Article pubs.acs.org/crt
Understanding of the Photoallergic Properties of Fluoroquinolones: Photoreactivity of Lomefloxacin with Amino Acids and Albumin Sonia Soldevila, M. Consuelo Cuquerella, and Francisco Bosca* Instituto Universitario Mixto de Tecnologia Quimica (UPV-CSIC), Universitat Politecnica de Valencia, Avenida de los Naranjos s/n, 46022 Valencia, Spain ABSTRACT: Although the phototoxic and photoallergic properties of fluoroquinolone antibiotics (FQ) are remarkable, the mechanisms involved in these processes are not completely understood. For this reason, it is considered worthwhile to study in detail the photochemical interactions of lomefloxacin (LFX) and its N-acetyl derivative ALFX, two 6,8-dihalogenated fluoroquinolones, with the most abundant protein in human plasma (human serum albumin, HSA) to analyze their covalent binding. Fluorescence measurements and laser flash photolysis experiments performed in this work have revealed that N-acetylation of the LFX piperazinyl moiety produces an important increase of the drug affinity to albumin. Thus, while the association constant (Ka) for the LFX···HSA complex is below 103 M−1, the Ka for the HSA···ALFX complex resulted in ca. 5 × 103 M−1. Interestingly, LFX is mainly located at site I of HSA, while ALFX shows no preference for site I or II. A high reactivity between the aryl cations generated from (A)LFX dehalogenation and Trp and Tyr together with the generation of covalent adducts between the FQ and these amino acids was observed. However, the interactions between the FQ singlet excited state and albumin in FQ···HSA complexes seem to be the key process of FQ covalent binding to albumin. Moreover, our findings have shown a correlation between the photobinding properties of dihalogenated fluoroquinolones to HSA and their FQ···HSA association constants.
■
INTRODUCTION Fluoroquinolones (FQs), drugs with a quinolinic main ring and an aminoalkyl substituent (Chart 1), are widely used as antibacterial agents. This family of compounds develops its pharmacological activity preventing the replication and repair of bacterial DNA by the gyrase (topoisomerase II) enzyme.1 During the past years, the antitumoral activity of some FQs has raised much attention.2−6 In vitro and in vivo studies have confirmed the anticancer effects of these drugs supported by the reduction of all-cause mortality among cancer patients.7 The FQ antitumor effect has been associated with the inhibition of mammalian DNA topoisomerase I, topoisomerase II, and DNA polymerase. This genotoxic effect is enhanced by UV irradiation8 and confers FQs with a potential property to be photochemotherapeutic agents. In this context, 6,8-dihalogenated FQs such as fleroxacin, BAY y3118, and lomefloxacin (LFX, Chart 1) have shown not only a remarkably photoinduced genotoxicity but also photoxicity and photoallergy.9−17 These adverse side effects have prompted a large number of studies concerning the photophysical and photochemical properties of dihalogenated FQ during the past years.12,18−22 Recently, an aryl cation with alkylating properties, generated by an unusual heterolysis of the strong C8-halogen bond of FQ,18−22 has been proposed as the origin of the most important FQ adverse side effects.21,23−28 Thus, it has been suggested that a covalent binding between a dehalogenated FQ and DNA or proteins could be the key process in the photoallergy and photogenotoxicity induced by these drugs.18−22 However, monohalogenated FQs, which do not © 2014 American Chemical Society
give rise to this type of reactive intermediate, also have photohaptenic properties.29−31 In fact, molecular bindings between monohalogenated FQs and bovine serum albumin have been observed after UV irradiations.29 In spite of the important reactivity of triplet excited states of some monohalogenated FQ with tyrosine and tryptophan,32,33 no research on the photochemistry of dihalogenated FQ in the presence of proteins or amino acids has been performed. In the present study, the photoreactivity of LFX will be analyzed in the presence of a model protein such as human serum albumin (HSA) and some of its most reactive linkers (tryptophan and tyrosine) to determine the main pathway involved in the generation of drug−protein covalent bonds. Moreover, as acetylation of the piperazinyl ring of FQ produces changes in the photophysical and/or photochemical behavior of these drugs,18,34−36 the acetylated form of LFX (ALFX, Chart 1) has been included in the present study. Chart 1. Structure of LFX and Its N-Acetyl Derivative ALFX
Received: October 8, 2013 Published: February 14, 2014 514
dx.doi.org/10.1021/tx400377s | Chem. Res. Toxicol. 2014, 27, 514−523
Chemical Research in Toxicology
Article
Competitive reactions between 2 × 10−4 M (A)LFX and ibuprofen (IBP) or warfarin (WAR) to bind HSA were carried out using 10−3 M PB aqueous solutions. The experiment consisted of registering FQ fluorescence at 355 nm, after the addition of increasing amounts of IBP (5 × 10−5 M to 6 × 10−4 M) or WAR. To avoid absorption at 355 nm, WAR was added only up to 2 × 10−4 M. Laser Flash Photolysis Experiments. A pulsed Nd:YAG laser was used for the excitation at 355 nm. The single pulses were of ∼10 ns duration, and the energy was from 10 to 1 mJ/pulse. A pulsed xenon lamp was employed as the detecting light source. The laser flash photolysis apparatus consisted of the pulsed laser, the Xe lamp, a monochromator, and a photomultiplier made up of a tube, housing, and power supply. The output signal from the oscilloscope was transferred to a personal computer. Aqueous solutions of 10−4 M (A)LFX were prepared in 10−3 M NaHCO3, and the experiments registered under anaerobic conditions bubbling N2O. The samples containing albumins needed special manipulation due to the impossibility of bubbling the solutions to remove oxygen. Thus, the anaerobic media were generated by flowing N2O during 20 min above the solutions without generating bubbles while stirring the solution. Transient absorption spectra at different times after the laser pulse were obtained for each sample in the presence and absence of HSA, paying special attention to intersystem crossing quantum yield changes and to the generation of new intermediates. The Trp and Tyr concentrations ranged between 10−4 and 10−2 M. The quenching experiments were carried out keeping constant the pH at 7.4 throughout the experiment. Aryl cation quenching rate constants by biomolecules were determined using the Stern−Volmer eq 3:
Emission studies, laser flash photolysis, and product analysis using gel filtration and ultraperformance liquid chromatography with high resolution mass spectrometry detection (UPLCHRMS) performed in this work suggest that the main process involved in FQ photobinding to albumin arises from the singlet excited state of complexed FQ (1(A)LFX···HSA).
■
MATERIALS AND METHODS
General Materials. Human serum albumin fatty acid free (HSA), ibuprofen (IBP), L-tyrosine methyl ester (Tyr), L-tryphophan (Trp), lomefloxacin (LFX), norfloxacin (NFX), and warfarin (WAR) were commercial products obtained from Sigma-Aldrich Chemical Co. Sephadex G-25 columns were from GE Healthcare UK limited. Sodium phosphate buffer (PB) and sodium bicarbonate buffer were prepared from reagent-grade products using milli-Q water; the pH of the solutions was measured through a glass electrode and adjusted with NaOH to pH 7.4. Other chemicals were of reagent grade and used as received. The samples of FQs were prepared with different phosphate buffer (PB) concentrations starting from a mother solution of 300 mM PB adjusted at pH 7.4 with a Crison pH-meter. Acetyl LFX (7-(4-acetyl-3methyl-1-piperazinyl)-1-ethyl-6,8-difluoro-1,4-dihydro-4-oxoquinoline3-carboxylic acid (ALFX)) was prepared as previously described from a solution of LFX (300 mg, 0.94 mmol) in Ac2O (50 mL) that was refluxed for 7 h.18 The solution was cooled to room temperature and concentrated. Afterward, the residue was dissolved in water, neutralized to pH ca.7.4, extracted with CH2Cl2, and concentrated to dryness. Absorption and Emission Measurements. Ultraviolet spectra were recorded on a UV/vis scanning spectrophotometer (Cary 50). Fluorescence emission spectra were recorded on a Photon Technology International (PTI) LPS-220B fluorimeter. Lifetimes were measured with a time-resolved spectrometer (TimeMaster fluorescence lifetime spectrometer TM-2/2003) from PTI by means of the stroboscopic technique, which is a variation of the boxcar technique. A hydrogen/ nitrogen flashlamp (1.8 ns pulse width) was used as the excitation source. The kinetic traces were fitted with monoexponential decay functions. Measurements were done under aerated conditions at room temperature (25 °C) in cuvettes of 1 cm path length. The excitation wavelength used to register the fluorescence lifetimes was 320 nm. The fluorescence quantum yield of quinine bisulphate in 1 N H2SO4 (ϕF = 0.546) was used as the standard. Albumin Fluorescence Quenching by Fluoroquinolones. Phosphate buffered (10−3 M, pH ca. 7.4) aqueous solutions containing 10−5 M HSA display a fluorescence band centered at 344 nm, after excitation at 295 nm. The fluorescence quenching of this band was monitored after the addition of increasing amounts of (A)LFX (from 10−6 to 1.2 × 10−5 M). Before analyzing the data, the so-called inner filter effect correction (IFE) was applied because both FQs absorb light at the excitation and emission wavelengths. The IFE correction was applied using eq 1:37−39
Fcorr = Fobs × 10(Aex + Aem)/2
1/τ = 1/τ0 + k[Quencher]
Lifetimes of the aryl cations of LFX and ALFX were determined from their decay traces obtained at 480 and 600 nm, respectively. Analysis of the Covalent Binding of Fluoroquinolones to HSA Induced by Light. The study was performed using a 1/1 molar ratio of drug/protein for UV−vis measurements. (A)LFX and norfloxacin (NFX) were added to 2 × 10−4 M HSA and allowed to incubate in the dark for 30 min. Photolysis was performed using a Rayonet photochemical reactor equipped with eight black light phosphor lamps emitting in the 310−390 nm range, with a maximum at 350 nm.45 Samples were then irradiated for various time periods (20 to 100 s with light doses from 40 to 280 mJ/cm2) and then kept in darkness during 24 h prior to analysis (controls included drug−HSA mixtures kept in the dark, HSA with and without irradiation, and irradiated drug added to HSA). The samples were diluted by a factor of 2, adding water before protein separation. Subsequently, to determine whether FQ is covalently linked to HSA, the solutions were chromatographed on Sephadex G-25 columns equilibrated with 2/8 ethanol/aqueous 10 mM PB, as described previously.29 The first fraction of each sample contained albumin (alone or covalently linked to a drug), while in the next one, the remaining free drug was found. Thus, the first fraction was analyzed by UV−vis spectrometry. The amount of drug linked to albumin was determined from absorption spectra of these samples, while the remaining free drug (second fraction of each sample) was determined by HPLC on a Hitachi apparatus equipped with a Spherisorb column (ODS-2, 10 mm packing), an L-6250 intelligent pump, and an L-400 fixed-wavelength UV detector at a wavelength of 325 nm. Acetonitrile/water/ trifluoracetic acid mixtures of 20/79.9/0.1 for LFX and NFX and 35/64.9/0.1 for ALFX were used as the mobile phase. The experiments were performed three times. Photoproduct Analysis of the Photolysis of (A)LFX in the Presence of Trp and Tyr. Photolysis of deaerated aqueous solutions of (A)LFX (10−4 M) at pH 7.4 were carried out in the absence and in the presence of tryptophan (Trp) and tyrosine methyl ester (Tyr) 10−2 M. The corresponding photoproducts were identified by UPLCHRMS. Briefly, the chromatography was performed on an Acquity UPLC system (Waters Corp.) with a conditioned autosampler at 4 °C. The separation was carried out on an Acquity UPLC BEH C18
(1)
where Fcorr and Fobs are the corrected and observed fluorescence intensities, respectively, and Aex and Aem are the absorbance values at the excitation and emission wavelengths, respectively. Fluorescence Quenching of Fluoroquinolones by Albumin. This fluorescence quenching study was performed by adding HSA concentrations from 10−5 to 10−4 M FQ buffered aqueous solutions (10−3 M PB, pH ca. 7.4) and after excitation at 355, 348, and 330 nm. These data were also analyzed using the eq 2 to establish the drug− biomolecule interactions. Equation 2 was selected to determine the drug−protein interactions from the fluorescence quenching data:40−44
F0/F = 1 + K sv[Q]
(3)
(2)
where F0 and F are the fluorescence intensities in the absence and presence of the quencher respectively, [Q] is the quencher concentration, and Ksv is the Stern−Volmer quenching constant. 515
dx.doi.org/10.1021/tx400377s | Chem. Res. Toxicol. 2014, 27, 514−523
Chemical Research in Toxicology
Article
column (50 mm × 2.1 mm i.d., 1.7 μm). The column temperature was maintained at 40 °C. The analysis was achieved with gradient elution using methanol and water (containing 0.01% formic acid) as the mobile phase. The Waters ACQUITY XevoQToF Spectrometer (Waters Corp.) was connected to the UPLC system via an electrospray ionization (ESI) interface. The ESI source was operated in positive ionization mode with the capillary voltage at 3.0 kV. The temperature of the source and desolvation was set at 100 and 400 °C, respectively. The cone and desolvation gas flows were 100 L h−1 and 800 L h−1, respectively. All data collected in Centroid mode were acquired using Masslynx software (Waters Corp.). Leucine− enkephalin was used as the lock mass generating an [M + H]+ ion (m/z 556.2771) at a concentration of 2 ng/mL and a flow rate of 50 μL/min to ensure accuracy during the MS analysis.
indicative that the association constant (Ka) of LFX···HSA is lower than 103 M−1 in agreement with that obtained by capillary electrophoresis.47 In this context, it is worth mentioning that there is a divergence between the present LFX result and previous data concerning the LFX association with human and bovine serum albumins.48−50 This is probably due to the lack of application of the inner filter effect corrections in previous emission data.32 (A)LFX Fluorescence Quenching by HSA. The results obtained above, as well as those described in a recent study for other FQ,32 are in agreement with low affinities of most of the FQ to HSA. The study of FQ fluorescence quenching by HSA was carried out by the addition of HSA (up to 10−4 M) to 10−4 M FQ solutions (λexc 320 nm) with the aim of determining more accurately the HSA···FQ association constants for (A)LFX and norfloxacin (NFX) as a monohalogenated FQ. Figure 2 shows (A)LFX fluorescence
■
RESULTS AND DISCUSSION Study of the (A)LFX Interactions with HSA by Emission Studies. Protein Fluorescence Quenching by FQs. Quenching of the emission of the Trp unit present in HSA is often used to obtain information about drug−protein interactions. The Trp fluorescence band in HSA is centered at ca. 344 nm (Figure 1), and it is exclusively exhibited by the
Figure 2. Emission spectra (λexc = 320 nm) of 10−4 M LFX (dash lines) or ALFX (solid lines) in 10−3 M phosphate buffer aqueous solutions, in the absence (black) and in the presence of 10−4 M HSA (red). Inset: Stern−Volmer plots corresponding to the fluorescence quenching of 10−4 M ALFX (empty symbols) and LFX (solid symbols) phosphate aqueous solutions by HSA.
Figure 1. Absorption spectra displayed by ALFX (dash), LFX (dash dot), HSA (red solid line), and tryptophan emission bands in HSA (black solid line).
intensities in the presence and absence of 10−4 M albumin. It is remarkable that ALFX emission is quenched more efficiently than that of LFX. In the case of NFX, an emission quenching similar to that obtained for LFX was detected (data not shown). The data of the fluorescence quenching assays were applied to eq 2 to determine Ksv, which is a simple model to obtain the (A)LFX···HSA association constants (Ka). It is assumed that the drug bound to HSA (FQ···HSA) does not emit51 and that there is no involvement of a dynamic process in emission quenching. This latter condition was confirmed since the fluorescence lifetime (τF) of both FQs does not change by the presence of HSA (τF for LFX and ALFX is 1.1 and 1.7 ns, respectively). Hence, the Ksv determined for the generation of the LFX···HSA complex was ca. 1 × 103 M−1, in agreement with those determined using other techniques.47 The Ksv of ALFX was ca. 5 × 103 M−1, a value very close to that obtained from protein fluorescence quenching assays. When the excitation wavelength employed was 355 nm instead of 320 nm, similar results were obtained. The use of specific probes to displace a protein ligand is a well-established method for binding site assignment. Ibuprofen (IBP) and warfarin (WAR) were chosen as stereotypical ligands for site II and I, respectively.52 It is important to note that the most probable candidate for complex formation at site I is
46
protein after excitation at 295 nm. In this context, experiments were conducted on 10−5 M HSA in 10−3 M PB aqueous solutions (pH ca. 7.4) in the presence of FQ (between 10−6 and 1.2 × 10−5 M). Since the quenchers show absorption at the excitation and emission wavelengths (Figure 1), correction for the inner filter effect (IFE) was applied before analysis of the results (more details are provided in the Materials and Methods section). Lomefloxacin did not significantly influence the intensity of HSA emission. By contrast, quenching of HSA fluorescence was observed after the addition of increasing amounts of ALFX. The FQ···HSA interactions were analyzed through the Ksv obtained from eq 2, assuming that in this process the quencher totally suppresses the intrinsic protein fluorescence. Thus, the Ksv for ALFX···HSA calculated from the slope of the plot F0/F versus [Q] was ca. 4 × 103 M−1. In the case of LFX···HSA association, no Ksv could be determined due to low fluorescence quenching. In static quenching processes, the Ksv values correspond to the total association constant of the quencher to the protein when the quencher totally suppresses the intrinsic protein fluorescence.42 Therefore, when these conditions apply, the resulting Ksv is a good approximation for FQ···albumin association constants.40 Hence, the results are 516
dx.doi.org/10.1021/tx400377s | Chem. Res. Toxicol. 2014, 27, 514−523
Chemical Research in Toxicology
Article
Figure 3. Fluorescence spectra (λexc = 355 nm) of LFX (top) and ALFX (bottom) in aqueous 1 mM PB before and after the addition of 10−4 M HSA (blue) and increasing amounts of ibuprofen (up to 6 × 10−4 M) or warfarin (up to 2 × 10−4 M).
Figure 4. (A) Transient absorption spectra of 10−4 M LFX in aqueous 1 mM NaHCO3 and 15 × 10−3 M Trp 10 ns after laser excitation (red line) and 0.4 μs (blue line). (B) Stern−Volmer plots corresponding to the quenching of the LFX aryl cation by Trp (■) and Tyr (○). (C) Transient absorption spectra of 10−4 M ALFX in aqueous 1 mM NaHCO3 and 9 × 10−3 M Trp 10 ns after laser excitation (green line) and 0.4 μs (blue line). (D) Decay and growth traces at 600 (green) and 520 nm (blue), corresponding to the absorption maxima of the ALFX aryl cation and Trp radical, respectively.
517
dx.doi.org/10.1021/tx400377s | Chem. Res. Toxicol. 2014, 27, 514−523
Chemical Research in Toxicology
Article
Trp214 in HSA, while in the case of site II it is the hydrophobic amino acid Tyr411.53 The experiments were performed on 1/1 FQ/HSA aqueous PB (10−3 M) solutions, and titration was conducted by the addition of increasing amounts of IBP (up 6 × 10−4 M) or WAR (up to 2 × 10−4 M) after excitation at 355 nm (Figure 3). The rise of LFX emission is higher in the presence of WAR than in the presence of IBP, which indicates qualitatively that LFX is mainly complexed to site I. As the probe association constants are 3.3 × 105 M−1 and 2.7 × 106 M−152 for WAR··· HSA and IBP···HSA, respectively, at concentrations higher than 2 × 10−4 M, more than 95% of the drugs are associated to the albumin. Taking this into account, the percentage of LFX··· HSA at site I was ca. 8% and at site II ca. 3%, under our experimental conditions. Different behavior was observed measuring the fluorescence of aqueous solution of 1/1 ALFX/HSA. In this case, the addition of the two probes resulted in a similar increase of the emission, which was indicative that ALFX···HSA was ca. 20% at site I and another 20% at site II. Thus, N-acetylation of LFX produces an increase of the affinity predominantly for site II. This is fully consistent with the preferential affinity of nalidixic acid (a quinolone lacking a piperazinyl ring) for site II of HSA.54 It is reasonable to think that the cationic piperazinyl ring increases the affinity of FQ for aqueous media. Reactivity of (A)LFX Aryl Cations with HSA. Photolysis of LFX and ALFX under N2O, at pH 7.4, generates aryl cations of LFX (λmax = 490 nm and τ ca. 200 ns) and ALFX (λmax = 600 nm and τ c a. 340 ns).18 Their transient absorption spectra are shown in Figure 4A and C. Laser excitation of (A)LFX (10−4 M) in NaHCO3 solutions (10−4 M, pH ca. 7.4) under N2O atmosphere displayed the same transient absorption species in the absence and the presence of HSA (up to 1.5 × 10−4 M). In addition, the aryl cation lifetimes of both FQs remained nearly unaltered. However, the generation of these intermediates decreased when HSA was added, this effect being most important in the case of the ALFX aryl cation. Figure 5 shows
K a = [FQ···HSA]/([HSA] × [FQ])
Applying the Ka previously determined by fluorescence quenching in eq 5 and taking into account the initial concentration of drugs and biomolecules in the LFP experiments, the increase in the amount of complexed drugs is coincident with a decrease in the generation of the corresponding aryl cation. Reactivity of (A)LFX Aryl Cations with Tryptophan (Trp) and Tyrosine (Tyr). The reactivity of (A)LFX aryl cations with Trp and Tyr was also addressed by laser flash photolysis. The bimolecular rate constants (kq) for these interactions were determined using the Stern−Volmer equation, eq 3, and they are listed in Table 1. The aryl cation Table 1. Reactivity of LFX and ALFX Aryl Cations with Biomolecules Trp Tyr HSA
the lowering in the absorption of the (A)LFX aryl cation at its λmax by the addition of HSA. Considering the biomolecule− drug equilibrium in eq 4: Ka
Kd
kq(LFX)/109 (M−1 s−1)
kq(ALFX)/109 (M−1 s−1)
1.2 0.9
