Interrogating surface versus intracellular transmembrane receptor populations using cell-impermeable SNAP-tag substrates

Employing self-labelling protein tags for the attachment of fluorescent dyes has become a routine and powerful technique in optical microscopy to visualize and track fused proteins. However, membrane permeability of the dyes and the associated background signals can interfere with the analysis of extracellular labelling sites. Here we describe a novel approach to improve extracellular labelling by functionalizing the SNAP-tag substrate benzyl guanine (“BG”) with a charged sulfonate (“SBG”). This chemical manipulation can be applied to any SNAP-tag substrate, improves solubility, reduces non-specific staining and renders the bioconjugation handle impermeable while leaving its cargo untouched. We report SBG-conjugated fluorophores across the visible spectrum, which cleanly label SNAP-fused proteins in the plasma membrane of living cells. We demonstrate the utility of SBG-conjugated fluorophores to interrogate class A, B and C G protein-coupled receptors (GPCRs) using a range of imaging approaches including nanoscopic superresolution imaging, analysis of GPCR trafficking from intra- and extracellular pools, in vivo labelling in mouse brain and analysis of receptor stoichiometry using single molecule pull down.


Introduction
Membrane receptors, including ligand-gated ion-channels, G protein-coupled receptors (GPCRs), receptor-linked enzymes and, to an extent, transporters, sense extracellular stimuli and convert them into intracellular signals that control cellular function in myriad ways 1 . As such, these proteins are a major focus of drug discovery programs, with GPCRs serving as the largest class of targets 2 . Through an array of approaches, it has become clear that receptor signaling is not restricted to the cell surface but is fine-tuned by a dynamic interplay of receptors both on the surface and in a variety of intracellular compartments 3-6 . Developing techniques for dissecting the relative properties of these distinct pools is an emerging challenge for receptor biology.
Fluorescence microscopy is a powerful technique for direct observation and analysis of molecular processes within a living cell that has been applied extensively to the study of membrane receptors. The continuing development of bright and stable synthetic dyes 7-10 along with the engineering of self-labelling suicide enzymes, such as SNAP, CLIP and Halo-tags 11 , has spurred the application of targeted, high-resolution imaging in a number of biological contexts [12][13][14][15][16][17][18][19][20][21] .
Organic dyes that are covalently linked to proteins offer superior brightness, photostability and flexibility compared to fluorescent proteins 10,22 . Many organic dyes are cell permeable and therefore suitable for intracellular labeling. However, this permeability is undesirable when cell surface targeting is required, since confounding background signals can arise from labelled, untrafficked proteins or accumulation of the unlabeled dye in membranes and intracellular compartments. Similarly, fluorescent protein-tagged membrane proteins tend to give high background signals when fused to membrane proteins, since they are expressed, translated and trafficked within the cell. While membrane impermeable fluorophores exist (e.g. Alexa, Atto or Abberior dyes), many of these fluorophores have been shown to accumulate at the membrane 23 , while the recently developed bright and stable Janelia Fluors 24 and MaP dyes 8 are engineered to be membrane permeable. So far, the membrane permeability of a probe has been considered a feature of the fluorophore, with the consequence that imaging only the extracellular protein pool requires changing to a spectrally and photophysically distinct dye as shown recently 25 . While generally useful for qualitative analysis, such an alteration makes quantitative comparisons difficult. Thus, a strategy for rendering dyes impermeable without altering their intrinsic photophysical or spectral properties is needed.
Herein, we describe a subtle, yet powerful modification of O 6 -benzylguanine (BG), the substrate for the SNAP-tag, by installing a sulfonate on the leaving group's C8 position (termed SBG), rendering them impermeable towards the lipid bilayer while conserving reactivity with SNAP. Our general approach allows clean surface labelling of GPCRs in living cells with improved membrane localization and resolution by STED nanoscopy, as well as enhanced signalto-noise ratio and spread in vivo. Moreover, SBG-linked fluorophores open up the possibility to pulse-chase receptors in different compartments, as well as to perform single molecule pulldown (SiMPull) of surface versus intracellular receptor populations. We anticipate that, with this facile strategy, the majority of linked substrates can be rendered impermeable for studies of membrane protein dynamics at the cell surface.

Results
As a proof-of-principle, we first set out to design and synthesize membrane impermeable versions of SNAP-Cell® TMR-Star and SNAP-Cell® 647-SiR, two popular commerciallyavailable fluorophores for SNAP-tag labeling ( Fig. 1a; Scheme S1). Based on previous studies, which report that alterations at guanine position C8 are tolerated for enzymatic SNAP labelling 26 , we hypothesized that substituents on the BG would alter the permeability of the entire compound without interfering with labeling. Conveniently, this moiety would also be liberated upon SNAP labeling, thus removing any potential alterations to the photophysical properties of the protein bound fluorophore itself. As such, three moieties were examined and prepared as their TMR-and SiR-bearing reagent, namely the parent BG (with H at C8: "BG-TMR" and "BG-SiR"), a previously described 26 carboxylate CBG (with a linked COOH at C8: "CBG-TMR" and "CBG-SiR") and, finally a sulfonate (with a linked SO3H at C8: "SBG-TMR" and "SBG-SiR") (Fig. 1b,   c). Bearing in mind that sulfonates display a pKa < 0, SBG will be permanently negatively charged in physiological buffers, thereby unable to cross the lipid bilayer membrane and, presumably, repelled further by the negatively charged surface. Accordingly, SBG-TMR and SBG-SiR were obtained by coupling sulfonate-containing taurine to a protected CBG precursor, which after deprotection allowed straightforward amide coupling to NHS-activated fluorophores (see SI).
For initial assessment of labelling properties we chose TMR since covalent binding of a non-fluorogenic dye can easily be observed using fluorescence polarization. As expected, BG-TMR, CBG-TMR and SBG-TMR showed complete SNAP labeling in vitro as assessed by full protein mass spectrometry 27 (Fig. S1-S4) with labelling kinetics ~3-times slower for SBG-TMR (t1/2 = 51.3 seconds) versus BG-TMR (t1/2 = 17.8 seconds) yet complete within minutes (Fig. S5a).
An advantage of using a charged residue is increased solubility, and as such, SBG-TMR can be readily dissolved in pure PBS at a concentration >1.5 mM, while BG-and CBG-TMR need to be dissolved in DMSO (>1 mM) before dilution in PBS for usage. More importantly, BG-TMR (~80 µM in PBS, 1% DMSO) was not stable in solution at room temperature, precipitating within minutes to leave a steady-state concentration of ~17 µM in the supernatant (Fig. S5b). Notably, SBG-TMR remained in solution at ~70 µM without the addition of DMSO over three days.
We next tested the ability of modified BGs to label intracellular SNAP by expressing either a cytosolic-or nuclear-targeted SNAP-tag before applying BG, CBG or SBG-conjugated fluorophores. While labeling with 500 nM BG-SiR for 45 minutes at 37 °C produced clear fluorescence for both cytosolic and nuclear SNAP-tags, labeling with 500 nM CBG-SiR or SBG-SiR did not produce any substantial signal with either construct in transiently transfected HEK293 cells (Fig. 1d, e). Notably, background fluorescence in untransfected cells was highest for BG-SiR, lower for CBG-SiR and undetectable for SBG-SiR (Fig. 1d, e). As such, we decided to continue our characterization with SBG-SiR because it showed a robust decrease in membrane permeability compared to CBG-SiR. We next labeled cells expressing a SNAP-tagged GPCR, metabotropic glutamate receptor 2 ("SNAP-mGluR2"), with either BG-SiR or SBG-SiR. Both compounds produced clear fluorescence over a similar range of labeling concentrations (Fig. 1f-h), but signals from SBG-SiR were more confined to the plasma membrane (Fig. 1f, right) and showed less background labeling in untransfected cells (Fig. 1g, h). Together, these data validate the idea that addition of an anionic sulfonate group to BG can render an attached fluorophore membrane-impermeable for targeting of surface proteins with reduced non-specific labeling.
Due to their distinct spectral and photophysical properties, different fluorophores are required for multimodal applications. Based on the desire to prevent membrane permeability of different fluorophores at will, we asked if this approach was generalizable to a family of fluorophores spanning visible to far-red wavelengths. To do this, we synthesized and tested SBGconjugated Oregon Green (OG), TMR, Janelia Fluor 549 (JF549) and Janelia Fluor 646 (JF646) ( Fig.   2a-d), the latter showing superior brightness and photostability over their tetramethyl and silicon rhodamine counterparts 24 . In all cases, SBG-conjugated fluorophores clearly label surface receptors, with minimal labeling of intracellular SNAP-tags ( Fig. 2a-d).
To test if membrane-localized SNAP-tag labels offer advantages for cell biology, we turned to nanoscopic STED imaging. A dye with outstanding far-red performance in STED microscopy with respect to photostability and brightness is JF646 28 . Accordingly, we used JF646 SNAP-tag probes to target SNAP-mGluR2 in transiently transfected HEK 293 cells. Similar to what was observed by widefield microscopy (Fig. 2d), we observed mainly intracellular staining in fixed cells with BG-JF646 (Fig. 2e). This intracellular fluorescence is likely due to a mixture of immature proteins that have not yet trafficked to the cell surface and surface receptors that have been internalized. By instead using SBG-JF646, we obtained images where the dye remained solely at the cell surface, and furthermore, were able to resolve membranes with a lateral resolution of 91±23 nm using STED nanoscopy (n=42; cf. FWHMconfocal = 295±85 nm, n=35) (Fig. 2f).
We next asked if SBG-conjugated fluorophores could allow for superior labelling of GPCRs in vivo. We recently established SNAP-tag labeling in vivo in the frontal cortex of living mice using local injection of BG-conjugated fluorophores 29 . Presumably, the high solubility and reduced cell permeability of SBG-conjugated fluorophores should lead to improved tissue staining.
Based on our prior study, we virally-delivered SNAP-mGluR2 into the medial prefrontal cortex (mPFC) of adult mice before injecting BG-JF549 or SBG-JF549 8 weeks later at the same coordinates ( Fig. 3a). Clear labeling was observed with both compounds (Fig. 3b-g), but we observed a larger spread of SBG-JF549 fluorescence in transduced brains ( Fig. 3d; Fig. S6) and untransduced brains showed a 2-fold higher background for BG-JF549 than its SBG counterpart (Fig. 3g).
Having established efficient surface-targeted labeling with SBG-conjugated fluorophores, we next asked if we could use BG-and SBG-conjugated fluorophores to separate the respective intra-and extracellular pools of a membrane receptor. We employed glucagon-like peptide-1 receptor N-terminally fused to SNAP ("SNAP-GLP1R"), and used two spectrally separated dyes to pulse-chase label different receptor pools (Fig. 4a). GLP1R is involved in glucose homeostasis 30 and is known to undergo rapid endocytosis and trafficking upon activation with the agonist Exendin4(1-39) (Ex4). Live tracking of surface-exposed receptors has previously been achieved by using BG-Alexa or BG-Atto dyes 31,32 or by the use of specific antibodies and fixation at given time points 30,32 . However, we wanted to test the ability to simultaneously analyze the surface pool and the intracellular pool that has been translated but not yet trafficked to the membrane. As such, we incubated cells with SBG-TMR (500 nM) to label SNAP-GLP1R at the cell surface, followed by BG-SiR (500 nM) to label the intracellular, residual and newly trafficked (i.e. during the wash step) surface receptor pools (Fig. 4a). After washing, SNAP-GLP1R was clearly labelled at the surface with TMR and intracellularly with SiR (Fig. 4b). No bleedthrough was apparent in controls that used only one dye (Fig. S7). Surface TMR:SNAP-GLP1R was activated by Exendin4(1-39) (Ex4; 25 nM), before tracking of TMR and SiR-labelled receptor pools in live cells using highresolution confocal microscopy ( Fig. 4b). After 15 minutes of agonist treatment, cells were washed and the antagonist Exendin4(9-39) (Ex4; 500 nM) was applied to allow the internalized and cytosolic receptors pools to be sorted and re-trafficked to the surface (Fig. 4b). As expected, TMRlabelled GLP1R readily internalized following ligand binding, before trafficking and degradation upon washout and application of antagonist (Fig. 4c), as evaluated by mean fluorescence intensity at the membrane and in the cytosol (Fig. 4c, d). Interestingly, we noticed a cytosolic pool of SiR:SNAP-GLP1R, which either remained static and did not undergo trafficking, or alternatively, was degraded before being replenished by the portion of surface receptor labelled by SiR (Fig. 4c,   d). Thus, GLP1R present at intracellular sites immediately before orthosteric activation are unlikely to contribute to ligand-induced receptor turnover.
We next wondered if our technique could be used to probe the stoichiometry of GPCR populations inside the cell versus at the plasma membrane. Fluorescence-based methods have been widely used for assessing GPCR dimer-and oligomerization but rarely distinguish between surface and intracellular pools which may lead to confounding results and discrepancies across studies. This is especially critical as GPCR homo-and hetero-multimerization remains a controversial topic that may have major implications for general physiology and drug discovery 33 . We decided to use our labeling probes in conjunction with single molecule pulldown ("SiMPull") a strategy which allows single receptor complexes to be isolated and imaged for analysis of stoichiometry via counting of fluorophore bleaching steps 34 . To probe a prototypical class C GPCR, reported to form constitutive dimers by most studies to date 35,36 , we used HA-SNAP-mGluR2. Conversely, an HA-SNAP-beta-2 adrenergic receptor construct ("HA-SNAP-ß2AR") was used as a prototypical class A GPCR, which has been found as a monomer or dimer or higher order oligomer depending on experimental conditions 34,[37][38][39][40][41] . Each construct was labeled with either SBG-JF549, to label only surface receptors, or SNAP-Surface® Block followed by BG-JF549 (see methods for details) to isolate intracellular receptors. Cell imaging showed distinct fluorescence patterns for each receptor depending on the labeling paradigm (Fig. 5a) and labeling controls indicated that the BG-surface block prevented >95% of labeling of surface receptors without altering the efficiency of labeling intracellular receptors (Fig. S8). Following labeling, cells were lysed and detergent-solubilized GPCRs were isolated for single molecule imaging at a low density on passivated coverslips using an anti-HA antibody as previously described 42 (Fig. 5b). Single molecules were imaged using TIRF microscopy to allow for stepwise fluorophore bleaching which could be used to measure receptor stoichiometry (Fig. 5c). SBG-JF549 labeled HA-SNAP-mGluR2 showed ~55-60% 2-step photobleaching, consistent with previous studies indicating the formation of strict mGluR dimers 35,36,42 . However, intracellular receptors labeled with BG-JF549 showed reduced 2-step bleaching, indicating reduced dimerization in this population (Fig. 5d). These data suggest that a portion of the intracellular receptors are immature and monomeric.
We next performed the same experiment with HA-SNAP-ß2AR (Fig. 5e, f). Consistent with our previous SiMPull study 42 , we found weak dimerization of surface receptors labeled with SBG-JF549 (Fig. 5g, h). However, when we targeted intracellular receptors a small, but significant increase in apparent dimerization was observed (Fig. 5h). Together these data demonstrate the suitability of SBG and BG dyes for isolating surface versus intracellular receptor pools for experiments that take place in vitro following cell lysis. In addition, they indicate that different receptor pools may have different distributions of monomeric and multimeric receptors, emphasizing the importance of identifying which pool is being probed in a given study.
Taken together, we report the design and use of novel highly-soluble and membrane impermeable probes for the examination of different GPCRs from the purified protein to live cells to the whole organism.

Discussion
The use of SNAP-tags to specifically label and interrogate membrane-spanning proteins has proven a powerful technique in cell biology. To further improve this method, we have rationally designed a modification to the BG leaving group to yield sulfonated-BG (SBG) that renders a range of a priori permeable fluorophores impermeable towards the plasma membrane. As such, the fluorophore remains identical after labelling, without alteration of its spectral or other photophysical properties. It should be noted that derivatizing the leaving group is possible for nucleobases as used for SNAP, but not for the leaving group of the Halo-tag, being a chloride atom. Recent approaches have used charged moieties synthetically introduced between the leaving group and the dye 43 , with the need to test for influences on binding kinetics and fluorogenicity, the latter which is optimized for the protein surface it is exposed to. Another approach is the use of inherently impermeable dyes, such as some Alexa, ATTO or Abberior, which display properties different to the dyes we aimed to use. Other impermeable modifications, such as relatively large quenchers custom tailored for the fluorophore, have been reported for no-wash labelling of charged fluorophores 44 . In contrast, we describe a minimal alteration, independent of the cargo that should be generally applicable.
Using our cell-impermeable SNAP substrates, we showcase fast and clean membrane staining of SNAP-mGluR2, accompanied by STED nanoscopy. By using SBG-linked bright and photostable dyes, we could restrict labelling to the lipid bilayer for different dyes in the visible spectrum. We furthermore obtained highly resolved images of SNAP-mGluR2 residing at the membrane using STED in fixed cells, which proved to be impossible after using the BG-version due to high intracellular background staining. The most stable and widely used far-red STED dyes (Atto 647N, STAR RED and STAR635/P) are neither membrane permeable nor fluorogenic and hence cannot be used in cases where a comparison between intracellular and extracellular ligands is needed. On the other hand, the best performing dyes for live STED imaging (SiR, JF646, CP 610) have been designed to ensure membrane permeability 9,24,45 and, therefore, they also do not allow for a comparison between intracellular and membrane protein pools, which necessarily requires the use of SBG-ligands.
JF549 showed superior behavior when applied as its SBG-version in vivo. After application via injection, SBG-JF549 showed a 2-fold reduction in background when compared to its BG-congener.
In contrast, its spread in virally-infected brains was markedly increased. These results, in addition to the ability to solubilize dyes without a co-solvent (i.e. DMSO), demonstrate the power of our simple chemical modification for use in living animals.
We were also able to stain a SNAP-GLP1R fusion construct at the membrane with SBG-TMR and the remaining, mostly intracellular, pool with BG-SiR. Separating pools of the same protein has been achieved before, for instance by using fluorogen activated protein (FAP) 46,47 probes or the fluorescence-activating and absorbance-shifting tag (FAST) 48 . While these previous studies rely on non-covalent labelled protein tags fused to the BK channel or a transmembrane helix, respectively, we report a SNAP-fusion to a blockbuster targeted GPCR. In addition, our system does not rely on Förster Resonance Energy Transfer (FRET), which adds another layer of complexity and need for additional control experiments, as has been shown for malachite green conjugates. Furthermore, our approach allows for the use of different colors in the same experiment, while the FAST system uses charged and non-charged forms of the same fluorophores.
As such, we show that only GLP1R present at the membrane before ligand stimulation undergoes trafficking in response to activation. GLP1R which is already present inside the cell prior to stimulation never reaches the membrane. As such, two pools of GLP1R likely exist in the unstimulated state: 1) surface-exposed receptor which is trafficking-competent in the presence of ligand; and 2) internalized, cytoplasmic, newly-synthesized or incorrectly processed GLP1R which slowly traffics to the membrane in the absence of ligand or is, alternatively, degraded and replenished by labelled residual membrane receptor. Since peptide ligand cannot enter the cell, it is unlikely that the internalized GLP1R pool contributes meaningfully to intracellular (e.g. endosomal) signaling responses. What is the relevance of such observations for GLP1R function? Firstly, the same GLP1R pool might constantly turnover with continuous ligand stimulation, with the intracellular pool never making it to the membrane within the imaging timescale used here (i.e. measuring dynamic changes after activation). Secondly, ligands that specifically target the intracellular GLP1R pool might further increase efficacy of GLP1R agonists used in the treatment of metabolic disease. As such, SBG-fluorophores as well as other SBG-conjugated species are likely to open up interesting avenues of investigation for both GLP1R biology and the broader study of protein trafficking.
Finally, we also demonstrate the value of the SBG approach for chemically-tagging surface receptors for subsequent biochemical isolation. We use this to show that SBG-targeted surface GPCRs can display different stoichiometries than BG-targeted intracellular GPCRs. In the case of the class C GPCR mGluR2 intracellular receptors, presumably immature proteins, we were able to show reduced dimerization compared to the strict dimerization of the cell surface population.
This result suggests that dimeric receptors are preferentially trafficked to or maintained on the cell surface. In contrast, intracellular ß2AR showed enhanced dimerization compared to surface pools.
Critically, the ability to use the same fluorophore (JF549) for each condition, removes any possibility that differences in photobleaching pattern is due to differences in dye photophysics.
Future work will be needed to dissect the determinants of the differential dimerization of these populations, their sensitivity to different stimuli and to assess this phenomenon across a range of GPCRs and other membrane proteins. The flexible control afforded by SBG-conjugated fluorophores will be critical for such studies.
In conclusion, we report the design and use of novel highly-soluble and membrane impermeable probes for the interrogation of different GPCRs from the purified protein level to live cells to the whole organism.

Synthesis
Chemical synthesis (Supporting Scheme 1, 2) and characterization of compounds is outlined in the Supporting Information. All compounds were >95% purity according to HPLC if not stated otherwise.

SNAPf expression, purification, and mass spectrometry after labelling
SNAPf was expressed and purified as described previously. 27  Fluorescence intensity values from multiple images were then averaged. Each condition was tested in at least 2 separate transfections. Slices were wet mounted to glass slides and secured with coverslip using VECTASHIELD HardSet Antifade Mounting Medium with DAPI (Vector Laboratories). Glass slides were imaged using an Olympus Confocal FV3000 and images were processed and analyzed in ImageJ.
Control experiments were performed in either SBG-TMR-or BG-SiR-labelled cells to exclude trafficking artefacts due to spectral overlap.

Single molecule pulldown assay
Single molecule pulldown (SiMPull) was performed using HA-tagged GPCRs isolated on glass coverslips as previously described using a biotinylated anti-HA antibody 42  x g, 4 °C for 1 min, cells were lysed using 1.2% IGEPAL detergent for 1 hour at 4 °C. Next, cells were centrifuged at 16,000 x g for 20 min at 4 °C and supernatant was collected and stored in ice until used. The cell lysate samples were then diluted using a dilution buffer containing 0.1% IGEPAL and introduced to the flow chamber. After obtaining an optimal number of spots in the field of view, the chamber was washed with the dilution buffer to remove unbound proteins.
Single molecule imaging was done using a 100x objective (NA 1.49) on an inverted microscope (Olympus IX83) in total internal reflection (TIR) mode at 20 Hz with 50 ms exposure times with an scMOS camera (Hamamatsu ORCA-Flash4v3.0). Samples were excited with 561 nm lasers and imaged using an emission filters of 595±25 nm. Data analysis was performed using custom made LabVIEW program as previously described 49 . Data was collected across at least 2 separate experimental days and then averaged to produce bar graphs in Fig. 5d and 5h.

Competing Interests
The authors declare no competing interests.

References:
(     Schematic showing single molecule pulldown configuration where an anti-HA antibody is used to isolate a sparse surface of SNAP-tagged mGluR2 following fluorophore labeling. c) Representative image of single molecules for SNAP-mGluR2, with representative bleaching traces for a 1-step and 2-step example (bottom). Note: >95% of spots bleached in either 1 or 2-steps. d) Summary of the proportion of 2-step bleaching steps for each labeling configuration. Each point represents one independent movie and bars show mean ± SEM. * indicates statistical significance (unpaired t test, p=0.0005). e-h) Same as a-d but with SNAP-ß2AR. * indicates statistical significance (unpaired t test, p=0.008). Scale bars are 10 µm.