N-Methyl deuterated rhodamines for protein labelling in sensitive fluorescence microscopy

Rhodamine fluorophores are setting benchmarks in fluorescence microscopy. Herein, we report the deuterium (d12) congeners of tetramethyl(silicon)rhodamine, obtained by isotopic labelling of the four methyl groups, show improved photophysical parameters (i.e. brightness, lifetimes) and reduced chemical bleaching. We explore this finding for SNAP- and Halo-tag labelling in live cells, and highlight enhanced properties in several applications, such as fluorescence activated cell sorting, fluorescence lifetime microscopy, stimulated emission depletion nanoscopy and single-molecule Förster-resonance energy transfer. We finally extend this idea to other dye families and envision deuteration as a generalizable concept to improve existing and to develop new chemical biology probes.


Introduction
Fluorescence microscopy is the technique of choice in modern biomedical research to elucidate structures or to interrogate function. Efforts to improve performances resulted in the development of super-resolution microscopy (nanoscopy), which is dened to obtain resolution higher than the diffraction limit described by Abbe's law. [1][2][3] While optics and instruments have been advanced constantly, the elaboration of synthetic molecular dyes has driven the eld to the current state-of-theart. 4 Given their small size, and the possibility to target them to cellular organelles by linking them to molecular targeting scaffolds (e.g. ER-Tracker™ or MitoTracker™) or to selflabelling protein (e.g. SNAP/Halo-tag fusions) substrates, is making them an attractive choice for the visualization and interrogation of biomolecular function. 5,6 Several chemical scaffolds, for instance nitrobenzodioxazoles (NBDs) 7 and coumarins, 8 remain interesting synthetic targets, yet xanthene dyes, which include tetramethylrhodamine (TMR) and silicon rhodamine (SiR), experienced a renaissance in terms of novel modications to tune and boost important parameters, such as color, brightness, lifetime and reactivity. [9][10][11][12][13] Isotopic labelling was recently among this, 14-16 yet remained underexplored.
We next turned to SNAP labelling and imaging in live cells on targets that are expressed extracellular or intracellular to The isotopic deuterium incorporation on the four methyl groups leads to d12 variants of TMR (X ¼ O) and SiR (X ¼ SiMe 2 ). Derivatization on the 6-carboxylate allows synthesis of BG-, CA-and Mal-congeners for SNAP-tag, Halo-tag and thiol labelling, respectively. (B) Excitation and emission spectra of TMR(-d12) (top) and SiR(-d12) (bottom). (C) Fluorescence polarization (mP) assay of BG-TMR(-d12) (top) and BG-SiR(-d12) (bottom) when incubated with SNAP f to determine labelling kinetics. (D) In vitro FRET assay of a BG-TMR(-d12) and CA-SiR(-d12) labelled SNAP-Halo protein shows an increase of acceptor/donor emission ratio for deuterated compounds. n ¼ 3 measurements. a In PBS; b in EtOH + 1% TFA; c in activity buffer (containing in mM: NaCl 50, HEPES 50, pH 7.3); d R ¼ BG and SNAP f in vitro in activity buffer; e SNAP bound in cells.
determine their tagging and permeability characteristics. 6 First, we employed CHO-K1 cells stably expressing SNAP-tagged glucagon-like peptide 1 receptor (SNAP-GLP1R:CHO-K1), 17 a cell line intensely used to study the physiology of this class B G protein-coupled receptor (GPCR), which is involved in glucose homeostasis and a drug target in diabetic patients, 18,19 as a benchmark for d12 performances. As such, cells were labelled with 1 mM BG-TMR/SiR(-d12) for 30 min, before washing and live imaging by confocal microscopy, revealing staining of SNAP-GLP1R with all deuterated and parental dyes tested ( Fig. 2A). Secondly, having established labelling on the outer plasma membrane, we investigated intracellular staining in live HeLa cells that stably express SNAP-tagged Cox8A (SNAP-Cox8A:HeLa) in the inner mitochondrial membrane (Fig. 2B), which has been used to study mitochondrial ultrastructures in live cells. 20,21 As for SNAP-GLP1R, we observed clean labelling with all dyes, and for both colors with an observable increase in brightness for the d12 derivatives. With this enhanced performance in microscopy, we wanted to quantify brightness by uorescence activated cell sorting (FACS) to obtain robust values over large sample sizes. Accordingly, we labelled SNAP-GLP1R:CHO-K1 and SNAP-Cox8A:HeLa cells with both, BG-TMR(-d12) and BG-SiR(-d12) to compare red and far-red color intensities by subsequent sorting (Fig. 2C). Histograms of labelled SNAP-GLP1R:CHO-K1 cells showed a right-shi in uorescence intensity when dyes were deuterated (Fig. 2C, le). In line with this, labelled SNAP-Cox8A:HeLa cells exhibited a pronounced shi to higher intensities for SiR-d12 compared to its nondeuterated congener (Fig. 2C, right), while TMR(d12) only displayed a subtle change. By normalizing intensities and comparison, we calculate higher mean intensities for our deuterated dye versions (Fig. 2D). While no large increase was observed in SNAP-Cox8A:HeLa cells for TMR-d12 (2%), mean intensity was markedly increased in SNAP-GLP1R:CHO-K1 cells (24%). Furthermore, SiR-d12 outperformed SiR on SNAP-GLP1R and SNAP-Cox8A with an intensity increase of 29% and 50%, respectively. In addition, uorescent lifetime confocal microscopy (FLIM) revealed longer uorescent lifetimes for d12 congeners compared to their counterparts (s(TMR) ¼ 2.3 vs. s(TMR-d12) ¼ 2.6 ns; s(SiR) ¼ 2.9 vs. s(SiR-d12) ¼ 3.5 ns) ( Fig. 2E and Table S1 †). Accounting for a higher chemical stability, TMR-d12 was not as susceptible to bleaching as TMR, while SiR and SiR-d12 exhibited similar, and compared to TMR, more photostable trends of bleaching in this microscopic setup (Fig. 2F). Setting the stage for more imaging opportunities, we endowed our recently reported LUXendin651, a SiR-linked antagonistic peptide with high affinity and selectivity towards GLP1R, 22,23 with the SiR-d12 congener via cysteine conjugation to Mal-SiR-d12 (ESI Fig. 2A and B †). While this peptide targets the orthosteric site of GLP1R (ESI Fig. 2C †), we observed higher signal intensities and longer lifetimes (ESI Fig. 2D, E and Table  S2 †) of LUXendin651-d12 in xed SNAP-GLP1R:CHO-K1 cells when compared to its rst generation LUXendin651. These results demonstrate that rhodamines with CD 3 bearing amines are not only applicable to live cell imaging but outperform nondeuterated uorophores, which is in line with our in vitro data of the unbound dyes.
With these encouraging results, we decided to test our deuterated probes in stimulated emission by depletion (STED) microscopy, a state-of-the-art imaging technique to reveal cellular dynamics and structures. 4 As such, and with SiR being one of the most successful far-red dyes for nanoscopy, 24 we investigated super-resolution images acquired in live SNAP-Cox8A:HeLa cells and included JaneliaFluor646 (JF 646 ) as an additional benchmark of dyes in the far-red regime. Aer incubation with 1 mM BG-SiR, BG-JF 646 or BG-SiR-d12, we recorded images of mitochondrial cristae under the same conditions, and while all three dyes displayed labelling, SiR-d12 was able to resolve cristae sharper with less background (Fig. 3A). While this can have multiple reasons that may not only be attributed to dye performance, we targeted the cytoskeleton by labelling and imaging live Tubb5-Halo 25 stably transfected COS7 cells (Tubb5-Halo:COS7) with homogenous expression levels (Fig. 3B). Microtubules resemble a classical benchmark to demonstrate the power of nanoscopy due to their constant diameter of $25 nm. Aer incubation with 1 mM CA-SiR, CA-JF 646 or CA-SiR-d12, we observed microtubular ne structures with a full width half-maximum (FWHM) of $78 nm for all farred dyes (Fig. 3C), notably with a marked increase in uorescence intensity for SiR-d12 of 30% and 22% compared to SiR and JF 646 , respectively (Fig. 3D). Taken together, our deuterated d12 silicon rhodamine displayed augmented brightness in nanoscopic experiments while retaining resolution.
Single-molecule FRET (smFRET) 26 has become a wellestablished method to study (dynamic) conformational changes and heterogeneity of biomacromolecules. 27,28 Alternating laser excitation 29 (ALEX) describes one implementation of smFRET that allows the study of freely-diffusing molecules in solution at room temperature. Here, FRET efficiency is determined during short diffusional transits (on the timescale of milliseconds) of individual donor-acceptor-labelled molecules through the excitation volume of a confocal microscope. The technique allows observation of relative distance changes 29 but also absolute distances 30 with a spatial and temporal resolution limited by the available photon budget (count-rate). 27, 28 We thus tested whether higher count rates are available from deuterated uorophores that are specically attached to cysteine residues in proteins. Our test system was SBD2, the soluble extracellular substrate domain of the amino acid importer GlnPQ from Lactococcus lactis. 31 The resulting E-S histograms of all different donor-acceptor pairs showed the expected donor-only (S > 0.8, D-only), acceptoronly (S < 0.3, A-only) and a donor-acceptor-containing species (S between 0.3-0.8, DA). The abundance of the DA population was on average >30% (ESI Fig. 3 †), which generally facilitated fast data acquisition within $30 min. Analysis of the DA-species revealed a low FRET apo and high-FRET holo state (Fig. 4C), which is consistent with our previous investigations 31, 32 and the idea that the protein changes from the open to its closed state dependent on the glutamine concentration in the buffer (Fig. 4A). Also, the biochemical properties of SBD2 were preserved aer uorophore labelling indicated by equal population of both conformational states at a glutamine concentration close to the dissociation constant K d of the protein of around 1 mM (Fig. 4D).
Further inspection of the data revealed differences in the photophysical properties of the dyes. While bleaching artefacts, i.e., bridges between the three major populations, were almost absent for all pairs at the chosen laser powers of 40 mW green and 15 mW red excitation, the overall sum count-rate of donorbased donor-emission (DD), donor-based acceptor-emission (DA) and acceptor-based acceptor-emission (AA) was distinct for all dye combinations. In both apo and holo state Cy3B-ATTO647N was by far brighter in comparison to TMR-SiR with a signicant number of molecules with count-rates >200 kHz (Fig. 4E). In agreement with results from Fig. 1-3, deuteration results in enhanced count-rates for TMR-d12-SiR-d12 over TMR-SiR. In summary, our results from smFRET investigations show that deuteration of rhodamines is a simple solution to improve spatial and temporal resolution in solution-based experiments via enhanced count rates. It was particular useful to see that both TMR-SiR (in deuterated and non-deuterated form) did not require addition of photostabilizers to the imaging buffer to obtain high-quality E-S histograms. Furthermore, the dye pair TMR-SiR has not been characterized and apparently shows a Förster radius similar to Cy3B-ATTO647N, which is larger than for the most commonly-used pairs in the eld (e.g., $5.0 nm for Alexa Fluor 555/647). 27,28 Following the results of deuterated rhodamines in various sensitive state-of-the-art uorescence applications, we next asked if our deuteration approach is limited to rhodamine scaffolds or a general concept to enhance uorescent dye properties. As such, we deuterated Coumarin 461 to obtain Coumarin 461-d3 and -d6 (Fig. 5A, Scheme S2A and Table S3 †), which showed similar extinction coefficients (3(Coumarin 461) ) and similar maximal excitation and emission wavelengths (l Ex/Em (Coumarin 461(-d3/6)) $ 372/470 nm) (Fig. 5B). As expected, quantum yield increased successively by deuteration level, giving rise to 43% higher brightness of Coumarin 461-d6 vs. Coumarin 461. Along these lines, CD 3 installment on NBDs (Fig. 5C, D and Scheme S2B †) and methylene blue (Fig. 5E, F and Scheme S2C †) to give NBD-d6 and methylene blue-d12 followed the same trend: no change in excitation and emission wavelengths ( Fig. 5D and F), yet brightness was enhanced in both cases by 4% and 37% for NBD and methylene blue, respectively, stemming from the product of extinction coefficient and quantum yield (3(NDB) ¼ 16 300 vs.
. This is encouraging towards the exploration of deuteration as a general approach to boost desired photophysical properties.
Finally, we were curious to nd some mechanistic insights of how deuterium incorporation improves uorescent rhodamines. Firstly, we determined excited-state lifetimes for TMR(-d12) and SiR(-d12) by transient absorption spectroscopy, which agree with the lifetimes obtained by uorescence lifetime imaging (Fig. 6A-D). This indicates that the decay of the luminescent state correlates with recovery of the electronic ground state. Particularly, the decay of the emissive state does not yield a long-lived triplet state, which would appear in a long-lived transient absorption signature. This led us to further investigate uorescence lifetime, which depends on the rotation of the alkyl amine group, and as such is temperature sensitive, decreasing at elevated temperatures. 33 If deuteration of the methyl groups affects rotational movements, we would be able to observe smaller changes in lifetime at increasing temperatures. For this reason, we acquired uorescent lifetimes of SiR and SiR-d12 bound to Tubb5-SNAP and Tubb5-Halo-tags (Fig. 6E) in a temperature dependent manner by FLIM (Tables  S4 and S5 †). Indeed, when comparing lifetimes at 20, 30 and 40 C, we found that both Tubb5-SNAP:SiR-d12 and Tubb5-Halo:SiR-d12 retained signicant longer lifetimes at 40 C when normalized to lifetimes at 20 C. We therefore reason that non-radiative decay of the excited singlet state via rotation around the dimethyl amino group is suppressed due to the stronger and heavier nature of deuterium (Fig. 6F).

Discussion
In our study, we synthesized and tested deuterated uorophores with enhanced uorescent properties in a set of applications. The use of deuterium to improve uorescence emission properties has been addressed in the past by using deuterated solvents. 34 In a preprint 14 uploaded at the same time to our preprint, 15 the Lavis laboratory reported on a similar approach, supporting our ndings: by equipping the rhodamine nitrogen atoms with deuterated ring systems, they describe Halo-tag substrates with enhanced uorescent properties (e.g. brightness and single particle tracing) for protein labelling in live cells. 16 In this work, we install labelling moieties for SNAP-and Halo-tag conjugation and apply them in different experimental setups. This led to the nding that our TMR-d12 and SiR-d12 improve by swapping hydrogen with deuterium on the methyl groups, enhancing photophysical parameters, such as brightness and lifetime, while reducing critical chemical parameters, such as bleaching. We employed quantitative 1 H NMR to assess concentrations of dyes in solution to precisely determine extinction coefficients, and measured quantum yields directly by use of an integrating sphere, and found in both systems signicant higher values for d12 uorophores.
While more mechanistic reasons for the enhanced properties may exist, we argue the following: (i) affecting the rotation around the aromatic carbon-nitrogen bond (in our case due to higher mass of the CD 3 groups) has marked effects on uorescent properties, 35 which could suppress non-radiative decays and in turn enhances quantum yield and lifetime; 9 (ii) a lower zero-point energy of the C-D vs. C-H bond results in slower reaction kinetics, as an higher energy barrier has to be overcome, 36 and this would reduce bleaching through for example generated reactive oxygen species. This is backed up by the Lavis laboratory, 16 since (i) it was observed that deuteration of azetidines (which are rotationally more locked) 9 does not lead to a large increase in quantum yield, and (ii) light-induced demethylation of deuterated uorophores is slower when compared to their non-deuterated congeners. Both of these arguments describe lower quantum yields of non-radiative decays, and result in an improved quantum yield for emission. In particular, the phenomenon in question is twisted intramolecular charge transfer (TICT), which is known to be temperature-dependent, 37 and has been explored in many ways to enhance quantum yield of uorophores by introducing steric demand and/or donor engineering. 38 This is further supported by the Lavis lab, as it was shown that deuteration alters quantum yield on dyes depending on the TICT donor/ acceptor. 16 More experimentation is needed, ideally in combination with in silico calculations, that, for instance, have been performed on dyes under acidic H/D exchange 39 where "a close examination of the low-lying singlet and triplet electronic states along the torsional motion of the amino groups revealed that the key to the isotope effect is changes in non-radiative channels". 39 As such, tunneling rates and intersystem crossing differences may contribute to our observed changes in uorophore behavior. Keeping this in mind, we showcase deuterated dyes that outperform their parent molecules in multiple experiments.
The enhancements are signicant and broadly applicable, ranging from in vitro FRET, to live cellular labelling and sorting, lifetime and super-resolution microscopy on SNAP-and Halotags and smFRET using maleimide-thiol chemistry. We observed "sharper" imaging on SNAP-Cox8A:HeLa cells with SiR-d12 compared to JF 646 and SiR. As these experiments were performed in live cells, the subjective perception of enhanced imaging may be attributed to different expression levels, more or less healthy mitochondria, cell cycle phase, and the dynamic change in cristae thickness. Although our observations were consistent in three independent experiments, it remains difficult to quantify, and as such, we aimed to include an unambiguous experiment to determine the STED performance of SiR-d12. For this reason, we chose live Tubb5-Halo:COS7 cells, where the distribution of Halo-tagged microtubules is homogeneous and the diameter of the ne-structure is constant at 25 nm. Indeed, in this setup, we found outstanding performance of deuterated Halo:SiR-d12 compared to JF 646 and SiR with a marked increase in brightness and no signicant difference in resolution. In times where photon counts in sophisticated imaging experiments (e.g. MINFLUX 40 ) are becoming increasingly more important, we anticipate that uorophore deuteration provides a method to advance in the eld as was shown by applications of deuterated dyes in single molecule experiments.
Furthermore, the concept was expandable to other dye scaffolds, such as coumarins, NBDs, and the thiazene containing dye methylene blue, giving enhanced brightness for all deuterated species. It should be noted here that extinction coefficients are within a close margin, however, quantum yields are signicantly increased for all dyes tested.
We anticipate this concept (i) to be generalizable to other xanthenes (e.g. SNARFs, and quenchers like QSY7) at N-C a -H positions for improving and ne-tuning spectroscopic properties; (ii) to be further explored with other isotopes, such as 13 C, 15 N or even 3 H that can be used as an additional, orthogonal radioactive tracer; (iii) to be used in different labelling approaches, such as the attachment to sulfonated BG (SBG) scaffolds allowing the separation of SNAP-tagged receptor pools, 41 to biomolecule targeting probes, [42][43][44] to "click chemistry" reagents (e.g. cyclopropenes, cyclooctenes) 45 or to photoswitchable ligands, 46,47 and (iv) to serve as multimodal dyes for isotope labelled mass spectrometric analysis, correlative light-electron microscopy (CLEM) 48 and confocal Raman microscopy. 49,50 Such efforts are of ongoing interest in our laboratories.

Synthesis
Chemical synthesis (ESI Schemes 1 and 2 †) and characterization of compounds is outlined in the ESI. † Purity of all dyes was determined to be of >95% by UPLC-UV/Vis traces at 254 nm and dye specic l max that were recorded on a Waters H-class instrument equipped with a quaternary solvent manager, a Waters autosampler, a Waters TUV detector and a Waters Acquity QDa detector with an Acquity UPLC BEH C18 1.7 mm, 2.1 Â 50 mm RP column (Waters Corp., USA).
Absolute quantum yields were determined for TMR(-d12) and SiR(-d12) by rst measuring steady-state UV-vis absorption spectroscopy on a Jasco V780 spectrophotometer and a Specord S600 (Analytik Jena) in 1 cm cuvettes. Steady-state emission spectroscopy: emission spectra were measured on an Edinburgh FLS980 emission spectrouorometer in a 1 cm cuvette at 90 angle. The solutions were prepared to have an absorbance of 0.1 at 510 nm for TMR and at 600 nm for SiR in PBS and quantum yields were determined on the same instrument equipped with an integrating sphere.

In vitro photobleaching
Solutions (20 mM) of TMR and TMR-d12 were prepared in activity buffer (containing in mM: NaCl 50, HEPES 50, pH 7.3 + 0.1% BSA). An aliquot of 10 mL was transferred into a 1.5 mL Eppendorf vial and spun down to form an homogenous aqueous drop at the bottom of the plastic tube. An aliquot was exposed to a white light beam in order to bleach the uorophore with an Hg (Xe) arc lamp (LOT-QuantumDesign GmbH, Darmstadt, Germany) for 0, 1, 2, 3, 4, or 5 minutes. Aer bleaching, each aliquot was diluted by addition of 90 mL activity buffer + 0.1% BSA and carefully mixed by pipetting up and down, and 50 mL of this solution were transferred into a 10 Â 3 mm black quartz cuvette with a side window (Hellma, Jena, Germany). Fluorescence spectra were recorded with a Jasco spectrouorometer (FP-6500) at 25 C from 300-750 nm with an excitation wavelength of 544 nm over 1 nm steps (BW (Ex) 5 nm, BW (Em) 1 nm, PMT 475 V). The experiment was performed in triplicate for each illumination time, the data points averaged and plotted versus time.

SNAP f and SNAP-Halo expression and purication
SNAP f was expressed and puried as described previously 16 and complete amino acid sequences for constructs used can be found in the ESI. † SNAP-Halo with an N-terminal Strep-tag and C-terminal 10xHis-tag was cloned into a pET51b(+) expression vector for bacterial expression and complete amino acid sequences for constructs used can be found in the ESI. † For purication, SNAP-Halo was expressed in the E. coli strain BL21 pLysS. LB media contained ampicillin (100 mg mL À1 ) for protein expression. A culture was grown at 37 C until an OD 600 of 0.6 was reached at which point cells were induced with IPTG (0.5 mM). Protein constructs were expressed overnight at 16 C. Cells were harvested by centrifugation and sonicated to produce cell lysates. The lysate was cleared by centrifugation and puried by Ni-NTA resin (Thermosher) and Strep-Tactin II resin (IBA) according to the manufacturer's protocols. Puried protein samples were aliquoted in activity buffer (containing in mM: NaCl 50, HEPES 50, pH 7.3), ash frozen and stored at À80 C.

Protein labelling for FRET and full protein mass spectrometry
For protein labelling, 1 mL of the corresponding dye(s) (200 mM in DMSO) were diluted in 220 mL of a 227 nM solution of SNAP-Halo in activity buffer (10 mg mL À1 BSA was added to controls where no SNAP-Halo protein was present). This resulted in a $4fold excess of labelling substrate and mixing was ensured by carefully pipetting the solution up and down. The reaction mixture was allowed to incubate at r.t. for 1 h before 20 mL were removed for QToF MS analysis to ensure full labelling. The remaining solutions were subjected to spin column purication (Sartorius Vivaspin 500 30 kDa MWCO PES, #VS0122) for three times by adding 500 mL of activity buffer for each cycle. Finally, the solutions were reconstituted in activity buffer, of which 200 mL were transferred into a Greiner black at bottom 96 well plate and emission spectra were recorded on a TECAN INFIN-ITE M PLEX (TMR: l Ex ¼ 510 AE 10 nm; l Em ¼ 550-800 AE 20 nm; 25 ashes; 20 ms integration time; SiR: l Ex ¼ 610 AE 10 nm; l Em ¼ 650-800 AE 20 nm; 10 ashes; 20 ms integration time) to observe FRET. Donor or acceptor only labelled constructs, and donor plus acceptor with the addition of 10 mg mL À1 BSA served as controls, and in these cases acceptor emission was not observed. FRET efficiency was calculated from the sum of maximal emission values derived from the raw spectra I TMR (572-584 nm) and I SiR (660-672 nm) according to eqn (3): Cell culture and FACS of SNAP-GLP1R:CHO-K1 and SNAP- , a temperature controlled chamber and operated by LAS X. TMR(-d12) were excited using l ¼ 561 nm and emission signals were captured at l ¼ 576-670 nm. SiR(-d12) were excited using l ¼ 640 nm and emission signals were captured at l ¼ 655-748 nm. The confocal images were collected using a time gated Hybrid detector (0.5-6 ns). FLIM images of sufficient signal were acquired without gating on Hybrid detectors within 512 Â 512 pxl of 114 nm/pxl with 10 repetitions. Fluorescence lifetime decay curves from selected regions with clear plasma membrane staining were tted with two exponential functions and the mean amplitude weighted lifetime is reported for each region.
Confocal uorescence microscopy for photobleaching experiments were performed on living cells expressing SNAP-GLP1R in PBS at room temperature on LSM710 or LSM780 (Carl Zeiss) operated by Zen Black Soware using a 63Â (1.40 NA oil) objective. TMR(d12) were excited using l ¼ 561 nm and emission signals were captured at l ¼ 564-712 nm. SiR(d12) were excited using l ¼ 633 nm and emission signals were captured at l ¼ 637-740 nm. Emission signal were collected on 34 channel spectral detector (QUASAR, Zeiss), a typical gain of 750 over a scan area of 512 Â 512 pxl.
Confocal and STED microscopy experiments on living Cox8A-SNAP:HeLa cells were performed in Life cell imaging buffer (Gibco) at 37 C using a 100Â (1.45 NA lambda oil) objective on a Nikon TiEclipse operated by Micromanager and equipped with a STEDyCON (Abberior Instruments) with 405/ 488/561/640 excitation and a 775 nm depletion laser. TMR(-d12) were excited using l ¼ 561 nm and emission signals were captured at l ¼ 580-630 nm. SiR(-d12) were excited using l ¼ 640 nm and emission was collected at l ¼ 650-700 nm. Both Emission signals were collected by a time gated APD (0.5-8 ns) with 8Â signal accumulation and 100 nm pixel size for confocal images. STED images for SiR(-d12) were collected with 41% 775 nm depletion laser power and 15 nm pixel size.
Live STED images of microtubules in COS7 cells expressing Halo-Tubb5 labelled with Halo-SiR or Halo-SiR-d12 were acquired using l ¼ 640 nm excitation, 775 nm depletion and emission signals captured at l ¼ 655-700 nm using a time gated Hybrid detector (0.5-6 ns) with 4Â line averaging. Images of 1024 Â 1024 pixel had a pixel size of 18.9 nm. Line proles of 100 microtubules were selected in 4 cells and intensity, sigma and FWHM ¼ 2.35 Â sigma were determined aer Gaussian tting (Image J).
smFRET: labelling of SBD2 and analysis by msALEX spectroscopy We followed our published protocols for labelling and imaging of SBD2. 31,32,51 In brief, His-tagged SBD2 was incubated in buffer containing 1 mM DTT to keep all cysteine residues in a reduced state. Subsequently, SBD2 was immobilized on a Ni Sepharose 6 Fast Flow resin (GE Healthcare) and then incubated overnight at 4 C with 25 nmol of each uorophore dissolved in labelling buffer (50 mM Tris-HCl pH 7.6, 150 mM NaCl). Subsequently, SBD2 was washed with one column volume labelling buffer to remove unbound uorophores. Fluorophore labelled SBD2 was then eluted with 1 mL of elution buffer (50 mM Tris-HCl pH 7.6, 150 mM NaCl, 500 mM imidazole). SBD2 was then further puri-ed by size-exclusion chromatography (ÄKTA pure, Superdex 75 Increase 10/300 GL, GE Healthcare) to remove remaining uorophores and aggregates. For all proteins, the labelling efficiency was higher than 40% for each labelling site (ESI Fig. 3 †).
msALEX microscopy was performed on a home-built microscope that was described in Gebhardt et al. 52 ALEX experiments were carried out by diluting the labelled proteins to concentrations of $125 or $250 pM in 50 mM Tris-HCl pH 7.6, 150 mM NaCl supplemented with the ligand glutamine at concentrations described in the text. Before each experiment, the coverslip was passivated for 5 minutes with a 1 mg mL À1 BSA solution in imaging buffer. Data recording and analysis was described previously. 52 Single-molecule events were identied using an allphoton and dual-colour burst-search 53 algorithm with a threshold of 15, a time window of 500 ms and a minimum total photon number of 100.

Transient absorption spectroscopy
Excited-state lifetimes were measured by transient absorption spectroscopy using a home built set up. 54 Pump-pulses with a temporal duration of 110 fs are generated by converting the fundamental of a Ti:Sapphire laser by a TOPAS C (Light Conversion Ltd). The pump pulse energy was adjusted to 100 mW at the sample position. The excited-state dynamics were probed by a white light supercontinuum, generated by focusing a small portion of the fundamental laser into a CaF 2 plate. The polarisation between the pump beam and the probe beam was set to magic angle (54.7 ). The samples were placed in a 1 mm cuvette with PBS as solvent. The sample concentration was adjusted to yield an OD between 0.1 and 0.2 at the excitation wavelength. To ensure sample integrity, absorption spectra of the samples were measured before and aer each experiment.
The transient absorption data were analysed using the Python-based KiMoPack soware. Prior to analysis, the data was chirp corrected and globally t using a sum of exponentials.

Data availability
Raw data can be provided from the corresponding authors upon reasonable request.

Author contributions
JB conceived and supervised the study. JB designed, and with KR and PP synthesized and characterized chemical compounds. KR, CC and JB performed in vitro measurements. KCA, JE, HG and ML performed cell culture, cell sorting and microscopy. TC designed smFRET experiments. NW conducted smFRET experiments and data analysis. KR, AV and BDI recorded and evaluated quantum yields using an integrating sphere and timeresolved experiments. BJ provided reagents. JB wrote the manuscript with input from all authors.

Conflicts of interest
JB has a licensing deal with Celtarys Research for LUXendin distribution.