Effects of N-1 substituent on the phototoxicity of fluoroquinolone antibiotics: comparison of pefloxacin and difloxacin

Jian-Feng Zhao· Yan-Cheng Liu· Yu-Lie Xu · Wen-Feng Wang

Abstract The photophysics and photochemistry of pefloxacin (PEF), a 1-ethyl-substituted fluoroquinolone(FQ) antibiotic, were studied using transient, steady-state experimental methods and computational methods. The fundamental photoproperties of PEF and its phototoxicity toward lysozyme, a single-chain protein, were compared with those of a 1-fluorophenyl-substituted FQ antibiotic,difloxacin(DIF).The results showed that the phototoxicity was significantly decreased by the insertion of the bulky 1-fluorophenyl substituent (the phototoxicity of DIF was approximately one-quarter of that observed for PEF). This trend was attributed to the lowest lying singlet state with sizeable oscillator strength (f ≥0.1) being shifted from 319 nm in PEF to 266 nm in DIF upon the insertion of the bulky substituent at the 1-position,as investigated by using computational methods. In addition, 95% of the solar UV irradiation that reaches the earth’s surface has wavelength ＞315 nm. Therefore, reducing the most effective excitation wavelength by optimizing the substituent at the 1-position may be a promising strategy to alleviate the phototoxicity of FQ antibiotics. These findings may be applied to other FQ antibiotics because a large number of phototoxicity studies on FQ antibiotics with different substituents at the 1-position can prove these finding’s effectiveness. Delafloxacin, an FQ antibiotic bearing a chlorine and bulky substituent at the 8- and 1-positions,respectively, exhibits no phototoxicity is the most recent example reported to date. To the best of our knowledge,this is the first transient and steady-state study of the effect of the N-1 substituent on the photochemistry and phototoxicity of FQ antibiotics.These findings will be beneficial to the development of novel FQ antibiotics without phototoxicity.

Keywords Phototoxicity · FQ antibiotics · Laser flash photolysis · Difloxacin · Pefloxacin · Pulse radiolysis

1 Introduction

The fluoroquinolone (FQ) family of antibiotics comprises hundreds of aromatic compounds that share a similar quinolone structure, and these have been extensively used since they were first applied as clinical therapy [1, 2]. FQ antibiotics have some remarkable advantages, such as broad antimicrobial spectrum, strong antibacterial activity,low rate of drug resistance, and high-cost efficiency [3].However, they also exhibit some adverse effects [4] that should be eliminated or suppressed.

The adverse effects of FQ antibiotics include dysglycemia, hyperglycemia, tendinitis, central nerve system side effects[5],gene toxicity[6],and phototoxicity[7–11].Phototoxicity has been reported in most members of the FQ family of antibiotics, and some listed FQ antibiotics have even been discarded due to their severe phototoxic effects[12], e.g., induction of skin tumors [13]. Therefore,tremendous efforts have been undertaken by the scientific community to unveil the phototoxic mechanism of FQ antibiotics and alleviate their phototoxicity [14].

In general, FQ antibiotics show two types of phototoxicity: (1) direct photosensitization triggered by the excited state of the FQ antibiotic and its subsequent reactivity toward biomolecules and (2) reactive oxygen species(produced via the interaction of ground state oxygen with the excited state FQ antibiotic)-mediated oxidation of biomolecules.The excited state of an FQ antibiotic plays a central role in the two types of phototoxicity displayed by this family of compounds,that is,the characteristics of the FQ antibiotic itself are the dominant factors controlling its phototoxicity. Recent studies have shown that multiple halogen substituents, especially 8-fluorinated derivatives[15], are the main contributors to the severe phototoxicity displayed by FQ antibiotics [12] because the halogen and the aryl radicals generated during the dehalogenation reaction are highly toxic toward biomolecules [16–18].On the other hand, 8-methoxy-substituted FQ antibiotics, such as gatifloxacin [19] and moxifloxacin [20], are unlikely to exhibit phototoxicity. Although the 8-halogenated substituent has been recognized to be a decisive factor for the phototoxicity exhibited by FQ antibiotics, the structure–phototoxicity relationship is largely unknown due to the complex interactions of the substituents at each position.For example, 8-halogenated FQ antibiotics have been reported to exhibit severe phototoxicity [21], but delafloxacin bearing 8-Cl and 1-(6-amino-3,5-difluoropyridine)groups has been found to be non-phototoxic[22].Recently,substituents at the 5-position have also been reported to affect the phototoxicity in FQ antibiotics, such as antofloxacin and levofloxacin [23], and the mechanism was unveiled as the combined action of the position and electric effects in our previous research study [24]. 1-ethyl or 1-cyclopropyl FQ antibiotics [25] exhibit severe phototoxicity, whether they contain 8-halogen substituent or not[26]; this important phenomenon has no clear mechanistic explanation and is the origin of the present work. We predict that 1-ethyl and 1-cyclopropyl substituents are capable of attacking the adjacent 8-substituent, especially when the 8-substituent is a halogen, which will facilitate the dehalogenation reaction and the release of halogen and aryl radicals that have been proved to be highly toxic toward biomolecules. This is also the reason we chose a bulky substituent at the 1-position for comparison with 1-ethyl-substituted FQ antibiotics (DIF and PEF). A bulky substituent at the 1-position will hinder the cyclization reaction that connects the 1- and 8- positions as well as suppress the generation of toxic halogen radicals. These controversial results prove the importance of substituent effects and demonstrate the huge challenge toward understanding the structure–phototoxicity relationships of FQ antibiotics.

In this work, pefloxacin (PEF) and difloxacin (DIF),1-ethyl- and 1-fluorophenyl-substituted FQ antibiotics,were investigated in order to better elucidate the contribution from the substituent effects at the 1-position on the phototoxicity of FQ antibiotics and help clarify the overall scenario of the structure–phototoxicity relationships. The photochemistry of DIF has been previously reported by our group [27]. Therefore, we initially focused on the photochemistry and phototoxicity of PEF in this study,compared its phototoxicity with DIF, and explored the mechanism behind it. The generation of the triplet excimer of PEF(3PEF*) under laser excitation has also been reported and has been shown to exhibit genotoxic effects[28]due to the formation of thymine cyclobutane dimers via an energy transfer mechanism[29].However,the interactions of PEF with proteins and its reaction mechanism with DNA/protein under light irradiation as well as a comparison with DIF and the substituent effects at the N-1 position have not been studied to date. The chemical structures and UV–Vis absorption spectra of PEF and DIF are shown in Fig. 1.

2 Experimental

2.1 Materials

PEF (purity ＞99.0%) was purchased from J&K Chemical Ltd. DIF (purity ＞99.8%), naproxen (NP), 2′-deoxyguanosine-5′-monophosphate (dGMP), and tryptophan (Trp) were purchased from Sigma Chemical Co. All of these compounds met the required purity and were used as received. Sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) gels were purchased from Life Technologies. tert-Butanol and phosphate salts (analytical grade) were obtained from commercial suppliers and used without further purification. Water was purified using a Millipore-Q system.

2.2 Steady-state absorption measurements

Steady-state UV–Visible absorption experiments were conducted on a Hitachi spectrophotometer (U-3900 type,Japan).

2.3 Determination of the pKa values

An aqueous solution of PEF at a concentration of 1 × 10–5mol L–1was prepared. HClO4and NaOH were utilized to control the pH of the solution, and a glass electrode was employed to monitor the pH.The pKavalues were determined by observing and simulating the halfheight of the titration curves.

2.4 Laser flash photolysis (ns-LFP)

An Nd:YAG laser with a wavelength of 355 nm and pulse duration of 5 ns was used as the pump source to conduct our ns-LFP experiments. The energy employed in this work is ＜7 mJ per pulse. A detailed description has been provided in our previous report [30]. Unless stated otherwise, the ns-LFP experiments use carefully sealed neutral aqueous solutions saturated with ultrapure N2(＞99.999%).

2.5 Pulse radiolysis (ns-PRL)

Ns-PRL experiments were implemented using a linear accelerator [31–34] with 10 MeV energy to generate and transmit an electron pulse, and the duration of each pulse was 8 ns. A detailed description of the ns-PRL setup has been published elsewhere [35].

On the basis of equations (a)–(c) [36], the water molecules are split into several reactive intermediates under the attack of the electron beam,among which,·OH and eaq–are the crucial species due to their dominant yield and high reactivity. tert-Butanol was used to scavenge ·OH and eaq–were retained to create a reducing environment. Conversely, to create an oxidizing environment, N2O was introduced to saturate sample solution to convert eaq–into·OH creating a ·OH radical-dominated oxidizing environment. The influence of the remaining intermediates was negligible under the designed experimental parameters.

2.6 Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE)

A 500-W Xenon lamp was used as the light source. A cutoff filter (centered at 355 nm with a range of 315–375 nm) was used to block light at the undesired wavelengths, which focused the light on a quartz cell with dimensions of 10 × 10 × 40 mm3, yielding an irradiation power of 34.3 mW cm–2. SDS-PAGE (Bio-Rad Mini-PROTEIN®3 cell) was performed to analyze the freshly irradiated solutions. The gels were stained using the Coomassie method (Coomassie Brilliant Blue G-250 solution)and then destained using a mixed solution of ethanol and ethanoic acid. A Quantity One scanner (Bio-Rad) was employed to evaluate the extent of photodamage [37].

2.7 DFT calculations

Time-dependent density functional theory calculations were conducted at the B3LYP/6-311G* level using the Gaussian 09 program [38].

3 Results and discussion

3.1 Absorption properties

PEF exhibits an intense absorption in an aqueous solution, which consists of two bands (Fig. 2). The molar absorption coefficient (εmax) of the band located at 260–280 nm was determined to be ca.2.8–4 × 104M–1cm–1, while it was 1.1–1.6 × 104M–1cm–1for the weaker band observed at 310–340 nm. According to the plot of the λmax(310–340 nm) vs pH (Fig. 2, inset), we can see that the data can be modeled using two sigmoidal type curves and the pKavalues were determined to be 5.10 ± 0.02 and 9.20 ± 0.02, respectively (Fig. 3). The structure of PEF contains an alkaline nitrogen atom (the para-N atom in the 7-N’-methyl-piperazine ring) and a carboxyl group at the 3-position. This structure enables the zwitterion to be the dominant form in a neutral aqueous solution of PEF. A similar phenomenon was also observed with DIF, and its pKavalues were determined to be 5.91 and 9.89 [27].

According to previous reports[39],the short wavelength band observed at 260–280 nm is caused by the absorption of the aromatic ring, while the long wavelength band at 310–340 nm can be attributed to the n →π* (HOMO–LUMO) electronic transition caused by the intermolecular H bond formed between the FQ antibiotic and the solvent(H2O). The blueshift and decrease in the intensity of the short wavelength band were caused by the dissociation of the 3-carboxylic group when the solution changes from acidic to neutral conditions. We proposed this process based on the results and analysis described above,as shown in Fig. 3.

3.2 Laser flash photolysis (ns-LFP)

Figure 4 shows the time-resolved absorption spectra obtained for a neutral aqueous solution of PEF. The absorption band centered at 610 nm almost disappears within 5 μs; this decay reaction has a rate constant of 6.1 × 107M–1·s–1. The half-life (τ1/2) of the excited state of PEF in a N2-purged solution was estimated to be 1.5 μs.However, the decay reaction was significantly accelerated under an oxygen atmosphere with a rate constant of 6.2 × 108M–1s–1,which is one magnitude faster than that observed under a N2atmosphere. These results show that the transient species was sensitive to oxygen quenching(Fig. 4, left inset), which is a classic characteristic of a triplet state [40].

The energy transfer quenching pathway proposed in Eq. (1) was applied to verify that the aforementioned excited state of PEF is a triplet state(3PEF*)and was used to estimate the energy of3PEF*. Since the λmaxof3NP*is centered at ～430 nm[41]and its triplet energy is 259 kJ mol–1based on Eq. (2) (Sandros equation), the reaction rate constant (kET) is only dependent on the triplet energy gap(ΔET)between3PEF*and3NP*.The kmaxin Eq. (2)is the optimized rate constant for FQ antibiotics and NP,which was assumed to be 2.2 × 109M–1s–1. It is the average value of this diffusion controlled energy transfer reaction between the triplet excited state of some FQ antibiotics and the ground state of NP [29]. Therefore, the triplet energy of3PEF*can be estimated by tracking its decay–time dependence during the energy transfer process.

Figure 5 shows that at the end of the laser pulse, a new band was detected at ～610 nm, which then decayed to form a new band at 430 nm.As mentioned beforehand,the band at 430 nm can be assigned to3NP*and thus confirms the excited state of PEF was3PEF*; the kETvalue was estimated to be 1.66 × 109M–1·s–1(Fig. 5, right inset).Hence, the triplet energy gap (ΔET) between3PEF*and3NP*, and the energy of3PEF*was calculated as 6 and 265 kJ mol–1, respectively, according to Eq. (2), which is in good agreement with those reported in the literature(269 kJ mol–1) [29].

The reactions of3PEF*with Trp and dGMP were investigated in order to study the reactivity of3PEF*and its capability toward oxidizing biomolecules, including nucleic acid and amino acids. Trp is one of the most feasible sites for protein oxidation [42] and dGMP is a wellestablished model compound of DNA [43]. Hence, their oxidation can be treated as an indicator of the photosensitive damage to biomolecules. The triplet energy of dGMP is 317 kJ mol–1[44], which is much higher than that calculated for3PEF*(265 kJ mol–1). Therefore, it is unlikely to transfer energy from3PEF*to dGMP. The transient absorption spectra obtained for a mixed solution of PEF and dGMP are shown in Fig. 6. On this occasion,PEF was excited to generate its corresponding triplet state(3PEF*) and the direct excitation of dGMP can be neglected due to its extremely low absorption at this wavelength[45].This was further confirmed using a blank experiment, i.e., no absorption band was observed in the laser flash photolysis study using a neutral aqueous solution of dGMP under the same experimental conditions. The absorption band was instantly formed upon optical excitation, which decayed faster upon increasing the concentration of dGMP (Fig. 6, left inset). A band centered at 370 nm clearly emerged upon the decay of3PEF*. The absorption band of the dGMP cation radical (dGMP+·) is located at 310 nm [46]; however, in the present work, the absorption band corresponding to dGMP+·was not observed due to the strong ground state bleaching of PEF in the range of 310–350 nm. Nevertheless, the transient absorption of the PEF radical anion(PEF-·)was confirmed at 370 nm in our pulse radiolysis study, which will be discussed in the following section. Furthermore, the electron transfer observed from dGMP to ciprofloxacin, an analogue of PEF,has been unambiguously proved[10]and energy transfer pathway was excluded. Therefore, it is reasonable to conclude that the accelerated decay of3PEF*in the presence of dGMP can be attributed to the electron transfer reaction between dGMP and3PEF*. We also determined the rate constant of the reaction to be 2.8 × 107M–1s–1(Fig. 6, right inset).

We found that3PEF*can be quenched by Trp in a similar way. The quenching process of3PEF*is significantly accelerated with a quenching rate constant of 6.3 × 107M–1s–1upon the addition of Trp (Fig. 7, right inset). The absorption band observed at ～660 nm was assigned as hydrated electrons, which can be used as evidence for the occurrence of the electron transfer reaction between3PEF*and dGMP. According to the results and analysis described beforehand, the electron transfer reactions between3PEF*and dGMP (or Trp) can be described using Eqs. (3) and (4).

3.3 Pulse radiolysis (ns-PRL)

To confirm the species generated from PEF and dGMP at 370 nm was PEF·–, an ns-PRL experiment using PEF with hydrated electrons (eaq-) was carried out, as shown in Fig. 6. In consideration that·OH is the dominant reactive species formed during the degradation of FQ antibiotics[47] and the potential photoionization of3PEF*may also produce the radical cation of PEF (PEF·+) via photoionization, which can be compared with the electron transfer reaction between PEF and·OH, the reaction of PEF with·OH was also investigated.

3.3.1 Reaction with hydrated electrons (eaq-)

Ns-PRL experiments were performed using an ultrapure nitrogen-saturated mixed neutral solution containing phosphate(2 mM),t-BuOH(2 mM),and PEF(0.2 mM)in order to investigate the anion radical generated from PEF(PEF·–). In this system, both·OH and eaq-radicals can be generated due to the attack of water molecules by the high energy electron beam, in which·OH was then eliminated upon reaction with t-BuOH,as shown in Eq.(b).Therefore,under these experimental conditions, the obtained signals can be assigned to the reduction in PEF by eaq–. The transient spectra of eaq–have a characteristic broad band observed from 450 to 730 nm.However,a growth followed by decay profile was observed at 370 nm for PEF,which is obviously different from the simple decay profile observed for the blank experiment (Fig. 8, Inset, left). Thus, the maximum transient absorption of PEF·–can be assigned as the band observed at 370 nm. The reaction constant(1.5 × 1010M–1·s–1) was estimated for the reaction between PEF and eaq–by tracking the decay of band at 690 nm against the concentration of PEF. On account of these results, the process of this reaction was proposed using Eq. (5).

3.3.2 Reaction with·OH

If the system is a N2O-purged mixed neutral solution of PEF (0.1 mM) and phosphate buffer (PB, 2 mM), then after ns-PRL the·OH will be the primary reactive radical according to Eq. (c). The time-resolved spectra obtained 2.0 μs after the electron pulse are displayed in Fig. 9, and the rate constant between ·OH and PEF was deduced from the curve generated from the absorption band centered at 410 nm, which was estimated to be 6.5 × 109M–1s–1.This value is reasonable because most aromatic compounds react with·OH with rate coefficients in the range of 6–8 × 109M–1s–1[48]. It is known that ·OH reacts with aromatic molecules via three different pathways: addition,hydrogen abstraction, and electron transfer [49]; therefore,we attempted to identify the dominant reaction pathway between ·OH and PEF. The computational absorption spectra for the addition ([PEF-OH]·), H-abstraction (PEF(-H)·, and electron transfer (PEF+·) products were calculated and compared with the experimental absorption spectra. Unfortunately, none of them was in accordance with experimental results. On the contrary, a mixture of these three products was better suited to the experimental spectra (data not shown). Using the time-resolved spectra and computational calculations,the mechanism of·OH and PEF was proposed, as shown in Eqs. (6)–(8):

3.4 SDS-PAGE

3.4.1 The effects of the irradiation time and concentration

The gel electrophoresis results obtained using an aqueous solution of lysozyme (M.W. = 14.4 kDa [50] and constant concentration)containing different concentrations of PEF at various illumination times are shown in Fig. 10.The band intensity at 28–31 kDa is recognized as the dimer of the lysozyme monomers bound via Trp-Trp cross-links[51, 52]. The number of Trp-Tyr and Tyr-Tyr cross-links[53] increases significantly upon increasing the concentration of PEF and illumination time. These results can be rationally explained by both the longer illumination time and higher concentration of PEF resulting in a higher proportion of the excited states of PEF, which effectively react with Trp and generate the neutral radical of Trp via Eq. (4) as discussed beforehand, thus increasing the crosslinking of lysozyme.

3.4.2 A comparison of the phototoxicity of DIF and PEF

Because the phototoxicity of PEF has been confirmed in this work and the reaction between DIF and biomolecules has been previously reported [27], it is reasonable to assume that a comparison of the phototoxic effects of these FQ antibiotics will be useful toward unveiling the substituent effects at the 1-position. Aqueous solutions of lysozyme containing DIF and PEF with the same absorption intensity at 355 nm (OD355nm= 0.20) were illuminated, and the intensity of the dimers and trimers was compared with the blank sample. Figure 10 shows that under the same experimental conditions, the intensity of the dimers in the sample containing DIF increased slightly when compared with the blank sample, which suggests the weak phototoxicity of DIF.On the contrary,the intensity of the dimers in the sample containing PEF was much stronger.More specifically,the enhanced phototoxic effect brought by PEF (38.0%) was fourfold higher than that observed for DIF (9.2%, calculated from Fig. 11b). This result was in accordance with the conclusion of a previous study stating that the phototoxicity of PEF is comparable to lomefloxacin [54], a well-known phototoxic FQ antibiotic[55]. This is also in agreement with the reported phenomenon that 1-ethyl FQ antibiotics will be phototoxic no matter if the 8-position is halogenated or not [26].

Why do 1-ethyl FQ antibiotics exhibit severe phototoxic effects when compared with 1-fluorophenyl FQ antibiotics?To answer this,we should firstly review the mechanisms of their phototoxicity. There are two general types of photosensitization pathway observed for the photooxidation of biomolecules. Type I is the direct interaction between the excited photosensitizer (or photoexcited degradation products of the original photosensitizer)and biomolecules,while type II is the energy transfer reaction from the excited photosensitizer to ground state molecular oxygen(O2, triplet) to produce singlet oxygen (1O2), followed by1O2-mediated oxidation of the biomolecule [56]. The excited states of the photosensitizer(photoexcited PEF and DIF molecules)can be a singlet state,but more commonly,they are dominated by excited triplet states [57]due to the significantly longer life time of the triplet state.

Following these mechanisms, we systematically analyzed and compared the processes after the photoexcitation of PEF and DIF. The fundamental photoproperties related to PEF and DIF are displayed in Table S1 in the Supporting Information. We first analyzed the type I mechanism,which involves the excited singlet and triplet states of PEF and DIF, as well as potential secondary photosensitizers that are generated by the photodegradation of the singlet and triplet states of PEF and DIF.As shown in Fig. 1,PEF and DIF belong to 4′-N-alkylated FQ antibiotics and their primary pathway of photodegradation is the demethylation reaction [58, 59], in which the products are also two phototoxic FQ antibiotics,norfloxacin(NOR)and sarafloxacin(SAR) (Fig.S1),respectively.So the excited states of PEF and DIF, and their photoproducts should be compared separately.

Firstly,the ground states of PEF and DIF will be excited to their singlet states upon illumination with light. Based on the time-dependent density functional theory(TD-DFT)calculations performed at the B3LYP/6-311G* level of theory (Gaussian 09 program), the excited singlets of PEF and DIF with the lowest energy were almost the same(3.09 and 3.11 eV at 401 and 398.5 nm, respectively) and this calculation was also in accordance with the UV–Vis absorption spectra obtained for PEF and DIF, as shown in Fig. 1, which are almost identical at longer wavelengths.However, singlet states with both the lowest energy and sizeable oscillator strength(f ≥0.1)[60]were observed at 318.9 nm (3.89 eV) and 266.2 nm (4.66 eV) for PEF(f = 0.114)and DIF(f = 0.463),respectively(Table S2 and S3). The much lower energy of the singlet states with sizeable oscillator strength indicates the higher probability to transfer from the ground state to their corresponding excited singlet states.This suggests that the excited singlet generation process was much more feasible for PEF than DIF under UVA (315–400 nm) irradiation. Furthermore,＞95% of sunlight that arrives at the earth’s surface is located in the UVA region[61],which will facilitate the generation of singlet PEF (1PEF*), but not the generation of singlet DIF (1DIF*).

Secondly, except for the direct reaction with biomolecules,which should be negligible due to the rapid decay of the singlet states (τ ≤10-9s) [60], the generated1PEF*and1DIF* can transform into their triplet states via an intersystem crossing (ISC) process or photodegradation reaction, both of which can contribute to photosensitive damage. The lifetime of these singlet states is too short to trigger photosensitive damage,and thus,there is no need to consider their direct contribution in this work. Table S1 shows the energy difference between3PEF* and3DIF*,1.5 kJ·mol–1, which is insignificant, while the quantum yield of the intersystem crossing process (ΦISC) for PEF was 1.6-fold higher than that of DIF while the rate constant with biomolecules for DIF was approximately threefold higher than that observed for PEF. The higher rate constants indicate that DIF displays phototoxicity faster than PEF, but it is not the reaction rate that acts as the decisive factor in the severity of the phototoxicity because in the present work the value observed for lysozyme is 2.5-fold higher than that of the FQ antibiotics and one lysozyme molecule contains six reactive Trp residues [62], which offers enough oxidation sites for the photoexcited FQ antibiotic molecules. Furthermore, the biomolecules found in humans or animals are also found in excess when compared with the number of photosensitizers that exist in their body. The quantum yield of the triplet state of the photoproduct of PEF (NOR, Φ’ISC= 0.52) was also 1.5-fold higher than that observed for DIF(SAR,Φ’ISC= 0.35,Table S1). The lower quantum yield observed for3SAR*was similar to3DIF*because the bulky substituent in the1-position remains the same in SAR as that observed in DIF.

As for the type II mechanism, according to previous studies,the quantum yields of singlet oxygen for PEF,DIF,and their primary degradation products, NOR [63] and SAR [64], are comparable and as low as 0.06–0.10 in a neutral aqueous medium [54]. Therefore, the type II mechanism is not important when comparing the phototoxicity of PEF and DIF.

In one word, the insertion of a bulky substituent at the 1-position gives rise to a significant increase in the energy gap between the ground state and the lowest sizeable excited singlet state of FQ antibiotics,which further results in the lower quantum yield of their triplet states and lower phototoxicity.

This proposed mechanism may be applied to the whole family of FQ antibiotics because it has been proved using several other FQ antibiotics. To name a few, SAR, a 1-fluorophenyl FQ antibiotic, also has lower ΦISCand weaker phototoxicity than that of its counterpart NOR, a 1-ethyl FQ antibiotic. Moreover, it has been confirmed using several phototoxicity studies on other FQ antibiotics with various substituents at the 1-position [22, 26, 65],among which delafloxacin, a newly listed FQ antibiotic bearing a Cl and bulky substituent at the 8-and 1-position,respectively, has been shown to be free of phototoxicity[22]. FQ antibiotics bearing a halogen substituent at the 8-position usually exhibit broader antibacterial activity,improved oral bioavailability, but they are rarely commercialized due to their severe phototoxicity. The use of a bulky substituent at the 1-position may dramatically reduce the phototoxicity of FQ antibiotics and will be a promising synthetic strategy for the development of the next generation of FQ antibiotics.

4 Conclusion

Herein, the photochemistry and phototoxicity of PEF, a 1-ethyl-substituted FQ antibiotic, have been investigated and compared with its counterpart, DIF, a 1-fluorophenyl FQ antibiotic. The insertion of a bulky substituent at the 1-position enables the absorption band of the lowest sizeable(f ≥0.1)singlet to be blueshifted from 319 nm(PEF)to 266 nm (DIF) and decreases the quantum yield of their excited triplet states to some extent, which are the crucial intermediates in the photosensitization process and thus significantly reduce their phototoxicity. To the best of our knowledge, this is the first study combining transient,steady-state,and computational methods based on different substituents located at the 1-position of FQ antibiotics and proposes a mechanism of why the insertion of a bulky substituent at the 1-position can alleviate the phototoxicity.This study will be beneficial to the development of novel FQ antibiotics that are free of phototoxicity.

Acknowledgements The authors thank Dr.Xing-Xing Jiang from the Technical Institute of Physics and Chemistry, Chinese Academy of Sciences,and Dr.Xiting Zhang from the University of Hong Kong for their advice on our DFT calculations.