Introduction

Membrane proteins play important roles in all aspects of cellular signaling and are the interface through which a cell responds to extracellular cues. One of the most important subsets of the cellular membrane protein makeup are G-protein–coupled receptors (GPCRs), which sense a variety of external cues to orchestrate a broad range of cognitive and physiological responses and are targets for approximately 30% of all clinically-approved drugs1. Therefore, modulation of GPCRs is a powerful means to manipulate cellular signaling and phenotype, enabling a broad spectrum of research and clinical applications. Unfortunately, expression of functional membrane proteins can only been achieved through the introduction of DNA or RNA-based genetic material into cells, which can be technically challenging in primary cells and offers poor control over the amounts of protein being produced2.

To obtain pure and soluble GPCRs while maintaining proper folding and transitions between conformational states, individual receptor molecules need to be in a native-like environment, which can be provided with the use nanolipoprotein particles (NLPs), also known as nanodiscs3,4. Nanodiscs hold great promise as a platform for efficient cellular or in vivo delivery of membrane-associated molecules, due to their highly bio-compatible properties (e.g. stability, lack of toxicity, biodistribution)5. However, their use has been limited to the delivery of drugs6, phospholipids7 and personalized cancer vaccines8, while nanodelivery of functional membrane receptors has not been reported.

Traditionally, nanodisc-solubilized GPCRs have been produced using cell-based methods that require expression in cell hosts followed by detergent extraction and nanodisc reconsitution, which is time-consuming and labor intensive3,9. Cell-free synthesis provides a viable alternative for large-scale production of membrane proteins and is emerging as a competitive choice due to its increasing production yields (up to mg/ml reaction amounts in the most optimized systems) and to its lower expression costs9,10,11,12,13,14,15, especially with the adoption of alternative energy regeneration systems for protein synthesis16.

Here, we developed a platform integrating cell-free production and nanodelivery of functional GPCRs to the plasma membrane of cells. Using the well-studied β2 adrenergic receptor (β2AR) as a prototypical GPCR17, we validated the utility of this platform by demonstrating that membrane-delivered β2AR responds to ligand binding and triggers cAMP production to rescue the wound healing defects of β1/β2AR double knockout primary cells. We expect this platform to be readily adapted to the cell-free production and nanodelivery of a broad range of GPCRs and other membrane proteins for DNA and RNA-free manipulation of cellular functions.

Full-length production of β2AR-NLPs by cell-free, co-translational approaches

To achieve large-scale, cell-free expression of functional β2AR, we first codon-optimized the cDNA encoding human β2AR protein for expression using Expressway™ Maxi Cell-Free E. coli Expression System (Supplementary Fig. 1). To induce spontaneous insertion of β2AR into Δ49A1-induced NLPs, we co-expressed codon-optimized β2AR with human apolipoprotein A-1 lacking the amino-terminal 49 amino acids (Δ49A1) in the presence of 1,2-ditetradecanoyl-sn-glycero-3-phosphocholine (DMPC) (Fig. 1a), as previously described11,18.

Figure 1
figure 1

Cell-free production of β2AR-NLPs. (a) Co-translational production of β2AR-NLPs was achieved by incubating β2AR and ApoAI DNAs in the presence of synthetic lipid nanoparticles (1,2-dimyristoyl-sn-glycero-3-phosphocholine, DMPC) with the cell-free reaction mixture at 30 °C for 18 hours. (b) Optimization of the GPCR-NLPs in vitro translation system. β2AR was cell-free produced in the absence of Δ49A1 cDNA and titrated in the presence of Δ49A1 cDNA (10:1, 20:1, 40:1). FluoroTect™ GreenLys (Promega) was added for visualizing newly synthesized proteins. Cell-free reactions were centrifuged at 10,000 rpm for 10 minutes. 5 μl aliquots of sample before centrifugation (total fraction, T) and of the supernatant (soluble fraction, S) were used for SDS-PAGE. Gel images were taken using Molecular Imager® Gel Doc™ XR System from Biorad. (c) Band intensity ratio of the soluble vs. total cell-free produced β2AR was utilized to determine solubility. Results are shown as mean ± S.E. (d) The identity and molecular weight of cell-free expressed β2AR was verified by side-by-side comparison with cell-expressed β2AR, via western blot analysis using anti-flag labeling.

Full size image

We then maximized the production of NLP-solubilized full-length β2AR protein by varying the ratio of pβ2AR to pΔ49A1 plasmids in the cell-free co-expression reaction mix (Fig. 1b). We found that a DNA ratio of 10:1 for pβ2AR to p∆49A1 promoted about 70% solubilization of β2AR in the total translated protein fraction (Fig. 1c and Supplementary Fig. 2a). Using western blot analysis, we further confirmed the expression of full-length β2AR (~47 kDa) using an anti-flag antibody (Fig. 1d and Supplementary Fig. 2b).

To conclusively demonstrate β2AR association with NLPs, we first purified particles by affinity chromatography via the Δ49A1 6XHis tag, followed by characterization using size-exclusion chromatography and transmission electron microscopy (TEM). The chromatographic analysis showed two clearly distinct elution peaks representing β2AR integrated into NLPs and empty NLPs (Supplementary Fig. 3), suggesting stable association of β2AR with NLPs. TEM further confirmed that approximately 27.5% of the NLPs in the purified sample contained integrated β2AR (Fig. 2a,b). β2AR-integrated NLPs displayed a significantly larger diameter than empty NLPs when imaged by TEM (β2AR-NLPs: 33.0 ± 3.0, empty NLP: 21.5 ± 1.4, P = 1.8641E-15). These results indicate that cell-free expressed β2AR successfully integrates into the NLP-supported bilayer.

Figure 2
figure 2

Transmission Electron Microscopy (TEM) characterization of β2AR-NLPs. TEM images of stain-embedded empty NLPs with average diameter of 21.5 nm and β2AR-NLPs with average diameter of 23 nm. Scale bar: 50 nm. (c) Particle size distribution of empty and β2AR-NLPs from TEM data. Standard deviation for empty and beta-adrenergic NLPs are 1.4 and 3.0, respectively (p < 0.0001, paired t test).

Full size image

The β 2AR-NLP complex is functional in vitro

We next sought to investigate the functional integrity of NLP-bound β2AR. To do so we assayed the ligand-induced conformational changes of NLP-bound β2AR using a single-molecule pull-down assay (SiMPull)19,20, tapping into a β2AR conformational biosensor (Nb80-GFP) that specifically stabilizes the active conformation of β2AR21. Immobilized β2AR-NLPs or empty NLPs were first pre-incubated with either antagonist ICI118551 (ICI, 10 μM) or agonist isoproterenol (ISO, 10 μM) before pulling down purified Nb80-GFP. Only in the β2AR-NLP sample in the presence of the agonist we could detect green fluorescent spots, representing β2AR-bound Nb80-GFP (~313 molecules). We observed a small number of pull-down Nb80-GFP either in the presence of ICI (~31 molecules) or under the conditions when empty NLPs were immobilized (~17 molecules) (Fig. 3a,b). As an additional confirmation of the specificity of binding, we also pulled down mCherry-labeled Gs protein in the presence of ISO, but not ICI (Fig. 3c,d). Importantly, we observed different association rates of Nb80-β2AR binding in the presence of ligands with different pharmacological properties (Supplementary Fig. 4a). The association rates for two full agonists (ISO and epinephrine, EPI, 10 μM) and one partial agonist (dobutamine, DOB, 10 μM) are 5.3*108, 1.4*109 and 1.8*108 M−1 min−1, respectively, which correlate with the known efficacy of the drugs examined (Supplementary Fig. 4b)22. These results suggest that cell-free expressed β2AR maintains functional integrity when integrated into NLPs and can distinguish the pharmacological properties of different ligands.

Figure 3
figure 3

Functional characterization of β2AR-NLPs for binding to Nb80-GFP or cognate Gs-protein. (a,c) Purified β2AR-NLPs or empty NLPs were immobilized on a chip via anti-flag antibody (M2, Sigma). Nb80-GFP or Gs-mcherry-containing HEK cell lysates were flown over the immobilized β2AR in the presence of either the β2AR agonist isoproterenol (ISO, 10 μM) or the β2AR antagonist ICI118551 (ICI, 10 μM). (b,d) Average number of fluorescent molecules per imaging area, Nf. Results are shown as mean ± s.d. (n > 5; ***p < 0.0001).

Full size image

Because our TEM data allows us to determine only the percentage of NLPs-bound β2AR versus empty NLPs, we further evaluated the fraction of functional receptors out of the NLP-bound β2AR pool using SiMPull (Supplementary Fig. 5a). To do so, we compared the number of pulled down Nb80-GFP molecules in the presence of 10 μM ISO to the total number of immobilized β2AR-NLPs, visualized with an anti-6xHis-555 antibody (Supplementary Fig. 5b,c). Assuming that 75% GFP is fluorescent, as previously reported20,23, we estimated that the percentage of functional β2AR particles was 10% of the total purified sample, which corresponds to 30 ng (330 fmol; 2*1011 molecules) per microliter (ul) cell-free reaction.

Functional insertion of cell-free expressed β 2AR-NLPs in living cells

Recent studies have described the application of NLPs to both in vitro and in vivo delivery of hydrophobic drugs and membrane antigens5,6,8. Thus, we tested whether cell-free expressed β2ARs could be delivered via NLPs onto cell membranes and maintain their function. When DiO-labeled empty NLPs were incubated with HEK 293 cells in serum-free medium, green fluorescence could be readily detected on the cell membrane after 14 hr incubation, suggesting efficient transfer of the lipid component of NLPs onto cell membranes (Supplementary Fig. 6a). To test nanodelivery of β2AR, we applied purified NLP-bound β2AR onto 293 cells (100 nM) for different amounts of time, followed by live-cell staining and imaging with an anti-flag antibody conjugated to Alexa-546 (Fig. 4a). After just 6 hr of incubation, red fluorescence intensity could already be detected at the cell membrane, indicating β2AR incorporation onto the cell membrane and continued to increase up to 18 hr of incubation, when it reached >10-fold higher intensity compared to the 6 hr timepoint (Fig. 4b).

Figure 4
figure 4

Nanodelivery of in vitro translated β2AR on cells. Time-course of β2AR-NLPs insertion onto HEK293 cell membranes was performed by live-cell anti-flag staining at various time points of incubation with β2AR-NLPs. (a) Representative fluorescence images showing flag staining intensity after 18 hours incubation. (b) Analysis of flag cell staining indicates efficient membrane insertion after overnight incubation (18 hours) (n = 20 cells, p < 0.0001). Results are shown as mean ± S.E. (c) Isoproterenol induced membrane-relocalization of Nb80-GFP in HEK293 cells after β2AR nanodelivery (18 hours, n = 8 cells). (d) Representative traces of fluorescence from line indicated in c.

Full size image

The conformational biosensor Nb80-GFP has been previously used for direct visualization of β2AR activation in living cells21. We thus used this approach to verify that nanodelivered β2AR maintained functional integrity as demonstrated by efficient cytosol-to-membrane recruitment of Nb80 and production of cyclic AMP (cAMP) upon agonist exposure (Fig. 4c). After 18 h of β2AR nanodelivery on 293 cells expressing Nb80-GFP, we performed dual-color, time-lapse imaging of flag staining and Nb80 relocalization in response to 10 μM ISO. We observed the rapid relocalization of Nb80-GFP from the cytoplasm to the membrane, confirmed by a marked increase in green fluorescence intensity on the cell membrane (~75%), while the red fluorescence intensity on the membrane remained constant (Fig. 4d and Supplementary Fig. 6b).

These results suggest that cell-free–produced β2AR can be functionally delivered to the cell membrane, resulting in a conformational response to agonist stimulation. More importantly, when stimulated by agonist, nanodelivered β2AR triggered a downstream signaling cascade comparable to that of endogenous β2AR receptors (Fig. 5a). We measured cAMP production with the fluorescent resonance energy transfer (FRET)-based cAMP biosensor ICUE3 in living cells24. As isoproterenol is a nonselective β1AR/β2AR agonist and most cell types endogenously express βARs, we isolated primary neonatal cardiac myocytes from both β1/β2AR double-knockout mice (dKO) and wild-type mice. Upon isoproterenol application, we observed cAMP production demonstrated by 15% FRET changes only in wild-type, but not in dKO myocytes (Fig. 5b). When β2AR was nanodelivered onto dKO primary myocytes for 14 hr, we observed 14% FRET changes in response to ISO, which was similar to that observed in wild-type cells. In contrast, 14 hr incubation of dKO myocytes with empty NLPs failed to generate any detectable FRET changes (Fig. 5c).

Figure 5
figure 5

Nanodelivered β2AR can drive cellular signaling. (a) Schematic of GPCR nanodelivery to cellular membranes to alter cellular properties. (b,c) ICUE measurement of cAMP production in primary β1/β2AR double-knockout neonatal cardiac myocytes (dKO myocytes) shows that nanodisc-delivered β2AR is capable to elicit cellular signaling in response to the agonist isoproterenol at a level comparable to endogenous β2AR (**p < 0.01, unpaired t test). Results are shown as mean ± S.E.

Full size image

Receptor nanodelivery alters cellular phenotype

The β2AR signaling has been directly linked to fibroblast migration during wound-healing processes25,26. Thus, we next asked whether β2AR nanodelivery could be used to induce a pro-migratory phenotype in cultured fibroblasts. To do so, we nanodelivered β2AR onto mouse embryonic fibroblasts (MEFs) from our dKO mice and then assayed their wound healing properties using an in vitro scratch assay. At 24 and 48 hr after scratch application, dKO MEFs displayed 33% and 57% gap closure, respectively. Upon β2AR nanodelivery, both with and without agonist stimulation (10 μM ISO), the wound-healing properties of these cells were significantly increased (Fig. 6a). β2AR-insertion without agonist application resulted in 50% and 79% gap closure at the two timepoints, indicating that basal activity of the receptor is sufficient to elicit a significant wound-healing effect (p = 0.014 after 24 hr). Agonist stimulation in addition to β2AR nanodelivery led to a further increase in gap closure, reaching 66% and 95% at the 24 and 48 hr time points, respectively (Fig. 6b, p = 0.0002 after 24 hr and p = 0.008 after 48 hr, n = 3). These results validate the feasibility of GPCR nanodelivery and prove that this is a suitable strategy to selectively modulate cellular signaling and phenotype.

Figure 6
figure 6

Nanodelivered β2AR increases rescues wound healing defects of β1/β2AR dKO mouse embryonic fibroblasts (MEFs). (a) Representative images of dKO MEFs treated with β2AR-NLPs or empty NLPs with or without isoproterenol and subjected to scratch assay for cell migration. Wound healing was analyzed at 24 and 48 hours. Red lines represent the cell front. (b) Wound healing was quantified as percentage of gap closure. Results are shown as mean ± S.E. (*p < 0.05, **p < 0.01, n = 3, unpaired t test).

Full size image

Discussion

In summary, we demonstrated for the first time a method for nanodelivery of membrane proteins, in which an in vitro one-step, cell-free system can be used to synthesize functional nanodisc-solubilized GPCRs that can be further delivered onto plasma membranes of cells to trigger signaling cascades and alter cellular function. To demonstrate the feasibility of GPCR nanodelivery and provide proof of concept for the therapeutic potential of this new delivery system, we cell-free produced nanodisc-solubilized β2AR and fully characterized its function both prior and after delivery on cell membranes. We validate nanodelivery of β2AR in three different cell types, including hard to transfect primary cells, where we show that nanodelievered β2AR triggers intracellular cAMP production in a similar manner to naturally occurring endogenous β2AR and imparts novel functional properties to the receiving cells, as evidenced by increased cell migration in an ex vivo wound-healing model using β1/β2AR double knockout cells.

Our method greatly expands the existing toolbox for direct protein transduction to living cells independent of DNA or RNA-based techniques2 and affords the benefits of cell-free, large-scale, in vitro protein expression (e.g., speed, purity, lower costs and no need for detergents).

Since a decade ago when membrane protein synthesis in cell-free systems started27, the yield-on-cost ratio has been continuously increasing. Membrane protein synthesis in cell-free systems (e.g. E. coli and wheat germ based cell-free systems) has started to reach levels comparable to soluble proteins15,28, especially with the development of a continuous exchange cell-free dialysis system (CECF) that provides a prolonged reaction time and freshly supplied reaction components14. Though the CECF reaction is approximately 10 times more expensive compared to a typical batch reaction, with the increased productivity, it is possible to achieve a significant reduction of the total costs by a factor of 10. The main costs of cell-free synthesis systems arise through the addition of T7 RNA polymerase, energy in the form of adenosine and guanosine triphosphate and an energy regeneration system. The costs of exogenous added T7 RNA polymerase might be decreased by using in house production. The development of alternative energy-rich components and the energy regeneration systems is already in process16. The yield-on-cost ratio of cell-free system for membrane protein synthesis will continue to increase given the increasing productivity of cell free systems in combination with the reduced costs for lysates and the energy regeneration system.

In the current study, we adopted an innovative approach to estimate the amount of functional GPCR product in the final expression product, which takes advantage of a protein binding assay (SiMPull). Because of the intrinsically different experimental procedures used to determine the yield of functional GPCR between this and previous studies, an exact comparison among yields produced by the different systems would not be accurate. However, the relatively overall lower efficiency of functional GPCR expression using in vitro translation (ng per ul cell-free reaction) compared to the broadly used insect cell-baculorivus expression system (<1 mg/L total purified proteins)28,29 is possibly due to the intrinsic limitations of the system. In fact, the tRNAs, amino acids, ribosomes and other components of the mixture are limited in an in vitro translation reaction, which further limits the time length of reaction efficacy. In addition, selection of different lipid types may further increase the solubility of the protein, though extensive characterization has been performed in the past15,28. Successful strategies to further optimize cell-free synthesis of GPCRs and membrane proteins in general have been done on a variety of lipid structures like liposomes, micelles, bicelles and nanodiscs27,30,31. Future work will focus on further increase the yields by applying modifications to the cell-free reaction conditions, such as integrating continuous exchange cell-free dialysis system and systematic optimization in selection of lipid types.

Given the efficiency of membrane protein incorporation strictly depends on the concentration of nanodisc-solubilized proteins applied onto the cells, our system also provides a much higher degree of precision in fine-tuning the membrane levels of delivered receptors than those achieved by DNA expression. The high simplicity and flexibility of this system may allow nanodelivery of a broad range of membrane proteins, such as other GPCRs, optogenetic actuators and reporters in various cell lines. Furthermore, It is promising that this system may allow nanodelivery of GPCRs to living organisms for both research and clinical uses, such as revision restoration32,33. However, extensive research focusing on substantial optimization, such as in vivo stability of the receptor complex, the bioavailability in certain tissue types and the specificity of the nanodelivery, is necessary to realize the full potential of this system for in vivo applications.

Methods

Cell-free, in vitro translation reaction for GPCR-NLP self-assembly

Cell-free reactions using the ExpresswayTM Maxi Cell-Free E. coli Expression System (Life Technologies) were carried out as previously described18. The gene for β2AR was codon-optimized for E. coli expression. A Flag tag was positioned after the start codon and NdeI and SmaI restriction sites were positioned at the beginning and the end of the construct, respectively. The sequence was ordered as a Geneblock (Integrated DNA Technologies) and cloned using the NdeI-SmaI sites into pIVEX2.3d and pIVEX2.4d (Roche). The pIVEX2.3d-β2AR construct was used to prepare β2AR-NLPs for cell-membrane insertion, followed by live staining, which required the flag tag to be exposed at the very beginning of the receptor N-terminus; the pIVEX2.4d-β2AR construct resulted in a protein product with an N-terminal 6xHis tag preceding the flag tag and was used for all other applications. The base changes that were made during codon optimization are shown in Supplementary Fig. 3. Small, unilamellar vesicles of DMPC (liposomes) were prepared by sonication of a 25 mg/ml water–DMPC solution on ice until optical clarity was achieved (usually 10 min), followed by 2 min of centrifugation at 14,000 rcf to remove metal contamination from the sonication probe tip. DMPC small, unilamellar vesicles were added to the cell-free reaction mixture prior to starting the reaction at a final concentration of 2 mg/ml. Addition of FluoroTect™ GreenLys (Promega) was done at the beginning of the reactions to facilitate visualization and quantification of synthesized proteins. For membrane protein and NLP coexpression in a 200 μl reaction, 2.5 μg of plasmid DNAs were added to the lysate mixture. The ratio of β2AR to Δ49A1 plasmid DNA (containg a N-terminal HIS tag) was kept constant at 20:1. The reactions were incubated at 30 °C for 18 h. Empty NLPs were generated by omitting the β2AR plasmid DNA.

Affinity purification of cell-free–produced NLPs

Empty NLPs or β2AR-NLPs, both containing 6xHIS tags on their ApoAI component, were purified from their respective reaction mixtures by Ni/NTA affinity purification. Briefly, we used immobilized metal ion affinity chromatography (IMAC) tips on an automatic pipette (80 μl resin volume, Mettler-Toledo). The resin was first equilibrated with native buffer (Imidazole 20 mM, NaCl 300 mM, NaHPO4 50 mM, pH 8.0), then incubated with the NLP-containing reaction mixture and washed three times before elution of the NLPs in native buffer with 400 mM imidazole. Eluted NLPs were then dialyzed in 2 L of phosphate-buffered saline (PBS) buffer using 3.5 KDa molecular weight cutoff D-tube dialyzers (Millipore). Total protein concentration was measured by bicinchoninic acid (BCA) assay (Thermo) on a Synergy2 plate reader (BioTek), using a standard curve with known concentrations of BSA. SDS-PAGE and fluorescence imaging of the gels were performed as previously described18.

Nb80-GFP quantification

Recombinant enhanced green fluorescent protein (EGFP; 1 mg/ml) was purchased from Vector Laboratories. Nb80-GFP containing 5xHis-tag was expressed in HEK293T cells and purified using IMAC resin tips (Rainin). A series of dilutions of EGFP and Nb80-GFP was made using Tris-EDTA buffer (10 mM Tris-HCl, 10 mM EDTA, pH 8.0) as the diluent. Fluorescence was determined by a Synergy 2 fluorescence microplate reader (BioTek Instruments) using a 485 nm, 20 nm bandpass excitation filter and a 528 nm, 20 nm bandpass emission filter with an instrument sensitivity setting of 50. Quantitation of EGFP was determined using a calibration curve determined by linear regression (r2 = 0.998).

Transmission electron microscopy

Empty NLPs and β2AR-NLPs were diluted to a final concentration of 0.2 mg/mL. A deep-staining approach was utilized for sample embedding immediately prior to examination with a JEOL 1230 TEM. All samples were mixed with 16% ammonium molybdate and 0.1% trehalose and immediately transferred to a carbon-coated copper grid. The grids were further prepared for imaging with a single wash using PBS buffer and were air-dried. All TEM micrographs were captured at 60,000x magnification and processed for size distribution measurements in FIJI package. From 8 micrographs, 140 particles were measured by the long axis. The measurements for empty and β2AR-NLPs were plotted and standard deviations were calculated using Microsoft Excel.

Single-molecule pull-down assay (SiMPull)

For the β2AR functionality test, cells were plated onto 100 mm petri dishes (VWR) at a density of 200,000 per dish and grown for 24 h before being transfected with 3 μg of Gs-mCherry plasmid DNA. Cell lysates were prepared by cell titration in apyrase reaction buffer (NEB) supplemented with protease inhibitors cocktail (Sigma) (20 mM MES, 50 mM NaCl, 5 mM CaCl2, 1 mM DTT, 0.05% Tween-20, pH 6.5). Complete cell lysis was ensured by probe-sonicating on ice for 30 s. Cell debris was removed by 10 min of centrifugation at 14,000 rcf (4 °C). Nearly all guanine nucleotide triphosphates and diphosphates (NTPs and NDPs) were hydrolyzed to monophosphates (GMPs) by incubating the lysate with apyrase at 30 °C for 1 h. SiMPull chips were first washed twice with T50 buffer (50 mM NaCl, 10 mM Tris-HCl, pH 7.8). Next, NeutrAvidin (Thermo) was added and slides were incubated for 5 min at room temperature (RT). After washing twice with T50, biotinylated anti-flag antibody (10 nM) was added and incubated for 10 min at RT, then washed twice with T50 again. Subsequently, 1 μg of purified cell-free–produced NLPs (either containing flag-tagged β2AR or empty nanodiscs as a negative control) was added, followed by 10 min incubation at RT. Unbound samples were removed from the chamber by washing twice with T50 buffer. Cell lysates expressing Nb80-GFP or Gs-mCherry were 1:10 diluted into T50 buffer containing 10 μM isoproterenol (Sigma) or ICI (Sigma), then added to the SiMPull chamber and incubated for 10 min at RT, followed by washing twice with T50 with isoproterenol or ICI. Proteins immobilized on the slides were visualized using a prism-based, total internal reflection fluorescence microscope equipped with excitation laser 488 nm (GFP) and 561 nm (mCherry) and DV2 dichroic 565dcxr dual-view emission filters (520/30 nm and 630/50 nm). Mean spot counts per image and standard deviations were calculated from images taken from 5 to 10 different regions using a script written in Matlab. For drug-specific Nb80 to β2AR ON rate determination: 1 μg of cell-free–produced β2AR-NLP was immobilized on SiMPull chips and preincubated with 10 nM drug (isoproterenol, dobutamine, or epinephrine) for 15 min. Immediately after injection of 1 nM purified Nb80 (diluted in T50 buffer containing 10 nM drug), several images were taken at each time point (0, 30 s and 1, 2, 3, 4, 5, 10, 15, 20 and 25 min). τ1/2 was determined by fitting the data with a one-phase association curve (GraphPad Prism 6).

Cell culture, DNA constructs and transfection

HEK293 cells (ATCC #CRL-1573) were grown in Dulbecco’s modified Eagle’s medium (DMEM, Invitrogen) supplemented with 10% fetal bovine serum and 1 mg/ml penicillin-streptomycin. The pEGFP_N1 plasmid containing Nb80-GFP was a gift from Dr. Huang Bo (UCSF). The original rat Gs protein pcDNA was from Addgene (55793). The pcDNA plasmid containing Gs-mCherry was made by restriction cloning using a single BamHI site inserted after amino acid 71 in the Gs protein. The mCherry insert was amplified by PCR with the addition of two BamHI sites and two flexible linker regions (GGGS) on each side prior to restriction cloning. The correct insertion of mCherry was confirmed by sequencing. Cell transfections were performed with Effectene® Transfection Reagent (Qiagen) according to manufacturer’s protocol. All cell-culture reagents were from Life Technologies, unless otherwise noted.

Live-cell confocal imaging

For insertion of NLPs onto cell membranes, the cells were plated onto 35 mm glass bottom microwell dishes (MatTek) at a density of 10,000 per well and transfected with 100 ng of Nb80-GFP plasmid DNA. After 14 h, the medium was switched to serum-free medium with the addition of 10 μg of purified total NLPs (corresponding to a final concentration of approximately 55 nM). The removal of serum is meant to prevent absorption of NLPs onto serum albumin and increase the efficiency of NLP insertion onto cell membranes. To maximize the insertion onto cell membranes, cells were incubated with the NLPs overnight. For surface-labeling of cells with membrane-inserted receptors, cells were washed three times in HBSS (Life Technologies) and supplemented with 2 mM Ca2+ and 15 mM HEPES (pH 7.4), prior to being incubated at 37 °C for 15 min with M1 anti-flag antibody (1:1000, Sigma) conjugated to Alexa 546 (A-20002, Thermo Fisher). The cells were imaged in HBSS buffer with a 40 × oil-based objective on an inverted confocal microscope (Observer, Zeiss). Images were acquired with a CCD camera driven by the Zen software (Zeiss). Isoproterenol (Sigma) was carefully added directly onto the dish during the imaging session (final concentration 10 μM). Time-lapse images were analyzed on Image J and Nb80-GFP membrane relocalization was confirmed by drawing a line across the cell membrane and using the plot profile function. The analysis of green and red fluorescent signal change (membrane ΔF/F0) was done using a custom-made script written in Matlab. Briefly, the green and red fluorescence intensities were measured in a mask generated using the signal from the red channel, indicating the membrane location of flag-β2AR. Fluorescent signal change for each wavelength was calculated as (F-F0)/F0, with F0 being the fluorescence intensity prior to agonist stimulation.

Fluorescent resonance energy transfer (FRET) measurement

Wild-type or β1/β2AR dKO myocytes were infected with adenovirus overnight for expression of the cAMP biosensor ICUE324. dKO myocytes were further incubated with β2AR-NLPs or empty NLPs from 6 to 24 h before cells were used in FRET recording. Two adenoviruses expressing ICUE3 and flag-β2AR were used to co-infect dKO cells overnight, for the cell-expressed β2AR data. Cells were imaged on a Zeiss Axiovert 200 M microscope with a 40×/1.3NA oil-immersion objective lens and cooled CCD camera. Dual emission ratio imaging was acquired with a 420DF20 excitation filter, a 450DRLP dichroic mirror and two emission filters (475DF40 for cyan and 535DF25 for yellow). The acquisition was set with 200 ms exposure in both channels. Images in both channels were subjected to background subtraction and ratios of yellow-to-cyan color were calculated at different time points.

Scratch assay

β1/β2AR dKO MEFs were plated onto 6-well plates in DMEM medium containing 10% FBS and incubated at 37 °C to create a confluent monolayer. β2AR-NLPs or empty NLPs were added to the cells at 20 nM concentration and incubated for 8 h. The cell monolayer was scraped in a straight line with a pipet tip to create a gap. The debris was removed and the cells were washed once to smooth the edge of the scratch with 1 ml of the growth medium, then incubated with 4 ml of medium. Immediately following the scratch, isoproterenol or PBS vehicle was added to cells, with a final isoproterenol concentration of 10 μM. Images of the scratches were taken under a phase-contrast microscope. Additional pictures were taken at the 24- and 48-h time points. The images acquired at each time point were analyzed for percentage gap closure on ImageJ.

Statistical analysis

To quantify GPCR insertion efficiency into NLPs from β2AR TEM images, a semi-automated custom-made particle-selecting tool was used. To quantify the particles based on diameter distribution, particle analysis tools from Image J were employed. The statistical distribution of NLPs revealed that both the mean and the mode for the size of GPCR-filled NLPs were 23.5 nm, with about 27% of insertion efficiency. For quantification of β2AR insertion on the cell membrane, statistical analisis was performed using one-way ANOVA with Dunnett’s post-test, for comparison of every condition versus control condition (Empty-NLPs). During scratch assay, on the outer bottom of the dish an ultrafine ink marker was used to lay reference points close to the scratch to facilitate retrieving the same field for all image acquisitions. For TEM nanodisc size distribution statistical analysis was performed using paired t-test. All other statistical analyses were performed using unpaired t test.