Photolabile coumarins with improved efficiency through azetidinyl substitution† †Electronic supplementary information (ESI) available: Experimental details, ESI figures and NMR spectra. See DOI: 10.1039/c7sc03627b

The efficiency of photoactivatable coumarins in water has been enhanced by substitution with azetidine.


Introduction
Photochemical processes have been used extensively in a wide range of elds, including chemical biology, 1 organic synthesis, 2,3 and materials science, 3 to gain selectivity and spatiotemporal control over the studied phenomena. The successful application of photochemical strategies relies on the ability of researchers to control the outcome of photoinduced reactions, by tuning chromophore parameters like absorption maxima, reaction selectivity or quantum efficiencies. A commonly adopted approach to modulate the photochemical evolutions of molecules consists in exploring structural variations of an already functioning scaffold [4][5][6] to look for new motifs or substituents that enhance the desired properties.
Recently, Lavis and co-workers demonstrated that azetidinylsubstituted uorophores such as 1 (l abs ¼ 355 nm; l em ¼ 471 nm; Chart 1) display larger quantum yields of uorescence (f F ) compared to those substituted with open-chain analogues, such as dimethylamino or diethylamino (compound 2, l abs ¼ 381 nm; l em ¼ 468 nm), or larger cyclic amines, such as pyrrolidinyl or piperidinyl. 7 Subsequently, Xu and co-workers found that aziridinyl-substituted uorophores behave similarly. 8 This improvement in f F values has been attributed to a decrease in the rate of population of twisted intramolecular charge transfer (TICT) states 9 upon excitation. 7,8 We hypothesized that if these small heterocyclic rings suppress competitive decay channels that affect emission, the same effect could be exploited to improve the efficiency of other important, non-emissive photochemical processes.
4-Methylcoumarin derivatives are widely used as uorophores and photocleavable (also known as "caging") groups. 10,11 Electron-rich coumarins may undergo photoinduced heterolytic cleavage when C4 is substituted with a methylene bearing a good leaving group (Scheme 1). The quantum yields of photoactivation (f PA ) of such coumarins are usually moderate to low [11][12][13][14][15][16][17] and depend on the nucleofugality of the leaving group. 11 It would be advantageous to be able to improve the efficiency of these photocleavable groups, in particular in a manner that is independent from the nucleofugality of the leaving group. Herein we report that azetidinyl substitution improves the efficiency of f PA of a series of photocleavable coumarins with leaving groups of diverse nucleofugality.

Results and discussion
We synthesized 7-azetidinylcoumarin derivatives 3a-d, 7diethylaminocoumarin derivatives 4a-d and the julolidinefused derivatives 5a-d (Chart 2). We prepared the julolidine series because for these compounds the involvement of the TICT state is prevented by incorporation of the nitrogen into a system of fused rings. The f PA values of compounds 3a-d, 4ad and 5a-d were measured in mixtures of phosphate-buffered saline (PBS, pH 7.4) and MeCN (3/7, v/v). The rates of photolysis of 0.3 mM solutions were determined via HPLC by following the disappearance of the starting material upon irradiation at 405 nm (7.5 mW, see ESI †). Fig. 1 depicts the results of photorelease experiments, including kinetic traces for representative runs of compounds 3b, 4b and 5b (Fig. 1a) and a comparison of the measured f PA values of all the derivatives studied (Fig. 1b). In general, the f PA values of derivatives 3a-d and 5a-d are comparable and signicantly larger than the ones measured for derivatives 4a-d.
The most noticeable exceptions are compounds 3c, 4c and 5c, which have overall lower f PA , but even within this family, 3c and 5c seem to be photoreleased slightly more efficiently than 4c. For all other compounds, the 4or 5-fold increase in f PA observed is consistent with the increase observed between the f F of compounds 1 and 2. These results conrm our hypothesis that azetidinyl substituents can be used to improve the efficiency of photochemical processes beyond uorescence through inhibition of competitive deactivation pathways.
We then turned to investigate the nature of the decay channel inhibited by the azetidinyl substituent. Whereas it has been suggested that small-ring substituents inhibit the population of TICT states, 7,8 the existence of TICT states for 7substituted coumarins has never been rmly established. 9 As a starting point, we performed femtosecond (fs) transient absorption (TA, 400 nm excitation and 100 fs time resolution, see ESI †) and uorescence up-conversion spectroscopy measurements (FLUPS, 400 nm excitation, 1340 nm mixing pulse, 0.1 mm BBO, see ESI †) on the model uorophores 1 and 2, to look for spectral signatures of a TICT state. We chose to work on the uorophores because the measured f F values (0.92 for 1 vs. 0.06 for 2) suggest that switching to the azetidinyl substituent renders radiative decay the predominant excited state process, whereas for the esters 3-5 heterolytic bond cleavage contributes to the kinetics of excited state decay, which may obscure the analysis.
The FLUPS and TA data were analyzed globally assuming a series of exponential steps to obtain evolution associated emission or differential absorption spectra (EAES or EADS), respectively. 18 FLUPS data in H 2 O and DMSO (Fig. 2) could be described for both dyes assuming three successive exponential steps. In H 2 O (Fig. 2a-d), a two-step red shi was observed, with a magnitude twice as large for 2 than for 1 ($1200 and $200 cm À1 vs. $680 and $100 cm À1 , respectively). Furthermore, the longer time constant (reecting the uorescence lifetime, s F , also determined from time-correlated single photon counting experiments, see ESI †), is more than one order of  magnitude smaller for 2 than for 1 (0.40 ns vs. 5.0 ns). On the contrary, in DMSO ( Fig. 2e-f) the red shis (700 cm À1 for both compounds) and the s F (2.7 ns for 2 vs. 3.3 ns for 1) are comparable for the two dyes. The initial red shis reect the equilibration of the surrounding polar solvent molecules. 19 The larger shis observed with 2 in H 2 O point to substantial solvent reorganization and stabilization of the excited state, such as those associated with hydrogen-bond (H-bond) interactions. 20 The shorter s F of 2 in H 2 O and MeOH (see ESI †), but not in other highly polar solvents such as MeCN and DMSO (see ESI †), does not support involvement of a TICT state. The TA data did not reveal either any spectral signature attributable to any state other than the optically-populated excited state (see ESI †). Therefore, other deactivation pathways must be operative.
The shortening of s F for 2 in polar protic solvents strongly suggests the possibility of H-bond induced non-radiative decay (HBIND) as deactivation mechanism of the excited state, a phenomenon already reported for other dyes. [21][22][23] The efficiency of this process was shown to depend on both the H-bond donating strength of the solvent, described by the Kamlet-Ta parameter a, 24 and the ability of the solvent to make an H-bond network, quantied by the density of OH groups (r OH ), itself approximated as 1/V s , where V s is the volume of a solvent molecule. To test this hypothesis, we determined the nonradiative decay rates (k nr ) of 1 and 2 from the s F and f F values in a series of solvents of different H-bonding abilities (Fig. 3a, see ESI †). The correlation observed between k nr and ar OH supports the involvement of HBIND as deactivation channel for 2 in protic solvents, but does not explain why the same pathway is not active for 1, which is structurally very similar.
To shed light on the origin of these differences, we analyzed the solvatochromism of the absorption and emission bands of 1 and 2 in solvents of different polarities, both protic and aprotic ( Fig. 3b and c, see ESI †). From these measurements it is possible to estimate the ground (m g ) and excited (m e ) state electric dipoles of the two dyes (see ESI †). 25,26 We observed that the dipole moments of 1 (m g ¼ 3.9, m e ¼ 9.5 D) are signicantly smaller than for 2 (m g ¼ 7.5, m e ¼ 11.8 D). Analysis of the solvatochromism of the absorption bands reveals further details. For 2 in protic solvents, a consistent bathochromic shi is observed (Fig. 3d), whereas for 1 both batho-and hypso-chromic shis are found. These shis can be rationalized considering the internal charge transfer (ICT) character of the rst singlet excited state of coumarins, in which the substituent at C7 donates electron density to the carbonyl group at C2. 25,27 Hbonding to the nitrogen atom stabilizes more the ground than the excited state resulting in a blue shied absorption, whereas H-bonding to the carbonyl group has the opposite effect. The observed shis of the absorption maximum (Fig. 3d) suggest that for dye 1, H-bonding is happening at both ends of the molecule; by contrast, in the case of compound 2, H-bonding seems to involve mainly the carbonyl moiety. This observation suggests that in both states the carbonyl of 2 has higher electron density than in 1, and is thus more prone to H-bonding. The smaller dipole moments of 1 relative to 2 point to weaker electron-donating strength of the azetidinyl substituent. This observation is in agreement with the higher ionization potential  of phenylazetidine (7.61 eV) 28 compared to that of N,N-diethylaniline (6.95 eV). 29 To conrm that this mechanism also applies to the photolabile esters, we measured the f PA of compounds 3a-5a using npentanol (PeOH) and EtOH as co-solvents (see ESI †) instead of PBS because both display lower H-bonding ability than H 2 O. In EtOH mixture (for PeOH results, see ESI †), the f PA of 3a, 4a and 5a became very similar (f PA ¼ 0.007, 0.0037, and 0.01, respectively). These efficiencies are in general lower than in aqueous mixture, which can be attributed to the reduced stabilization (and hence formation) of the ion pair in less polar solvents. These observations further support our working hypothesis that azetidinyl substituents shut down HBIND-associated decay of coumarin excited states, leading to increased uorescence or photoactivation efficiency in H 2 O.
Finally, the applicability of this improved photoremovable group was conrmed in live cells. We chose to prepare compound 6 (Fig. 4, for synthesis see ESI †), which is a photoactivatable uorescein probe. Photoactivatable dyes are useful for a number of applications including cell tracking, 30 intracellular diffusion experiments 31 and super-resolution microscopy. 32 Live human cervical cancer (HeLa) cells were incubated with compound 6 in growth medium for 30 min. Aer this time, imaging in the green channel (l ex ¼ 488 nm; l em ¼ 525 AE 25 nm) gave nearly no signal, conrming that the uorescein fragment of compound 6 remains in the dark, spirolactone form (Fig. 4) and the coumarin is not excited at that wavelength. A region of interest comprising a single cell (Fig. 4) was irradiated with a 405 nm laser (120 mW, 50% power), which induced a 7-fold increase in uorescence intensity. This experiment conrms that the azetidinylcoumarin protecting group can be removed with light in live cells.

Conclusions
To the best of our knowledge, azetidinyl substituents have been used, in the context of photochemistry, only to increase the f F of uorophores. 7,8 We have demonstrated, however, that this simple substitution leads to the enhancement of other useful photoprocesses in the presence of H 2 O. We also propose a mechanism of quenching that does not invoke TICT states, but rather HBIND as the unproductive decay channel that is shut down by azetidinyl donors in coumarins. Albeit beyond the scope of this work, we propose that azetidinyl substitution could enhance other important photochemical processes in H 2 O; for instance, would an azetidinylated version of methylene blue be more efficient in photocatalysis 33 or photodynamic therapy? 34 Further studies are required to investigate the origin of the increase in f F observed for other uorophores and determine whether azetidinyl substitution affects the population of TICT states, as currently thought, or it deactivates other excited state decay pathways as it seems to be happening for the coumarins reported here. Testing these hypotheses might reserve interesting challenges and surprises.

Conflicts of interest
The authors declare no competing nancial interests.