Sulfonated rhodamines as impermeable labelling substrates for cell surface protein visualization

Sulfonated rhodamines that endow xanthene dyes with cellular impermeability are presented. We fuse charged sulfonates to red and far-red dyes to obtain Sulfo549 and Sulfo646, respectively, and further link these to SNAP- and Halo-tag substrates for protein self-labelling. Cellular impermeability is validated in live cell imaging experiments in transfected HEK cells and neurons derived from induced pluripotent stem cells (iPSCs). Lastly, we show that Sulfo646 is amenable to STED nanoscopy by recording membranes of SNAP/Halo-surface-labelled human iPSC-derived neuronal axons. We therefore provide an avenue for rendering dyes impermeable for exclusive extracellular visualization via self-labelling protein tags.


INTRODUCTION
Fluorescent microscopy is often the method of choice to visualize and interrogate cell biology. [1,2] Two major methods can be distinguished, that is the use of genetically-encoded fluorescent proteins or the use of small molecule fluorophores. [2] The latter can be targeted by chemical fusion to a selective and tight small molecule binder, or by means of self-labelling protein tags. [3][4][5] A plethora of fluorescent small molecules are available for microscopy, spanning different photophysical and chemical properties. [6,7] Desirable properties are brightness, resistance to photobleaching, and cellular permeability. [8][9][10][11][12] Depending on imaging modality, other properties might be desirable such as blinking or fluorogenicity. However, very few fluorescent dyes exist for exclusive SNAP-and Halo-tag labelling on cell surface proteins, best typified by transmembrane receptors. Within that repertoire, even fewer are suitable for stimulated emission depletion (STED) nanoscopy, [13] since higher laser powers are required that may lead to photobleaching, although impermeable dyes were used in the beginning [14] . While robust and permeable dyes can be targeted to SNAP-or Halo-tag, non-specific background labelling may become problematic, especially in live cell applications. [15] Moreover, this might be the case for cell surface receptor localization ( Figure 1A), and as such it remains difficult to isolate surface and intracellular pools of receptors not only for nanoscopy. [16] To restrict a priori fluorogenic dyes to the cell surface, we set out to synthesize rhodamines endowed with a cell impermeable sulfonate moiety. To achieve this, the rhodamine scaffold of fluorogenic and bright JaneliaFluor (JF) dyes [8] was extended with a short sulfonate-containing linker to provide red and far-red colors (Sulfo549 and Sulfo646), which can be targeted to cell surface tags and receptors ( Figure 1B). The versatility of this approach is highlighted in live cell imaging with various tagged constructs in different cell types including human induced pluripotent stem cells (iPSCs)-derived neurons, and in fixed neurons by STED nanoscopy to obtain high-definition resolution of axonal membranes.

RESULTS
We set out to synthesize impermeable dyes by the synthetic addition of charged sulfonates, which remain deprotonated and therefore not only become cell impermeable, but may also increase their solubility in aqueous media. With our primary aims in mind, i.e. i) impermeability, ii) labelling of SNAP-and Halo-tags, and iii) usage in STED nanoscopy, we decided to use xanthene dyes as a blueprint for our design, which are known to exist in two states, an open (fluorescent) and a closed (non-fluorescent) form ( Figure 2A) and are among the most stable towards photobleaching. The recently reported JaneliaFluor dyes are rhodamine-based fluorophores, showing higher brightness and fluorogenicity than their tetramethyl rhodamines congeners, due to installment of azetidines as nitrogen containing moieties. [8] As such, we synthesized Sulfo549 and Sulfo646 congeners by introducing a carboxylate handle on the 3-position of the azetidine, which was further derivatized to a sulfonated head group via peptide coupling to taurine (Figure 2A, Scheme S1). A carboxylate in the 6-position served as a position to install O 6 -benzylguanine (BG) or a chloroalkane (CA) group, which act as substrates for the self-labelling SNAP-and Halo-tag, respectively, and thereby obtained four molecules displaying two colors and two labelling modalities ( Figure 2B, Scheme S1). In a first set of experiments, we assessed the excitation and emission profiles of our dyes in their unbound (i.e. BG-and CA-linked) and in their bound (i.e. SNAP-and Halo-tag reacted) states ( Figure 2C). Fluorogenicity was reduced as expected when charges are added in close proximity to the dye, however, all dyes still showed high brightness ( Figure S1, S2). Labelling was confirmed in vitro by incubation of Sulfo dyes with recombinant SNAP-and Halo-tag and subsequent mass spectrometry (see Supporting Information). Furthermore, we assessed kinetics of BG-Sulfo549 (t1/2 = 28.0 sec) labelling on SNAP-tag versus BG-TMR (t1/2 = 8.9 sec) and BG-JF549 (t1/2 = 15.3 sec) by means of fluorescent polarization, and found a slight decrease in rate of labelling by a factor of ~3.14 and ~1.83, respectively ( Figure S3). Still, full labelling of SNAP:Sulfo549 was achieved within minutes and, interestingly, with enhanced polarization.
Next, we wanted to test our molecules in a cellular setting for protein labelling in fluorescent microscopy. To prove that our impermeabilization strategy is applicable, we expressed a SNAP-Halo-construct with a nuclear localization signal (NLS; SNAP-Halo-NLS [9] ) in HEK293T cells, before titrating 100-5000 nM of permeable JF646 or its impermeable counterpart Sulfo646. Clear concentration-dependent nuclear signals were detected for JaneliaFluor dyes, but not for the Sulfo probes ( Figure S4). To confirm surface labelling, we cloned two constructs containing: i) an IgK trafficking signal for the plasma membrane; and ii) a SNAP-tag and Halo-tag separated by a single pass transmembrane (TM) domain (see Supporting Information). As such, the construct should be labelled exclusively on the surface when using impermeable dyes, while a permeable dye would lead to background staining from proteins residing in the cell. Indeed, by titration of 100-5000 nM of JF646 or Sulfo646 as before, we observed an increased background when using permeable dyes, and solely surface labelling when using Sulfo646, irrespective of the tag used ( Figure S5). Furthermore, due to the installation of two orthogonal tags on the same construct, we were able to dual-color label with a red and far-red dye, testing for permeability and localization properties.
Having established that 100 nM of substrates lead to sufficient labelling, SNAP-TM-Halotransfected HEK293T cells were incubated with a combination of BG-Sulfo646/CA-JF549 at each 100 nM concentration ( Figure 3A). Widefield fluorescent imaging revealed surface localized staining for SNAP:Sulfo646 in combination with intracellular signals presumably stemming from non-surface trafficked or nascent Halo:JF549 ( Figure 3B). This could be further resolved by plotting a line through a transfected cell ( Figure 3C), which depicts plasma membrane and intracellular signals from the two colors. Using the same construct and settings, we then switched the dye colors To show the utility of the Sulfo dyes for labelling cell surface receptors, we transfected AD293 cells with N-terminal SNAP-and Halo-tagged glucagon-like peptide-1 receptor (GLP1R), a class B GPCR and target for the incretin-mimetic class of anti-diabetic therapy ( Figure 5). In its nonstimulated state, GLP1R displays minimal constitutive activity and is largely present at the cell surface. Notably, differences between JF549 and JF646 and their Sulfo derivatives were present for SNAP labelling, and became even more prominent for Halo labeling, where a large improvement in membrane resolution and brightness was detected.
Given the performance of the Sulfo dyes so far in heterologous cell systems, we sought to extend studies to more complex cell populations where background signal can make accurate protein localization difficult. To allow this, human cortical neurons were derived from iPSC and cocultured with murine primary astrocytes [17,18] , before transfection with our SNAP-TM-Halo and Halo-TM-SNAP constructs. Labeling was performed with the respective impermeable far-red Sulfo646 and permeable red JF549, before live imaging was conducted by confocal microscopy.
Using the SNAP-TM-Halo construct labelled with BG-Sulfo646 and CA-JF549, clear labelling was observed for BG-Sulfo646 ( Figure 6A). Likewise, Halo-TM-SNAP labelled with CA-Sulfo646 and BG-JF549 ( Figure 6B) showed clearest labelling for CA-Sulfo646. Thus, for both self-labelling tags, the Sulfo dyes demonstrated excellent performance for cell surface protein visualization in more complex cell types.
Lastly, we fixed the cultures before STED nanoscopy on Halo:Sulfo646 and SNAP:JF549 labelled neurons ( Figure 7A). An accumulated line profile along an axon (white box) of the confocal and STED images revealed that Sulfo646 is amenable to nanoscopy ( Figure 7B). By two-Gaussian fitting, we obtained sharper full width half-maximal values for STED versus confocal (FWHMconfocal = 268.5 and 292.3 nm; FWHMSTED = 157.5 and 223.1 nm) microscopy. Consistent with our previous results, Halo:Sulfo646 generated a more pronounced signal along the axonal membrane, while SNAP:JF549 could be observed in the intra-axonal compartment ( Figure 7C).

DISCUSSION
The need for custom-tailored dyes is in high demand as the range of microscopy modalities, experimental techniques and labelling strategies increases. Recent developments [9,[19][20][21][22] have focused mostly on boosting brightness, fluorescent lifetimes, chemical stability and/or fluorogenicity, the latter being a cause for cellular permeability. For the interrogation of cell surface proteins that are genetically fused to self-labelling protein tags (e.g. SNAP-and Halo-tag), however, cell impermeable dyes are desirable. Rendering dyes impermeable is usually achieved by introduction of sulfonates, which remain negatively charged in biological systems and are therefore not able to cross the plasma lipid bilayer. While many sulfonated dyes exist, such as Alexa488/568/647, or LD555/655 [23] , their application for enzyme self-labelling and STED nanoscopy has so far not reached the performance bar set by permeable dyes. A recent study, however, showed great performance of LD dyes for FRET measurements of dimer formation on SNAP-tagged GPCRs and in TIRF microscopy for single molecule FRET recovery after photobleaching. [24] Rhodamine dyes have attracted some attention in recent years since the emergence of JaneliaFluor dyes, which rely on the exchange of azetidine groups for dimethylamines on various molecular scaffolds [8] . Indeed, one impermeable version has been described, JF635i, which retains some of its fluorogenicity and has been used to observe Halotagged transferrin receptor recycling. [25] Nevertheless, it has only been described as a Halo-tag substrate and has not been subjected to super-resolution STED nanoscopy. We aimed to expand this palette, by using azetidine containing rhodamines for red and far-red imaging, which show minimal fluorogenicity and therefore maximal brightness. Accordingly, we synthesized Sulfo549 and Sulfo646 based on JF549 and JF646, each bearing two sulfonate groups, and further linked them to SNAP-and Halo-tag substrates BG and CA, respectively. Our probes add favorably to a previous study, where we reported on a strategy to limit any dye to cell surface exposed SNAP tags by altering the BG substrate to a sulfonate itself (termed SBG) [16] . With a simple chloride anion being the leaving group for the Halo-tag, introducing a charged sulfonate is not tolerated.
Therefore, the impermeable characteristics need to instead be provided by the dye, for which we provide a solution herein.
To test our approach, we cloned constructs for cellular transfections that bear a SNAP-and Halotag separated by a transmembrane domain, thereby placing each tag either side of the plasma membrane. These constructs allowed screening of permeability parameters in live HEK293 cells.
Sulfo dyes performed well in these systems, where no background signals from intracellular spaces were detected. Line scans revealed clean plasma membrane staining of Sulfo dyes and intracellular staining of the permeable JF dyes.
Encouraged by this, we next used SNAP-and Halo-tagged GLP1R to provide performance benchmarking in a more relevant cell surface signaling molecule. Comparable to previous findings with the SNAP/Halo constructs, we obtained clean surface labelling for both colors and both protein tags, with more prominent effects for Halo, which is in line with in vitro measurements.
Given that GLP1R labelling was performed at 37 °C, where constitutive activity of GPCRs can be increased, intracellular signal from Sulfo dyes can be observed, and we note that their strength might differ due to different expression and endocytosis levels. While we have shown dual color GLP1R trafficking in a previous report on a SNAP-tagged construct [16] , our results herein set the stage for flexible use of colors on the Halo-tag not only on GLP1R, for instance for post endocytic protein trafficking [25,26] and the interrogation of cell surface receptor ensembles [27][28][29] .
We were also able to extend this approach to more complex cell types, by staining live somatodendritic and axonal compartments in human cortical neurons derived from iPSC, cocultured with astrocytes. By transfecting neurons with SNAP-TM-Halo or Halo-TM-SNAP, we aimed to benchmark far-red staining of the outer membrane with both tags, as this color was our choice for later super-resolution imaging. As such, we observed surface labelling with both constructs and the use of the respective Sulfo646, with signals markedly brighter when bound to Halo. In contrast, CA-JF549 clearly stained the intracellular Halo-tagged protein pool of neurons, while BG-JF549 non-specifically accumulated in co-cultured astrocytes. Self-labelling tags have been employed in "brainbow" labelling, where they offer more flexibility in terms of colors available, more straightforward applicability than antibodies, and better survival of the rather harsh clearing conditions with respect to fluorescent proteins. [30] Indeed, we anticipate Sulfo dyes to be a favorable addition to such studies and their performance in whole tissues [31,32] , and with this observed trend in mind, we chose to continue with Halo:Sulfo646 in our preparations.
For this reason, we fixed the astrocyte/neuron co-culture and tested SNAP:JF549 and Halo:Sulfo646 for STED nanoscopy. As expected, JF549 was not amenable to the depletion laser, but gratifyingly, Sulfo646 was able to improve full width half maximal values when a broad line plot was applied along an axon in STED imaging. In addition, the resolution, i.e. the distance of the two maxima, of the membranes were resolved to be ~300 nm in both confocal and STED. Previous studies resolved the median axonal diameter of organotypic GFP expressing CA1 neurons to be 203 nm by STED [33] , and to be 242 nm by super-resolution shadow imaging (SUSHI) [34] .

SUMMARY
We have designed and synthesized sulfonated fluorescent rhodamines dyes (Sulfo549 and Sulfo646) that are based on the JaneliaFluor scaffolds to obtain bright and impermeable dyes in the red and far-red. By linking these dyes to substrates recognized by the SNAP and Halo-tag, we were able to achieve exclusive cell surface labelling in HEK293/AD293 cells and in human induced neurons by means of widefield and confocal microscopy. Lastly, we employed STED nanoscopy on Sulfo646-labelled neuron axons and showcase their performance by resolving the axonal membranes. We anticipate that these and other sulfonated rhodamines will be useful for visualizing cell surface proteins using a range of imaging approaches spanning widefield through super-resolution.

Chemistry, cloning and in vitro protein labelling
Chemical Schemes, synthetic protocols, protein labelling in vitro, and characterization can be found in the Supporting Information. Plasmids were cloned using Gibson assembly cloning Kit (NEB), primer were designed using the NEBuilder assembly tool. Plasmids were isolated using a mini prep kit (Thermo Fisher). DNA concentration was measured on a NanoDrop (Thermo Fisher) and verified by Sanger sequencing (see Supporting Information).

In vitro fluorescence spectroscopy
Purified SNAPf and Halo was obtained as previously described. [20] Labelling dyes were dissolved

Protein mass spectrometry
Labelling substrates were dissolved in DMSO to a concentration of 1 mM and diluted in activity buffer (containing: 50 mM NaCl, 50 mM HEPES, pH=7.3 + 4 µg/mL BSA) to 20 µM. Protein was diluted in activity buffer to a concentration of 2 µM. 25 µL of each protein and labelling agent were combined in a mass spec vial and allowed to incubate at r.t. for 1h, before full protein mass was acquired. In case for non-labelling control, 25 µL of activity buffer was mixed with 25 µL of protein.

Human iPSC-derived neurons
Human induced pluripotent stem cells (iPSCs) engineered to express mNGN2 under a doxycycline-inducible system in the AAVS1 safe harbor locus were used for the i 3 Neuron differentiation protocol, as described previously. [17,18] In brief, iPSCs were seeded on Matrigel        is largely confined to the intra-axonal compartment (STED signals). Scale bar = 1 µm.