High photostability in non-conventional coumarins with far-red/NIR emission through azetidinyl substitution

Replacement of electron-donating N,N-dialkyl groups with three- or four-membered cyclic amines (e.g., aziridine and azetidine, respectively) has been described as a promising approach to improve some of the drawbacks of conventional fluorophores, including low fluorescent quantum yields (ΦF) in polar solvents. In this work, we have explored the influence of azetidinyl substitution on nonconventional coumarin-based COUPY dyes. Two azetidine-containing scaffolds were first synthesized in four linear synthetic steps and easily transformed into far-red/NIR-emitting fluorophores through N-alkylation of the pyridine moiety. Azetidine introduction in COUPY dyes resulted in enlarged Stokes' shifts with respect to the N,N-dialkylamino-containing parent dyes, but the ΦF were not significantly modified in aqueous media, which is in contrast with previously reported observations in other fluorophores. However, azetidinyl substitution led to an unprecedented improvement in the photostability of COUPY dyes, and high cell permeability was retained since the fluorophores accumulated selectively in mitochondria and nucleoli of HeLa cells. Overall, our results provide valuable insights for the design and optimization of novel fluorophores operating in the far-red/NIR region, since we have demonstrated that three important parameters (Stokes' shifts, ΦF, and photostability) cannot be always simultaneously addressed by simply replacing a N,N-dialkylamino group with azetidine, at least in nonconventional coumarin-based fluorophores.


INTRODUCTION
In recent years we are witnessing a new resurgence in the use of fluorophores based on small organic molecules on advanced biological imaging techniques. 1 Current research efforts are focused on (i) the development of novel far-red and near-infrared (NIR) fluorophores by using modern chemical synthetic tools, (ii) fine-tuning their photophysical properties (e.g., absorption and emission wavelengths, fluorescence quantum yields and lifetimes, Stokes shifts and photostability) and (iii) improving their physicochemical properties (e.g., aqueous solubility and cell permeability), which are all together required for the applications of fluorescence microscopy in biological systems. 2 Although numerous fluorophores are currently available, in vivo applications such as fluorescence-guided surgery urgently demand novel efficient, cell-permeable fluorescent probes based on low molecular-weight scaffolds operating in the optical window of the tissues. 3 Since the discovery of umbelliferone (7-hydroxycoumarin) at the end of the nineteenth century, 1d coumarins have been traditionally used as fluorescent organic molecules owing to their well-stablished photophysical properties and good cell permeability, and nowadays some coumarin derivatives (e.g., Alexa Fluor 350) are still being used in fluorescence microscopy.
However, the fact that conventional coumarins such as Coumarin 1 (compound 1 in Scheme 1) require UV excitation hampers most in vivo applications due to the toxicity of this radiation and its low capacity of penetration in biological tissues. For this reason, great efforts have been dedicated in the last decades to red-shift absorption and emission of coumarins into the blue-green-red region of the visible spectrum by introducing electron-withdrawing groups (EWG) at the coumarin skeleton, by extending the conjugation system through position 3,4 or by fusion of aromatic cycles, 5 including other fluorescent scaffolds. 6 Although the majority of the abovementioned coumarin derivatives maintain the electronwithdrawing lactone moiety of the original umbelliferone, which creates a push-pull effect with electron-donating groups incorporated at position 7 (e.g., N,N-dialkylamino or hydroxy/alkoxy), in recent years some groups have demonstrated that absorption and emission of these chromophores can also be red-shifted by modifying it. Indeed, green light emission was accomplished either through thionation of the carbonyl group or by extending the conjugation of the system at position 2 with a dicyanomethylene group, being the applications of dicyanomethylene-caged oligonucleotides and cyclic RGD-containing particularly appealing. 7 Very recently, we have reported the synthesis of novel coumarins in which one cyano group of dicyanocoumarin derivatives was replaced by a phenyl ring containing EWGs 8 or by the electron-deficient pyridine heterocycle 9 with the aim of further 4 increasing the push-pull character of the coumarin chromophore. The later coumarin scaffolds (compounds 2 and 3 in Scheme 1) allowed us to develop a new family fluorophores, nicknamed COUPY, whose photophysical properties can be easily tuned by selecting the appropriate combination of the N-alkylating group at the pyridine moiety and the substituent at position 4 of the coumarin skeleton (compounds 4-7 in Scheme 1). 9 Taking into account their low molecular weight, COUPY dyes offer many attractive features such as large Stokes' shifts, brightness and emission in the far-red/NIR region, as well as aqueous solubility and good cell permeability. Such novel fluorophores opened the way to solving some of the main drawbacks associated with conventional coumarin-based fluorophores. Recently, Lavis and collaborators described that replacement in conventional fluorophores of electron-donating N, N-dialkyl groups (e.g., rhodamines, coumarins, naphthalimides, acridines, etc.) with azetidine allows to increase the fluorescent quantum yield ( F ) in polar solvents. 10 A similar behavior was described by Liu, Xu and co-workers when the threemembered aziridine ring was used. 11 These structural modifications, as well as other based on 5 cyclic amines, 12 have been postulated to suppress twisted intramolecular charge transfer formation (TICT), which is one of the major non-radiative de-excitation pathways in many fluorophores containing N,N-dialkyl motifs. Indeed, twisting of the N,N-dialkylamino group out the fluorophore plane upon photoexcitation forms a non-emissive and short-lived reactive chemical species, which can be avoided through structural modifications and rigidification.
Rivera-Fuentes and co-workers have demonstrated that azetidinyl substituent can also be exploited to modulate photochemical processes in coumarin-based caging groups. 13 In addition, they have proposed a mechanism of fluorescence quenching for coumarin 1 that does not involve TICT states but rather H-bond induced non-radiative decay (HBIND). 14 Such unproductive decay channel was inhibited in 1Az analogue (Scheme 1) by azetidinyl substitution.
Taking into account all the antecedents on azetidinyl-substituted fluorophores, we set out to study how this modification could influence the photophysical and physicochemical properties of coumarin-based COUPY dyes (4-7, Scheme 1). With this idea in mind, herein we describe the synthesis and photophysical characterization of azetidinyl analogues of COUPY fluorophores (4Az-7Az, Scheme 1), in which diethyl-or dimethylamino groups at position 7 of the coumarin skeleton have been replaced by the four-membered azetidine ring.
Surprisingly, the fluorescent quantum yield of the resulting 7-azetidinyl-COUPY fluorophores was not significantly modified in aqueous media with respect their dialkylamino analogues.
However, azetidinyl substitution led to a substantial improvement in photostability and larger Stokes' shifts in polar solvents relative to the parent N,N-dialkylamino COUPY dyes, while emission was kept in the optimal far-red/NIR region. High cell permeability was retained after azetidine incorporation and the fluorophores accumulated selectively in mitochondria and nucleoli of HeLa cells as occurred with their parent N,N-dialkylamino dyes. Overall, our results provide valuable insights for the design and optimization of fluorophores operating in the far-red/NIR region.

Design, synthesis and characterization of 7-azetidinyl-coumarin fluorophores
The synthesis of azetidinyl-coumarin scaffolds 2Az and 3Az was first explored by following an straightforward methodology recently developed by us for the preparation of the parent N,N-dialkylamino-COUPY scaffolds 2 and 3, 9 which makes use of thiocoumarin precursors (Scheme 2). Unfortunately, the reaction of azetidinyl-coumarin 1Az 10 with Lawesson's reagent (LW) did not afford the expected compound (8) but several side-products, most of them resulting from the reaction of the thionating reagent with the azetidine moiety, as inferred by HPLC ESI-MS analysis (Scheme 2). Azetidine ring-opening occurred even under milder conditions (e.g., 60 ºC).
Taking into account the high reactivity of azetidine with Lawesson's reagent, an alternative synthetic route for the preparation of 2Az and 3Az was explored (Scheme 3) in which thionation and the formation of the exocyclic C=C double bond would occur before the incorporation of the four-membered azetidine heterocycle. Starting from commercially available 7-hydroxycoumarins (10a and 10b), reaction with trifluoromethanesulfonic anhydride in pyridine conduced to the expected triflate derivatives (11a and 11b, respectively) in near quantitative yield. 15 In this case, thionation with LW under standard conditions (reflux in toluene, overnight) afforded the expected 7-triflylthiocoumarins (12ab). The next step involved condensation of 12a and 12b with 4-pyridylacetonitrile to provide coumarins 13a (44%) and 13b (57%), respectively, after silica column chromatography.
As previously found with the parent COUPY scaffold 2 (Scheme 1), 9 the 1 H NMR spectrum of 2Az showed two sets of proton signals in an ∼93:7 ratio ( Figure 1). 1 H− 1 H NOESY experiments (Figures 1 and S1) confirmed the existence of E and Z interconverting rotamers in solution around the exocyclic carbon-carbon bond and that the E rotamer was the one preferred. In the case of the 4-CF 3 analogue (3Az), the E rotamer was the major species identified ( Figure S2).
Having at hand the two azetidinyl-COUPY scaffolds, we synthesized the corresponding Nalkylated pyridinium derivatives following our previously described methodology. 9 As shown in Scheme 3, reaction of 2Az and 3Az with methyl trifluoromethanesulfonate afforded Nmethylpyridinium-coumarin dyes 4Az and 6Az, respectively. Similarly, the use of 2,2,2trifluoroethyl trifluoromethanesulfonate allowed to obtain azetidinyl-COUPY dyes 5Az and 7Az. The four new coumarin derivatives were isolated after silica column chromatography as purple (4Az-6Az) and dark blue (7Az) solids, and their purity was assessed by HPLC ( Figure   S3). Characterization was carried out HR ESI-MS and 1D ( 1 H, 13 C, and 19 F) and 2D NMR.
Similarly to parent COUPY dyes, the E rotamer was the major species for the corresponding 7-azetidinyl derivatives 4Az, 5Az and 7Az, while around 7% of rotamer Z was identified for  The spectroscopic and photophysical properties of 7-azetidinyl-COUPY fluorophores (4Az-7Az) were investigated to assess the effect of replacing the 7-N,N-dialkylamino group in COUPY dyes by the four-membered azetidine ring. The UV-Vis absorption and emission spectra of the compounds were collected in six solvents of different polarity and viscosity (see Figure 1 and Figures S8-S11). The photophysical properties of 4Az-7Az are summarized in Tables 1, S1 and S2 and compared with those of their parent 7-N,N-dialkylamino-COUPY dyes (4-7, respectively). 9 As shown in Figure 2, the new 7-azetidinylcoumarins reproduced the trend previously found with 4-7 and provided intense coloured solutions due to the existence of an intense absorption band in the visible spectrum, with absorption maxima ranging from 515 nm (4Az) to 592 nm (7Az) in aqueous solution (PBS buffer), and from 566 nm (4Az) to 630 nm (7Az) in DCM. Such bathochromic effects with respect conventional coumarins (e.g., compare with 7-azetidinylcoumarin 1Az:  abs = 355 nm in H 2 O) 13 are a consequence of the lengthening of the conjugated chromophore and the increased push−pull character of the conjugated system in COUPY dyes. The absorption maximum of 4-CF 3 9 derivatives (6Az and 7Az) was significantly red-shifted with respect the 4-CH 3 coumarins (4Az and 5Az) (from 25 to 47 nm, depending on the compounds and on the solvent).
Moreover, the incorporation of a second strong electron-withdrawing CF 3 group moiety led to an additional red-shift in the absorption maximum (e.g., compare  abs = 541 and 607 nm for 6Az and  abs = 592 and 630 nm for 7Az in PBS and DCM, respectively).
As previously found with the parent N,N-dialkylamino dyes 4-7, the 7-azetidinyl-COUPY analogues showed negative solvatochromism since an increase in the solvent polarity led to a significant blue-shift in their absorption maxima (e.g., for 4Az, λ abs = 515 nm in PBS and 566 nm in DCM). However, this effect was much pronounced in azetidinyl-containing dyes than in their corresponding parent compounds (e.g., compare data for 4Az with that for 4: λ abs = 545 nm in PBS and 569 nm in DCM), which suggests that replacement of the dialkylamino group by azetidine is accompanied by an increase of the difference between ground and excited state dipole moments ( Figure S10 and Table S2). This blue-shift in absorption maxima in H 2 O upon azetidinyl substitution was also previously found for conventional coumarin 1 (e.g., compare  abs = 381 nm for 1 vs  abs = 355 nm for 1Az). 13 By contrast, the molar absorption coefficients (ε) of 7-azetidinyl-COUPY dyes were not significantly influenced by the polarity of the solvent, which is the opposite trend that we previously found for 7-N,N-dialkylamino-COUPY dyes in which a slight hyperchromism was showed in lesspolar solvents. In general, the azetidine-containing COUPY dyes exhibited ε values smaller than those of the parent 7-N,N-dialkylamino dyes. Tables 1 and S1, the compounds' emission maxima showed a slight blue-shift with respect the N,N-dialkylamino series in polar solvents (ca 4-8 nm; e.g.,  em = 675 nm for 7Az and  em = 682 nm for 7 in PBS), which parallels the solvent effect on the compounds' absorption maxima. As a result, 7-azetidinyl-COUPY dyes (except 7Az) exhibited larger Stokes' shifts in polar solvents compared with their 7-N,N-dialkylamino parent dyes (e.g., 112 nm for 6Az vs 92 nm for 6 in PBS buffer). Such large Stokes' shifts could find application in Förster-type resonance energy-transfer (FRET) experiments where a good separation between excitation and emission bands result in better contrast. As previously found with the parent 7-N,Ndialkylamino-COUPY dyes, the absorption and emission maxima for the 7-azetidinyl derivatives (4Az-7Az) was not significantly modified by changing the pH ( Figure S11). Table 1. Photophysical properties of 7-azetidinyl-COUPY dyes 4Az-7Az in different solvents. The data for coumarin derivatives 4-7 has been included for comparison purposes. 9 As shown in Tables 1 and S1, 4-CH 3 azetidinyl dyes (4Az and 5Az) exhibited excellent Φ F in less-polar solvents (e.g., 0.82 and 0.64 in toluene, respectively) whereas the fluorescence quantum yields for the 4-CF 3 analogues (6Az and 7Az) were moderate (e.g., 0.24 and 0.15 in toluene, respectively). To our surprise fluorescence quantum yields (Φ F ) of all the azetidinyl dyes both in polar protic and nonprotic polar solvents were similar (e.g 0.21 for 4Az and 0.18 for 4 in ACN) or even slightly lower than those of the parent N,N-dialkylamino-containing dyes (e.g., 0.06 for 7Az and 0.12 for 7 in ACN). As previously stated, the stabilization of the TICT excited state is a major cause for fluorescence quenching in polar solvents for many conventional fluorophores, although HBIND has been recently postulated as a deactivation mechanism of the excited state of coumarin 1. 13 Replacement of the 7-N,N-dialkylamino group in conventional coumarins with azetidine lead to a clear improvement in the fluorescent quantum yield of the fluorophore. (e.g.,  F = 0.06 for 1 and  F = 0.92 for 1Az in H 2 O). 13 Interestingly, Φ F values in glycerol (Table S1) were slightly lower in the azetidinyl-COUPY dyes (except for 4Az, which was higher) compared with N,N-dialkylamino-containing parent dyes, and similar fluorescence lifetime ( F ) values were obtained in this viscous polar solvent by time-resolved fluorescence spectroscopy ( Figure S12). Overall, all these observations indicate that fluorescence quenching of non-conventional coumarin-based COUPY fluorophores in polar media cannot be exclusively attributed to the formation of a TICT excited state or to HBIND, which should had been prevented in their 7-azetidinyl-COUPY analogues, but to other competing deactivation channels.
Very importantly, the photostability of azetidinyl-COUPY dyes was considerably increased in aqueous media relative to the N,N-dialkylaminocounterparts. As shown in Figure 3, the photostability of coumarin 4Az was similar to that of 4, while 5Az was photostable up to light fluences 3.3-fold higher than in the case of 5. The large photostability of the 4-CF 3 series is particularly appealing since these compounds are photostable up to light fluences larger than 200 (7Az) and 400 (6Az) J/cm 2 , fluences that are more than 10-and 20-fold, respectively, higher than those used for cellular imaging experiments. 17 13 In summary, the replacement of N,N-dialkylamino groups with azetidine in conventional coumarins (e.g., 1 vs 1Az) and in non-conventional coumarins such as COUPY dyes (e.g., 4-7 vs 4Az-7Az) leads to significant differences in their photophysics ( Figure 4). The larger blueshifts in the absorption spectra of 4Az-7Az in polar solvents indicate that their ground state is more polar than the excited state when compared with 4-7. However, azetidinyl substitution does not significantly modify the position of the emission wavelength. As a result, azetidinyl-COUPY dyes show enlarged Stokes shifts in polar solvents compared with their parent COUPY dyes.

Fluorescence imaging of 7-azetidinyl-COUPY dyes in living cells
Having stablished how azetidine substitution influences the photophysical properties of the azetidinyl derivatives of COUPY dyes, we investigated their cellular uptake in living cells by using confocal microscopy. Compounds 4Az to 6Az were irradiated with a yellow light laser ( ex = 561 nm) and 7Az with a red one ( ex = 633 nm). As shown in Figure 5, fluorescence was clearly observed in all cases in different cellular organelles after 30 min of incubation with the compounds, thereby confirming a good uptake by HeLa cells. Moreover, azetidinyl-COUPY dyes reproduced the same pattern of staining than their N,N-dialkylamino parent fluorophores (4-7) since mitochondria and nucleoli were clearly stained, together with intracellular vesicles. 9 This subcellular localization of 4Az was confirmed with colocalization experiments using commercially available specific markers for labelling mitochondria (MitoTracker Green FM, MTG), lysosomes (Lysotracker Green FM, LTG) and nuclei (Hoechst 33342). In addition, accumulation into nucleoli was further investigated through enzymatic digestion with RNase A. calculate the intensities of one channel co-localizing with the other. This coefficient ranges from 0 to 1 and is a good indicator even when the intensities between both channels clearly differ. Our results showed a clear correlation between MTG and 4Az signals with a Pearson's coefficient being equal to 0.778 on average. Moreover, the Mander's coefficients also confirmed that 4Az was located in the mitochondria. The amount of co-localization of 4Az over MTG (M1) was 0.54 on average, whereas that of MTG over 4Az (M2) was 0.77, which indicates that there is more MTG signal that co-localizes with 4Az than 4Az co-localizing with MTG. This was already expected as the 4Az staining was also located in the cytoplasm, vesicles and nucleoli where no MTG staining is observed. Similarly, the fluorescence observed on some vesicles along the cytoplasm was predominantly associated with lysosome accumulation, as inferred from co-localization experiments with Lysotracker Green ( Figure   S14). Two additional experiments were carried out to investigate accumulation of the compound inside the nucleoli. An indirect evidence of nucleoli localization was first obtained when costaining 4Az with Hoechst 33342 since the fluorescence emission of the coumarin dye inside the nuclei was only observed in the spots where lacked Hoechst staining (Figure 6, arrowheads). In order to further confirm the origin of the staining of nucleoli by 4Az, we performed a ribonuclease (RNase A) digestion experiment following a previously reported procedure. 18 As shown in Figure S15, the fluorescence signal of coumarin 4Az at the nucleoli 16 was completely lost after treatment with RNase, which indicates that RNA is the target of coumarin 4Az in the nucleoli.
Overall, confocal microscopy experiments in living HeLa cells indicate that azetidinyl-COUPY fluorophores retain the cell membrane permeability of COUPY dyes, allowing visualization of specific cellular organelles, mainly mitochondria, lysosomes and nucleoli, after incubation during 30 min.
Finally, the photostability of coumarins 4Az-7Az was also evaluated in HeLa cells by continuous irradiation with the laser beam of the confocal microscope ( ex = 561 nm for 4Az to 6Az and  ex = 633 nm for 7Az). As shown in Figure

Materials and Methods
Unless otherwise stated, common chemicals and solvents (HPLC grade or reagent grade quality) were purchased from commercial sources and used without further purification.
Aluminium plates coated with a 0.2 mm thick layer of silica gel 60 F 254 were used for thinlayer chromatography analyses (TLC), whereas flash column chromatography purification was carried out using silica gel 60 (

Synthesis of 7-azetidinyl-COUPY scaffold 3Az
4 . The published method with some modifications was followed to synthesize compound 10b. 15 Trifluoromethanesulfonic anhydride (1.7 mL, 7.7 mmol) was added dropwise to a solution of 7-hydroxy-4trifluoromethyl-coumarin (1.62 g, 7.0 mmol) in pyridine at 0 °C. After stirring for 5 h at room temperature, AcOEt (50 mL) was added to the reaction mixture, and the resulting organic phase was washed with saturated NaCl (3 × 50 mL). The organic layer was washed with 5 % aqueous HCl (5 × 40 mL) and saturated NaCl (2 × 40 mL), dried over anhydrous Na 2 SO 4 and filtered. The solvent was removed under reduced pressure to give 2.41 g (95 % yield) of a white solid, which was used without further purification in the next step.

Cell culture and treatments
HeLa For co-localization experiments with Lysotracker Green (0.2 M), the same procedure as for Mitotracker Green was used but without Hoechst 33342 treatment.
For digestion experiments with RNase enzyme, HeLa cells were treated with coumarin 4Az (5 M) for 30 min at 37 ºC in non-supplemented DMEM. After removal of the medium and two washes with DPBS, cells were fixed with 4% paraformaldehyde (Sigma) in PBS for 10 min at RT. Then, cells were permeabilized (2 x 5 min) with PBS buffer containing glycine (20 mM) and treated with 0.05% Saponin in PBS buffer containing glycine (20 mM) for 10 min at room temperature. After permeabilization, cells were washed again (2 x 5 min) with PBS buffer containing glycine (20 mM For photostability studies, HeLa cells were incubated with the compounds (5 M, 30 min at 37 ºC) and kept in low glucose DMEM without phenol red for fluorescence imaging.

Fluorescence imaging
All microscopy observations were performed using a Zeiss LSM 880 confocal microscope equipped with a 405 nm laser diode, an Argon-ion laser, a 561 nm laser and a 633 nm laser.
The microscope was also equipped with a full enclosure imaging chamber (XLmulti S1, Pecon) connected to a 37 ºC heater and a 5% CO 2 providing system. Cells were observed using a 63X 1.2 multi immersion objective. Coumarins 4Az to 6Az were excited using the 561 nm laser and detected from 570 to 670 nm. Coumarin 7Az was excited with the 633 nm laser and detected from 650 to 750 nm. In co-localization studies, Mitotracker Green FM and Lysotracker Green were observed using the 488 nm laser line of the Argon-ion laser whereas the 405 nm laser diode was used for observing Hoechst 33342.
In photostability studies cells were continuously irradiated every 5 s with the 561 nm laser at 15.8 W for 5 min. Laser power was measured using a photodetector (model 818-UV, Newport) connected to an optical power meter (model 840-C, Newport).
Image analysis was performed using Fiji (1.51d). 23 Unless otherwise stated, images are colorized using Fire lookup table. In photostability studies mean intensity was measured at each time point in at least 4 different areas of 5 m 2 at each subcellular compartment (mitochondria, cytoplasm and nucleoli) and also in areas without cells to measure the background. Images were filtered with a median filter of radius 1 to reduce noise before intensity measurements.
In co-localization studies, images were processed using Fiji. First each channel was filtered using a median filter with a radius of 1 and a Gaussian filter with a radius of 2. Then background was subtracted with a rolling ball radius of 50. To measure the Manders' coefficients a threshold including all the cytoplasm and discarding the background was set.
On the other hand, the Pearson's coefficient measurements were performed after manually selecting and clearing the 4Az signal at the nucleoli. The same selection and clearing was applied on the Mitotracker channel. Both co-localization coefficients were measured using the JaCoP plugin 18 on the different stacks of images (n=4) with each stack containing 3 to 5 cells.