Azulene-Derived Fluorescent Probe for Bioimaging: Detection of Reactive Oxygen and Nitrogen Species by Two-Photon Microscopy

Two-photon fluorescence microscopy has become an indispensable technique for cellular imaging. Whereas most two-photon fluorescent probes rely on well-known fluorophores, here we report a new fluorophore for bioimaging, namely azulene. A chemodosimeter, comprising a boronate ester receptor motif conjugated to an appropriately substituted azulene, is shown to be an effective two-photon fluorescent probe for reactive oxygen species, showing good cell penetration, high selectivity for peroxynitrite, no cytotoxicity, and excellent photostability.


■ INTRODUCTION
Reactive oxygen species (ROS) and reactive nitrogen species (RNS) are important mediators in many physiological and pathological processes. 1 The one-electron reduction of O 2 in vivo leads to the formation of O 2 •− (superoxide), which in turn undergoes disproportionation catalyzed by superoxide dismutase to give O 2 and H 2 O 2 (hydrogen peroxide). Another fate of O 2 •− is to react with endogenously produced NO (nitric oxide) to form ONOO − (peroxynitrite) via a nonenzymatic process. Ordinarily, the flux of H 2 O 2 is tightly regulated; aberrant H 2 O 2 production or overexposure is implicated in the pathogenesis of many diseases such as cancer and neurodegenerative conditions. 2 Similarly, while ONOO − has roles in signal transduction, its strongly oxidizing and nitrating properties mean it can react in an uncontrolled manner with various biomolecules. 3 Elevated levels of ONOO − have been linked to cardiovascular, neurodegenerative, and inflammatory diseases as well as cancer. 4 In view of this, there is a significant need for tools and techniques to elucidate the roles of ONOO − , H 2 O 2 , and other ROS/RNS in biological systems.
Of the various techniques that have been employed for the study of ROS/RNS, the use of fluorescent probes has several advantages. They can enable realtime fluorescence imaging of ROS/RNS with high sensitivity, are nondestructive, and are usually operationally straightforward. Such fluorescent probes may be categorized as chemosensors (those which reversibly bind their analytes) and chemodosimeters (irreversible reaction-based probes). Probes of this second type often have high analyte selectivity and are applicable even to those ROS/RNS that have very short lifetimes. 5 We and others have reported many fluorescent probes for imaging ROS/RNS, 6 including H 2 O 2 7 and ONOO − . 8 In the field of fluorescence imaging, the use of two-photon excitation fluorescence microscopy (TPM) 9 brings particular advantages. Specifically, TPM allows higher spatial resolution, and the use of excitation wavelengths in the near-infrared (NIR) region allows much greater tissue penetration, since minimal absorption by hemoglobin and water occurs in the NIR window. 10 Furthermore, excitation by lower energy photons means photobleaching and damage to the sample are minimized, and the autofluorescence of intrinsic fluorophores is avoided. 11 The design and development of small-molecule based fluorescent probes that can be used in TPM has been an area of much recent research activity. 12 Among these, various two-photon fluorescent probes for ROS/RNS have been exploited, since one-photon excitation with short wavelength could generate artificial ROS/RNS. 13 For H 2 O 2 , two-photon fluorescent probes have been reported that utilize the reaction with boronic acids or their esters, with nitroxide radicals, or Baeyer−Villiger type rearrangements. 14 For ONOO − , twophoton fluorescent probes have been reported that also utilize the reaction with boronic acids or their esters, as well as ones which function by oxidation of electron-rich arenes, αketocarbonyls, selenides, or by nucleophilic addition to electron-poor alkenes and arenes. 15 These reported probes incorporate a variety of fluorophores, including well-known ones that have been extensively used in one-photon fluorescence, such as naphthalimides, rhodamines, coumarins, etc. In contrast, in the work presented here, we report a fluorescent probe for ROS/RNS that incorporates a previously unexploited class of fluorophore, namely an azulene.
Azulene is a nonalternant bicyclic aromatic hydrocarbon, isomeric with naphthalene, yet with appreciably different properties. 16 The S 0 − S 1 energy gap is smaller than in naphthalene, leading to an absorption in the visible region, and hence the blue color for which azulene is named. The nonradiative S 1 → S 0 transition is extremely fast, 17 whereas the nonradiative S 2 → S 1 transition is much slower. 18 Consequently, S 2 → S 0 fluorescence is the principal mode of emission, 19 in violation of Kasha's rule. 20 The predominance of emission from higher excited states has also been noted for many azulene derivatives. 21 These properties have led to azulenes being used extensively as both colorimetric 22 and fluorescent 23 probes for various analytes. Azulene derivatives have also been exploited as fluorescence sensitizers 24 and quenchers 25 as well as in photoswitches 26 and molecular logic gates. 27 In the context of bioimaging, 18 F-labeled azulene derivatives have been used for PET imaging, 28 but to our knowledge, azulenes have not previously been used for fluorescence imaging. In the specific context of ROS/RNS detection, no reports employing azulene-containing probes have appeared to date. The two-photon fluorescence of azulene 29 and some derivatives 30 has been studied, yet it has never been utilized in any sensing context (although it has been exploited for fluorescence lithography 31 and applications in optical data storage have been suggested 32 ).

■ RESULTS AND DISCUSSION
Design Principles. Our design strategy for an azulene chemodosimeter imaging probe is centered on the concept of an inversion of internal charge transfer (ICT) directionality upon reaction with the analyte. This exploits the inherent dipole moment of azulene of 1.08 D, unusually high for a hydrocarbon. Thus, the charge distribution in azulene is as shown in Scheme 1a, with the smaller ring having the higher electron density as a result of the contribution of the resonance structure shown, in which both rings (tropylium and cyclopentadienide) are discrete 6π aromatic systems. Our approach involves the use of a pinacol boronate ester as a receptor motif, whose reaction with ROS/RNS leads to the desired reversal of the direction of ICT. In many cases, boronate-based probes function by the carbon−boron bond undergoing oxidation by the ROS/RNS to give a carbon− oxygen bond, which in turn leads to the cleavage of a selfimmolative linker and a change in fluorescence response. In the present case, we reasoned that a boronate group appended directly to the azulene seven-membered ring could effect the desired perturbation of the fluorophore upon reaction with ROS/RNS (Scheme 1b). This approach exploits the fact that the boronic ester is electron-withdrawing (Scheme 1c), whereas the hydroxyl group is electron-donating by resonance (Scheme 1d). Our strategy anticipates that the electronic influence of the boronic ester will attenuate the inherent polarization of the azulene, leading to the Az-6-Bpin system having reduced ICT character overall. On the other hand, the effect of the hydroxyl group in the Az-6-OH system will reinforce the inherent polarization of the fluorophore, and hence lead to increased ICT character overall and enhanced fluorescence.
The simplest embodiment of the design strategy described above would be to synthesize and evaluate 6-azulenyl pinacolborane itself as a fluorescent probe. However, the reaction product, 6-hydroxyazulene, is reportedly unstable, which may be due in part to the fact that it exists in tautomeric equilibrium with its keto form. 33 We therefore refined our design strategy, introducing esters at the azulene 1-and 3positions, as this substitution pattern is known to stabilize 6hydroxyazulenic compounds. 34 Finally, we introduced an electron-donating group (amine) at the 2-position. The final probe structure 1 and its oxidation product 2 are depicted in Scheme 2a. The purpose of the amine substituent is to enhance the contribution of the resonance effect that conflicts with the inherent polarity of azulene, as shown in Scheme 2b, with an additional ammonium-boronate resonance contribution compared to Scheme 1c. Furthermore, the ester groups will not only stabilize the hydroxyazulene product 2 of reaction with ROS/RNS, they will also enhance the contribution of the resonance effect that reinforces the inherent polarity of azulene, as shown in Scheme 2c.
Probe molecule 1 and its oxidation product 2 were modeled using DFT as well as azulene itself. NBOs (natural bonding orbitals) were computed, and selected molecular orbitals (HOMO, LUMO, and LUMO+1) for each compound are shown in Figure 1.
The probe molecule 1 may be accessed in three steps from commercial materials (see ESI). 35 We have termed this substance "AzuFluor 483-Bpin" (to indicate the nature of the receptor motif and the emission maximum of its turn-on response). AzuFluor 1 was then reacted with H 2 O 2 (THF, 24 h, 95%) and the novel product isolated and fully characterized as the stable 6-hydroxyazulene 2. Crystals of 2 suitable for Xray diffraction were grown from THF/hexane and the solid state structure of 2 is shown in Figure 2.
Probe Characterization. The AzuFluor probe 1 itself exhibits near-negligible fluorescence, whereas the oxidation Scheme 1. Initial Concept for Azulene-Based Fluorescent Probe for ROS/RNS

Journal of the American Chemical Society
Article product 2 exhibits fluorescence emission at λ em = 483 nm, upon excitation at 350 nm, in a mixed aqueous buffer/ methanol system, as shown in Figure 3. Fluorescence excitation spectra were also acquired (Figures S1−S3), as well as UV−vis absorption spectra ( Figure S4). A value of λ max = 335 nm was observed in the absorption spectrum of 2; thus, the oxidation product displays an appreciable Stokes shift of 148 nm. Fluorescence emission of 2 was also examined in various solvents ( Figure S5), and we observed the solvatofluorochromism expected for an ICT fluorophore; the observed λ em value underwent a bathochromic shift from 480 to 491 nm with increasing solvent polarity.
A pH titration showed that the fluorescence intensity is greatest at pH 7−8 ( Figure 4; see also Figures S6−S8), which corresponds to the pH range at which a probe for cellular imaging would need to function in most cases. Fluorescence intensity decreases markedly at lower pH, and is also attenuated above pH 8. An NMR study of 2 at different pH values (Figures S9−S11) shows it to be stable upon prolonged standing in solution across a range of pH values, albeit with a downfield shift of the aryl resonances at low pH. Thus, we do not believe the decrease in fluorescence at extremes of pH is due to decomposition of 2. Rather, we ascribe these phenomena to changes in the ionization state of 2. Thus, at

Journal of the American Chemical Society
Article low pH the amino group in 2 would be protonated; in [2+H] + the ICT interaction between the amine and the esters would be abolished. Conversely, at high pH, deprotonation of the aryl hydroxyl group to give [2-H] − would also affect the fluorescence response. The computed pK a values for the amine and hydroxyl groups in 2 are 3.60 and 10.95, respectively (Figures S12−S13), which supports the above explanation.
It is well established in the literature 14,15 that boronate receptors respond to both ONOO − and H 2 O 2 , although the greater nucleophilicity of ONOO − leads to a reaction rate ∼10 6 greater than for H 2 O 2 . 36 We therefore evaluated 1 as a probe for both potential analytes, with the expectation of a faster and more sensitive response to peroxynitrite. Fluorescence titration experiments ( Figure S14) with 1 and H 2 O 2 established the limit of detection to be 1.72 μM. A time drive experiment ( Figure S15) showed the fluorescence response of 1 (500 nM) to H 2 O 2 to be slow, with fluorescence intensity continuing to increase after 30 min even at [H 2 O 2 ] = 500 μM. In contrast, 1 was much more sensitive to ONOO − as expected, with fluorescence titration experiments (Figure S16) establishing the limit of detection to be 21.7 nM. The selectivity of 1 for various ROS was then examined, and 1 was found to be highly selective for ONOO − and H 2 O 2 over HOCl, HO • , O 2 •− , ROO • , and 1 O 2 ( Figure S17). Furthermore, with equimolar concentrations of all ROS, 1 exhibited pronounced selectivity for ONOO − over H 2 O 2 at both 5 and 30 min time points (Figures 5, S18−S19). The UV−vis spectroscopic response of 1 to ONOO − and H 2 O 2 was also studied (Figures S20−S25; Tables S1−S2). Two-photon action spectra (Φδ, where Φ is the fluorescence quantum yield and δ is the two-photon absorption crosssection) of probe 1 in the presence and absence of H 2 O 2 or ONOO − were investigated ( Figure S26) to determine the maximum optical brightness. Upon addition of ROS, the spectroscopic maximum (Φδ max ) was red-shifted from 700 to 810 nm, along with an increase of the Φδ value (at 810 nm) from 1.2 to 3.2 GM. A similar result was observed with twophoton microscopy imaging ( Figure S27). The measured maximum two-photon optical brightness of Φδ max = 3.2 GM compares favorably with the reported values 37 for other commonly used fluorophores of similar or greater molecular weight, such as dansyl hydrazine, Lucifer yellow, Cascade blue, or Indo-1 (see SI, Table S3 for a tabulated comparison). The fluorescence quantum yield of 2 was determined to be 0.010 ( Figures S34−S36). Thus, at 810 nm the two-photon absorption cross-section of 2 (Φδ÷Φ) is 320 GM.
Probe Evaluation in Cell Studies. In vitro studies were performed to validate the ability of 1 in aqueous solution to detect endogenous and exogenous H 2 O 2 and ONOO − by TPM. MTT assays confirmed that 1 is not cytotoxic under the identified imaging conditions ( Figure S28). Two-photon excited fluorescence intensities were determined by staining HeLa cells with 1 at a concentration of 5 μM (by adding 5 μL of a 1 mM DMSO stock solution of 1 per 1 mL of total volume of medium) in the presence and then absence of ONOO − (100 μM), to find the optimal two-photon excitation wavelength ( Figure S27a). This was determined to be λ ex = 800 nm, at which wavelength the mean fluorescence intensity was nearly 4-fold higher after treatment with ONOO − ( Figure  S27b). Probe 1 showed high photostability, without any decay of fluorescence intensity even after irradiation for 1 h at 2 s intervals, both with and without pretreatment with ONOO − ( Figure S29). Co-localization experiments were performed with 1 and representative organelle markers (ER, mitochondria, and lysosomes). The Pearson correlation coefficients with the markers for the ER, mitochondria and lysosomes are 0.79, 0.70 and 0.41, respectively ( Figure S30). This indicates that 1 is distributed through the cytosol, including in these organelles, rather than any specific localization.
With ideal imaging conditions established, the ability of 1 to detect ROS in living cells was then demonstrated. Thus, the mean fluorescence intensity was measured for TPM images of 1 in RAW 264.7 macrophages pretreated with H 2 O 2 (2 mM) and ONOO − (100 μM). The fluorescence intensity of cells pretreated with ROS was shown to increase nearly 3-fold (Figure 6b-c), compared to negative control (Figure 6a, cells stained with probe 1 only). Next, RAW 264.7 macrophages were pretreated with endogenous ROS-inducing reagents to observe whether probe 1 could respond to the endogenous ROS of macrophages. Phorbol myristate acetate (PMA), 38 and lipopolysaccharide/gamma interferon (LPS/IFN-γ) 39 are known to stimulate macrophages to release ROS such as NO • , O 2 •− , H 2 O 2 , and ONOO − , whereas 3-morpholinosydnonimine (SIN-1) 40 is employed as a positive control (exogenous source of ONOO − ). Each cell was individually incubated with either PMA (1 μg mL −1 ), LPS (1 μg mL −1 ) + IFN-γ (50 ng mL −1 ), or SIN-1 (50 μM) and then stained with probe 1. As a result, 1 showed strong fluorescence intensity, increasing by approximately 2-fold, as compared with the control group (Figure 6d,e,g). ROS-induced fluorescence intensity did not rise above basal levels upon cotreatment with ebselen (50 μM), a commonly used scavenger of superoxide (the precursor of ONOO − and H 2 O 2 ; 41 Figure 6f,h). Overall, the obtained outcomes showed that 1 can directly detect the presence of H 2 O 2 and ONOO − in biological systems.
Finally, we investigated the ability of probe 1 to visualize ONOO − and H 2 O 2 in living tissues. A fresh slice of rat hippocampus was stained with 1 at a concentration of 50 μM (by adding 5 μL of a 10 mM DMSO stock solution of 1 per 1 mL total volume) for 1 h at 37°C. The TPM images obtained in the CA1 regions displayed only weak signals, as we observed in cultured cells (Figure 7a,e). However, bright fluorescence was observed upon pretreatment with PMA (10 μg mL −1 ) and SIN-1 (200 μM), increased by approximately 2 times

Journal of the American Chemical Society
Article compared to negative control (Figure 7b,c) whereas fluorescence did not increase above basal levels upon cotreatment with SIN-1 and ebselen (200 μM) (Figure 7d,f). This confirms that our probe can directly detect endogenous ROS in living tissues with TPM.

■ CONCLUSIONS
We have synthesized and evaluated the first two-photon fluorescent bioimaging probe containing an azulene fluorophore. The rationally designed probe, AzuFluor 483-Bpin (1), has been shown to possess high selectivity for peroxynitrite and hydrogen peroxide over other reactive oxygen species. Furthermore, 1 is cell-permeable, photostable, and noncytotoxic in vitro on the time scale of the two-photon fluorescence microscopy experiments. In view of the susceptibility of the azulene fluorophore to perturbation by attached substituents, the wide array of known receptor motifs that could be attached, and the availability of synthetic methodology to attach them, we anticipate that azulene fluorophores based on the design principles we have described here will find many further applications in two-photon fluorescence imaging.