Lipid Nanodiscs as a Tool for High-Resolution Structure Determination of Membrane Proteins by Single-Particle Cryo-EM

Abstract

The “resolution revolution” in electron cryomicroscopy (cryo-EM) profoundly changed structural biology of membrane proteins. Near-atomic structures of medium size to large membrane protein complexes can now be determined without crystallization. This significantly accelerates structure determination and also the visualization of small bound ligands. There is an additional advantage: the structure of membrane proteins can now be studied in their native or nearly native lipid bilayer environment. A popular lipid bilayer mimetic are lipid nanodiscs, which have been thoroughly characterized and successfully utilized in multiple applications. Here, we provide a guide for using lipid nanodiscs as a tool for single-particle cryo-EM of membrane proteins. We discuss general methodological aspects and specific challenges of protein reconstitution into lipid nanodiscs and high-resolution structure determination of the nanodisc-embedded complexes. Furthermore, we describe in detail case studies of two successful applications of nanodiscs in cryo-EM, namely, the structure determination of the rabbit ryanodine receptor, RyR1, and the pore-forming TcdA1 toxin subunit from Photorhabdus luminescens. We discuss cryo-EM-specific hurdles concerning sample homogeneity, distribution of reconstituted particles in vitreous ice, and solutions to overcome them.

Introduction

It is predicted that ~ 23% of human genes code for membrane proteins (Uhlén et al., 2015). Because of their surface accessibility and central importance in many cellular mechanisms, they are attractive drug targets. Thus, in 2006, membrane proteins constituted more than 60% of all known drug targets (Overington, Al-Lazikani, & Hopkins, 2006).

Structural information is key for understanding the interaction between membrane proteins, in particular receptors, and small drug molecules in atomic detail. However, structural investigations on membrane proteins are also important to understand essential cellular mechanisms, including signaling, transport, and energy conversion at the molecular level.

While some structures have been obtained by electron crystallography (Raunser & Walz, 2009) and solution NMR (Hiller & Wagner, 2009), the majority of membrane protein structures has been determined by X-ray crystallography (Vinothkumar & Henderson, 2010). Although the structure of the first membrane protein was solved in 1985 (Deisenhofer, Epp, Miki, Huber, & Michel, 1985), membrane proteins have remained a challenge for structural biology over the past decades. Besides difficulties in obtaining ordered crystals in the case of crystallography, major bottlenecks have been the expression, solubilization, and purification of sufficient amounts of recombinant and biologically active membrane proteins. In recent years, significant progress has been made to overcome these chalenges, and many structures of important membrane proteins have been solved. These include transporters, respiratory complexes, G-protein-coupled receptors (GPCRs), ion channels, and adventitious membrane proteins (Vinothkumar & Henderson, 2010).

However, in many cases, in particular for mammalian membrane proteins, crystals of sufficient quality have not been obtained. Recently, the introduction of a new generation of direct electron detectors (McMullan, Faruqi, Clare, & Henderson, 2014) and the parallel development of image-processing software (Brilot et al., 2012, Li et al., 2013) increased the resolution limit of single-particle electron cryomicroscopy (cryo-EM) to atomic resolution. Cryo-EM of the post-“resolution revolution” era, as Werner Kühlbrandt coined it (Kühlbrandt, 2014), allows now for the direct structure determination of proteins to near-atomic resolution (between 3 and 4 Å) without crystallization. Consequently, this new development has had a major impact on structural biology, and many structures of mammalian protein complexes that resisted crystallization so far have been determined in the last 3 years. Good examples are γ-secretase (Bai et al., 2015), respiratory complex I (Fiedorczuk et al., 2016, Zhu et al., 2016) and supercomplexes (Letts et al., 2016, Wu et al., 2016), ATP synthase (Allegretti et al., 2015), different ions channels (Liao et al., 2013, Wu et al., 2016), and transporters (Kim et al., 2015).

As for X-ray crystallography, pure protein samples are needed for single-particle cryo-EM studies. Typically, purification protocols for membrane proteins start with extracting the transmembrane proteins from membranes by solubilizing them in detergents. Protein purification by means of different techniques, mainly affinity and size-exclusion chromatography (SEC), are also performed in the presence of detergent to prevent the precipitation of the protein. Detergent-solubilized membrane proteins can be directly used for single-particle cryo-EM. Since cryo-EM allows observing individual unrestrained protein molecules embedded in a thin layer of amorphous ice, there are no restrictions on choosing a suitable detergent, which is usually not the case in crystallography. For the same reason, detergent-free systems can also be used to study the structures of membrane proteins. Among these systems are polymers such as amphipols (Popot et al., 2011) and styrene–maleic acid (SMA) copolymers (Scheidelaar et al., 2015), which can substitute detergents, or lipid-based systems, such as lipid nanodiscs (Schuler, Denisov, & Sligar, 2013), a saposin-lipoprotein nanoparticle system (e.g., Salipro®, EP 2745834 B1) (Frauenfeld et al., 2016), and even liposomes (Wang & Sigworth, 2010).

Since detergents must be used at concentration levels above the critical micelle concentration (CMC) to prevent precipitation of the protein, their concentration in the solution is high enough to decrease the contrast of cryo-EM images (Schmidt-Krey & Rubinstein, 2011). This is not the case when reconstituting membrane proteins in amphipols. Therefore, amphipols have been successfully used for the determination of the first high-resolution structures of the TRPV1 channel (Liao et al., 2013) and γ-secretase (Bai et al., 2015). In general, amphipols are popular polymers for reconstituting membrane proteins in single-particle cryo-EM (Popot et al., 2011). Although this process is more straightforward than reconstituting proteins into lipid environments, amphipols as well as detergents have the disadvantage that the system does not contain lipids. There are a number of reasons why studying the structure of membrane proteins in a lipid environment has advantages over examining them in detergent.

First of all, the lipid bilayer is the native environment for membrane proteins, and it has significantly different physical properties than its mimetic detergent micelle (Zhou & Cross, 2013). In contrast to planar lipid bilayers, detergent micelles typically have a spherical shape and differ considerably in their dielectric constant and electrostatic potential. In some cases these differences can have severe consequences, resulting in structures of nonnatively folded membrane proteins when crystallized from detergent solution (Cross et al., 2013, Zhou and Cross, 2013). However, the effects are normally milder and while the proper fold is preserved when the membrane protein is solubilized in detergent or reconstituted in polymers, the protein dynamics, which are more sensitive to the environment than the structure, are very often influenced or altered. For example, the absorption spectrum of bacteriorhodopsin shifts upon solubilization in detergent and its photocycle kinetics are changed (Lam & Packer, 1983), even if its structure remains virtually unaltered. A more recent vivid example is related to the functional dynamics of the rabbit ryanodine receptor 1 (RyR1). Solubilized in Tween-20 (Yan et al., 2015) or CHAPS (Zalk et al., 2015), the structure of the channel is similar to the one determined in a lipid environment (Efremov, Leitner, Aebersold, & Raunser, 2015). However, in Tween-20 it is locked in a closed state and does not open under conditions that induce the full opening of the channel in lipid membranes (Bai, Yan, Wu, Li, & Yan, 2016).

The function of membrane proteins often depends on the direct interaction with lipids (Phillips, Ursell, Wiggins, & Sens, 2009). A layer of annular lipids surrounds the proteins. These lipids are in direct contact with the hydrophobic surface of the protein and tightly seal the membrane at the protein–membrane interface. Often they take distorted arrangements to adapt to the rough surface of the protein (Lee, 2004) and play an important role in stabilizing membrane protein oligomers (Gupta et al., 2017). However, most of the annular lipids are removed during the purification process, and in most cases they are not reintroduced afterward for crystallography or single-particle cryo-EM with detergents or polymers. In many crystal structures (Yeagle, 2014) and also recent cryo-EM structures (Liao et al., 2013) though, some very tightly bound lipid molecules remain bound during purification and can be resolved in the structures.

The optimal way to purify a membrane protein together with its natural annular lipids is to use detergent-free purification procedures that make use of SMA copolymers (Dörr et al., 2014, Dörr et al., 2016) or nanodiscs (Civjan, Bayburt, Schuler, & Sligar, 2003). The development of these techniques is still in its infancy, and it has not yet been broadly applied. The most common approach is thus to purify the protein using detergents and reintroduce lipids afterward. In most cases, lipid composition can be varied and adjusted to mimic the natural membrane environment or to identify key lipids and lipid compositions. This has been done, for instance, for bacteriorhodopsin (Lee et al., 2015), the ABC transporter MbsA (Kawai, Caaveiro, Abe, Katagiri, & Tsumoto, 2011), or cytochrome P450 reductase (Das & Sligar, 2009).

Among the different available options for reconstituting membrane proteins in a lipid bilayer, the lipid nanodisc system currently presents the best-understood and -characterized method. Lipid nanodiscs were developed in the laboratory of Stephen Sligar (Bayburt & Sligar, 2003). They are composed of a patch of lipid bilayer, which is surrounded by membrane scaffold protein (MSP). MSP is a derivative form of apolipoprotein A-1 (Apo-A1) (Ritchie et al., 2009) and composed of short amphipathic helices. In an assembled nanodisc, two MSPs arrange in a parallel or antiparallel manner, forming a belt around the hydrophobic region of the lipid bilayer that includes a few hundred lipid molecules (Fig. 1). Constructs of MSP with various numbers of amphipathic helices have been designed so that different sizes of nanodiscs can be obtained reaching diameters between ~ 10 and ~ 17 nm. This corresponds to ~ 100–600 lipids per nanodisc (Ritchie et al., 2009, Schuler et al., 2013). While MSP defines the size limits of the nanodiscs, as can be observed by negative stain electron microscopy (EM), the actual dimensions of the reconstituted discs vary significantly (Nasr et al., 2017). A recently reported method to generate covalently circularized nanodiscs using sortase produced significantly more homogeneous nanodiscs with diameters of up to 80 nm (Nasr et al., 2017).

Nanodiscs have shown to be potentially useful in studying membrane protein complexes, in particular fragile temporary signaling complexes, like those formed by GPCRs (Leitz, Bayburt, Barnakov, Springer, & Sligar, 2006). Detergents can disrupt the hydrophobic interactions involved in protein–protein and protein–lipid–protein interactions. For example, the β₂-adrenergic receptor, where mainly two palmitic acid and two cholesterol molecules mediate the interactions between two protomers of the natural dimer, binds only to G-proteins when reconstituted in nanodiscs but not when solubilized in detergents (Leitz et al., 2006). Other examples are the proton pump bacteriorhodopsin (Johnson et al., 2014) the ABC transporters MsbA (Kawai et al., 2011) and MalK (Bao & Duong, 2014). It has been shown that the functional properties of these proteins reconstituted in nanodiscs are similar to those in native membranes (Bao and Duong, 2014, Johnson et al., 2014, Kawai et al., 2011). Nanodiscs have been successfully used to determine high-resolution structures of membrane proteins using solution NMR (Hagn & Wagner, 2015) and single-particle cryo-EM (Efremov et al., 2015, Frauenfeld et al., 2011, Gao et al., 2016, Gatsogiannis et al., 2016, Shen et al., 2016).

Although nanodiscs are capable of mimicking the situation in a native lipid membrane, some biophysical properties of the lipids are slightly different and have to be taken into account when working with nanodiscs. For instance, the phase transition temperatures of lipids are higher and the thermal expansion coefficients of lipids are lower than in liposomes (Denisov, McLean, Shaw, Grinkova, & Sligar, 2005). Molecular dynamics simulations showed that the lipid layer that is in direct contact with MSP is perturbed due to the hydrophobic mismatch between MSP and lipids. However, 15–20 Å away from MSP, corresponding to two layers of lipid molecules, the properties of the lipids were similar to those in a planar bilayer (Schuler et al., 2013, Shih et al., 2005). Indeed, it has been shown for the ABC transporter MsbA that its activity is higher in larger nanodiscs containing more lipids (Kawai et al., 2011). EPR experiments with labeled lipids indicated that lipids within nanodiscs are generally more ordered than in liposomes and to some extent resemble a lipid bilayer in the presence of cholesterol (Stepien, Polit, & Wisniewska-Becker, 2015).

Examples of single-particle cryo-EM structures solved of membrane proteins reconstituted into lipid nanodiscs include the early work on the complex between the bacterial ribosome and the SecYE translocon (Frauenfeld et al., 2011), the muscular isoform of the rabbit ryanodine receptor (RyR1) (Efremov et al., 2015), more recent high-resolution studies of the Tc toxin from Photorhabdus luminescens (Gatsogiannis et al., 2016), the transient receptor potential channels TRPV1 (Gao et al., 2016) and PKD2 (Shen et al., 2016) (Fig. 2; Table 1). These examples demonstrate the feasibility of high-resolution 3-D structural analysis of membrane proteins embedded in nanodiscs, even with relatively small extracellular domains, as in the case of the transient receptor potential channels. In spite of the heterogeneous shape of the nanodisc, the transmembrane region of the proteins is one of the best-resolved parts of the structures.

Below, we first describe general considerations and approaches that need to be taken into account when reconstituting membrane proteins into lipid nanodiscs for single-particle cryo-EM. Next, we provide detailed protocols of reconstitution and cryo-EM sample preparation for two proteins, the rabbit ryanodine receptor RyR1 and the TcdA1 subunit of the Tc toxin complex from P. luminescens.