Synthesis and Evaluation of New Potential Benzo[a]phenoxazinium Photosensitizers for Anticancer Photodynamic Therapy

The use of photodynamic therapy (PDT) and development of novel photosensitizers (PSs) for cancer treatment have received more and more attention nowadays. In the present work, five benzo[a]phenoxazinium derivatives have been prepared and evaluated for their in vitro anticancer photodynamic activity for the first time. They are red light absorbers and show low fluorescence quantum yield. Of these compounds, PS4 exhibited a higher quantum yield for reactive oxygen species (ROS) generation. The assays with cells in vitro showed that PS1 and PS4 were not significantly toxic in the dark, but was robustly toxic against the murine breast adenocarcinoma cells 4T1 and normal murine fibroblast cells NIH-3T3 upon photoactivation. More interestingly, PS5 was particularly selective towards 4T1 cancer cells and nearly non-phototoxic to non-cancerous NIH-3T3 cells. The results described in this report suggest that these new benzo[a]phenoxazinium derivatives are potential candidates as PSs for anticancer PDT. Further investigation of benzo[a]phenoxaziniums for anticancer PDT is warranted.


Introduction
Photodynamic therapy (PDT) is a minimally invasive protocol that has been used in anticancer therapy for a long time [1]. PDT is based on the focal photoactivation of photosensitizers (PSs), which can directly act on the target tissues and then elicit photochemical reactions that eventually lead to oxidative stress [2]. The main consequences of these events include direct cytotoxicity, collapse of the tumor microvasculature, and/or activation of immune response against tumor antigens [3]. A particularly important benefit of PDT as a cancer therapy is the possibility to restrict its effects to the irradiated site, sparing the normal tissues. Although PDT has been successfully applied in the treatment of skin, gynecological, gastrointestinal, and some head and neck cancers for a long time, only a few PSs (e.g., porfimer sodium, temoporfin, aminolevulinic acid and photofrin) have been put into the market [4,5]. Up to now, most current work focuses on the improvement of porphyrin and chlorin-type PS molecules and, especially, synthesis of new type of PSs with high water solubility, strong absorption in near-infrared region, and long half-life [2,6].

Design and Synthesis
The ligand-mediated targeting (LMT) strategy in PDT has been explored to increase the efficacy and reduce adverse effects of PSs [2]. In the present study, five benzo[a]phenoxazinium chlorides possessing different functional fragments at the 5-amino position were designed and prepared based on the LMT strategy. PS1 was a simple known benzo[a]phenoxazinium derivative with methyl propionate reported by Frade et al. [13,14,20], which was shown to have antimicrobial activity against Saccharomyces cerevisiae [14], but its anticancer PDT had not yet been investigated. Other compounds are new synthetic benzo[a]phenoxazinium derivatives: PS2 possesses a morpholinoethylamine moiety, which is a well-known ligand for targeting lysosome [21]; PS3 is equipped with a biotin moiety, which is a well-known tumor-targeting molecule [22] and has been used for selective delivery of PS to cancer tissues [23,24]; PS4 was a conjugate of benzo[a]phenoxazinium and pregnenolone. Pregnenolone, known as a precursor to most hormones, had been explored as carrier of anticancer drugs [25]. In addition, several types of pregnenolone derivatives were reported to have anticancer activity by our and other groups [26][27][28]. PS5 was synthesized as the first benzo[a]phenoxazinium dimer to determine the superimposed effect of the benzo[a]phenoxazinium core on anticancer PDT efficacy.

Design and Synthesis
The ligand-mediated targeting (LMT) strategy in PDT has been explored to increase the efficacy and reduce adverse effects of PSs [2]. In the present study, five benzo[a]phenoxazinium chlorides possessing different functional fragments at the 5-amino position were designed and prepared based on the LMT strategy. PS1 was a simple known benzo[a]phenoxazinium derivative with methyl propionate reported by Frade et al. [13,14,20], which was shown to have antimicrobial activity against Saccharomyces cerevisiae [14], but its anticancer PDT had not yet been investigated. Other compounds are new synthetic benzo[a]phenoxazinium derivatives: PS2 possesses a morpholinoethylamine moiety, which is a well-known ligand for targeting lysosome [21]; PS3 is equipped with a biotin moiety, which is a well-known tumor-targeting molecule [22] and has been used for selective delivery of PS to cancer tissues [23,24]; PS4 was a conjugate of benzo[a]phenoxazinium and pregnenolone. Pregnenolone, known as a precursor to most hormones, had been explored as carrier of anticancer drugs [25]. In addition, several types of pregnenolone derivatives were reported to have anticancer activity by our and other groups [26][27][28]. PS5 was synthesized as the first benzo[a]phenoxazinium dimer to determine the superimposed effect of the benzo[a]phenoxazinium core on anticancer PDT efficacy.
The synthetic routes for target compounds PS1 to PS5 are depicted in Schemes 1-5, respectively. Briefly, compound 2 was synthesized by alkylation of 1-naphthylamine (1) with 3-bromopropanoic acid, and nitroso derivative 4 was prepared from the nitration reaction of 3-(diethylamino)phenol (3). Then a cyclization reaction of 2 and 4 (in refluxing methanol) produced target compound PS1. The coupling reaction of 2 with 2-morpholinoethanamine, tert-butyl (2-aminoethyl)carbamate, and pregnenolone derivative 8 yielded intermediates 5, 6 and 9, respectively. Deprotection of N-Boc in 6 followed by coupling with Biotin-NHS gave compound 7 or followed by reaction with 2 produced 10. Starting from intermediates 5, 7, 9, 10 and using compound 4 once again, the target benzo[a]phenoxaziniums PS2 to PS5 were finally obtained in the last step using a similar synthetic protocol as that of PS1. The 1 H-, 13 C-NMR, LR-MS and HR-MS spectra for PS1-PS5 can be found in Supplementary materials. acid, and nitroso derivative 4 was prepared from the nitration reaction of 3-(diethylamino)phenol (3). Then a cyclization reaction of 2 and 4 (in refluxing methanol) produced target compound PS1. The coupling reaction of 2 with 2-morpholinoethanamine, tert-butyl (2-aminoethyl)carbamate, and pregnenolone derivative 8 yielded intermediates 5, 6 and 9, respectively. Deprotection of N-Boc in 6 followed by coupling with Biotin-NHS gave compound 7 or followed by reaction with 2 produced 10. Starting from intermediates 5, 7, 9, 10 and using compound 4 once again, the target benzo[a]phenoxaziniums PS2 to PS5 were finally obtained in the last step using a similar synthetic protocol as that of PS1. The 1 H-, 13  acid, and nitroso derivative 4 was prepared from the nitration reaction of 3-(diethylamino)phenol (3). Then a cyclization reaction of 2 and 4 (in refluxing methanol) produced target compound PS1. The coupling reaction of 2 with 2-morpholinoethanamine, tert-butyl (2-aminoethyl)carbamate, and pregnenolone derivative 8 yielded intermediates 5, 6 and 9, respectively. Deprotection of N-Boc in 6 followed by coupling with Biotin-NHS gave compound 7 or followed by reaction with 2 produced 10. Starting from intermediates 5, 7, 9, 10 and using compound 4 once again, the target benzo[a]phenoxaziniums PS2 to PS5 were finally obtained in the last step using a similar synthetic protocol as that of PS1. The 1 H-, 13  acid, and nitroso derivative 4 was prepared from the nitration reaction of 3-(diethylamino)phenol (3). Then a cyclization reaction of 2 and 4 (in refluxing methanol) produced target compound PS1. The coupling reaction of 2 with 2-morpholinoethanamine, tert-butyl (2-aminoethyl)carbamate, and pregnenolone derivative 8 yielded intermediates 5, 6 and 9, respectively. Deprotection of N-Boc in 6 followed by coupling with Biotin-NHS gave compound 7 or followed by reaction with 2 produced 10. Starting from intermediates 5, 7, 9, 10 and using compound 4 once again, the target benzo[a]phenoxaziniums PS2 to PS5 were finally obtained in the last step using a similar synthetic protocol as that of PS1. The 1 H-, 13

Absorption and Emission Studies
UV-vis absorption and emission spectra of 5 μM PS1 to PS5 in water were measured to investigate their optical properties (Table 1 and Figure 2A). The absorption maxima (λmax) for compounds PS1 to PS4 located at about 650 nm with the molar extinction coefficients (ε) between 22,600 and 51,800 M −1 cm −1 , which can be ascribed to the π-π* transition of the large π system of the benzo[a]phenoxazinium core. PS5 showed a shoulder peak at 649 nm (ε = 35,600 M −1 cm −1 ) from the benzo[a]phenoxazinium fluorophore. Meanwhile, PS5 also showed a main absorption at 604 nm with a higher ε of 65,000 M −1 cm −1 , which probably resulted from the π-π interaction of the two intramolecular benzo[a]phenoxazinium fluorophores as a result of the flexibility of the amide linker [29].
Under excitation at 600 nm, all these compounds exhibited near-infrared emissions at about 681 nm, with Stokes shifts of about 35 nm ( Figure 2B). In addition, the relative fluorescence quantum yields (φ) were measured in water using fluorescein as a standard (φs = 0.98) [30]. These compounds all showed relatively low fluorescence quantum yields (0.025 to 0.116), which is consistent with the result of previously reported benzo[a]phenoxazinium chlorides [13]. This result indicates that PS1-PS5 may not tend to decay back to the ground state by emitting fluorescence after excitation and have a great potential in undergoing intersystem crossing to form a relatively long-lived triplet state, which is necessary for acting as desirable PDT candidates.

Absorption and Emission Studies
UV-vis absorption and emission spectra of 5 μM PS1 to PS5 in water were measured to investigate their optical properties (Table 1 and Figure 2A). The absorption maxima (λmax) for compounds PS1 to PS4 located at about 650 nm with the molar extinction coefficients (ε) between 22,600 and 51,800 M −1 cm −1 , which can be ascribed to the π-π* transition of the large π system of the benzo[a]phenoxazinium core. PS5 showed a shoulder peak at 649 nm (ε = 35,600 M −1 cm −1 ) from the benzo[a]phenoxazinium fluorophore. Meanwhile, PS5 also showed a main absorption at 604 nm with a higher ε of 65,000 M −1 cm −1 , which probably resulted from the π-π interaction of the two intramolecular benzo[a]phenoxazinium fluorophores as a result of the flexibility of the amide linker [29].
Under excitation at 600 nm, all these compounds exhibited near-infrared emissions at about 681 nm, with Stokes shifts of about 35 nm ( Figure 2B). In addition, the relative fluorescence quantum yields (φ) were measured in water using fluorescein as a standard (φs = 0.98) [30]. These compounds all showed relatively low fluorescence quantum yields (0.025 to 0.116), which is consistent with the result of previously reported benzo[a]phenoxazinium chlorides [13]. This result indicates that PS1-PS5 may not tend to decay back to the ground state by emitting fluorescence after excitation and have a great potential in undergoing intersystem crossing to form a relatively long-lived triplet state, which is necessary for acting as desirable PDT candidates.

Absorption and Emission Studies
UV-vis absorption and emission spectra of 5 µM PS1 to PS5 in water were measured to investigate their optical properties (Table 1 and Figure 2A). The absorption maxima (λ max ) for compounds PS1 to PS4 located at about 650 nm with the molar extinction coefficients (ε) between 22,600 and 51,800 M −1 cm −1 , which can be ascribed to the π-π* transition of the large π system of the benzo[a]phenoxazinium core. PS5 showed a shoulder peak at 649 nm (ε = 35,600 M −1 cm −1 ) from the benzo[a]phenoxazinium fluorophore. Meanwhile, PS5 also showed a main absorption at 604 nm with a higher ε of 65,000 M −1 cm −1 , which probably resulted from the π-π interaction of the two intramolecular benzo[a]phenoxazinium fluorophores as a result of the flexibility of the amide linker [29].

ROS Production
The reactive oxygen species (ROS) generated in photoreactions are the key factor for PDT as they can induce cytotoxicity via damage to different biomolecules, including proteins, nucleic acids and lipids [31,32]. Therefore, compounds PS1 to PS5 were evaluated for their effects on the production of ROS by DPBF method [7]. Figure 3 shows the results of ROS production when the compounds were irradiated with different energy densities. All these molecules produced ROS in an energy-dependent fashion, indicating that they are potential candidates for anticancer PDT. Remarkably, the photoactivated production of ROS was most intense with PS4, suggesting that this compound has the highest quantum yield for ROS generation. PS3, on the contrary, showed the lowest production of ROS compared with others.

Photodynamic Activity Against Cells In Vitro
Finally, in vitro photodynamic activities of PS1 to PS5 against two cell lines, including murine breast adenocarcinoma cell 4T1 and normal murine fibroblast cell NIH-3T3, were tested by irradiated Under excitation at 600 nm, all these compounds exhibited near-infrared emissions at about 681 nm, with Stokes shifts of about 35 nm ( Figure 2B). In addition, the relative fluorescence quantum yields (ϕ) were measured in water using fluorescein as a standard (ϕ s = 0.98) [30]. These compounds all showed relatively low fluorescence quantum yields (0.025 to 0.116), which is consistent with the result of previously reported benzo[a]phenoxazinium chlorides [13]. This result indicates that PS1-PS5 may not tend to decay back to the ground state by emitting fluorescence after excitation and have a great potential in undergoing intersystem crossing to form a relatively long-lived triplet state, which is necessary for acting as desirable PDT candidates.

ROS Production
The reactive oxygen species (ROS) generated in photoreactions are the key factor for PDT as they can induce cytotoxicity via damage to different biomolecules, including proteins, nucleic acids and lipids [31,32]. Therefore, compounds PS1 to PS5 were evaluated for their effects on the production of ROS by DPBF method [7]. Figure 3 shows the results of ROS production when the compounds were irradiated with different energy densities. All these molecules produced ROS in an energy-dependent fashion, indicating that they are potential candidates for anticancer PDT. Remarkably, the photoactivated production of ROS was most intense with PS4, suggesting that this compound has the highest quantum yield for ROS generation. PS3, on the contrary, showed the lowest production of ROS compared with others.

ROS Production
The reactive oxygen species (ROS) generated in photoreactions are the key factor for PDT as they can induce cytotoxicity via damage to different biomolecules, including proteins, nucleic acids and lipids [31,32]. Therefore, compounds PS1 to PS5 were evaluated for their effects on the production of ROS by DPBF method [7]. Figure 3 shows the results of ROS production when the compounds were irradiated with different energy densities. All these molecules produced ROS in an energy-dependent fashion, indicating that they are potential candidates for anticancer PDT. Remarkably, the photoactivated production of ROS was most intense with PS4, suggesting that this compound has the highest quantum yield for ROS generation. PS3, on the contrary, showed the lowest production of ROS compared with others.

Photodynamic Activity Against Cells In Vitro
Finally, in vitro photodynamic activities of PS1 to PS5 against two cell lines, including murine breast adenocarcinoma cell 4T1 and normal murine fibroblast cell NIH-3T3, were tested by irradiated

Photodynamic Activity Against Cells In Vitro
Finally, in vitro photodynamic activities of PS1 to PS5 against two cell lines, including murine breast adenocarcinoma cell 4T1 and normal murine fibroblast cell NIH-3T3, were tested by irradiated (PDT) with 25.8 J/cm 2 of light (λ 660 nm) or not (dark) by MTT bioassay [33]. As shown in Figure 4, all the tested compounds did not display significant toxicity towards both NIH-3T3 and 4T1 cells in the dark with inhibition ratio less than 50% within the concentration range of 2.5 to 40 µM. The absence of toxicity in the dark is a requirement for a desirable PS, as it avoids that non-irradiated tissues become affected during a PDT protocol. Regarding the photodynamic activity, it is remarkable that both PS1 and PS4 were robustly phototoxic against 4T1 cancer cell in a concentration-dependent manner (p < 0.05). Compound PS3 presented no evident photodynamic activity against 4T1 cells. Interesting results were also obtained with PS5, which showed a significant phototoxicity against 4T1 cells but no activity against NIH-3T3 in PDT experiment, indicating that this compound might have a better selectivity towards the cancerous 4T1 cells tested in this study. However, further investigation on the selectivity of PS5 towards cancerous cells and the mechanism of action is warranted. (PDT) with 25.8 J/cm 2 of light (λ 660 nm) or not (dark) by MTT bioassay [33]. As shown in Figure 4, all the tested compounds did not display significant toxicity towards both NIH-3T3 and 4T1 cells in the dark with inhibition ratio less than 50% within the concentration range of 2.5 to 40 μM. The absence of toxicity in the dark is a requirement for a desirable PS, as it avoids that non-irradiated tissues become affected during a PDT protocol. Regarding the photodynamic activity, it is remarkable that both PS1 and PS4 were robustly phototoxic against 4T1 cancer cell in a concentration-dependent manner (p < 0.05). Compound PS3 presented no evident photodynamic activity against 4T1 cells.
Interesting results were also obtained with PS5, which showed a significant phototoxicity against 4T1 cells but no activity against NIH-3T3 in PDT experiment, indicating that this compound might have a better selectivity towards the cancerous 4T1 cells tested in this study. However, further investigation on the selectivity of PS5 towards cancerous cells and the mechanism of action is warranted.

General Spectroscopic Measurements
Absorption spectra were recorded in a Shimadzu UV-2600 Spectrophotometer (Shimadzu, Kyoto, Japan). Fluorescence measurements were performed using an Agilent Cary Eclipse (Varian, Palo Alto, California, USA). The concentrations of all the compounds were 5 µM. Fluorescence spectra were corrected for the instrumental response of the system. All solutions were prepared using Milli-Q grade water. The fluorescence quantum yields (ϕ) were determined according to the method Equation (1) below: where ϕ is fluorescence quantum yield; F is integrated area under the corrected emission spectra; η is the refractive index of the solution; A is the absorbance at the excitation wavelength; the subscripts u and s refer to the unknown and the standard, respectively. We chose fluorescein in water as reference, ϕ s = 0.98.

ROS Detection
To evaluate the capacity of the compounds to produce singlet oxygen, the DPBF method was used. Briefly, 200 µL-aliquots of compounds in DMSO (40 µM) were plated in 96-well plates. Then, 10 µL of a solution of DPBF in ethanol (0.22 mg/mL) were added to each aliquot. Controls consisted of each compound without DPBF, and DPBF alone. Then, the microplate was irradiated every 10 s using 660 nm LED (Light Emitting Diode, XL001WP01NRC660, Shenzhen S. O. Co, Shenzhen, China). The optical density of the DPBF solution at 414 nm was used as an index of ROS production, as the DPBF is degraded by ROS and its light absorption at this wavelength is thus decreased.

Phototoxicity Assay
The toxicity of different treatments against 4T1 and NIH-3T3 cells was measured by an MTT assay. Briefly, 4T1 and NIH-3T3, 1 × 10 4 cells per well, were treated with different concentrations of the compounds for 30 min, in RPMI and DMEM, respectively, and then washed twice with PBS. After, the microplates were: (1) maintained in the dark; or (2) irradiated with a light emitting diode (LED, λ 660 nm) at a final energy density of 25.8 J/cm 2 . The control consisted of cells that received only culture medium. Next, the cells were washed with PBS, cultured for further 24 h, and then the culture medium was replaced by a 0.5 mg/mL MTT solution in culture medium. The cells were then incubated for 2.5 h at 37 • C in a 5% CO 2 , humid atmosphere. The MTT solution was then discarded, the formazan produced by the viable cells was extracted with 200 µL DMSO, and the optical density was read at λ 595 nm with a microplate spectrophotometer. This experiment was performed in triplicate for each treatment, and the results were expressed as percentages relative to control.

Conclusions
In conclusion, four benzo[a]phenoxazinium derivatives PS1-PS4 bearing different functional groups in the amino side chain and the first benzo[a]phenoxazinium dimer PS5 were prepared. The investigation on optical properties of PS1 to PS5 in water indicated that they are red light absorbers with low fluorescence quantum yields (0.025~0.116). The ROS production study revealed that all these benzo[a]phenoxaziniums produced ROS in an energy-dependent fashion, with PS4 having the highest ROS quantum yield. Finally, the anticancer PDT activities of this series of benzo[a]phenoxaziniums were evaluated for the first time. The bioassay results indicated that PS1 and PS4 show significant photodynamic activities against 4T1 cancer cells and NIH-3T3 normal murine fibroblast cells, and PS5 showed intriguing anticancer PDT activity selectively towards 4T1 cancer cells over NIH-3T3 normal cells. Together with the optical properties and photodynamic bioassay results, this series of benzo[a]phenoxazinium derivatives can be highlighted as new PSs worthy of further investigation in anticancer PDT study.