Bioorthogonal labeling with tetrazine-dyes for super-resolution microscopy

Genetic code expansion (GCE) technology allows the specific incorporation of functionalized noncanonical amino acids (ncAAs) into proteins. Here, we investigated the Diels-Alder reaction between trans-cyclooct-2-ene (TCO)-modified ncAAs, and 22 known and novel 1,2,4,5-tetrazine-dye conjugates spanning the entire visible wavelength range. A hallmark of this reaction is its fluorogenicity - the tetrazine moiety can elicit substantial quenching of the dye. We discovered that photoinduced electron transfer (PET) from the excited dye to tetrazine is the main quenching mechanism in red-absorbing oxazine and rhodamine derivatives. Upon reaction with dienophiles quenching interactions are reduced resulting in a considerable increase in fluorescence intensity. Efficient and specific labeling of all tetrazine-dyes investigated permits super-resolution microscopy with high signal-to-noise ratio even at the single-molecule level. The different cell permeability of tetrazine-dyes can be used advantageously for specific intra- and extracellular labeling of proteins and highly sensitive fluorescence imaging experiments in fixed and living cells.

S ingle-molecule localization microscopy is a powerful method for subdiffraction-resolution fluorescence imaging of cells and tissue 1,2 . Since the density of fluorophores controls the achievable structural resolution 3 , efficient and specific labeling with fluorescent probes is a decisive factor in this super-resolution microscopy technique. Despite recent progress in the development of new fluorophores with higher fluorescence quantum yield, photostability, and intrinsic photoswitching in aqueous buffer [4][5][6][7][8] , specific and efficient labeling of the molecule of interest with minimal linkage error remains challenging. As the field moves toward ever higher spatial resolution using, e.g. expansion microscopy 9,10 , the effective size of the label (fluorophore, linker and affinity reagent) will be the main limiting factor of super-resolution microscopy.
Immunolabeling with antibodies is still the method of choice for fluorescence imaging of fixed cells, but the large size of antibodies introduces a displacement of 10-15 nm of the fluorophore from the molecule of interest 1 . This linkage error is smaller for fluorescent proteins and protein-based self-labeling tags but is still several nanometers 11 . Small (1.5 × 2.5 nm) camelid antibodies (so called nanobodies) directed against green fluorescent protein (GFP) or shorter peptide epitopes have been used successfully for super-resolution microscopy 12 but the palette of targets is small 13 . Even the most optimized labeling strategies using short peptide tags labeled with bivalent nanobodies or fluorescently labeled monomeric streptavidin yield linkage errors of~2 nm in direct stochastic optical reconstruction microscopy (dSTORM) experiments [14][15][16][17] . This problem demands the development of efficient labeling methods with small dyes, which can be site-specifically and quantitatively attached to a protein of interest with low linkage error 18 .
Genetic code expansion (GCE) technology enables the introduction of noncanonical amino acids (ncAAs) with small functional groups at any position in a target protein [19][20][21] . In this strategy, a native codon is replaced with a rare codon, such as the amber (TAG) stop codon, at a specific site in the gene of the protein of interest. The modified protein is then expressed in mammalian cells along with an additional tRNA-tRNA synthetase pair (tRNA-RS) that is orthogonal to the host translational machinery. The active site of the tRNA synthetase enzyme is engineered to only accept a specific ncAA, which is incorporated into a tRNA that recognizes the rare codon. The ncAA is simply added to the growth medium and thereby incorporated into the protein at a specific site [19][20][21] . A particularly promising type of ncAA include strained alkenes, such as trans-cyclooct-2-ene (TCO*), that can react with a 1,2,4,5-tetrazine in an ultrafast, specific, and bioorthogonal inverse electron-demand Diels-Alder reaction. In this strategy, TCO*-modified ncAAs, such as TCO*-L-lysine (TCO*-Lys), can be efficiently and directly labeled with organic dyes with minimal linkage error 22,23 . The high selectivity and rate of this click chemistry reaction has resulted in a large number of commercially available fluorophore-tetrazine conjugates allowing labeling of mammalian cells and whole organisms with organic dyes, even in living systems [24][25][26] .
Another interesting property of this labeling strategy is the potential for fluorogenicity. It has been reported that some tetrazine-functionalized dyes (tetrazine-dyes) function as fluorogenic probes, meaning these compounds substantially increase fluorescence intensity upon reaction with the strained dienophile such as TCO 27,28 . This makes tetrazine-dyes especially interesting for live-cell labeling and fluorescence imaging applications since the fluorogenic reaction could lower background and potentially eliminate the need for washing out excess fluorophore. Despite these apparent advantages however, the use of ncAAs and tetrazine-dyes to label proteins for super-resolution microscopy applications remain rare and unoptimized [29][30][31][32] .

Results
Spectroscopic characteristics of tetrazine-dyes. The absorption and emission maxima of the 22 tetrazine-dyes were measured in aqueous buffer (PBS, pH 7.4) before and after reaction with ncAA TCO*-Lys (Table 1). In TCO*-Lys isomerization to the less reactive cis-isomer is prevented and the compound is stable for several days under physiological conditions 33 . The majority of tetrazine-dyes are commercially available as 3-methyl-6-phenyl-1,2,4,5-tetrazine (Me-Tet) derivatives and only some of them are also available as 3-phenyl-1,2,4,5-tetrazine (H-Tet) derivatives. Those dyes which are not available as tetrazine derivatives were synthesized by reacting N-hydroxysuccinimide derivatives of the fluorophores with 3-(p-benzylamino)−1,2,4,5-tetrazine (H-Tetamine; Supplementary Fig. 1). The main difference between Me-Tet and H-Tet dye conjugates are the different reactivity and chemical stability: H-Tet exhibits an approximately 30-fold higher click reaction rate constant but a lower chemical stability 33 .
Almost all tetrazine-dyes investigated exhibited a strong fluorescence increase upon reaction with TCO*-Lys (Table 1, Fig. 1). The increase in fluorescence intensity is most pronounced for the shorter wavelength absorbing dyes ATTO425, ATTO465, and ATTO488 with turn-on ratios of 15-40. Due to the broad absorption spectrum of tetrazine peaking at~515 nm it can efficiently quench the fluorescence of dyes emitting at wavelengths ≤ 550 nm by fluorescence resonance energy transfer (FRET) (Supplementary Fig. 1) 27 . Hence, the tetrazine chromophore can act as both quencher and bioorthogonal click-reaction group. Indeed, the turn-on ratios of the tetrazine-dyes ATTO425, ATTO465, ATTO488, and ATTO532 after coupling to TCO*lysine scale according to the emission maxima of the dyes and their overlap with the absorption spectrum of tetrazine (Table 1, Supplementary Figs. 1 and 2).
The longer wavelength absorbing oxazine dyes ATTO655, ATTO680, and ATTO700, and the large Stoke-shift dyes AZ503 and AZ519 show turn-on ratios of 6-13 (Table 1, Fig. 1). These high turn-on ratios are not entirely reflected in the increase in fluorescence lifetime indicating that short fluorescence lifetime components are missed in our time-correlated single-photon counting (TCSPC)-experiments with a time resolution of approximately 40 ps.
Long-term spectroscopy studies revealed that some dyes such as ATTO465, ATTO488, and ATTO655 show a slight but highly reproducible decrease in fluorescence intensity after the initial strong increase upon addition of TCO*-Lys (Fig. 1), which can be attributed most likely to tautomerization of the TCO*-tetrazine conjugate resulting in the release of lysine by decarboxylative elimination 34 . Elimination of the dye moiety from the lysine results in the formation of a new heterocyclic compound (pyridazine), which potentially quenches the fluorescence of some dyes depending on their electronic properties.
Comparative click reactions demonstrate that the click reaction of H-Tet-ATTO488 and H-Tet-Cy5 with TCO*-Lys proceeds ~30-times faster than for the corresponding Me-Tet derivatives 35 but the turn-on ratios differ strongly especially for the two ATTO488 tetrazine derivatives ( Supplementary Fig. 3). In particular, H-Tet-ATTO488 is substantially less quenched than Me-Tet-ATTO488. Accordingly, the turn-on ratio is much higher for the Me-Tet derivative. Assuming similar FRET efficiencies for both ATTO488 conjugates the different turn-on ratios can only be explained by the reduced chemical stability of H-Tet 35 . Obviously, a fraction of H-Tet-ATTO488 decomposes spontaneously in physiological buffer because of the higher reactivity, which results correspondingly in a lower FRET-efficiency and turn-on ratio ( Supplementary Fig. 3). Table 1 Spectroscopic characteristics (absorption maxima, λ abs (nm), emission maxima, λ em (nm), fluorescence lifetimes, τ (ns)) of flexibly linked tetrazine-dyes after click reaction with TCO*-lysine and turn-on ratios measured in PBS, pH 7.4. Click reactions were performed by adding 25 µM TCO*-Lys to 1 µM dye solutions a The quality of the fit was judged by the reduced χ 2 values and the randomness of the weighted residuals. For all fits, χ 2 values between 0.900 and 1.200 were obtained. Most clicked fluorescent dyes showed nonexponential fluorescence kinetics, which could be best described using a biexponential model. Therefore, average fluorescence lifetimes τ av = a 1 τ 1 + a 2 τ 2 are given. The lifetime values in brackets were measured after clicking with TCO*-lysine b The turn-on ratio was determined from the minimum and maximum fluorescence intensity value recorded before and after addition of 25 µM TCO*-Lys to the tetrazine-dyes (  Turn-on ratio Turn-on ratio Turn-on ratio Turn-on ratio Higher quenching efficiencies with turn-on rates of >1,000 have been achieved by using through-bond energy transfer (TBET) in tetrazine-BODIPY derivatives 28 . However, the TBET quenching is decreasing with increasing absorption wavelength of the dyes 36 . Therefore, bodipy and coumarin dyes have been used mostly in studies, which are less suited for super-resolution microcopy 27,28,37 . Flexibly-linked tetrazine-dyes as used here are easier to access from a synthetic standpoint but exhibit lower turn-on ratios after reaction with dienophiles (Table 1) 29 .
In addition, strong static quenching was observed for the two oxazine dyes ATTO655 and ATTO700 as well as for the Sirhodamine dye JF 646 (Fig. 2) suggesting that non-or only weakly fluorescent ground-state complexes between the longer wavelength absorbing oxazine and Si-rhodamine dyes and Me-Tetamine form in aqueous solutions, where the lifetime of the complexes is below the time-resolution of our TCSPC setup.
Me-Tet-ATTO655, Me-Tet-ATTO680, Me-Tet-ATTO700, and Me-Tet-JF 646 show a high turn-on ratio (Table 1) and are therefore efficiently quenched by Me-Tet-amine also in intramolecular quenching experiments. This finding indicates that the dye and the tetrazine moiety, when connected by a flexible linker, can adopt a nearly coplanar stacking conformation, which is required for efficient fluorescence quenching.
Our results indicate that the underlying quenching mechanism in longer wavelength absorbing oxazine and Si-rhodamine dyes is only efficient at very short distances, i.e. within complexes. This is also supported by the observation that the fluorescence intensity of ATTO655, ATTO680, and ATTO700 tetrazine conjugates increases strongly upon addition of denaturing agents such as guanidinium chloride while ATTO488 and Cy5 show only small effects if at all (Fig. 3). Hence, it can be concluded that hydrophobic interactions between the longer wavelength absorbing oxazine and Si-rhodamine dyes and Me-Tet with a stacked arrangement of the conjugated π-electron systems play an important role in the formation of ground-state complexes with ultrafast fluorescence quenching. Since tetrazine exhibits a very high electron affinity 40,41 , the excited longer-wavelength absorbing dyes are most likely efficiently quenched in their stacked conformation via photoinduced electron transfer (PET). The carbocyanine dyes such as Cy5 exhibit a higher water solubility and are thus less prone to form complexes with Me-Tet. In the case of Me-Tet-ATTO488 fluorescence quenching is dominated by through-space FRET where complex formation between dye and tetrazine moiety does not play a dominant role.
Fluorescence quenching dynamics. Knowing that ATTO655 and similar dyes form nonfluorescent complexes with tetrazine in aqueous solutions driven by hydrophobic interactions, we can assume an equilibrium between two states for the dye in conformationally flexible tetrazine-dyes: an 'open' fluorescent state A open and a 'closed', complexed nonfluorescent state B closed . Both states are populated according to the closing and opening rate constants, k closing and k opening . To investigate this equilibrium in more detail, we performed fluorescence correlation spectroscopy (FCS) experiments (Fig. 4). FCS analyzes temporal fluorescence fluctuations from highly diluted samples (nanomolar concentrations), probing molecules diffusing through a confined detection volume (typically 1 femtoliter with 1-20 molecules) by Brownian motion 42 . Characteristic time scales of molecular processes that result in fluctuating fluorescence emission can be measured under thermodynamic equilibrium conditions with nanosecond time resolution [43][44][45] .
Under moderate excitation conditions, no photophysical process, such as intersystem crossing, is apparent in the submillisecond time domain of FCS-curves recorded from Me-Tet-ATTO655 after reaction with TCO*-Lys (Fig. 4a). The FCS curve shows a millisecond decay corresponding to the molecules' diffusion through the confocal excitation/detection volume. However, the FCS curve recorded from Me-Tet-ATTO655 before addition of TCO*-Lys displays an additional fast decay occurring on the nano-to microsecond time scale together with a change in amplitude (Fig. 4a). Bimolecular experiments with ATTO655 and 25 mM tetrazine reveal a similar FCS decay (Fig. 4b) confirming the idea of diffusion-driven formation of non-fluorescent complexes. By fitting an analytical FCS curve representing twostate fluorescence intermittency to the bimolecular data with a mixture of free ATTO655 and tetrazine we estimated k closing = (0.04 ± 0.01) × 10 9 s −1 and k opening = (0.015 ± 0.002) × 10 9 s −1 . These values represent an association rate constant of (1.5 ± 0.5) × 10 9 M −1 s −1 and an association constant of (100 ± 23) M −1 , close to the values from Stern-Volmer analysis (a sphere-ofaction contribution in the Stern-Volmer plots of ATTO655 renders the Stern-Volmer association constant slightly larger) 38 . Concentration estimates from the amplitude of FCS curves confirm that all fluorophores contribute to the observed bimolecular FCS decays and no population of fluorophores exist, that are quenched over time periods longer than~1 ms (the diffusion-limited observation time) (Fig. 4c).
It is thus reasonable that the nano-to microsecond decay recorded for Me-Tet-ATTO655 reflects fluorescence fluctuations caused by intramolecular quenching of ATTO655 upon contact with Me-Tet mediated by the conformational flexibility of the linker. The time scale of the fast quenching fluctuations is well separated from that of translational diffusion. The calculated rate constants for opening and closing, k opening of~9.0 × 10 6 s −1 and k closing of~3.3 × 10 6 s −1 , demonstrate that ATTO655 stays on average 27% of the time in its quenched complexed conformation. The fast decay thus accounts for about one fourth of the observed turn-on ratio of 13 ( Table 1). The remaining turn-on ratio is indeed reflected in a concentration difference by about a factor of four that is observed from the amplitude of FCS curves (Fig. 4c). This observation of multiple decay components on time scales above and below the FCS diffusion time of~1 ms, which are not detected in intermolecular quenching experiments (Fig. 4b) indicates that the intramolecular linker introduces additional conformational constraints for the process of complex formation.
Whereas Me-Tet-ATTO680 and Me-Tet-ATTO700 show similar behavior, the FCS-curves recorded from Me-Tet-Cy5 and Me-Tet-Cy5B in the absence and presence of TCO*-Lys are identical demonstrating that quenched complexes are not formed. The observed fast decay for Cy5 is due to well-characterized cis/ trans-isomerization ( Supplementary Fig. 5) 46 . Accordingly, we observed that Cy5B, which is not capable of isomerization due to the constrained conjugated bond structure, does not show the isomerization decay (Fig. 4d).
Super-resolution microscopy with tetrazine-dyes. In order to enable testing of different tetrazine-dyes for super-resolution microscopy applications in an experimentally easy and comparable way, we followed a recently introduced method 37 and synthesized phalloidin-TCO, for labeling of the actin skeleton of fixed cells and subsequent attachment of tetrazine-dyes by click chemistry. Confocal fluorescence images demonstrate that all tetrazine-dyes are well-suited for high-end microscopy ( Fig. 5, Supplementary Fig. 6). However, comparative fluorescence labeling and imaging experiments with unmodified phalloidin show that some tetrazine-dyes exhibit the tendency to bind nonspecifically to intracellular structures. Especially some of the yellow to red absorbing ATTO-dyes (e.g. Me-Tet-ATTO550, Me-Tet-ATTO565, and Me-Tet-ATTO590) show substantial nonspecific background signal ( Supplementary Fig. 6).
Next, we tested their performance in super-resolution microscopy by re-scan confocal microscopy (RCM) 47,48 and structured illumination microscopy (SIM) 49 providing a theoretical resolution improvement factor of 1.4 and 2.0, respectively (Fig. 6,  Supplementary Figs. 7 and 8). Most tetrazine-dyes investigated, including the new dyes HMSiR 5 , JF 646 6,7 , and Cy5B 8 , exhibit a low tendency to bind nonspecifically to cellular structures and provide super-resolved actin images with high signal-to-noise ratio. Furthermore, fluorogenic tetrazine-dyes such as Me-Tet-ATTO488, Me-Tet-ATTO655, and Me-Tet-ATTO680 with turn-on ratios ≥ 6 ( Table 1) enable wash-free super-resolution fluorescence imaging (Fig. 6b) 30,31 . In addition, also other tetrazine-dyes such as H-Tet-Cy3 and H-Tet-Cy5 with lower turn-on ratio allow wash-free fluorescence imaging (Supplementary Fig. 9). This finding demonstrates that not only the turn-on ratio but likewise a high water-solubility and low tendency of unspecific binding to cellular components in combination with superior click reactivity can promote wash-free imaging of intracellular structures.
Next, we performed single-molecule localization microscopy experiments to verify that tetrazine-dyes can be used as well for single-molecule sensitive fluorescence imaging (Fig. 7). The obtained super-resolution images demonstrate, that H-Tet-Cy5 coupled to phalloidin-TCO can be reliably photoswitched in the presence of millimolar concentrations of thiols enabling dSTORM imaging 14,15 of the actin skeleton with superior spatial resolution (Fig. 7a). In addition, we tested also the bridged carbocyanine dye Cy5B 8 for PALM-like imaging under reductive imaging conditions (Fig. 7b) and the spontaneously blinking dye HMSiR 5 in PBS, pH 7.4 without addition of photoswitching buffer (Fig. 7c) and achieved similar image qualities; or to be precise, the dyes exhibit slightly different localization precisions of~12 nm (Cy5 The TNFR superfamily is a very important class of signal transduction molecules in the immune system 51 . TNFRs are single-membrane-spanning proteins that contain an extracellular TNF-binding region and a cytoplasmic tail. There are several antibodies commercially available that recognize the extracellular domain of TNFRs. However, antibody binding often interferes with biological function of TNFRs 52 . To enable fluorescence imaging of TNFR1 and GluK1 in living cells with organic dyes, we selected five different extracellular positions for ncAAs incorporation into GluK2 and TFNR1. The positions were selected to ensure easy accessibility for bioorthogonal click labeling with tetrazine-dyes based on available structural data (GluK2: S47TAG, S272TAG, K306TAG, S309TAG, S343TAG; TNFR1: S42TAG, Q77TAG, R121TAG, S137TAG, K161TAG). For identification of the best mutant we used the following selection criteria. First, with tdEOS and eGFP, respectively, attached C-terminally after the amber suppression site, the fluorescence signal is only detectable when the ncAA is incorporated properly into the protein of interest. Hence, only those mutants are selected, which exhibit a strong fluorescence intensity (expression level of the fluorescent protein). Second, the fluorescent proteins should be localized at the expected cellular position (here, the plasma membrane) to ensure selective extracellular labeling with tetrazine-dyes. Third, bioorthogonal labeling with plasma membrane impermeable tetrazine-dyes (H-Tet-Cy5) should show selective membrane labeling. Based on these experiments, we identified GluK2 S343TAG -eGFP and TNFR1 S42TAG -tdEOS as the best candidates for further experiments.
To improve the incorporation efficiency of ncAAs and reduce the background signal we used a recently optimized pyrrolysinebased CGE system for click chemistry 53 . By adding a strong nuclear export signal (NES) to the N-terminus of the pyrrolysine tRNA synthase (PylRS AF ) sequence (NESPylRS AF ) (AF refers to a previously described PylRS mutant that accepts bulky side-chain moieties), unspecific nuclear labeling is reduced 53 . In combination with cell-impermeable dyes such as H-Tet-Cy5 this approach enables selective and efficient labeling of extracellular domains of correctly incorporated membrane receptors in living cells First, we exchanged serine at position 42 by TCO*-Lys in the extracellular domain of TNFR1 (TNFR1-tdEOS) in HEK293T cells by amber suppression. We tested different H-Tet-Cy5 concentrations ranging from 10 nM to 5 µM and found that labeling with 1.5 µM H-Tet-Cy5 for 10 min provides the best labeling efficiency 54 . Cells were labeled with H-Tet-Cy5 and imaged by confocal laser scanning microscopy. Live-cell fluorescence images clearly show selective labeling of TNFRs in the plasma membrane of cells (Fig. 8a).
Next, we labeled the kainate receptor GluK2 (GluK2 S343TAG -eGFP) in HEK293T cells by GCE (amber suppression) with TCO*-Lys and H-Tet-Cy5. Live-cell confocal fluorescence images show again exclusively extracellular fluorescence labeling of GluK2 (Fig. 8b). Corresponding dSTORM images demonstrate a b c that GluK2 is expressed at high expression rates but homogeneously distributed in the plasma membrane of fixed HEK293T cells (Fig. 8c).
To verify the labeling efficiency, we performed immunofluorescence labeling of kainate receptor GluK2 and TNFR in HEK293T cells. Since commercially available anti-GluK2 antibodies are directed against the intracellular domain cells have to be fixed and permeabilized before labeling. Confocal fluorescence images demonstrate that different antibodies can be used successfully to visualize kainate receptors in fixed cells (Supplementary Fig. 10). Staining of plasma membrane TNFRs with two commercial antibodies showed specific membrane labeling but resulted in negligible or substantially lower labeling efficiency as compared to bioorthogonal click labeling ( Supplementary Fig. 11). These results demonstrate that bioorthogonal click chemistry with tetrazine-dyes can be used advantageously for labeling of membrane receptors. Site-specific incorporation of ncAAs into the extracellular domain of plasma membrane receptors enables live-cell visualization of plasma membrane receptors that cannot be labeled by antibodies. In addition, it enables higher labeling densities than standard immunolabeling with antibodies. This suggests that the extracellular labeling sites of plasma membrane receptors is more accessible to small tetrazine-dyes such as H-Tet-Cy5 than to larger IgG antibodies with a molecular size well above 10 nm. Furthermore, antibodies may crosslink to a certain degree during the fixation process and thus induce apparent clustering of receptors in the plasma membrane 54 .
Live-cell bioorthogonal labeling. So far intracellular live-cell bioorthogonal labeling by GCE and fluorescence imaging has been hampered by the high fluorescence background resulting from the excess of unincorporated ncAAs 33 . Even water-soluble tetrazine-dyes such as H-Tet-SiR 4 or tetrazine-dyes with high turn-on ratio do not automatically permit high-end live-cell fluorescence imaging of intracellular target proteins. In general, incorporation of TCO*-Lys into proteins in response to an inframe amber stop codon does not alter cell viability 20,25 . Depending on the incorporation site the expression level and cellular location might, however, be influenced. This off-target binding further increases the fluorescence background.
To demonstrate live-cell bioorthogonal labeling and imaging with tetrazine-dyes we selected tubulin as an important cytoskeleton component of cells in our experiments. Microtubules are dynamic structures composed of αβ-tubulin molecules that are constantly integrated or degraded as the microtubules grow and shorten. Microtubule dynamics can be easily monitored in live cells using fluorescently labeled tubulin and video microscopy. Since overexpression of α-tubulin can potentially influence the dynamics of endogenous microtubules after incorporation and generate a high fluorescent background 30 , we used two alternative live-cell tubulin labeling strategies. First, we introduced TCO*-Lys into the microtubule-associated protein (MAP) ensconsin (E-MAP-115), which is known to be better tolerated by cells at high expression levels 55 and second, we synthesized Docetaxel-TCO (see Supplementary Methods), a microtubule-stabilizing antimitotic drug [56][57][58] (Fig. 9). It has been shown that cells expressing four to ten times the physiological level of endogenous MAP exhibited microtubule dynamics indistinguishable from those of untransfected cells indicating that E-MAP-115 most likely serves to modulate microtubule functions or interactions with other cytoskeletal elements 55 . We inserted TCO*-Lys into the microtubule binding domain (EMTB) of E-MAP-115, which is C-terminally tagged with three GFPs (EMTB K87TAG −3xGFP) 59 for live-cell bioorthogonal labeling and fluorescence imaging (Fig. 9). Thus, monitoring of the GFP signal allowed us to select strongly expressing cells and change to TCO*-Lys free medium one day before bioorthogonal labeling with cellpermeable H-Tet-SiR 4,30 in order to minimize the non-specific background signal. In addition, we used again the tRNAPyl/ NESPylRS AF pair to improve nuclear export of tRNA-Synthetase 53 .
Using this optimized approach, we were able to perform fluorescence imaging of microtubule dynamics in living cells by re-scan confocal microscopy (RCM) 47,48 (Fig. 9d, Supplementary  Movies 1 and 2). However, additional single-molecule localization microscopy experiments of fixed EMTB K87TAG −3xGFP cells with H-Tet-HMSiR illustrate that the fluorescence background of unbound EMTB still limits the achievable image quality (Fig. 9c). Next, we tested whether Docetaxel-TCO specifically binds to microtubules and enables live-cell labeling of the cytoskeleton with H-Tet-SiR. Live-cell standard confocal as well as superresolution microscopy images recorded by RCM and SIM 49 demonstrate specific labeling of the cytoskeleton by click chemistry (Fig. 9d). Microtubules are polar filaments exhibiting a dynamic instability, i.e. extensive phases of microtubule growth are followed by rapid disassembly and regrowth 60 . To investigate microtubule dynamics, we performed live-cell re-scan confocal time lapse microscopy showing that EMTB K87TAG −3xGFP labeled microtubules exhibit distinct polymerization and degradation events (Supplementary Movies 1 and 2, Supplementary  Fig. 12a). Microtubules labeled with 10 µM Docetaxel-TCO show damped dynamics lacking microtubule polymerization and degradation events (Supplementary Movie 3 and Supplementary  Fig. 12b). This demonstrates that the microtubule-stabilizing antimitotic drug Docetaxel-TCO is not suited for studying microtubule dynamics in live cells if used at concentrations > 100 nM 57 .

Discussion
With continuously increasing spatial resolution of superresolution microscopy the development of small and efficient fluorescent probes generating minimal linkage errors becomes particularly important. Site-specific introduction of a ncAA such as TCO*-Lys into the amino acid chain of the protein of interest followed by bioorthogonal click chemistry with tetrazine-dyes represents a broadly useful possibility to overcome current limitations and enable high-end fluorescence imaging with organic dyes. In addition, a continuously increasing number of organic dyes is commercially available as tetrazine-dyes. Whereas FRET is responsible for fluorescence quenching in blue-and greenabsorbing dyes, we discovered PET from the excited dye to tetrazine as the main quenching mechanism in red-absorbing oxazine and Si-rhodamine derivatives. We show that a stacked arrangement of the conjugated π-electron system of the dye and tetrazine is required for strong fluorescence quenching. Upon reaction with dienophiles quenching interactions are reduced resulting in a considerable increase in fluorescence intensity.
The resulting fluorescence increase can be used advantageously for wash-free fluorescence imaging experiments. The two large Stokes-shift dyes Me-Tet-AZ519 and Me-Tet-AZ503 and the redabsorbing dyes H-Tet-Cy5, H-Tet-Cy5B, and H-Tet-HMSiR are particularly interesting for autofluorescence-free super-resolution fluorescence imaging by re-scan confocal microscopy or structured illumination microscopy, and single-molecule localization microscopy, respectively. On the other hand, some tetrazine-dyes tend to bind nonspecifically to intracellular structure highlighting the importance of tetrazine-dye selection and suited control experiments for each specific application. Our data also demonstrate that ncAAs can be incorporated site-specifically into even small extracellular domains of plasma membrane proteins for which antibodies are not available. Due to the smaller size of tetrazine-dyes they react also with protein sites that are difficult to access by much larger antibodies and thus enable higher labeling efficiencies. Because tetrazine-dyes provide linkage error free stoichiometric labeling of proteins at well-defined extracellular and intracellular positions they are ideally suited for singlemolecule fluorescence imaging and tracking as well as superresolution microscopy experiments in fixed and living cells.
DNA constructs and plasmids. Plasmid amplification was performed via transformation in E. coli (XL1-Blue) and DNA isolation via MIDI-prep (NucleoBond® Xtra Midi, Macherey & Nagel, #740410). The respective amber stop mutants were generated by introducing a TAG codon through PCR-based site-directed mutagenesis. The plasmid for expression of the TNF-receptor pTNFR1-tdEOS was a gift from Mike Heilemann (Addgene plasmid #98273). The plasmid for the expression of the Kainate receptor (pcDNA3-GluK2) was a gift from Dr. Peter Seeburg (Max-Planck Institute for Medical Research, Heidelberg, Germany). An AgeI restriction site was introduced at the C-terminus of the GluK2 coding sequence in front of the STOP-codon. Subsequently, GluK2 was cloned into the MCS of the pEGFP-N1 plasmid (Clontech #U55762) using the restriction enzymes EcoRI and AgeI. The plasmid for expression of the microtubule-associated protein Ensconsin (EMTB-3xGFP) was a gift from William Bement (Addgene plasmid #26741). The plasmid for the expression of the tRNA/tRNA-synthetase pair (pCMV tRNA Pyl / NESPylRS AF ) was kindly provided by Edward Lemke 53 .
Transfection. Cells were transfected at 60-80% confluence with a commercial transfection reagent (JetPrime, Polypus, #114-01) with the suitable plasmid/ reagent mixture (according to manufacturer´s recommendations). The total DNA amount per well was 500 ng in a 1:1 ratio between the DNA for the respective POI and pCMV-tRNAPyl/NESPylRSAF. The ncAA (TCO*A; SiChem GmbH, Bremen, #SC-8008) was supplemented separately at a final concentration of 250 μM, diluted 1:4 with 1 M HEPES. After 6-8 h the medium was exchanged to fresh cell growth medium. Cells were incubated~12-36 h before labeling and fixation/imaging.
Bioorthogonal labelling. For bioorthogonal labeling of clickable receptors the transfected HEK293T or COS-7 cells were incubated with 1.5 μM of the respective tetrazine-dye in cell growth medium for 10 min at room temperature or on ice. After incubation the cells were washed twice and fixed at room temperature with a mixture of 4% formaldehyde and 0.25% glutaraldehyde for 10 min. The cells were again washed three times with Hank's Balanced Salt solution (HBSS, Sigma, #55037 C) before imaging. Alternatively, the cells were imaged live without fixation. For labelling of actin filaments via pre-labeling with phalloidin-TCO the cells were first fixed with 4% formaldehyde in cytoskeleton buffer (10 mM MES pH 6.1, 150 mM NaCl, 5 mM EGTA, 5 mM glucose, 5 mM MgCl 2 ) at room temperature for 10 min. Cells were then washed 2-3 times with HBSS (ThermoFisher, #14025092) and then incubated for at least 1 h with phalloidin-TCO (~0.5 μM) at room temperature or alternatively for overnight at 4°C. Afterwards the cells were washed again 2-3 times with PBS and stored until further labeling/imaging. Fluorescence labeling was performed by adding the respective tetrazine-dye (1-3 μM in PBS) for 10 min at room temperature. Cells were then rinsed with 2-3 times PBS and imaged or, for wash-free applications, imaged without removing the tetrazine-dye solution.
Measurements of absorbance/emission spectra. Time-dependent fluorescence intensities were measured in quartz glass cuvettes using a FP-6500 spectrofluorimeter (Jasco). Fluorescence was excited at the absorption maxima. The sample temperature was adjusted to 25°C using a Peltier thermocouple.
Intermolecular fluorescence quenching experiments. To interpret quenching efficiencies of tetrazine and fluorophores, bimolecular quenching experiments with ATTO655, ATTO700, JF 646 , and Cy5 were performed with Me-Tet-amine in phosphate-buffered saline (PBS) at pH 7.4. Bimolecular static and dynamic quenching constants, K stat and K dyn , were determined from time-resolved and steady-state fluorescence quenching experiments using Stern-Volmer analysis: where τ 0 and F 0 are the fluorescence lifetime and intensity in the absence of a quencher, τ and F are the fluorescence lifetime and intensity in the presence of the quencher Q with the concentration [Q], and K stat and K dyn denote the static and dynamic Stern-Volmer constant, respectively. Dynamic quenching is attributed to collisional interactions at a dynamic quenching rate constant of k dyn . Static quenching originates in bimolecular complexes that are very efficiently quenched and K stat can thus be interpreted as an association constant K a . Experimental estimates are provided as results from linear regression together with standard error from the fitting procedure.
Fluorescence-correlation-microscopy. PET-FCS was performed using a custombuilt confocal fluorescence microscope setup as described previously 39,40 equipped with a 640 nm diode laser (OBIS640, Coherent Europe B.V., Utrecht, Netherlands). Samples were diluted in PBS at pH 7.4 and filtered through a 0.2 μm syringe filter, transferred onto a microscope slide and covered by a cover slip. Samples with concentrations of 1 nM yielded an average of ∼20 molecules in the detection focus of the microscope setup. Sample temperature was adjusted to 25°C using a custombuilt objective heater. For each sample, at least three individual autocorrelation functions were recorded of at least 5 min measurement time each. PET-FCS data was fitted to an analytical model for diffusion and a two-state equilibrium with offand on-rate constants k closing and k opening : where N is the number of detected molecules; τ D is the diffusion time; A = k closing / k opening is the equilibrium constant and τ rel = (k closing + k opening ) −1 is the exponential time constant.
Confocal microscopy. Confocal fluorescence images were recorded on a commercial LSM700 (Zeiss; ×63/1,4 oil objective). Single plane images were acquired with a pixel size of 90 nm and suitable settings for the respective dye. Images were adjusted for brightness and contrast (only linear changes).
Statistics and reproducibility. All experiments were performed at least three times unless otherwise stated. Statistical analysis of spectroscopic data was performed using OriginPro (OriginLab, MA). Experiments with one coherent dataset were performed successively to exclude instrumental variations.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.