Bis-Rhodamines Bridged with a Diazoketone Linker: Synthesis, Structure, and Photolysis

Two fluorophores bound with a short photoreactive bridge are fascinating structures and remained unexplored. To investigate the synthesis and photolysis of such dyes, we linked two rhodamine dyes via a diazoketone bridge (−COCN2−) attached to position 5′ or 6′ of the pendant phenyl rings. For that, the mixture of 5′- or 6′-bromo derivatives of the parent dye was prepared, transformed into 1,2-diarylacetylenes, hydrated to 1,2-diarylethanones, and converted to diazoketones Ar1COCN2Ar2. The high performance liquid chromatography (HPLC) separation gave four individual regioisomers of Ar1COCN2Ar2. Photolysis of the model compound—C6H5COCN2C6H5—in aqueous acetonitrile at pH 7.3 and under irradiation with 365 nm light provided diphenylacetic acid amide (Wolff rearrangement). However, under the same conditions, Ar1COCN2Ar2 gave mainly α-diketones Ar1COCOAr2. The migration ability of the very bulky dye residues was low, and the Wolff rearrangement did not occur. We observed only moderate fluorescence increase, which may be explained by the insufficient quenching ability of diazoketone bridge (−COCN2−) and its transformation into another (weaker) quencher, 1,2-diarylethane-1,2-dione.


■ INTRODUCTION
The possibility to modify two fluorophores (and change the emission parameters of two dye residues) in the course of one photochemical reaction is intriguing and remained unexplored. If we consider two masked (caged) fluorophores bound with a linker (Scheme 1), the assembly may include two photoconvertible caging groups, one for each fluorophore (Scheme 1A). In this case, the photoactivation is stepwise, and the whole structure represents only a bare aggregate of two caged dyes. Alternatively, if a single photoreactive group efficiently suppresses the emission of the whole compound, and this group can be transformed into a nonquenching state, then both fluorophores may be activated in one step (Scheme 1B). This option is particularly challenging, as the quenching efficiencies of energy or electron transfer strongly depend on the distance. Therefore, we have chosen a potential fluorescence quencher and used it as a linker directly connecting two (identical) fluorophores.
The literature survey revealed that the fluorescein derivatives incorporating benzil fragments (Ar 1 COCOAr 2 ) are essentially nonfluorescent (due to photoinduced electron transfer). 1−3 Therefore, we applied photoconvertible 2-diazo-1,2-diarylethanones Ar 1 COCN 2 Ar 2 closely related to Ar 1 COCOAr 2 , prepared bis-fluorophores bridged with a diazoketone linker, and studied their photolysis. Our motivation was to clarify whether the short diazoketone bridge (COCN 2 ) incorporated between two dyes will suppress their emission, and whether a Wolff rearrangement will take place. As fluorophores, we have used N,N′-bis(2,2,2-trifluorethyl)-substituted rhodamines, 4 which have absorption and emission spectra very similar to those of fluorescein. The structures of newly prepared compounds are given in Figure 1. Synthesis. The synthesis of bromorhodamines 6a,b from aminophenol 5 4 is given in Scheme 2. In the condensation reaction leading to compounds 6a,b, we compared two sets of conditions (see legend to Scheme 2). Higher yields (43−47%) were achieved when the first step was carried out without a solvent. Due to high temperature (160°C) and the presence of water in the gas phase, the partial cleavage of the 2,2,2trifluoroethyl amino group and the formation of the rhodol byproducta dye with the hydroxyl group instead of one CF 3 CH 2 NH residuewere observed. Under drastic condensation conditions, the undesired reaction was inevitable; it decreased the yields of the target compounds and complicated the isolation of pure dyes 6a,b. For isolation of compounds 6a,b, we applied chromatography on reversedphase (C 18 silica gel) because crystallization or chromatography on regular silica was not successful. The mixture of bromides 6a and 6b was stable by storing at −18°C but slowly decomposed at room temperature. A high degree of purity (>95% HPLC area) was required for the success of the next coupling step (Scheme 3). Only by applying highly pure bromides 6a,b, we were able to obtain acetylenes 7a−c in synthetically useful amounts.
The acetylene-bridged systems consisting of two fluorescent dyes linked directly through the triple bond belong to the family of through-bond energy transfer cassetten (TBET-C). 5−7 The reaction conditions in Scheme 3 (for details, see the Experimental Section) may be applied for the synthesis of other TBET-Cs.
The reported conditions of hydration reaction (Scheme 4) were first checked with diphenylacetylene (tolane (9) in Scheme 5) as a model. Transformation of tolane to deoxybenzoin 10 catalyzed by Nafion NR50, 8 Ga(F 3 CSO 3 ) 3 , 9 or CF 3 SO 3 H in CF 3 CH 2 OH 10 proceeded smoothly and with good yields. However, under all of these conditions, hydration of acetylenes 7a−c was sluggish. With HSO 3 F (magic acid), 11 Nafion NR50, Nafion 117, or p-toluenesulfonic acid, ketones did not form at all. Only by using great excess of water, trifluormethanesulfonic (TfOH, reagent), and propionic (solvent) acids at 140°C, we managed to detect the formation of regioisomeric ketones (Scheme 4). The combinatorial fashion of the reaction sequence 6a,b−7a−c−8a−d increased the number of regioisomers on each step. The hydration reaction proceeded through the corresponding vinyl esters formed from acetylenes and TfOH. Further optimization was required, to fully hydrolyze these esters to ketones 8a−d. The HPLC analysis was difficult, due to numerous peaks with similar retention times. However, we managed to isolate a mixture of 8a−d and then separate it to individual components 8a [5(CH 2 ),5 (CO)], 8b [6(CH 2 ),5(CO)], 8c [5(CH 2 ),6-(CO)], and 8d [6(CH 2 ),6(CO)] so that the overall yield was about 80%. For that, we used preparative HPLC on reversed phase with a gradient of acetonitrile in the basic aqueous buffer.
Bis(rhodamine)diazoketones 1a−d ( Figure 1) were prepared according to the modified and optimized procedure of M. Regitz using p-toluene sulfonyl azide and DBU as a base   12,13 Diazoketones 1a−d were sensitive to acids and decomposed under acidic conditions. They were isolated in milligram amounts and purified by means of preparative HPLC with acetonitrile and basic aqueous buffers (e.g., AcONH 4 at pH 8.6). The overall preparative yield of all compounds 1a [5(N 2 ),5 (CO)], 1b [6(N 2 ),5(CO)], 1c [5(N 2 ),6(CO)], and 1d [6(N 2 ),6(CO)] was about 40%. To avoid decomposition, the products were stored at −18°C in the dark. 14 Structure Elucidation of Diazoketones 1a−d. The regularities of 1 H NMR spectra reported for 5-and 6substituted (in the pendant phenyl ring) rhodamines 15 allowed us to assign structures to compounds 1a−d ( Figure 1). Additionally, we used gCOSY and gHMBCAD spectra showing 1 H− 1 H and multibond (optimized for three bonds) 1 H-13 C correlations, respectively. In the proton spectra, we observed six 1-proton multiplets corresponding to two 3substituted benzene rings: one with CO and one with CN 2 group. For isomer 1 (lowest retention time in HPLC), these signals were 8.09, 8.07, 7.92, 7.73, 7.56, and 7.20 ppm. In the gCOSY spectrum, we did not observe cross-peaks between 8.09 and 7.73 ppm, but all other cross-peaks required for two sets of three protons were present. We could conclude that the signal at 8.09 ppm belongs to the same set as the multiplets at 7.73 and 7.20 ppm, and the signals at 8.07, 7.92, and 7.56 ppm belong to another aromatic ring. In the gHMBCAD spectrum of this compound, we found that the 13 C resonance in CO of the diazoketone has cross-peaks with multiplets at 7.56 and 7.92 ppm. Therefore, the signals at 8.07, 7.92, and 7.56 ppm belong to the ring linked with CO in COCN 2 , and the group of signals with δ = 8.09, 7.73, and 7.20 ppmto the ring bound with CN 2 . In each set, the most high-field signal belongs to H-7(7′)the proton nearby the fluorophore. 15 This proton is shielded by the π-system of the fluorophore. The molecule is twisted, and H-7(7′) is out of the plane of the three fused sixmembered rings. Thus, in the ring with CO, H-7′ is found at 7.56 ppm (weak splitting, 6-CO isomer), and for the ring with CN 2 , the signal at 7.20 ppm belongs to H-7 (strong splitting, 5-CN 2 isomer). To confirm that there was no rearrangement (exchange of the oxo and diazogroups in the course of diazotransfer in Scheme 4), we isolated the precursor of compound 1c (isomer 1). This ketone is named 8c in Scheme 4 and Table 1. The structure of 1,2-diarylethanone-1 8c was established using the principles mentioned above, and 8c was shown to be the "true" precursor of 1c: [5(CH 2 ),6(CO)]-8c.
Photolysis of Azibenzil PhCOCN 2 Ph (11) and Bis-Rhodamines 1a−d Having Diazoketone Bridge. The main reactivity pattern of α-diazoketones and, in particular, azibenzil 11 (Scheme 5), which we used as a model compound, includes elimination of dinitrogen and formation of highly reactive carbenes. 16 The reactions can be induced thermally, photochemical, or catalytically (acids, heavy metal oxides, and salts). The synthetically useful and well-studied reaction path includes the formation of carbene, its rearrangement into ketene, and the reaction with a nucleophile (e.g., water, alcohol, or amine); the overall transformation known as Wolff rearrangement (Scheme 5). 17 The photochemically induced Wolff rearrangement discovered by Horner 14 is advantageous because the photolysis is the most "ketenerich" reaction path, while thermal or catalytic reactions lead mostly to the products of C−H insertion. 17,18 Azibenzil (11) 19−21 was prepared from tolane (9) as given in Scheme 5. The photolysis of azibenzil 22,23 was performed under irradiation with 365 nm light in acetonitrile−water mixtures (80/20; v/v) in the presence of HEPES (pH 6.5) or HCOONH 4 buffer (pH 7.3−7.4). The reaction mixtures were analyzed by means of HPLC with a UV−vis absorption (diode array) spectrometer and a mass spectroscopic detection (LC-MS). The expected product of the photolysis (in the absence of amines in the reaction solution)diphenylacetic acid (13) 24 was detected along with deoxybenzoin (10), benzil (14), and traces of diphenylmethane (15) (Scheme 5). These compounds were identified by comparison with commercial reference substances (retention times, UV, and mass spectra).
In some experiments, we also detected products with higher masses: an oxazole formed upon [2 + 3] cycloaddition from acetonitrile and ketene 12b, 25 as well as small amounts of 3,3,6,6-tetraphenyl-1,2,4,5-tetroxane, the peroxide related to the photocyclization product of diphenylacetic acid. 26 Photolysis of the solutions containing aqueous HEPES buffer provided complex mixtures with diphenylacetic acid (13) as one of the main products ( Figure S1). Irradiation in the presence of aqueous HCOONH 4 was found to be "cleaner" (Figure 2) and resulted in the formation of diphenylacetic acid amide (16; Scheme 5). Azibenzil 11 and amide 16 had the same retention times under conditions of HPLC separation. Unlike azibenzil (11) and benzil (14), amide 16 did not display the absorption maximum at about 320 nm. The composition of amide 16 was confirmed by HRMS data obtained for the reaction mixture (see Figure S2). The origin of amide 16 is obvious: it formed from ketene 12b and ammonia, as the strongest nucleophile present in the Table 1 equilibrium in aqueous ammonium formate (2 mM) at pH 7.3−7.4 (the initial concentration of azibenzil was 0.1 mM.) At physiological pH, ammonia may be considered as an analogue of biogenic amines, 27 which have basicity similar to ammonia. Having in mind the encouraging results obtained with model diazoketone 11, we performed the photolysis of diazoketones 1a−d (12 μM) in aqueous acetonitrile (acetonitrile/water = 80/20; v/v) in the presence of ammonium formate buffer (pH 7.3-7.4) (Scheme 6). Surprisingly, in this solvent, diketones Ar 1 COCOAr 2 were the main products formed upon full conversion of the starting diazoketones 1a−d. The LC-MS data ( Figure S3a) indicated that the molecular masses of the photolysis products were always 12 Da lower than the molecular masses of diazoketones 1a−d. A mass difference of −12 Da corresponds to the elimination of nitrogen (−28) and the addition of one oxygen atom (+16). For diazoketones 1a−d, the Wolff rearrangement is disfavored, probably because the migration ability of the bulky and heavy dye residue is reduced. The fluorescence signals (and their quantum yields) of diazoketones 1a−d and the mixture of products obtained from their photolysis are given in Figure 3. The emission efficiencies of compounds 1a−d vary in the range of 0.09− 0.24. Their emission is reduced, compared with the parent rhodamines, which are highly fluorescent, 4 but not completely quenched by the presence of the diazoketone bridge. The diazoketone residue turned out to be an inefficient quencher, at least for these rhodamine dyes. The comparison of the absorption spectra recorded before and after photolysis is given in Figure S4b. Compounds 1a−d have 3−4 times higher absorption at 365 nm (irradiation wavelength) than the parent fluorophoreN,N′-bis(2,2,2-trifluoroethyl)rhodamine. 4 The presence of the azibenzil chromophore (Figure 2, abs. max. 325 nm) is masked by the relatively strong absorption of the parent dye with a maximum at 290 nm ( Figure S4b). The photolysis of compounds 1a−d was accompanied by an increase in emission by 20−240% (Figures 3, S4a, and S5). On the other hand, the relative absorption intensity at 300−310 nm decreased, after the photolysis was complete. The absorption spectra of the products and the parent rhodamine dye are much more similar to each other than the absorption spectra of diazoketones 1a−c, which differ from each other considerably ( Figure S4b). As expected, isomers 1 and 3 (compounds 1b and 1c in Figure 1) gave the same diketone 5-ArCOCOAr-6. The products' retention times ( Figure S3a) and emission gains were very similar: 30 and 20%, respectively ( Figure 3). For all diazoketones, the photoactivation ratios (1.2−2.4) are moderate, if compared with dyes having two 2nitrobenzyloxycarbonyl residues attached to the nitrogen atoms in one fluorophore, 28 photoactivatable rhodamine spiroamides, 29 or rhodamines incorporating the spiro-diazoketone fragment. 30 This result may be explained if we assume that the quenching ability of diazoketone COCN 2 is higher than that of α-diketone COCO, but the former does not completely inhibit the emission, while the latter does not allow to unfold the full fluorescence signal pertinent to two fluorophores. In addition, the quenching ability of the COCN 2 residue toward "left" (Ar 1 ) and "right" (Ar 2 ) aryl groups in Ar 1 COCN 2 Ar 2 is expected to be different, and may also depend on the substitution pattern of the aromatic ring (i.e., 5′ or 6′). The Wolff rearrangement is unfavored because the migration ability of the very bulky dye residue is low.

■ CONCLUSIONS
We prepared and studied the photolysis of assemblies consisting of the two identical fluorophores directly bound with a short, compact, and photoconvertible diazoketone bridge (−COCN 2 −). Structurally, this approach to compounds in which two fluorophores can be activated with one photon is simpler than the design of sophisticated assemblies containing one photoconvertible unit (FRET acceptor) bound with two fluorescent dyes (FRET donors). 31 In the course of photolysis, we observed only a moderate fluorescence increase. However, this method may be easily extended to compounds with other, more efficient quenchers linking two fluorescent dyes and undergoing photoconversion into another, essentially nonquenching state.    Figure S3). Photochemistry. Irradiation experiments were performed in a home-build setup, 32 using a 365 nm LED as irradiation source (M365-L2, Thorlabs), a deuterium/xenon lamp (DH-2000-BAL, Ocean Optics) as an illumination source (for recording absorption spectra), and a diode array spectrometer (FLAME-S-UV−VIS−ES, Ocean Optics). The intensity of the irradiation light was calibrated with a chemical actinometer (Azobenzene in MeOH). The samples were kept at 20°C and continuously stirred with a Peltier-based temperature controller (Luma 40, Quantum Northwest, Inc.). The absorption of the samples was recorded at a right angle with respect to the irradiation source, at fixed irradiation intervals until complete conversion to the final product. At fixed intervals, a small sample was extracted to perform LC-MS experiments (Shimadzu LCMS-2020).