Synthetic Biology-Based Solution NMR Studies on Membrane Proteins in Lipid Environments

Abstract
Although membrane proteins are in the focus of biochemical research for many decades the general knowledge of this important class is far behind soluble proteins. Despite several recent technical developments, the most challenging feature still is the generation of high-quality samples in environments suitable for the selected appli- cation. Reconstitution of membrane proteins into lipid bilayers will generate the most native-like environment and is therefore commonly desired. However, it poses tremen- dous problems to solution-state NMR analysis due to the dramatic increase in particle size resulting in high rotational correlation times. Nevertheless, a few promising strat- egies for the solution NMR analysis of membrane inserted proteins are emerging and will be discussed in this chapter. We focus on the generation of membrane protein samples in nanodisc membranes by cell-free systems and will describe the characteristic advantages of that platform in providing tailored protein expression and folding envi- ronments. We indicate frequent problems that have to be overcome in cell-free synthe- sis, nanodisc preparation, and customization for samples dedicated for solution-state NMR. Detailed instructions for sample preparation are given, and solution NMR approaches suitable for membrane proteins in bilayers are compiled. We further discuss the current strategies applied for signal detection from such difficult samples and describe the type of information that can be extracted from the various experiments. In summary, a comprehensive guideline for the analysis of membrane proteins in native-like membrane environments by solution-state NMR techniques will be provided.

1.INTRODUCTION
Within the last 20 years, techniques to characterize membrane proteins improved dramatically and especially examples of their structural investiga- tion increased nearly exponentially (membrane proteins of known structure database: http://blanco.biomol.uci.edu/mpstruc/). The majority of known structures have been determined by X-ray crystallography and solution NMR spectroscopy, but with recent developments in detector technology the determination of high-resolution structures (Ku€hlbrandt, 2014) by cryo- EM is catching up. The drastic growth of structural data can be mainly attrib- uted to improvements in measurement techniques and methodological advancements regarding protein production, purification, and sample prep- aration. High-end protein engineering (Serrano-Vega, Magnani, Shibata, & Tate, 2008), stabilization by small molecules or antibody fragments (Kim et al., 2015; Rasmussen et al., 2011, 2007), in meso crystallization (Caffrey, 2015), or lipid nanodiscs (Bayburt & Sligar, 2003) are only some of these developments. A general tendency is to focus on natural-like membrane envi- ronments for membrane protein research as more and more studies focus on the role of lipids for membrane protein structure and stability (Barrera, Zhou, & Robinson, 2013; Gupta et al., 2017; Henrich et al., 2017; Saliba, Vonkova,& Gavin, 2015). In particular, the use of nanolipid particles as solubilization environment for membrane proteins in structural studies has set new standards (Denisov & Sligar, 2016; Efremov, Gatsogiannis, & Raunser, 2017; Nikolaev et al., 2017). This is not surprising as detergents can have a dramatic influence on the overall structural integrity or dynamics of membrane proteins (Dehez, Schanda, King, Kunji, & Chipot, 2017). Interestingly, the most preferred detergents for membrane solubilization and protein purification like n-dodecyl β-D-maltoside (He, Wang, & Yan, 2014) preserve a high amount of lipids still attached to the extracted membrane protein (Ilgu€ et al., 2014).

In this respect sample preparation techniques that avoid detergent con- tacts and allow subsequent analysis within natural membranes or membrane mimetics are desirable. Here, the cotranslational insertion of membrane pro- teins into defined membranes by synthetic approaches offers a remarkable advantage as any detergent contact is avoided during the direct folding of the nascent protein chain into a provided lipid bilayer (Harris et al., 2017). Nanodiscs are extremely well suited for this production process via cell-free expression systems (Roos et al., 2014). The extraordinary features and physicochemical properties of nanodiscs streamline protein purifica- tion and enable a sophisticated analysis by many state-of-the-art techniques (Denisov & Sligar, 2017). Control of the bilayer composition allows infor- mation about the effect of specific lipids (Dawaliby et al., 2016; Henrich et al., 2016; Rues, Do€tsch, & Bernhard, 2016) or about the influence of general bilayer characteristics on membrane protein structure or stability (McClary, Sumida, Scian, Pac¸o, & Atkins, 2016) to be obtained. In solu- tion NMR spectroscopy, short-chain detergents have almost exclusively been used to provide a hydrophobic environment as they produce reasonable small particles that allow sufficiently fast tumbling. Bicelles were the first bilayer containing environment that gave high-resolution spectra of mem- brane proteins (Morrison et al., 2012). While they are widely used in solution NMR (Du€rr, Gildenberg, & Ramamoorthy, 2012), their classical preparation still requires treatment of the solubilized membrane protein with detergent (Table 1). Nanolipoprotein particles only slowly found their entrance into the NMR field (Glu€ck et al., 2009) since the first nanodiscs with a diameter of ~10 nm still were of unfavorable size (Denisov, Grinkova, Lazarides, & Sligar, 2004). Several recent improvements in nanodisc design (Hagn et al., 2013; Nasr et al., 2017) and sample preparation (Laguerre et al., 2016) initiated a broader usage of nanodiscs for solution NMR (Table 1).In this chapter, we describe the technical details of a complete pipeline from cell-free membrane protein production to NMR analysis in membrane environment. This involves the setup of customized cell-free protein pro- duction (Schwarz et al., 2007) allowing specialized labeling strategies (Hein, Lo€hr, Schwarz, & Do€tsch, 2017; Laguerre et al., 2016; Lo€hr et al., 2012; Lo€hr, Tumulka, Bock, Abele, & Do€tsch, 2015), the cotranslational
membrane protein insertion into defined nanodiscs (Henrich et al., 2016; Rues et al., 2016), and their transition into small bicelles for solution NMR measurements (Laguerre et al., 2016). The presented workflow will serve as a guideline to alleviate solution NMR characterizations of mem- brane proteins in native-like environments and potentially critical steps will be indicated.

2.BASIC CELL-FREE EXPRESSION SYSTEM
Expression systems based on cellular lysates have continuously been optimized for approximately the past 20 years. Systems utilizing lysates from different organisms including prokaryotic as well as eukaryotic sources (Kigawa et al., 2004; Madin, Sawasaki, Ogasawara, & Endo, 2000) are described and to some extent commercially available. Commercial systems are widely used and enable the quick access to cell-free technology, but they restrict options to optimize production and quality of the synthesized pro- teins. Individually prepared systems offer a more diverse variety of opportu- nities to adjust the expression conditions, as all supplemented components are known and under control (Schwarz et al., 2007). In addition, recent pro- teomics studies identified the residual protein components of Escherichia coli cell-free lysates obtained by standard preparation procedures (Foshag et al., 2018; Hurst et al., 2017). For economic reasons, tunable and most efficient protein synthesis, systems based on E. coli (Henrich, Hein, Do€tsch, & Bernhard, 2015) or wheat germ (Harbers, 2014) cell lysates are thus cur- rently recommended. E. coli is preferred for the on-site preparation of cell-free lysates as the preparation protocol and the quality of the lysate source are better controllable and more reliable.Various strains can be used as lysate source and the most common are E. coli A19 and BL21. Special applications may require engineered strains having, for example, eliminated release factors for the efficient incorporation of non- natural amino acids (Hong et al., 2014; Peuker et al., 2016). A standard S30 lysate preparation procedure out of a 10-L fermenter is described, which is able to provide 60–100 mL of cell-free lysate yielding 0.3–1.5 g of expressed protein (Foshag et al., 2018; Schwarz et al., 2007), but it should be noted that a variety of protocol modifications exist (Kigawa et al., 2004; Kim et al., 2006; Kwon & Jewett, 2015; Shrestha, Holland, & Bundy, 2012; Yang, Patel, Wong, & Swartz, 2012; Zubay, 1973):1. Strike out E. coli A19 cells from a glycerol stock on a fresh agar plate and incubate the plate for 6–8 h. Inoculate two flasks containing 150 mL LB medium (10 g tryptone, 10 g NaCl, 5 g yeast extract/L) with the freshly grown cells and incubate them overnight at 37°C under vigorous shaking.

The cultures usually grow to an OD600 of 6–7 overnight.Inoculate 10 L of culture medium (16 g tryptone, 5 g NaCl, 10 g yeast extract/L) supplemented with 100 mM glucose and potassium phosphate buffer (22 mM KH2PO4, 40 mM K2HPO4) 1:66–100 with the pre- culture. At this stage, we recommend performing the cultivation in a fer- menter to achieve higher cell densities and obtain a better yields and a higher quality of extract. It is also possible to use conventional culture flasks, but in our hands the lysate performance is reduced drastically. Cultivate the cells at 37°C under high oxygen saturation and stir until they reach mid-log phase (~4–5 OD600) before cooling them down to ~18–20°C. The cooling step should not exceed 30 min. A growth curve should be recorded beforehand to assess the mid-log phase period in a given fermenter setup. he fast and efficient T7 RNA-polymerase is commonly employed for tran- scription within a coupled transcription and translation cell-free system (Schwarz et al., 2007). Since E. coli lysates still contain the endogenous RNA-polymerase (Foshag et al., 2018), cell-free reactions can alternatively be carried out with standard E. coli promotors and without the addition of T7 RNA-polymerase (Shin & Noireaux, 2010), but a significant drop in expression yield may occur. The following section illustrates the preparation of T7 RNA-polymerase by overexpression in E. coli cells (Schwarz et al., 2007). The yield of T7 RNA-polymerase out of 1 L E. coli culture is usually sufficient to supplement 0.5–1 L of cell-free lysate.

Fig. 1 General L-CF workflow including nanodisc formation (yellow), cell-free produc- tion (green) as well as sample purification (blue). Pictograms illustrate molecules or devices and workflow steps are indicated.The reconstitution of “empty” nanodiscs (i.e., nonmembrane protein containing) as they are provided in cell-free reactions (Katzen et al., 2008; Roos et al., 2012) has been well established for more than 10 years (Bayburt, Grinkova, & Sligar, 2002; Denisov et al., 2004). Over the years engineered nanodiscs were designed for particular tasks, as, for example, nanodiscs with a smaller diameter that are more suitable for solution NMR experiments (Hagn et al., 2013). Particular comments and guide- lines for successful nanodisc formation are given.For lipid stock solutions dedicated for nanodisc reconstitutions 50 mM is a reasonable concentration (Roos et al., 2014). The sufficient solubilization of the lipids is crucial for the successful reconstitution of a homogeneous population of nanodiscs. As nanodiscs are formed upon detergent removal, detergents with a relatively high critical micellar concentration (cmc) are suitable as they can be easily removed from the mixture. Commonly the bile acid salt sodium cholate is used (Bayburt et al., 2002). In general, it is also possible to use other detergents, but detergents with a very low cmc can become problematic. Most lipids can be completely solubilized by 100 mM sodium cholate in water, but increased detergent concentrations should be considered if the solution still stays turbid. In this respect especially lipid mixtures extracted from natural sources can become problematic. Lipid stocks can be stored at —20°C.The MSP can be produced by conventional expression in E. coli cells and purified via the N-terminal His6-tag (Denisov et al., 2004). Thereby the same protocol is used for all different variants of MSP:volumes of MSP-buffer with 1% Triton X-100, MSP-buffer at pH 8.9 containing additional 50 mM cholic acid, MSP-buffer, and MSP-buffer with addition of 50 mM imidazole. The protein is eluted by five column volumes of MSP-buffer containing additional 300 mM imidazole.For some experiments it can be beneficial to label the MSP with heavy iso- topes (Li, Kijac, Sligar, & Rienstra, 2006; Peetz et al., 2017). For labeling with 15N and 13C isotopes, LB overnight cultures are harvested by centrifugation (7000 × g, 10 min, 4°C) and resuspendedin minimalmedia (2 g/L 13C-glucose, 1 g/L 15N-NH4Cl, 0.2 g/L FeCl3, 0.67 g/L ZnCl2, 1 mM MgSO4, 1 mM CaCl2, 0.01 g/L thiamine, 0.1 g/L ampicillin, 3 g/L KH2PO4, 12.8 g/L Na2HPO4, 0.5 g/L NaCl). Expression is induced after 2 h at 37°C and 200 rpm shaking and the cultures are incubated for additional 3 h before harvesting (Peetz et al., 2017).

For the formation of nanodiscs the MSP is first mixed with detergent- solubilized lipids (Fig. 1). It is crucial to consider previously defined MSP to lipid ratios for the formation of homogeneous discs (Marty, Wilcox, Klein, & Sligar, 2013; Roos et al., 2012). Due to the dilution of the lipid stocks it should be verified that the final detergent concentration is not reduced below its cmc. In that case either additional sodium cholate or n-dodecylphosphocholine (DPC) could be added to reach concentrations above cmc (Roos et al., 2012) or higher concentrated MSP stocks may be used. If the appropriate ratio of the selected lipid/MSP types is not known, a corresponding prescreen on an analytical scale (~100 μL samples) should be carried out. The samples are incubated with gentle shaking for about 1 h at temperatures above the lipid transition temperature, preferentially room temperature. Nanodisc formation is then induced by detergent removal with biobeads or by dialysis. For dialysis, a ~250 × volume (i.e., 5 L for 20 mL of mixture) of DF-buffer (40 mM Tris–HCl, pH 8.0, 100 mM NaCl) is recommended with incubation for 3 days and one buffer exchange each day. To separate the nanodiscs from soluble aggregates or aggregated lipids, the solution is centrifuged (22,000 × g, 20 min, 4°C) and the super- natant can be concentrated with DF-buffer equilibrated Centriprep tubes (10 kDa MWCO) by multiple centrifugation steps (20 min, 2000 × g, 4°C) to a final concentration of 0.5–1 mM (Fig. 1). The concentrated solutions can be shock frozen by liquid nitrogen and stored at —80°C until usage.

The homogeneity of nanodisc stocks and the MSP:lipid ratio screens is ana- lyzed by size-exclusion chromatography: 50 μL ofa 20–50 μM nanodisc solu- tion is separated with an analytical size-exclusion column (e.g., Superdex 200 increase) in DF-buffer (Roos et al., 2012). Size and shape of prepared nanodiscs may be further evaluated by negative stain cryo-EM (Henrich et al., 2017). To prevent aggregation or extensive stacking on the grids, the nanodisc concentration should be kept ~20 μM.In general, cell-free production approaches can be carried out in three major expression modes (Schwarz et al., 2007). The P-CF mode is per- formed without any supplied hydrophobic environment, but the resulting membrane protein precipitates can readily be solubilized in detergents. In the D-CF mode, detergents are supplied to the reaction leading to an imme- diate solubilization of the expressed protein. L-CF expressions contain bilayer structures such as liposomes, bicelles, or nanodiscs for membrane protein solubilization. Due to its simplicity and efficiency P-CF expression is usually the first choice for the production of a protein in a cell-free system, whereas the L-CF mode offers the opportunity to synthesize solubilized membrane proteins without any detergent contacts. Depending on the nature of the synthesized membrane protein, liposomes, bicelles, or nanodiscs might have different efficiencies for its proper cotranslational insertion and folding.To setup a cell-free reaction, a variety of compounds responsible for energy supply (e.g., acetyl phosphate), a suitable chemical environment (e.g., Mg2+, HEPES-buffer), precursors (e.g., amino acids), and preservation agents (e.g., protease inhibitors) are mixed together in a feeding mixture (FM), ensuring supply with nutrients as well as dilution of toxic byproducts. The FM is sep- arated by a membrane from the reaction mixture (RM) holding the com- ponents responsible for transcription (e.g., T7 RNA-polymerase) and the protein production machinery (e.g., cell lysate).

We highly recommend the usage of such a continuous exchange setup with an RM to FM ratio of 1:15–20 to achieve an efficient production of material required for NMR experiments. The RM volume should be adjusted to the total amount of desired protein in the NMR sample and is usually between 1 and 10 mL. In Table 2 all ingredients with stock solution and final concentrations for the expression of an unlabeled protein are displayed and the required vol- umes for a 1-mL reaction with a 1:20 ratio of reaction to FM and a Mg2+ concentration of 18 mM are indicated. For preparative scale expressions in the mL range, the utilization of Slyde-A-lyzer devices with an MWCO of 10 kDa (Thermo Scientific) as container for the RM is very convenient. These are placed in special plastic containers that hold the FM (Schwarz et al., 2007). For screening purposes, smaller containers holding ~55 μL of RM are sealed with a dialysis membrane with an MWCO of 12–14 kDa fixed with a Teflon ring and then placed into standard 24-well cell culture plates holding 825 μL of FM. The basic steps for setting up nanodiscs of only ~8 nm. This size is usually sufficient to accommodate most membrane proteins and even the formation of large assemblies such as the pentameric complex of the seven-transmembrane helix containing proton pump proteorhodopsin with a total molecular mass of ~135 kDa can be achieved (Peetz et al., 2017). On the Sodium cholateother side, the limited mem- brane space can also hamper the insertion of specific membrane proteins or complexes. Thus, the expression of the target into different nanodisc types should be tested to evaluate the formation of soluble membrane protein/nanodisc complexes.