we illustrate how single molecules of a representative of the seven-transmembrane helical proteins can be incorporated into soluble nanoscale phospholipid bilayers with controlled stoichiometry. Membrane proteins incorporated into these discoidal bilayer structures can avoid the aggregation and purification problems often inherent in structural and functional investigations of integral membrane proteins. In addition to maintaining functional integrity and stability in a solubilized phospholipid bilayer environment, it is possible to direct these nanobilayer structures to surface assembly for high throughput screening or biophysical investigations.

While investigating the structure and function of the human lipoproteins, we were struck by the beauty of a self-assembly process wherein phospholipid and apolipoproteins could form small disk-like complexes upon removal of detergent (Carlson et al. 1997). Such phospholipid disks have been studied extensively in the field of plasma lipoprotein research. The structure of the complex is a discoidal bilayer surrounded at its edges by amphipathic α-helices as determined by electron microscopy, small angle scattering, and atomic force microscopy (Forte et al. 1971; Atkinson et al. 1976; Wlodawer et al. 1979; Brouillette et al. 1984; Carlson et al. 1997). Realizing that the human apolipoproteins are not optimized for assembling disk structures and contained undesired sequences encoding receptor affinity, enzyme binding, and plasticity for the stabilization of the common spherical form of the lipoproteins, we engineered a class of amphipathic membrane scaffold proteins (MSPs) that were optimized for self-assembly of discoidal bilayers. A detailed description of several MSPs and their self-assembly was recently published (Bayburt et al. 2002). The MSP-assembled particle consists of a phospholipid bilayer core ∼70 Å in diameter containing 160 phospholipids stabilized at its edges by the amphipathic α-helical MSP. Upon further considering this simple process for generating a soluble monodisperse and uniform-sized phospholipid bilayer, which we term Nanodisc, we hypothesized that if an integral membrane protein target were included in the self-assembly protocol, one might effect the solubilization of this target into a native-membrane-like environment that is homogeneous at the single molecule level. A further advantage is that the MSP can be engineered with tags or chemically reactive groups for physical manipulation or attachment to various matrices.

We chose bacteriorhodopsin (bR) from Halobacterium salinarum as a target for incorporation into Nanodiscs. bR has a rich and extensive base of spectroscopic and biophysical characterizations of structure and activity and, as such, is an excellent model for the general class of seven-transmembrane receptors insofar as its transmembrane region consists of seven helices. Many G-protein–coupled receptors of the seven-transmembrane class of integral membrane proteins are thought to have functionally relevant multimeric forms. The ∼7-nm diameter phospholipid domain size of the Nanodiscs could allow one to control the oligomeric state of the target, and the spectroscopic delineation of monomeric bR versus the native purple membrane trimeric state is easily made (Heyn et al. 1975).

Results and Discussion

bR was incubated with dimyristoylphosphatidycholine (DMPC), cholate and a MSP containing eight 22-mer amphipathic helices together with an N-terminal histidine tag separated by a factor X protease cleavage site described as MSP1 in Bayburt et al. (2002). The histidine tag was not cleaved from the MSP1 used in this work. Detergent removal initiated the self-assembly process. The resulting mixture was eluted from a calibrated gel filtration column (Fig. 1A). The majority of the bR elutes as a single peak, while a minimal amount elutes as larger aggregates. Such elution profiles are typical of Nanodiscs produced in the absence of bR, in which the larger aggregates reflect the a priori mismatch between the amount of lipid in the mixture and the stoichiometry in a disk (Bayburt et al. 2002). At the correct stoichiometry in the reaction mixture, a homogeneous single size of discoidal bilayer is obtained (Bayburt et al. 2002).

The absorbance at 280 nm follows the total protein, whereas at 550 nm it monitors the presence of bR. The total protein peak trails that of Nanodiscs containing bR due to the slightly larger size of the bR-containing species. The peak can be reinjected and eluted as a homogeneous particle size (Fig. 1A, inset), indicating that the bR Nanodiscs exist in a stable complex. Omission of the MSP from the self-assembly mixture results in large aggregates that do not pass through a 0.22-μm filter. The self-assembly is a highly reproducible process that produces bR Nanodiscs in high yield, as all steps in self-assembly and subsequent purification had yields in the range of 70%–90%.

bR particles formed with MSP and phospholipids have a diameter of 92 Å as determined from calibrated size-exclusion chromatography (Fig. 1A). To directly observe the structure of these bilayer disks, we performed atomic force microscopy under aqueous buffer conditions. Clearly evident from Figure 1B is a homogeneous population of structures that are ∼10 nm in overall diameter.

We also studied bR incorporated into Nanodiscs by transmission electron microscopy. Phosphotungstate-stained samples of discoidal phospholipid bilayer structures formed by using purified plasma apolipoproteins and synthetic phospholipids typically exist as rouleaux, that is, stacked disk structures having a repeat distance equivalent to that of a phospholipid bilayer (Forte et al. 1971; Brouillette et al. 1984). Similar rouleaux are formed by negative-stained bR samples (Fig. 1C). The repeat distance for the bR disk structures has a narrow distribution centered at 5.9 nm. In contrast to the ∼10-nm sizes obtained by size-exclusion chromatography and atomic force microscopy (AFM), the diameters of rouleaux in EM images are heterogeneous and average 16 nm. The negative staining procedure might cause fusion of the disks, leading to larger overall diameter and heterogeneity. Interestingly, a disk fusion process on a surface has been documented by using atomic force microscopy of phospholipid bilayer disks formed with apolipoprotein A-I, which also resulted in formation of a 16-nm-diameter disk population from a 10-nm population (Carlson et al. 1997). Although the exact nature of the disk fusion and increase in diameter is unknown, the diameters of disks composed of apolipoproteins A-I are known to be a function of the number of phospholipids and associated lipoprotein molecules in the complex (Swaney 1980).

To quantitate the physical composition of the resulting target incorporated bilayers, bR Nanodiscs were purified by affinity chromatography for determination of the stoichiometry of bR, DMPC, and MSP. The extinction coefficient of bR solubilized into Nanodiscs at 550 nm was determined to be 52,800 ± 600 M−1cm−1 by the method of retinal titration (Rehorek and Heyn 1979). This value agrees well with the 19% decrease in absorbance observed upon monomerization of bR with Triton X-100, which results in an estimated extinction of 51,000 M−1cm−1 (Dencher and Heyn 1978). Based on the change in absorbance at 280 nm upon retinal binding (50,100 ± 300 M−1cm−1) and the calculated molar absorption coefficient at 280 nm of retinal-free protein and the MSP based on amino acid content (24,700 M−1cm−1 in disk buffer), the purified Nanodiscs have a stoichiometry of 1.9 ± 0.2 MSPs to one bR. Radiolabeled DMPC was used to measure 163 ± 8 molecules of DMPC per two MSP molecules. The natural lipids of the purple membrane were present during disk formation. Approximately seven lipids per bR are found in purple membrane (Kates et al. 1982) and may exist in the disks, possibly in association with bR.

An excess of MSP and phospholipid relative to bR was intentionally used to allow, in theory, only one bR molecule per disk. The retinal chromophore of bR provides a convenient indicator of the association state of the photoreceptor. The peak of dark-adapted bR in MSP-assembled Nanodiscs is 550 nm, which shifts slightly to higher wavelength upon light adaptation. A similar absorbance maximum and light adaptation was found for bR incorporated into liposomes. The absorbance maximum in DMPC bilayers is blue-shifted relative to the crystalline form in purple membrane, indicative of a monomeric species. A better indication of the oligomerization state of bR can be obtained from circular dichroism spectroscopy. Circular dichroism spectra of bR in Nanodiscs do not show the characteristic bilobed spectrum of purple membrane (Fig. 2A; Heyn et al. 1975), indicating that bR is in a monomeric state in Nanodisc structures. A disordered trimeric structure would also lack a bilobed spectrum, but the presence of a trimer would be incompatible with the measured stoichiometry and the known composition of Nanodiscs.

Critical to the utility of Nanodiscs in solubilizing integral membrane proteins is the resulting activity of the incorporated target. bR in its native form displays a photointermediate characterized by an absorbance maximum at 410 nm. The light minus dark spectrum of bR Nanodiscs shows an absorbance maximum near 410 nm that is similar in magnitude to bR in liposomes (Fig. 2B).

We also measured the binding of all-trans-retinal to bacterioopsin Nanodiscs and found a dissociation constant of 3 × 10−7 M, very close to the value of 3.6 × 10−7 M measured for purple membrane (Rehorek and Heyn 1979). According to the fit of Figure 3, 93% of bR could be reconstituted based on the initial concentration of bacterioopsin. Thus, the majority of bacterioopsin in Nanodiscs is functional with respect to cofactor binding, and the cofactor binding site is not significantly perturbed in the Nanodisc environment.

To determine the absolute orientation of the phospholipid bilayer in Nanodiscs, fluorescence-detected linear dichroism (LD) measurements were made by using fluorescent lipid probes with known binding orientation. The first probe, DiI, is known to orient with the long axis of its indocarbocyanine group nearly parallel to the bilayer surface (Axelrod 1979). Measurements of the tilt angle from membrane normal by using arachidate-doped films have yielded values of ∼75 degrees (Edmiston et al. 1996; Tronin and Blasie 2001). A second probe containing a BODIPY group should orient with its long axis more normal to the plane of the bilayer (Karolin et al. 1994).

The organization of the phospholipids present in the disk particles is expected to be the major influence on the orientation of lipidic probes. Nanodiscs, and phospholipid bilayers in general, are known to interact with and sit flat on planar surfaces such as glass or mica. LD measurements were made by using the fluorescent probes in a glass-supported phospholipid bilayer and adsorbed Nanodiscs on a glass surface (Fig. 4). The values of the dipole tilt angle of DiI in a planar bilayer and in adsorbed disks were 74° and 76°, respectively, in good agreement with published values, and support the model of a disk structure with a bilayer domain oriented on a glass surface. A similar value for the dipole tilt angle of DiI is found for bR Nanodiscs (76°; Fig. 4) also consistent with this model. Thus, DiI maintains the same orientation in phospholipid bilayers and bR Nanodiscs. Likewise, the tilt angles of the BODIPY probe are the same in bilayers, Nanodiscs, and bR Nanodiscs (49, 51, and 50°, respectively). The fact that the probes have the same orientation in bR disks and planar bilayers is evidence of a bilayer organization of phospholipid in bR Nanodiscs.

Through the use of electron microscopy, AFM, size-exclusion chromatography, and bilayer probe orientation measurements, we have demonstrated that bR will self-assemble with phospholipid and MSP to form a discoidal nanobilayer structure 9 to 10 nm in overall diameter, similar in size and stoichiometry to a characterized Nanodisc with no protein incorporated (Bayburt et al. 2002). Earlier work with plasma lipoproteins indicate that the lipoprotein is arranged at the periphery of a disk-like phospholipid core (Atkinson et al. 1976; Wlodawer et al. 1979). This evidence indicates the structure depicted in Figure 5.

The self-assembly process reported in this article describes a broadly applicable and gentle means for the solubilization and purification of integral membrane proteins. Such methodologies would greatly aid the isolation of functional protein components of cellular membranes for biochemical and biophysical study. Nanodiscs provide a vehicle for research into membrane proteins, and the technology is well suited to the needs for physical manipulation, single biomolecule research, and adaptability to drug screening applications.

Materials and methods

Formation of bR Nanodiscs

MSP was produced as described and stored at 4°C (Bayburt et al. 2002). The extinction coefficient for MSP at 280 nm in buffer (10 mM Tris at pH 7.4, 0.1 M NaCl, and 0.01% NaN3) was determined to be 24,700 M−1cm−1 by using a known concentration based on calculated extinction (Gill and von Hippel 1989). bR was obtained from Sigma as a lyophilized powder, or was grown and purified as described (Oesterhelt and Stoeckenius 1974) and stored as a lyophilized powder. bR in purple membrane was measured by using an extinction coefficient of 62,700 M−1cm−1 in the light-adapted form (Rehorek and Heyn 1979). bR was initially solubilized with 4% w/v Triton X-100 according to published procedures (Dencher and Heyn 1982). MSP1 stock solution (typically ∼200 μM) and a DMPC/cholate mixture (50/100 mM in buffer) were added to bR (typically ∼80 μM) to give mole ratios of 2 : 160 : 0.2. After 1 h at room temperature, detergent was removed by treatment for 3–4 h with 400 mg wet Biobeads SM-2 (BioRad) per milliliter of solution, with gentle agitation to keep the beads suspended (Levy et al. 1990). The disk preparations were then purified by gel filtration chromatography on a Superdex HR 10/30 column to remove small amounts of aggregates. In some cases, biotinylated bR (see below) and 3H-DMPC, specific activity 25.4 mCi/mole and synthesized as described (Hanel et al. 1993), were used to simplify purification and for determination of the stoichiometry of lipid, bR, and MSP per Nanodisc.

Biotinylation of bR and affinity chromatography

bR was biotinylated in Triton X-100 solubilized form by using a 20-fold excess of succinimidyl-6′-(biotinamido)-6-hexanamido hexanoate (Pierce) in 25 mM phosphate (pH 6.9) at room temperature for 1 h followed by dialysis against 25 mM phosphate (pH 6.9) containing 2% w/v Triton X-100. Biotinylated bR was found to contain one to two biotin groups by using a 2-hydroxyazobenzene-4′-carboxylic acid/avidin assay (Green 1970). Affinity chromatography was performed with immobilized monomeric avidin (Promega) by using the manufacturer's protocol.

Retinal reconstitution

Retinal titrations were performed according to published procedures after photobleaching of bR Nanodiscs in the presence of hydroxylamine (Rehorek and Heyn 1979). bR (19.7 μM) was present well above the dissociation constant, and the initial linear titration values were used to obtain the extinction coefficient (ε550) for bR in Nanodiscs.

Atomic force microscopy

Images of bR Nanodiscs were obtained with a Digital Instruments Nanoscope IIIa using scanner “A”. Samples were prepared by incubating 10 μL of an affinity-purified sample (∼1 μM bR) and 10 μL of imaging buffer (10 mM Tris at pH 8, 0.15 M NaCl, and 10 mM MgCl2) on freshly cleaved mica for 10–20 min, after which the sample was positioned in the fluid cell and rinsed with imaging buffer. Imaging was done in contact mode with sharpened microlevers (Veeco Instruments) by using a rectangular cantilever with a nominal spring constant of 0.02 N/m.


UV/Vis spectra were recorded on Varian Cary Bio 300 at ambient temperature. Circular dichroism measurements were recorded on a Jasco J-720 spectropolarimeter at ambient temperature. Spectra taken with steady-state illumination were measured by using a diode array spectrophotometer with excitation by a 543.5-nm 0.3-mW HeNe laser (Melles Griot). The beam was expanded and directed through the sample perpendicular to the measuring beam. Liposomes were formed as for disks with the omission of MSP. The disk sample contained 9.5 μM bR, and the liposome sample contained 4 μM bR. Spectra were measured at 15°C.

Transmission electron microscopy

Samples were prepared by dialyzing into 125 mM ammonium acetate (pH 7.4). A 375-nM bR disk sample was negative-stained by mixing 1 : 1 with 2% phosphotungstate (pH 7.0) and placing a drop on carbon-formvar–coated grids for 30 sec (Forte and Nordhausen 1986). Electron micrographs were obtained by using a Philips CM200 transmission electron microscope operated at 200 kV.

Fluorescence-detected LD measurements

The chopped beam of a 543.5-nm HeNe laser (Melles Griot, model 05-LGP-173) was passed through a 544-nm interference filter to remove glow discharge radiation. Polarization was rotated with 1° resolution by an achromatic half-wave plate (Melles-Griot). A hemicylindrical glass prism coupled to a glass slide was used as the total internal reflection element. Slides were cleaned by soaking in pirhana (3 : 1 H2SO4-H2O2). A CoverWell perfusion chamber (Grace Bio-Labs) attached to the sample side was used as a transparent cell for the sample buffer. An incident angle of 80° was used. Emitted light was collected with a 10×, 0.25 N.A. objective and selected with a 580-nm interference filter (Oriel, model 59390). The photomultiplier output was fed into a lock-in amplifier (Stanford Research, SR850 DSP). The fluorescent probes BODIPY 558/568 C12 and DiI(C16) were obtained from Molecular Probes and added to samples from concentrated ethanol stocks to give 0.01 to 0.02 mole fraction probe. Samples were allowed to adsorb to the glass slide for ∼30 min and rinsed with buffer before measurements were taken. The experimental geometry and equations for evanescent light absorbtion are as described (Thompson et al. 1984; Thompson and Burghardt 1986). Fluorescence intensity as a function of incident polarization was corrected for overall intensity variation due to reflection from the optics by using measurements of randomly oriented rhodamine in solution.

Figure Figure 1..

Gel filtration chromatography of bR Nanodiscs after self-assembly. (A) The chromatogram of the MSP disk formation mixture after detergent removal shows absorbance at 550 nm (solid line) and 280 nm (dotted line). K indicates the partition factor (Velution − Vvoid) / Vinternal. The inset shows reinjection of MSP self-assembled Nanodiscs with monitoring the absorbance at 550 nm. (B) An AFM image of purified bR Nanodiscs on mica. (C) TEM micrograph of phosphotungstate-stained bR Nanodiscs.

Figure Figure 2..

Optical spectroscopy of bR in Nanodiscs. (A) CD spectra were obtained for disks formed with MSP (solid line) and, for reference, native purple membrane (broken line). (B) Difference spectrum (steady-state illumination minus dark) of bR Nanodiscs and bR proteoliposomes. The absorbance of the bR liposome sample (dashed line) has been normalized to the concentration of bR in the disk sample (solid line, 9.5 μM).

Figure Figure 3..

Titration of bacteriopsin disks with all-trans-retinal. The initial linear portion of the curve was used to obtain the extinction coefficient at 550 nm. The curve is a fit to a quadratic binding equation, yielding a dissociation constant of 0.3 μM.

Figure Figure 4..

Fluorescence-detected linear dichroism of probes in surface-oriented assemblies as a function of angle of polarization. Circles indicate DMPC bilayer; squares, DMPC disks; and diamonds, bR DMPC disks. Closed symbols indicate DiIC16(3); open symbols, BODIPY 558/568 C12. Fits of the probe transition dipole polar angle (θ) for bilayer, disk, and bR disk are 74°, 76°, and 76° (DiIC16) and 49, 51° and 50° (BODIPY), respectively.

Figure Figure 5..

Model of the organization of a Nanodisc. Cartoon shows a cutaway view of a 10-nm-diameter Nanodisc composed of phospholipid (blue) and MSP (orange) containing a molecule of bR (purple). bR was constructed by using coordinates obtained from the Protein Data Bank (PDB ID 1C3W; Luecke et al. 1999; Berman et al. 2000).


This work is supported by grants from National Institutes of Health GM33775 and GM63574. We thank J.K. Lanyi (University of California, Irvine) for a gift of purple membrane preparation and S. C. Hartsel (University of Wisconsin, Eau Claire) for H. halobium JW-3 cells. The CD measurements reported in this paper were performed at the Laboratory for Fluorescence Dynamics (LFD) at the University of Illinois at Urbana-Champaign (UIUC). The LFD is supported jointly by the Division of Research Resources of the National Institutes of Health and UIUC. The publication costs of this article were defrayed in part by payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 USC section 1734 solely to indicate this fact.