The use of fluorescence correlation spectroscopy to monitor cell surface β2‐adrenoceptors at low expression levels in human embryonic stem cell‐derived cardiomyocytes and fibroblasts

The importance of cell phenotype in determining the molecular mechanisms underlying β2‐adrenoceptor (β2AR) function has been noted previously when comparing responses in primary cells and recombinant model cell lines. Here, we have generated haplotype‐specific SNAP‐tagged β2AR human embryonic stem (ES) cell lines and applied fluorescence correlation spectroscopy (FCS) to study cell surface receptors in progenitor cells and in differentiated fibroblasts and cardiomyocytes. FCS was able to quantify SNAP‐tagged β2AR number and diffusion in both ES‐derived cardiomyocytes and CRISPR/Cas9 genome‐edited HEK293T cells, where the expression level was too low to detect using standard confocal microscopy. These studies demonstrate the power of FCS in investigating cell surface β2ARs at the very low expression levels often seen in endogenously expressing cells. Furthermore, the use of ES cell technology in combination with FCS allowed us to demonstrate that cell surface β2ARs internalize in response to formoterol‐stimulation in ES progenitor cells but not following their differentiation into ES‐derived fibroblasts. This indicates that the process of agonist‐induced receptor internalization is strongly influenced by cell phenotype and this may have important implications for drug treatment with long‐acting β2AR agonists.


| INTRODUCTION
The β 2 -adrenoceptor (β2AR) is a prototypical member of the class A family of G protein-coupled receptors (GPCRs). 1 Crystal structures of the active β2AR bound to the partial agonist salmeterol or the native hormone adrenaline, with either a Gs protein or active state-stabilizing nanobodies, have greatly advanced understanding of the pharmacology and activation of β2ARs. [2][3][4][5] Upon ligand stimulation β2ARs preferentially bind to the heterotrimeric G-protein Gαs, stimulating the activation of adenylyl cyclase to catalyze the formation of cyclic AMP from ATP. 6 Cell surface signaling is terminated following the phosphorylation of the C-terminal domain of the β2AR by G-protein receptor kinases (GRK), signposting the receptor for beta-arrestin mediated internalization. [7][8][9] This can lead to a second phase of signaling from intracellular endosomes. 10,11 The β2AR undergoes ligand-directed internalization via a clathrin-mediated and Rab 5-regulated constitutive endocytic pathway for recycling alongside transferrin, but can also localize to lysozymes if directed toward a degradation pathway. [12][13][14][15][16] The β2AR has been shown to display distinct trafficking profiles dependent on the presence of single-nucleotide polymorphisms (SNPs). [17][18][19][20] While there are a number of SNPs present in the human β2AR only four result in altered amino acids; Arg16Gly, Glu27Gln, Val34Met, and Thr164Ile. [21][22][23] Within the human population, Arg16Gly and Glu27Gln are most common and these are present in the N-terminus region. 18,19 The relative frequency of these genotypes within the human population varies with race 24 ; however, linkage disequilibrium leads to the Arg16Glu27 haplotype being extremely rare. 25 Cell phenotype has been shown to be important in determining the nature of the molecular mechanisms underlying β2AR function and regulation. 7,14,26 For instance, responses to βAR agonists differ dramatically between primary vascular smooth muscle cells and recombinant HEK293 systems. 26 Similarly, clustering of β2ARs has been demonstrated in cardiomyocyte-like rat H29C cells but not in HeLa or CHO cell lines. 14 Cell type specificity is likely to be dictated by the expression levels of both the β2AR and its effector proteins, as observed in the case of the GRK subtype responsible for β2AR phosphorylation. 7 To determine these molecular mechanisms, single-molecule imaging techniques have been employed which have the sensitivity necessary to expose the spatial and temporal characteristics of these short-lived effects. [27][28][29][30][31] Coupling of single-molecule imaging techniques with the use of human embryonic stem (ES) cells, that can be induced to differentiate into multiple cell phenotypes, provides a potential approach to study the impact of cell phenotype on receptor pharmacology. [32][33][34] An important consideration is the need to work at low receptor expression levels similar to those seen endogenously to limit the impact of receptor over-expression. 35,36 A drawback of imaging endogenous GPCRs using standard light microscopy within their native environment; however, is that low levels of endogenous receptor expression can lead to poor signal to noise ratios. 37,38 One approach with the sensitivity to monitor receptors in membrane microdomains at very low receptor expression levels is fluorescence correlation spectroscopy (FCS). [39][40][41][42][43][44] Here, we have generated haplotype-specific SNAP-tagged β2AR human ES cell lines and applied FCS to monitor cell surface receptors in differentiated fibroblasts and cardiomyocytes.

| Creation of SNAPβ2AR construct
To create the pSIN-B2-BSD(RQ) lentiviral construct we PCR amplified the Arg16Gln27 (RQ) variant of SNAPβ2AR from the pcDNA3.1(Neo+)sig.SNAP(Met > Leu mutant) expression vector 45 with the following primer pairs: Fwd1-CTTAAaCTaGtTACCGCCACCATGCGGCTCTGC and Rev1-TCTGCAGAATTCTTACAGCAGTGAGTCATTTG, and after digestion with SpeI and EcoRI restriction enzymes cloned it into the SpeI/EcoRI pSIN-BSD lentiviral vector backbone produced from the pSIN-GFP-BSD plasmid. 46 All remaining haplotypes variants were created by a PCR-splice mutagenesis technique 47  5′-ATGGCTTCCATTGGGTGCCAG, for substituting A for G at position 48 of β2AR. The following two primers were used as flanking primers for all PCR-splice reactions: Fwd2-CATTCTCAAGCCTCAGACAGTGG and Rev2 5′-AAATCACATATAGACAAACGCACAC. The final PCR products for each corrected SNP variant were SpeI/EcoRI digested and cloned into the pSIN-BSD SpeI/ EcoRI backbone as described above. In total three vectors were produced from pSIN-B2-BSD(RQ) which were: pSIN-B2-BSD(RE), pSIN-B2-BSD(GQ), and pSIN-B2-BSD(GE) which represent the four different haplotypes involving the single-nuclear polymorphisms at Arg16Gly and Glu27Gln. In order to generate lentiviral particles, each construct was transfected together with packaging vectors in to a BL15 cell line and the particles produced were purified on streptavidin paramagnetic beads (Promega) as described in detail earlier. 48,49 Purified viral particles were used to transduce HEK293T and HUES7 cell lines at MOI of 10. Twenty-four hours following transduction, growth media was exchanged for fresh medium and after 48 hours Blasticidin (Thermo Fisher Scientific) was added to the media at a concentration of 2.5 μg/mL for HEK293T and 2 μg/mL for HUES7 cells. A population of antibiotic-resistant cells was expanded in culture for 3 weeks before being banked for further experimental use.

HEK293T cells
Generation of constructs for CRISPR/Cas9 genomeengineering of the N-terminal region of the β2AR genomic locus was performed as described previously 16 with the following modification: Following synthesis of the β2AR donor repair template, sig-SNAP 45 was ligated into the repair template using the restriction enzymes KpnI and BamHI. A mutation introduced during the synthesis of the donor repair template to eliminate an internal KpnI restriction site was then corrected by site-directed mutagenesis using primers described previously. 16 The donor template, therefore, resulted in cells expressing genome-edited sig-SNAPβ2AR with the start codon (Met) of the β2AR deleted.
CRISPR/Cas9 genome-engineering of the β2AR genomic locus in HEK293T was performed as described previously. 50 Briefly, HEK293T cells were seeded into six well plates and incubated for 24 hours at 37°C/5% CO 2 . HEK293T cells were cultured in DMEM supplemented with 10% fetal bovine serum (FBS). At 60% confluency, cells were transfected with px459 sgRNA/Cas9 expression constructs 16 and the donor repair template encoding sig-SNAPβ2AR using FuGENE transfection reagent (Promega). Twenty-four hours later cells were cultured with puromycin (0.3 µg/mL, Sigma-Aldrich) for 3 days. Following selection, cells were cultured without puromycin and seeded into clear flat bottom 96-well plates at 1 cell per well and allowed to expand for 2-3 weeks. Single colonies were then seeded into duplicate 48 well plates and allowed to expand. To screen for positive clones, genomic DNA was extracted using a Wizard Genomic DNA Purification Kit (Promega) and PCR performed using the following primers 5′-CGAGTGTGCTGAGGAAATCA-3′ and 5′-CGCCAGAGCTAGACACCAC-3′. The primers used were designed to anneal outside the left homology arm of the donor repair template and the SNAP tag, respectively. Positive clones were expanded prior to use.

| Cell culture
Transgenic and non-transgenic HEK293T cells were maintained in growth media at 37ºC 5% CO 2 ; Dulbecco modified eagles medium (DMEM) supplemented with 10% fetal bovine serum (FBS). In the case of transgenic lines, the growth medium also contained blasticidin (2.5 μg/mL). Once confluent, cells were dislodged from the flask surface by gentle shaking after incubation in 0.25% trypsin and a cell pellet formed following 5 minutes 1000 g centrifugation. For subsequent assays, cells were re-suspended in DMEM supplemented with 10% FBS and seeded at a density of 40-50 000 cells/well on poly-D-lysine treated clear-bottomed white-walled (radioligand binding), black-walled (internalization assay) 96 well plates or 8-well Nunc Lab-tek chambered coverglass (No. 1.0 borosilicate glass bottom) (imaging). Cells were incubated at 37ºC 5% CO 2 overnight prior to assay.
The human embryonic stem cell line, HUES7 (ES), was gifted from Harvard University under Material Transfer Agreement 51 and approved for use by the UK Stem Cell Steering Committee (Ref no 17/12/2007). Undifferentiated cells were cultured under feeder-free conditions on Matrigelcoated surfaces in the presence of MEF (mouse embryonic fibroblast) conditioned medium supplemented with 4 ng/ mL of bFGF (basic fibroblast growth factor), as previously described. 34,49 Transgenic haplotype variant lines were maintained in the same conditions in presence of 2 μg/mL of blasticidin in order to maintain the expression level of transgenes. For monolayer differentiation into cardiomyocytes, cells were first transitioned to the E8 culture medium (Life Technologies) and banked, and then differentiated as previously described. 52 Fibroblast differentiation was as follows: Transgenic cell lines were first seeded at 40 000 cells/ cm 2 in a Matrigel-coated T25 flask and cultured in Essential 8 medium until reaching 80% confluence. The culture was then treated with StemPro34 (Life Technologies) medium supplemented with 10 ng/mL of BMP4 (bone morphogenetic protein 4, R&D systems) and 8 ng/mL of Activin A (Life Technologies) and cultured for 48 hours; on day 2, the medium was switched into RPMI medium added with B27 supplement and 10% FBS; on day 4, the medium was changed into the fibroblast culture medium, which containing DMEM supplemented with 10% FBS, 1X GlutaMAX and 1X NEAA (non-essential amino acids). Differentiation into specific cell types was confirmed by immunohistochemistry to identify cell type-specific markers (Supporting Information Figure S1).

| Radioligand binding
Cells were plated at densities of 50 000 (HEK293T) or 70 000 (ES) cells per well in white-walled, clear-bottomed 96 well plates in growth media and left overnight. Growth media was aspirated and cells were incubated in 100 µL serum-free media (HEK293T:DMEM; ES: RPMI conditioned media supplemented with B27 supplement (Life Technologies)) in the presence or absence of 10 pM-10 µM of CGP 12177, CGP 20712A, ICI 118551 or propranolol (for competition binding assays), and the presence or absence of 10 µM propranolol (for saturation binding assays). The addition of 100 µL of serum-free medium or serumfree medium containing 0.01-20 nM (saturation) 3 H-CGP 12177 immediately followed and cells were incubated for 2 hours at 37°C 5% CO 2 . Cells were washed twice in PBS (Phosphate Buffered Saline) and 200 µL Microscint 20 added to each well and a white bottom and a topseal added to each plate. Plates were then left overnight and readings taken (counts per minute; CPM) on a TopCount Microplate Scintillation Counter (Packard Instrument, CT).

| SNAP labeling
The SNAP-tag labeling system allows real-time live-cell imaging of the protein of interest following labeling of the 20 kDa SNAP-tag which can be engineered to be expressed on the N-or C-terminus of the protein. The SNAP-tag represents O6-alkylguanine-DNA alkyltransferase that reacts specifically and rapidly with benzylguanine (BG) derivatives resulting in an irreversible covalent linking with a fluorescent probe. 53 Cells were incubated in growth medium in the presence of 0.5 µM (imaging), 0.

| Confocal microscopy
Cells were plated at varying densities on 8-well Nunc Labtek chambered coverglasses in growth media and left overnight. Slides were pre-coated with poly-D-lysine (HEK293T) or Matrigel (ES progenitor cells and cardiomyocytes) or used without pre-coating in the case of fibroblasts. The following day the media was removed and cells were SNAP labeled with SNAP-Surface Alexa Fluor 488 (SNAP-488) as described above. Cells were imaged on either the Zeiss LSM710 or Zeiss LSM880 confocal microscopes, both with a Zeiss Axio Observer Z1 stand (Carl Zeiss, Germany) using 488 nm Argon laser excitation (to image SNAP-488), a 493-630 nm emission range, and a plan-apochromat 40× 1.2NA oil immersion objective. Pinhole was set at 1 airy unit and gain and offset settings were kept constant within the experimental day. Cells were imaged in the presence or absence of formoterol and processed using Zen Black (2012) software (Carl Zeiss, Germany).

| Immunocytochemistry
To confirm the cell type of ES-derived fibroblasts and ESderived cardiomyocytes, the dissociated fibroblasts and cardiomyocytes were seeded into Matrigel-coated 96-well plate at 55,000 cells/well and maintained for at least 3 days. The culture was fixed with 4% paraformaldehyde (diluted in PBS) for 12 minutes prior to being stored at 4°C for up to one week. On an experimental day, cells were incubated with 0.01% Triton X-100 (diluted in PBS) for 15 minutes and washed twice in 0.05% Tween 20 (diluted in PBS) to perforate membranes. Cells were then incubated in 5% human serum (diluted in PBS) for 30 minutes prior to antibody treatments. The fibroblast cultures were incubated for 1 hour with primary sheep anti-human CD90 antibody (No. AF2067, 1:50 dilution, R&D systems) and then 1 hour with secondary donkey anti-sheep AlexaFluor633 (No. A21448, 1:500 dilution, Thermo Fisher Scientific). The cardiomyocytes were incubated for 1 hour with primary mouse anti-human α-actinin antibody (No. A7811, 1:800 dilution, Sigma-Aldrich) and 1 hour with secondary goat anti-mouse AlexaFluor 488 (No. A11001, 1:1000 dilution, Thermo Fisher Scientific). Both cell types were incubated for 15 minutes with 0.5 µg/mL of DAPI (No. D9542, 1:500 dilution, Sigma-Aldrich) to stain nuclear regions. Immunofluorescence images were captured with a 20x NA 0.45 long working distance objective on an Operetta High-Content System (PerkinElmer, UK) with a 300 W Xenon lamp (360-640 nm continuous spectrum) laser excitation.

| Ligand-induced receptor internalization
Cells were plated at densities of 40 000 (HEK293T) or 70 000 (ES progenitor) cells per well in poly-D-lysine/ Matrigel-coated black-walled, clear-bottomed 96 well Greiner plates in growth media and left overnight. The following day, the media was removed and cells were labeled with SNAP-488 as described above. Following labeling the cells were washed once in imaging buffer (HBSS or modified ES:HBSS), and then incubated for 1h in imaging buffer in the presence or absence of 100 µL formoterol (10 pM-1 µM), salbutamol (100 pM-100 µM) or salmeterol (1 pM-10 µM). Buffer was then removed and cells washed once more in buffer and then fixed by incubating in 10% ice-cold formalin solution (in PBS) for 15 minutes prior to 10 minutes incubation in ice-cold PBS and a 15 minutes incubation in PBS containing 2 μg/mL of Hoechst 33342 stain. Cells were then washed once in PBS and then 100 µL PBS replaced per well and the 96well plate was wrapped in foil and stored at 4°C for at least 24h. Four confocal images were recorded per well on the Molecular Devices ImageXpress Ultra confocal plate reader (Molecular Devices, CA) equipped with a Plan Fluor 40x NA 0.6 extra-long working distance objective using 405 nm laser excitation with a 477/60 band pass filter (to image Hoechst 33342) and 488 nm laser excitation with a 525/50 band pass filter (to image SNAP-488). The images were analyzed using the granularity analysis algorithm within the MetaXpress software (Molecular Devices, CA) which identifies fluorescent species, "granules," defined by a fluorescence intensity threshold and size limits. Size limits detailing SNAP-labeled receptor or Hoechst 33342 stained nuclear regions, for both cell types are as follows. HEK293T; receptor 7-15 µm in width and nuclear regions 6-9 µm in width. ES; receptor 1-3 µm in width and nuclear regions 5-15 µm in width. As the intensity of these regions depended on the efficiency of their labeling, threshold background levels were set per plate and data were normalized to allow comparison across experiments.

| Fluorescence correlation spectroscopy
Cells were plated on 8-well Nunc Lab-tek chambered coverglass (No. 1.0 borosilicate glass bottom) pre-coated with Matrigel (ES progenitor cells, cardiomyocytes), poly-D-lysine (HEK293T) or without coating (fibroblasts) in growth media and left at least 24h before analysis. Cells were labeled with SNAP-488 (0.1 µM) for 30 minutes at 37ºC and then washed twice in warm imaging buffer followed by a longer 20 minutes wash before finally being left in 180 µL imaging buffer at 37ºC until use. Cells were then incubated in the absence or presence of the β2AR agonist, formoterol (10 minutes, 10 nM, 37°C) before being allowed to cool to 24°C. FCS readings were taken on either a Zeiss LSM880 microscope on a Zeiss Axio Observer Z1 stand (HEK293T) or on a Zeiss LSM510NLO ConfoCor 3 microscope (ES progenitor cells, fibroblasts, cardiomyocytes) both fitted with 40X c-Apochromat 1.2 NA waterimmersion objectives using an Argon laser at 488 nm excitation with emission collected through a 505-610BP filter (LSM10NLO) or a 508-691 BP (LSM880) and the pinhole set to 1 Airy unit as previously described. 54 Briefly, the detection volume was positioned in x-y using a live confocal image and then in z following an intensity z-scan ±2 µm (in 0.25 µm intervals) from the approximated membrane position. For HEK293T cells, the detection volume was placed at the peak of the intensity z-scan and fluorescent fluctuations collected for 1 × 20-30s at ~0.5 kW/cm 2 laser power. For ES and progenitor cells sequential 1 × 20-30s reads at ~0.5 kW/cm 2 laser power were taken at successively decreasing z positions (0.25-0.50 µm intervals) on the same cell and the membrane trace was identified from the first trace to show typical membrane fluctuation patterns and the associated autocorrelation curves.
Autocorrelation curves were analyzed within Zen Black (2012) software and fitted with a diffusion model consisting of two components, the first depicting a 3D (free SNAP) and the second a 2D (receptor) diffusing component. 42 The autocorrelation function (G(τ)) in relation to a fluorescent species diffusing through a 3D Gaussian detection volume can be defined as follows: where; G ( ) is the normalized intensity autocorrelation function which describes N number of fluorescent particles where f i is the fraction of the species, i, of a total m, having a dwell time of τ Di . A is a pre-exponential factor to account for the photophysics of the fluorphore where T which is the percentage of molecules in the triplet state and τ T is the lifetime of the triplet state. 39 structure parameter that represents the ratio of the vertical to the radial axis of the confocal volume and can be determined experimentally. 54 Considering that the diffusion of the SNAP-tagged β2AR is limited to the 2D plane of the plasma membrane, S → ∞ and Equation (1) simplifies to; We used a combination of Equations (1) and (3) to model a diffusion profile to describe a first component depicting free SNAP-488 (confined to that of free SNAP-488, τ D1 = 20-80 µs) and a second component that represented the membrane receptor-bound SNAP-488 (τ D2 ). Where necessary, an offset, k, was included to allow the autocorrelation curve to be fitted with an asymptote of >1. This is commonly seen where there is global photobleaching in the fluorescence fluctuations, for instance, due to a high immobile receptor fraction, or membrane movement within the volume, which results in a curve transposed in y. On average the offset was 0.16 ±0.04% of the measured amplitude, reads were excluded if the offset was > 5% of the amplitude, or if an asymptotic value had not been reached. For the trace reads collected from the HEK293T lines, the initial 5 s were discarded prior to modeling to compensate for the bleaching of the fluorophore.
The precise radial and vertical axes of the confocal observation volume were determined on each experimental day by measuring the diffusion of 20 nM Rhodamine 6G as previously described. 54 Fluorescence fluctuation measurements were also analyzed using PCH (photon counting histogram) within Zen Black (2012) software. PCH analysis reports the average molecular brightness of the measured population and can be used to define whether single (one-component) or multiple brightnesses were present. 55 PCH uses a frequency distribution of photon counts within a defined time bin. A time bin of 0.1 ms was applied to all raw traces, appropriate for the diffusion speed of the membrane protein SNAPβ2AR being studied and analyzed. PCH histograms were fitted to either a one-or two-component PCH model which was determined by goodness of fit criteria at higher photons per bin values, with the data displayed on a linear-log scale. A systemdependent first-order correction was used, which accounts for photons detected outside of a true Gaussian detection volume, determined from the calibration data using Rhodamine 6G and was set at 0.3 for the Zeiss LSM510NLO ConfoCor 3 microscope and 0.6 for the Zeiss LSM880.

| Data analysis
Determination of agonist potency, antagonist affinity, and equilibrium dissociation constants were made by fitting data within GraphPad Prism version 7 for Windows (GraphPad Software, San Diego California USA, www.graph pad.com).
Statistical significance was determined, where appropriate, using Mann-Whitney, one-way/two-way ANOVA, and Kruskal-Wallis with Tukey or Dunn's multiple comparison test if P < .05 (statistically significant). A minimum of three independent experiments was chosen for all experimental work. This was based on the variance of the data obtained and power analysis that predicted a high probability of observing small differences in measured parameters. For example, the probability of detecting a change in pEC 50 of 0.5 (3-fold) with n = 3 is 0.89. For n = 4 this increased to 0.96.
The HUES7 cell line was a kind gift by Chad Cowan and Doug Melton. 51 The BL15 was provided by David Darling. 56 The HEK293T cell line was obtained from the American Type Culture Collection (CRL-11268).

| Characterization of ES cell progenitor cells expressing SNAPβ2AR
Relative expression levels and equilibrium dissociation constants determined from radioligand-binding with 3 H-CGP 12177 were similar for the four ES progenitor cell lines expressing different haplotypes of SNAPβ2AR (P > .05 oneway ANOVA, Figure 1, Table 1). The non-transgenic ES cells displayed negligible-specific binding of 3 H-CGP 12177 (data not shown). Whole-cell competition binding displayed typical β2AR pharmacology in all four ES cell lines with high affinity for propranolol, CGP 12177, and the β2AR selective antagonist ICI 118551 and low affinity for the β1AR selective antagonist CGP 20712A ( Figure 2, Table 1). Similar results were observed when the SNAPβ2AR constructs were stably expressed in HEK293T cells (Table 1). Confocal imaging of the SNAPβ2AR expressed within the ES cell lines showed that membrane localization of the receptor could be clearly defined and underwent agonist-induced internalization (Figure 3). A membrane-impermeable SNAPtag substrate was used to label the receptor and therefore any intracellular receptor must have predominantly originated from the cell surface. Unstimulated ES progenitor cells expressed the SNAPβ2AR at the cell surface but there was also clear evidence of constitutive receptor internalization ( Figure 3A,C). Internalized receptor appeared to exist as clustered multiple small granules in contrast to the larger clusters of merged granules observed within the transgenic HEK293T cell lines (Supporting Information Figure S2). Agonist-induced internalization was clearly observed following 30 minutes treatment with 1 µM formoterol although high levels of receptor also remained on the cell membrane ( Figure 3B,D). Very similar receptor localization was observed across the four transgenic ES cell lines in the presence and absence of the βAR agonist formoterol. Non-transgenic ES cells displayed no discernible SNAP labeling (data not shown).

| Quantitative evaluation of agonistinduced receptor internalization using high content screening
To determine agonist concentration-response relationships for ligand-induced internalization of the SNAPβ2AR, we undertook high content confocal imaging of the four haplotypes as described in Section 2.8 ( Figure 4). Agonists displayed the same rank order of potency in the four ES transgenic lines with the two partial agonists salbutamol and salmeterol producing lower maximal responses than the full agonist formoterol (Table 2; Figure 4). Internalization in response to salmeterol  (Table 1) and salbutamol was quite varied and, apart from the cell line expressing the Arg16Gln27 (RQ) haplotype of SNAPβ2AR, there were some experiments where there was no response to salbutamol, salmeterol or both ( Table 2). The EC 50 values for salmeterol and salbutamol were only determined from those experiments that produced a measurable response and these were very similar across the haplotypes (two-way ANOVA, P > .05). Basal levels of internalization per cell were similar across all haplotypes for the separate cell types (P > .05, oneway ANOVA). The internalization achieved with salbutamol was greatest for the RQ haplotype in both ES cells (Figure 4; Table 2; P < .05, two-way ANOVA) and HEK293T cells (Supporting Information Figure S3, Table 2).

| Reduced expression of SNAPβ2ARs following the differentiation of ES progenitor cells into fibroblasts and cardiomyocytes
Following the differentiation of human ES cells into fibroblasts or cardiomyocytes (confirmed with the fibroblast marker Thy-1 (CD90) and the cardiomyocyte marker α-actinin; Supporting Information Figure S1), the expression of SNAP-labeled receptors was markedly reduced ( Figure 5). Only diffuse membrane labeling was observed in differentiated fibroblasts ( Figure 5A-D), while SNAPβ2AR was not detectable within the CM populations using confocal imaging ( Figure 5E,F). Fibroblasts did, however, display constitutive receptor internalization ( Figure 5C). However, following treatment with formoterol (10 nM, 20 minutes) no major difference in the extent of membrane expression or intracellular granules was observed in the differentiated fibroblasts ( Figure 5B,D).

| The use of FCS to study GPCR dynamics at low receptor expression levels
In order to quantify the expression of the SNAPβ2AR fusion protein at the cell membrane and study its dynamics in these low-expression systems, we employed fluorescence correlation spectroscopy (FCS). FCS monitors fluorescently labeled  (Table 1) species as they pass through a small confocal volume (0.2 fl, 0.2 μm 2 of membrane) 40 which can be placed within a defined region of the cell membrane. This method allows not only the quantification of such parameters as particle number per area (N/µm 2 ), but also provides a measurement of their average diffusion coefficient (D, µm 2 s −1 ). FCS works particularly well in low expressing systems where greater sensitivity is imparted by the larger fluctuations in fluorescence caused by low fluorescent particle numbers. 40 As a proof of principle, we optimized our protocol by studying the SNAPβ2AR in a low expression system using a CRISPR/Cas9 genome engineered HEK293T cell line that expressed the SNAP-tagged β2AR under endogenous promotion. These data were compared with those obtained in HEK293T cells overexpressing the SNAPβ2AR following lentiviral transfection. Similar to the SNAPβ2AR in differentiated cardiomyocytes, the SNAPβ2AR within the CRISPR/Cas9 edited cells could not be visualized by standard confocal imaging (data not shown). It was possible, however, to measure fluorescence fluctuations on the cell membrane using FCS ( Figure 6A). The countrate (Hz) was 20 to 30-fold greater in the overexpressing cell line compared to the CRISPR/Cas9 genome-edited cells and this resulted in significant bleaching in the overexpression line ( Figure 6B). As a consequence, the first 5 seconds of the fluctuation trace were excluded from all measurements before the analysis of autocorrelation curves. Autocorrelation curves were constructed from these fluctuation traces ( Figure 6C,D) and illustrate the difference in expression level where G(0) is inversely proportional to N. Fitting of these curves showed that two distinct diffusing populations could be resolved with different dwell times in the measurement volume ( Figure 6C,D). The faster-diffusing component (dwell time τ D1 ; 20-80 µs) was consistent with free unbound SNAP-488 dye diffusing within the detection volume. 43 The slower diffusing component (τ D2 ) represented the SNAPβ2AR expressed on the cell membrane. The diffusion coefficient (D) of this second component, however, was very similar between cell lines ( Figure 6F, CRISPR: 0.12 ± 0.01, n = 24; OE: 0.10 ± 0.01 µm 2 s −1 , n = 30, respectively). The particle number recorded from the CRISPR/Cas9 edited cells for the SNAPβ2AR (37.4 ± 8.4 N/µm 2 , n = 24) was significantly lower than that seen in the over-expressing (OE) cells (349.1 ± 39.0 N/µm 2 , n = 30, Figure 6E, unpaired t test P < .0001). Fluctuations traces were then analyzed by photon counting histogram analysis (PCH) which reports the average molecular brightness of the measured population and can be used to define whether a single (one-component) or multiple brightnesses are present. An increase in brightness can indicate clustering or oligomerization of the taggedspecies potentially as a result of membrane reorganization prior to GPCR internalization. 43 All HEK293T traces fitted F I G U R E 3 Confocal imaging of ES progenitor cells expressing the Gly16Gln27 (GQ) SNAPβ2AR haplotype following labeling with 0.5 µM surface-SNAP-Alexafluor488. Constitutive internalization can be observed in unstimulated cells (A) and inset (C). Increased internalization is observed following stimulation with 1 µM formoterol, 30 minutes (B), and inset (D). Multiple intracellular granules are observed which appear to form clusters (arrows) but high membrane localization is also still observed. Images are representative of images collected from at least three independent experiments to a one-component photon counting histogram (PCH) with the overexpression (OE) cell line exhibiting a significantly higher molecular brightness (ε, counts per molecule (cpm) per second; Supporting Information Figure S4, CRISPR: 20 728 ± 2730 cpm.s −1 , n = 24; OE: 34 620 ± 3322 cpm.s −1 , n = 30; P < .01 un-paired t test).

| FCS measurements in ES progenitor cells and ES-derived fibroblasts and cardiomyocytes
The use of FCS in human ES cells and derived cell types required additional steps to locate the cell membrane (Figure 7). Typically the upper membrane of a cell expressing a SNAPlabeled membrane receptor can be identified as the peak of a z-intensity scan; however, this was not the case for the ES and derived cell types. The non-uniform nature of these cells coupled with the short sampling time meant that fluctuation traces which most closely described the apical membrane could be located up to 1 µm above this peak. Primary cells are routinely non-uniform and in the case of the cells, we have studied here, very thin. While the waist of the confocal volume is around 0.25 µm in diameter the vertical axial range is closer to 1 µm. A confocal volume, therefore, placed on a thin primary cell could theoretically encompass both apical and basal membranes. To ensure we could capture the trace that most closely represented that of the apical membrane we took sequential 20-30 seconds trace reads every 0.25 µm through the cell starting above the cell membrane ( Figure 7B,D). The first trace showing clear membrane-related fluctuations was used to determine membrane diffusion properties of the receptor. Two distinct decay populations could be resolved relating to the SNAP-488 within these progenitor cells similar to those observed in HEK293T cells ( Figure 7E). Single point FCS was carried out on ES progenitor cells, differentiated ES-derived fibroblasts, and ES-derived cardiomyocytes expressing the GQ variant of SNAPβ2AR to determine relative membrane particle number and diffusion properties of the SNAPβ2AR (Figure 8). Fluorescence fluctuations could be detected in all three cell types and traces representing the apical membrane could be readily identified ( Figure 8A-C). Two distinct diffusing species could be detected in all three transgenic cell types ( Figure 8D-F). The drop in receptor expression level observed using confocal microscopy was confirmed using FCS where the average particle number recorded from the ES-GQ cells (186.8 ± 33.2 N/µm 2 , n = 14) was significantly greater than that seen in differentiated GQ fibroblasts (60.6 ± 10.9 N/ µm 2 , n = 28) and GQ cardiomyocytes (26.1 ± 4.6 N/µm 2 , n = 23) ( Figure 8G, Kruskal-Wallis, fibroblasts P < .01, cardiomyocytes P < .0001). The diffusion coefficients (D) were, however, very similar between cell types ( Figure 8H; ES cells, 0.054 ± 0.005 µm 2 s −1 (n = 14); fibroblasts, 0.055 ± 0.004 µm 2 s −1 (n = 28) and cardiomyocytes, 0.052 ± 0.005 µm 2 s −1 (n = 23); Kruskal-Wallis, P > .05). The majority of traces, from all cell types (ES; 92.9%, fibroblasts; 75%, cardiomyocytes; 95.7%), fitted to a one-component fit with a brightness consistently <16 000 cpm.s −1 ( Figure 8I). The minority of traces fitted to a multiple component fit consisting of two distinct brightnesses, the first being <16 000 cpm.s −1 and the second being >16 000 cpm.s −1 ( Figure 8I). The average molecular brightness (ε) was unaffected ( Figure 8I; ES cells 6577 ± 920 cpm.s −1 (n = 14); fibroblasts 5197 ± 396 cpm. s −1 (n = 28); cardiomyocytes 6074 ± 866 cpm.s −1 (n = 23); Kruskal-Wallis, P > .05).

| Monitoring receptor internalization using FCS
While high-content imaging provides a powerful methodology to study agonist-internalization of SNAPβ2ARs in cells, FCS analysis allowed us to investigate the effect of an agonist in a membrane micro-domain of single living cells. ES-derived fibroblasts did not show agonist-induced internalization when monitored using confocal microscopy. To confirm this finding we undertook FCS studies on ES-derived fibroblasts using the SNAPβ2AR-RQ haplotype. The RQ haplotype was chosen since it consistently responded to agonist challenge with the highest level of receptor internalization in ES progenitor cells (Figure 4). In keeping with the observations made using confocal microscopy ( Figure 5), agonist treatment (10 minutes, 10 nM formoterol) had no significant effect on SNAPβ2AR-RQ particle number (N/µm 2 ) in ES-derived fibroblasts ( Figure 9A, Control; 173.2 ± 23.8, n = 21. Formoterol; 150.3 ± 24.6, n = 22, Mann-Whitney, P > .05). The diffusion coefficient (D, µm 2 s −1 ) was also unaffected by agonist treatment (Figure 9B. Control, 0.041 ± 0.003, n = 21; Formoterol, 0.043 ± 0.004, n = 22, Mann-Whitney, P > .05). Finally, the traces were also subject to PCH analysis to determine the average molecular brightness (ε). All traces fitted to a one-component PCH with no significant change observed following formoterol treatment ( Figure 9C, Control; 6939 ± 891 cpm.s −1 , n = 21. Formoterol; 6901 ± 633 cpm.s −1 , n = 22).
To confirm that we were able to observe agonistinduced changes in the parental ES progenitor cells within F I G U R E 4 Quantification of agonist-induced internalization of SNAPβ2AR in ES progenitor cells expressing the four different haplotypes: A, Arg16Gln27 (RQ); B, Arg16Glu27 (RE); C, Gly16Gln27 (GQ) and D, Gly16Glu27 (GE). A-D, Single experiments (n = 6-9) were performed with duplicate determinations and plotted data represent mean ± SEM of the number of SNAP-488 labeled granules expressed as a percentage of that obtained with 1 µM formoterol after 1 hour stimulation with βAR agonists; formoterol (red circle), salmeterol (green square) and salbutamol (blue triangle). EC 50 and E MAX values alongside the number of separate experiments performed are provided in Table 2. Derived Emax values (mean ± SEM, n = 6-9) for salbutamol (E, blue) and salmeterol (F, green) are also shown. Raw data were subject to 2-way ANOVA with Tukey's multiple comparison test; *P < .05, **P < .01 the membrane using single point FCS, we studied the ES-SNAPβ2AR-RQ progenitor cell line which exhibited agonistinduced internalization using both confocal imaging and high content imaging (Figures 3 and 4). ES-SNAPβ2AR-RQ cells subject to formoterol treatment (10 minutes, 10 nM formoterol) displayed significantly reduced SNAPβ2AR particle numbers ( Figure 9D, 143.8 ± 23.1 N/µm 2 , n = 27) compared to those treated with vehicle (341.9 ± 60.0 N/µm 2 , n = 27. Mann-Whitney, P < .01). The diffusion coefficient (D, µm 2 s −1 ) of the SNAPβ2AR was, however, not affected by agonist treatment (Figure 9E; control, 0.045 ± 0.004, n = 27; formoterol treatment, 0.039 ± 0.003, n = 27). In contrast, the molecular brightness (ε) of all traces fitted to a onecomponent PCH and was significantly increased following formoterol treatment ( Figure 9F; control; 3061 ± 415 cpm. s −1 , n = 26; formoterol, 4121 ± 531 cpm.s −1 , n = 27. P > .05). These data suggest that the SNAPβ2AR-RQ is clustering on the cell membrane in ES progenitor cells prior to internalization following agonist treatment.

| DISCUSSION
In the present study, we have used a combination of FCS and ES cell technology to investigate the impact of cell differentiation on receptor mobility and internalization in human ES progenitor cells and in ES-derived cardiomyocytes and fibroblasts. We generated four different ES progenitor cells expressing SNAP-tagged β 2 -adrenoceptors with one of the four haplotypes involving single-nucleotide polymorphisms at position 16 (arginine or glycine) and position 27 (glutamine or glutamate). All expressed the SNAPβ2AR to a similar level and all four haplotypes could be internalized in response to the full agonist formoterol, as determined by high content confocal imaging. Salmeterol and salbutamol acted as partial agonists for this response, and in the case T A B L E 2 Agonist potencies (pEC 50 ) and maximal responses (Emax) obtained for formoterol, salbutamol and salmeterol-mediated production of intracellular granules as a measure of agonistinduced internalization. If wells did not respond to agonist, Emax was set to zero and data were not included in pEC 50 [17][18][19][20] In agreement with the data presented here, previous in vivo cardiovascular experiments in human volunteers have linked the RQ haplotype with enhanced receptor desensitization, while the GE haplotype produces a blunted susceptibility to agonist-induced desensitization. 57,58 A striking feature of the initial experiments performed with ES-derived cardiomyocytes and fibroblasts was that the expression of SNAP-tagged β 2 -adrenoceptors dropped dramatically following the differentiation process. Although the receptor was still detectable in ES-derived fibroblasts, this was not the case with ES-derived cardiomyocytes where the SNAPβ2AR fluorescence could not be distinguished from background autofluorescence. One approach that has the sensitivity to monitor receptors in membrane microdomains at very low receptor expression levels is FCS. 40 FCS monitors fluorescently labeled species as they pass through a small confocal volume (0.2 fl, 0.2 μm 2 of membrane) 40 and allows the quantification of receptor expression in terms of fluorescent particle number (N/µm 2 ) and their diffusion characteristics (D, µm 2 s −1 ). FCS works particularly well in low expressing systems where greater sensitivity is imparted by the larger statistical variation caused by low fluorescent particle numbers. 40 As a consequence, the autocorrelation function G(τ) at τ = 0 is inversely proportional to the particle number. 40 As a first step to apply FCS to study cell surface receptor expression in membrane microdomains of single ES-derived cells, we undertook a comparison of SNAPtagged β 2 -adrenoceptors in HEK293T cells overexpressing F I G U R E 5 Confocal imaging of SNAPβ2AR in (A-D) human ES-derived fibroblasts expressing the Arg16Glu27 haplotype or (E,F) human ES-derived cardiomyocytes expressing the Arg16Gln27 haplotype. All cells were labeled with 0.5 µM surface-SNAP-Alexafluor488. E, No visible SNAP labeling could be observed through standard confocal imaging of ES-derived cardiomyocytes; F, brightfield image is included to illustrate the presence of cells. While SNAPβ2AR could be imaged in ES-derived fibroblasts (A) and inset (C) it was not possible to identify agonistinduced internalization of receptor following formoterol (10 nM; 20 minutes) treatment (B, and inset (D)). A-D, Diffuse membrane expression could be observed, more strongly apparent at membrane-membrane boundaries (arrow). SNAPβ2AR following lentivirus transfection or in CRISPR/ Cas9 genome-edited HEK293T cells expressing SNAPβ2AR under endogenous promotion. In HEK293T cells, the endogenous expression of β 2 -adrenoceptors is very low 59 and SNAPβ2ARs were not detectable by confocal microscopy in CRISPR/Cas9 genome-edited HEK293T cells. The use of FCS was, however, able to detect the large difference in expression levels (CRISPR cells, 37.4 particles per µm 2 ; lentivirus over-expressing cells, 349.1 particles per µm 2 ) without a significant difference in the diffusion coefficient of the β 2 -adrenoceptor between the two cell types. Interestingly, the average molecular brightness of the diffusing fluorescent particles containing SNAPβ2ARs in over-expressing cells was significantly higher than that in the CRISPR/Cas9 genome-edited cells. This perhaps indicates that the clustering of the β 2 -adrenoceptor is occurring at higher expression levels similar to the observation recently made for CXCR4, which forms transient homodimers at high receptor densities. 60 The difference in brightness of SNAPβ2ARs between overexpressed cells and CRISPR/Cas9 genome-edited HEK293 cells is approximately twofold. It has been previously reported that β 2 -adrenoceptors have a propensity to form dimers and higher order oligomers 36 but at the present time, we cannot conclude whether the changes observed for F I G U R E 6 Fluorescence correlation spectroscopy of CRISPR/Cas9 edited HEK293T cells. Representative FCS traces taken on the cell membrane of (A) HEK293T cells that have been CRISPR/Cas9 genome-engineered to express SNAPβ2AR under endogenous promotion or (B) HEK293T cells overexpressing (OE) SNAPβ2AR. C, Autocorrelation analysis for CRISPR/Cas9 genome-edited cells and (D) OE cells. Raw (blue) and fitted data (green) are displayed. E, Individual mean particle number N/µm 2 and (F) diffusion coefficient (D; µm 2 s −1 ) measurements are plotted for CRISPR (closed circles) and OE (open circles) cell lines. In (E) please note the different scales for the CRISPR and OE particle numbers. In (F) the mean ± SEM of all measurements (dotted line) are displayed for each cell line F I G U R E 7 Positioning the confocal volume on the cell membrane of ES or differentiated cell type. A, The FCS detection volume, depicted by the red crosshair intersect, is positioned on the live high-zoom confocal image of an ES-derived fibroblast expressing the Gly16Gln27 haplotype of SNAPβ2AR labeled with 0.1 μM surface-SNAP AlexaFluor488. The z-position of the upper membrane is approximated by eye, and bright clusters are avoided as these could represent immobile receptor. B, An intensity z-scan is recorded at low laser power at the crosshair, at 2 µm above and below the approximated position of the upper cell membrane (position 0.0 µm) in 0.5 µm steps. The FCS detection volume is moved to +1.5 µm above the peak intensity reading (eg, moved to position 0.5 µm). C, 1 × 30 seconds FCS trace reads are collected at experimental laser power every 0.25 µm moving toward the position of peak intensity. FCS trace reads obtained at the z positions displayed in (B) are shown. D, Autocorrelation curves are constructed for individual traces within Zen Black (2012) software. Curves are modeled with a 1 × 3D + 1 × 2D autocorrelation model which includes a triplet component and an offset to allow for membrane drift (see Section 2.9). Curves can be excluded for several reasons, including if the 2D component is too fast/slow to depict a membrane protein (purple trace), if the curve fails to reach a lower plateau or allowed offset is too great (red trace), or if the curve cannot be described by the model. The blue and green traces contain a membrane component; however, the blue trace is selected as this trace has a greater proportion of 2D vs 3D fluctuations. E, Autocorrelation curve and model for the selected blue trace. The 3D dwell time, τ D1 = 74 µs and represents free surface-SNAP-AlexaFluor488, the 2D dwell time, τ D2 = 146 ms and represents SNAP labeled SNAPβ2AR. Blue-raw data, green-fit model the β 2 -adrenoceptor reflect changes in monomer-dimer or dimer-higher-oligomer equilibria.
FCS was able to quantify the large difference in β 2adrenoceptor expression levels between ES progenitor cells expressing the Gly16Gln29 (GQ) haplotype (186.8 particles per µm 2 ) and the equivalent ES-derived GQ fibroblasts (60.6 particles per µm 2 ) and ES-derived GQ cardiomyocytes (26.1 particles per µm 2 ). Diffusion coefficients were similar across all three ES cell lines (0.05 µm 2 s −1 ) and approximately half the value of the diffusion coefficients obtained in HEK293T cells (0.1 µm 2 s −1 ) indicative of their different membrane compositions and organizations. The particle number detected by FCS in ES-progenitor cells expressing the Arg16Gln27 (RQ) phenotype (341.9 particles per µm 2 ) was higher than in the equivalent GQ cells and following differentiation to ES-derived fibroblasts, the particle number (173.2 particles per µm 2 ) was also higher than that detected in the equivalent GQ haplotype, but significantly lower than in the (RQ) ESprogenitor cells.
The RQ haplotype was chosen for studies of agonistinduced receptor internalization of SNAPβ2ARs using FCS since it consistently exhibited a robust agonist-induced receptor internalization in ES progenitor cells when monitored by high content confocal imaging.
Following agonist treatment of ES progenitor cells (10nM formoterol, 10 minutes), the SNAPβ2AR-RQ haplotype displayed a significant reduction in cell surface particle number from 341.9 particles per µm 2 to 143.8 particles per µm 2 F I G U R E 8 Cell membrane fluorescence correlation spectroscopy of human ES progenitor cells and differentiated fibroblasts and cardiomyocytes. Representative traces are shown for (A) ES progenitor cells, (B) ES-derived fibroblasts, and (C) ES-derived cardiomyocytes expressing the SNAPβ2AR with theGly16Gln27 haplotype. Autocorrelation traces (blue) and the resulting autocorrelation fit (green) are shown for (D) ES progenitor cells, (E) fibroblasts, and (F) cardiomyocytes. Individual experimental (G) membrane particle number N/µm 2 , (H) diffusion coefficients D (µm 2 s −1 ) and (I) molecular brightness ε, (cpm.s −1 ), are plotted with the mean (dotted line) ± SEM for ES (black closed circles), FIB (black open circles) and CM (orange closed circles). G, Average particle number recorded for ES progenitor cells is significantly greater than that recorded in ES-derived fibroblasts (P < .01) and ES-derived cardiomyocytes (P < .0001; Kruskal-Wallis). I, data were fitted to either a one-component or two-component PCH analysis and the first (1, left) and second (2, right) brightness values are plotted. For data that fit to a twocomponent analysis the second, brighter, population represented a small percentage of the total measured population; 5.1% (ES; n = 1), 5.7% (CM, n = 1) and 3.1% ± 0.9 (FIB, n = 7) which was consistent with agonist-induced receptor internalization. The diffusion coefficients of the receptor-containing complexes were not affected by agonist treatment but their molecular brightness (ε) was significantly increased, suggesting that there maybe clustering of β 2 -adrenoceptor on the cell membrane prior to internalization. It is notable that the brightness values observed in ES progenitor cells (and also fibroblasts and cardiomyocytes) were very much lower than those measured in HEK293 cells and this probably reflects the impact of the local cellular environment and fluorescence quenching. The role of receptor clustering on the cell membrane has been reported previously in association with the internalization of particular GPCRs including the β 2 -adrenoceptor 43,61 and the 1.35-fold increase in brightness observed in ES progenitor cells probably equates to an increase in the proportion of dimeric/oligomeric species as receptors are sequestered at sites of receptor endocytosis.
In marked contrast to the data obtained in ES progenitor cells, no significant reduction in particle number was evident in ES-derived fibroblasts following formoterol treatment in cells expressing either the Arg16Gln27 (RQ) or Gly16Gln27 (GQ) haplotypes. These data were, however, consistent with the observations made in ES-derived fibroblasts expressing the RQ haplotype using confocal microscopy ( Figure 5). Taken together these studies indicate that susceptibility to agonist-induced receptor internalization is dependent upon cell phenotype and can be changed following the differentiation of ES progenitor cells into fibroblasts. This may have important implications for drug discovery where attempts are being made to introduce signaling bias into agonist molecules that can stabilize specific conformations of GPCRs and alter their ability to couple to G proteins or β-arrestin molecules. 62 Reduced internalization will reduce the extent of signaling from intracellular endosomes. 10,11 Furthermore, the differential sensitivity of different cell phenotypes to agonist-induced internalization of the β 2 -adrenoceptor may also have implications for drug therapy with inhaled longacting β 2 -agonists. 63,64 In summary, the present study has demonstrated the power of FCS in investigating cell surface β 2 -adrenoceptors at the very low expression levels often seen in endogenously expressing cells. FCS was able to detect SNAP-tagged β 2 -adrenoceptor expression in both ES-derived cardiomyocytes and CRISPR/Cas9 genome-edited HEK293T cells, where the expression level was too low to detect them using standard confocal microscopy. Furthermore, the use of ES cell technology in combination with FCS allowed us to demonstrate that cell surface β 2 -adrenoceptors can be internalized in response to formoterol-stimulation in ES progenitor cells but not following their differentiation into ES-derived fibroblasts. This indicates that the process of agonist-induced receptor internalization is strongly influenced by cell phenotype and this may have important implications for drug treatment with long-acting β 2 -agonists.