Rational Design of bright and photostable dyes with long Stokes shift. Recently reported strategies offer two important routes to improve properties of regular rhodamines: symmetric xanthene with inhibition of TICT effect and asymmetric xanthene with vibronic structure (Fig. 1). However, none of the reported approaches can simultaneously improve brightness, photostability, and Stokes shift. Thus, we assumed that a combination of vibronic structure and TICT inhibition could open up an opportunity to develop a new type of bright and photostable fluorophores with long Stokes shift (Fig. 1).
We started from the asymmetric rhodamine to enlarge the Stokes shift and BDQF-1 was thus synthesized based on the reported work23, 24. Consistent with previous work, BDQF-1 displays a greatly increased Stokes shift (99 nm) than RhB (27 nm)23. However, only a weak fluorescence signal was collected from BDQF-1 in PBS buffer solution (Supplementary Fig. 4 and Supplementary Table 1), limiting its application in a cellular environment. Inspired by the reported approaches that optimize the brightness through tuning the electron density in xanthene13, 14, 19, we thus developed new dyes BDQF-(2-7) that contained various EWGs in quinoxaline moiety of the fluorophore scaffold. Such fluorophores can be easily synthesized by the condensation of 2-(4-diethylamino-2-hydroxybenzoyl)benzoic-acid and quinoxaline moieties BDQ-(1-6), which were generated via nucleophilic substitution (BDQ-(2-5)) or amidation followed by the reduction reaction (BDQ-6) (Fig. 2a and Supplementary Scheme 1).
In BDQF dyes, the emission maxima (λem) showed a clear hypochromatic shift from 673 nm to 590 nm as the electron withdrawing ability of quinoxaline substituents increases, which showed a linear correlation between λem and Hammett substituent constant (σp) (Supplementary Fig. 5). The inductive effect of the substituents was further demonstrated by the well-regulated changing of the electron density of xanthene scaffolds in DFT calculation (Supplementary Fig. 6). Meanwhile, we found that the stronger the electron withdrawing ability of quinoxaline in BDQF-(1-6) was, the higher brightness and smaller Stokes shift the fluorophores displayed (Fig. 2b). Considering the importance of both brightness and Stokes shift, we identified BDQF-6 with 2-(2,2,2-trifluoroethyl) octahydropyrrolo[1,2-a]pyrazine group as the best fluorophore. It displayed a 2.4-fold increase in quantum yield (0.74) and 2-fold enhancement in brightness (ε × Φ = 6.6 × 104 L·mol−1·cm−1) in aqueous solution compared to its parental fluorophore, RhB. Meanwhile, BDQF-6 showed a red-shifted excitation/emission spectra of 578 nm/634 nm, generating a long Stokes shift of 56 nm (Fig. 2c). In addition, BDQF-6 displayed significantly higher photostability and less photobluing over RhB under the illumination of light at 530 nm for 80 min (Fig. 2d and Supplementary Fig. 7-9). We also observed highly consistent fluorescence intensity and lifetime of BDQF-6 in various solvents or buffer solutions with abundant proteins, which cannot be achieved in RhB (Fig. 2e, 2f, Supplementary Fig. 10). Importantly, DFT calculation of BDQF-6 showed the asymmetric electron distribution in the HOMOs (Fig. 2g) and increased energy barrier to form TICT state (Fig. 2h), thus indicating the coincidence of vibronic feature24 and TICT inhibition14 in BDQF-6. The outstanding performance of BDQF-6 thus demonstrates the exceptional strength of the proposed synergistic strategy in the dye development.
Evolution of BDQF-6 derived probes for live-cell confocal imaging. We next incubated Hela cells with BDQF-6 to evaluate its cell permeability, brightness, and photostability in live-cell imaging. BDQF-6 showed a similarly fast cellular staining as RhB, while possessed a stronger fluorescence signal. The significantly increased fluorescence signal is assumed to be due to the combination of higher brightness and larger Stokes shift from BDQF-6 (Supplementary Fig. 11). Importantly, RhB can be easily photobleached within 10 min under continuous irradiation at 560 nm, while negligible signal change of BDQF-6 was observed, indicating the outstanding photostability of BDQF-6 in cellular environment (Fig. 3a and Supplementary Fig. 12-13). To achieve the specific labeling of target protein, we next synthesized BDQF-6 ligand (BDQF-9-Halo) from carboxyl-containing BDQF-6 (BDQF-9) for labeling HaloTag, a widespread protein tag31 (Fig. 3b). Incubation of live HeLa cells expressing
a HaloTag–histone 2B (H2B) fusion with BDQF-9-Halo offered a bright nuclear staining (Supplementary Fig. 14), providing 1.5-fold stronger fluorescence signal than RhB derivatives, RhB-Halo (Fig. 3c-3f). Interestingly, compared to RhB-Halo, BDQF-9-Halo displayed a much lower cytoplasmic background, which thus endowed a higher nuclei-to-cytosol signal ratio (Fcyt/Fnuc=18) without any washing steps (Fig. 3c-3f). Since it was reported that shifting the equilibrium from the zwitterionic form to the spirocyclic form could help optimize the cell permeability and fluorogenicity18, 19, we next transferred the carboxyl group in BDQD-9 to acyl 2,2,2-trifluoroethylamide, producing probe BDQF-10. D50, representing the dielectric constant at which half of the fluorophore population is in the zwitterionic form, has been commonly applied to evaluate the equilibrium between zwitterionic and spirocyclic form32. Based on the reported work, BDQF-10 with a D50 of 45 was thus expected to be a good candidate for generating fluorogenic probes (Supplementary Fig. 15). The subsequent synthesized BDQF-10-Halo HaloTag ligand showed a large increase in absorbance (23-fold) and fluorescence intensity (490-fold) upon binding to HaloTag (Supplementary Fig. 16). Consistently, the intracellular background signal of live Hela cells treated with BDQF-10-Halo is extremely low, thereby offering a superior signal-to-noise ratio in nuclear protein labeling (Fnuc/Fcyt = 106) (Fig. 3e). Notably, even the formation of spirolactam in BDQF-10 improves the fluorogenicity and cell permeability (Supplementary Fig. 17), it also reduces the brightness to some extent due to the incomplete recovery to the fluorescent zwitterionic state upon HaloTag binding (Fig. 3f)21.
We next extended this bright and photostable fluorophore BDQF-6 to prepare new probes for organelles staining in living cells. Probe BDQF-6-Mito and BDQF-6-Lyso were synthesized by one-step esterification and amidation reaction respectively (Supplementary Scheme S1). We succeeded to utilize these probes in fast and high contrast wash-free staining in mitochondria and lysosomes, as confirmed by the colocalization with MitoTracker Green and LysoTracker Green respectively (Fig. 3g-3j and Supplementary Fig. 18-20).
Superior photostability of BDQF-6 derivatives in STED imaging. STED microscopy allows the visualization of biological structures with high spatial and temporal resolution in living cells33, 34, 35, 36, 37, 38. However, the severe photobleaching of fluorophores has greatly limited the frame number collected in STED imaging. Till now, only a very few reported probes for covalently labeling of proteins can provide satisfying performance in STED imaging, especially 3D STED microscopy39, 40, 41. Rhodamine derivatives (e.g., SiR32, CPY42, JF60815) are the most popular fluorophores in STED imaging due to their good photophysical properties, cell permeability, and photostability3. We thus compared the performance of BDQF-, CPY- and JF608-derived probes, which share a similar emission wavelength, in STED microscopy with a 775 nm depletion laser. BDQF-9-Halo, CPY-Halo, and JF608-Halo were synthesized and applied for specific staining of vimentin fused with HaloTag in fixed U-2 OS cells. CPY-Halo and JF608-Halo provided STED images with full width at half-maximum (fwhm) resolution of 86±9 and 83±10 nm in the first frame. However, only 2-3 frames of STED images with >50% of the initial fluorescence intensity were obtained due to the rapid photobleaching under a 775 nm depletion laser (Fig. 4a, 4b and Supplementary Fig. 21-22). In contrast, under the identical conditions, BDQF-9-Halo offers 9 frames of STED images with >50% of the initial fluorescence intensity while remaining the full width at half-maximum (fwhm) resolution of 57±5 nm (Fig. 4a and 4b and Supplementary Fig. 23-24). When optimizing the imaging settings to obtain the highest resolution, BDQF-9-Halo enabled the visualization of vimentin filaments with an fwhm of 37±4 nm (Supplementary Fig. 25). The significantly increased frame numbers and resolution in STED imaging demonstrate the superior photostability of BDQF-9-Halo. We thus tried to utilize BDQF-9-Halo in 3D STED imaging, which is difficult to be achieved when rapid photobleaching of traditional fluorophores occurs during sequential xzy-scan38. Incubation of U-2 OS cells transiently expressing mitochondrial import receptor Tomm20-HaloTag with BDQF-9-Halo enabled the construction of 3D STED images of Tomm20 along the whole mitochondria (Fig. 4c and Supplementary Fig. 26). In addition, the excellent cell permeability and high contrast staining utilizing BDQF-9-Halo and BDQF-10-Halo allowed us to perform the wash-free live-cell STED imaging of vimentin filaments with a resolution of 59±7 nm (Fig. 4d and Supplementary Fig. 27).
Extension of the strategy to different types of fluorophores. Numerous classic fluorophore scaffolds contain the dialkylamino motif and often suffer from TICT, leading to the decreased quantum efficiency and photostability14, 15, 47, 48. Encouraged by the excellent performance of BDQD-6, we next extended the strategy to other widely-used fluorophores with different colors. Replacing the dialkylamine with 2-(2,2,2-trifluoroethyl)octahydropyrrolo[1,2-a]pyrazine in the xanthene of rhodol (BDQD-11) vastly increased the quantum yield from 0.21 to 0.62 and the brightness from 1.3 × 104 to 3.2 × 104 L·mol−1·cm−1. Meanwhile, BDQD-11 displayed a red-shifted absorbance/emission maxima from 518 nm/546 nm to 548 nm/612 nm with a long Stokes shift of 64 nm (Table 1 and Supplementary Fig. 28). The confocal imaging of live Hela cells also indicated the greatly improved photostability of BDQD-11 than its parental fluorophore (Supplementary Fig. 29). Importantly, similar improvements in the brightness, photostability, and Stokes shift were also found in other xanthene-containing fluorophores, such as pyronin (Table 1 and Supplementary Fig. 28, 30).
We next applied the strategy to coumarin and Boranil that possess very different fluorophore scaffolds. Consistently, the introduction of 2-(2,2,2-trifluoroethyl)octahydropyrrolo[1,2-a]pyrazine motif in BDQF-(13-15) also offers dramatically increased brightness (5.1-8.1 folds) and improved photostability (Supplementary Fig. 31-34 and Supplementary Table 4-5). Notably, BDQF-15 and BDQF-14 displayed a large Stokes shift of 136 nm and 92 nm respectively (Table 1). Chloroalkane was next conjugated with BDQF-16, an analog of BDQF-14, to produce HaloTag ligand BDQF-16-Halo (Fig. 5a and Supplementary Scheme 1). Similar to the optical properties of BDQF-14, BDQF-16-Halo also exhibited a large Stokes shift (110 nm) in aqueous solution (Fig. 5b). Interestingly, probe BDQF-16-Halo displayed a good nuclear protein labeling without any washing steps, while its parental probe Coumarin-Halo showed no specific staining under the same conditions (Fig. 5c and 5d). Furthermore, BDQF-16-Halo enables the imaging of vimentin filaments in live-cell STED microscopy using a depletion laser of 595 nm (Supplementary Fig. 35). Overall, these results demonstrate that replacing the dialkylamino motif with 2-(2,2,2-trifluoroethyl)octahydropyrrolo[1,2-a]pyrazine is generalizable to different fluorophore scaffolds, producing substantial improvements in brightness, photostability, and Stokes shift simultaneously.
Bright and photostable fluorophores are always highly desired to develop chemical sensors to avoid the false signal from fluorescence photobleaching43, 44. It is worth noting that the new strategy only modifies the dialkylamino motif on one side of the xanthene scaffold in Rhodol, which leaves the oxygen atom on the other side free for producing sensors45, 46. In proof-of-principle experiments, the phosphate group was conjugated to BDQF-11 to develop a sensor for qualitatively detecting alkaline phosphatase (ALP) (Fig. 5e). BDQF-11-ALP showed ultra-weak fluorescence in buffer solution, while the addition of alkaline phosphatase can remove the phosphate group, thereby resulting in a huge signal enhancement (Fig. 5f and 5g). BDQF-11-ALP also displayed good specificity towards ALP in the presence of various biomolecules (Fig. 5h). Hela cells incubated with BDQF-11-ALP showed a much stronger fluorescence signal than normal liver L02 cell line (Fig. 5i and Supplementary Fig. 36), probably due to the higher expression level of ALP in tumor cells49, 50. Meanwhile, the addition of Na3VO4 reduced the ALP level in Hela cells, thus producing a decreased fluorescence signal (Fig. 5i and Supplementary Fig. 36).