10520
J. Phys. Chem. B 2007, 111, 10520-10528
The Spectrophysics of Warfarin: Implications for Protein Binding Bjo1 rn C. G. Karlsson,† Annika M. Rosengren,† Per Ola Andersson,‡ and Ian A. Nicholls*,† Bioorganic and Biophysical Chemistry Laboratory, School of Pure and Applied Natural Sciences, UniVersity of Kalmar, SE-391 82 Kalmar, Sweden, and Department of EnVironment and Protection, FOI NBC Defence, SE-901 82 Umeå, Sweden ReceiVed: March 30, 2007
The photophysical behavior of the isomers of the anticoagulant drug warfarin in various solvents and solvent mixtures was investigated using absorption, 1H NMR, and steady-state and time-resolved fluorescence spectroscopies in conjunction with B3LYP-based theoretical treatments. Complex absorption patterns were observed, indicative of the presence of different isomers of warfarin in the various solvents studied. In alkaline aqueous solution, the deprotonated open side form of warfarin is highly dominant and only one S0fS1 singlet transition could be observed in the absorption spectrum centered at 320 nm. These observations were supported by theoretical density functional calculations (B3LYP) in which the geometries of nine isomers of warfarin were optimized and their respective eight lowest singlet and three lowest triplet excitation energy levels were predicted. Examination of the fluorescence excitation and emission spectra of the isomers in nonpolar and polar organic solvents showed the presence of the deprotonated open side chain form of warfarin in 2-propanol, ethanol, and acetonitrile. Time-resolved fluorescence experiments revealed a short decay time constant, τ1, in all solvents studied while in more polar environments a second longer one, τ2, was evident varying between 0.5 and 1.6 ns depending on solvent polarity. The variation of number and length of fluorescence lifetimes as a function of solvent environment has provided a tool for examining warfarin protein binding. Studies on the binding of warfarin to human serum albumin (HSA) have been undertaken, and different modes of binding were observed which are indicative of binding to the anion-selective Sudlow I and, second, a lower affinity mode of interaction.
Introduction Coumarin derivatives such as vitamin K play key roles in many biological processes.1,2 Coumarin and its derivatives are of significant interest as tools in the study of biological systems on account of their importance as pharmaceuticals and because of their capacity to fluoresce. Although coumarin itself demonstrates a fairly low fluorescence quantum yield in most organic solvents at room temperature,3 the attachment of different electron-withdrawing or accepting substituents on the coumarin core structure can greatly influence the fluorescence properties.4-10 These observations have led to the synthesis of a vast number of coumarin derivatives, demonstrating interesting photophysical properties such as, for example, intramolecular charge transfer (ICT)11,12 and twisted intramolecular charge transfer (TICT)13,14 relaxation processes. The nature and position of substituents is important for the fluorescence behavior of these compounds, and many spectroscopic and theoretical studies have been undertaken to characterize the photophysical behavior of these substances.15-17 A fluorescent coumarin derivative of great clinical significance is the highly potent anticoagulant warfarin.18,19 The binding of warfarin to a range of proteins, for example, human serum albumin (HSA)20-22 and CYP2C9,23 and model systems, such as cyclodextrins,24 has been undertaken using various spectroscopic studies. In the 1970s, Valente and co-workers25-27 * To whom correspondence should be addressed. E-mail: Ian.Nicholls@ hik.se. † University of Kalmar. ‡ NBC Defence.
reported a series of NMR studies demonstrating that warfarin in solution is present as a series of isomers, Chart 1. Interestingly, they observed that the position of the equilibrium was dependent upon the nature of the solvent polarity, hydrogen donating/accepting ability, and pH. For example, warfarin preferably adopts cyclic hemiketal structures, rather than the open side chain isomer, in low dielectric constant solvents.28 As the transfer from solution to a binding site on a protein involves changes in the environment of warfarin, it can potentially even effect changes in position of the various equilibria responsible for formation of the various isomers.23,26,28 Accordingly, more extensive study of the isomeric distribution of warfarin in organic solvents and solvent mixtures is highly motivated and may offer further knowledge concerning warfarin’s in vivo function. In this study, absorption, fluorescence, and NMR spectroscopic studies have been used, in conjunction with B3LYP theoretical treatments, to investigate the isomeric distribution in a range of organic solvents. The influence of solvent on the photophysical properties of the isomers of warfarin is investigated and discussed. Moreover, we report an unexpected fluorescence behavior in the deprotonated open side chain isomers 1(-) and 4(-) of warfarin. Furthermore, we propose that an understanding of the relationship between structure and spectroscopic behavior of warfarin’s various isomers can provide a tool for studying warfarin-protein binding. Experimental Section Chemicals. Racemic warfarin (3-(R-acetonylbenzyl)-4-hydroxycoumarin, min 98%) and human serum albumin (HSA,
10.1021/jp072505i CCC: $37.00 © 2007 American Chemical Society Published on Web 08/11/2007
Spectrophysics of Warfarin Implications CHART 1: Possible Solution Structures of Warfarin Previously Reported in the Literaturea 23,25
a Open side chain, 1; deprotonated open side chain, 1(-); intramolecularly hydrogen-bonded open side chain, 1(H); major, 2(M), and minor, 2(m), cyclic coumarin hemiketals; major, 3(M), and minor, 3(m), cyclic chromone hemiketals; open side chain, 4, and deprotonated open side chain chromone, 4(-), structures.
essentially fatty acid free, 99%) were obtained from SigmaAldrich (St. Louis, MO). Anthracene (g96%) was from Merck Schuchardt OHG (Hohenbrunn, Germany). Sodium hydroxide (g99%) was obtained from Merck KGaA (Darmstadt, Germany). NMR chemicals used include the following: acetic acidd4 (99.9 atom% D) and chloroform-d (99.9 atom% D, containing 0.03% v/v TMS) were from Aldrich Chemical Co. (Milwaukee, WI). The water used was of Millipore quality (Millipore, Bedford, MA). The buffer used was phosphate-buffered saline (PBS) consisting of 0.5 M Na2HPO4 (Scharlau)/KH2PO4 (Merck) and 0.1 M NaCl at pH 7.3. All solvents were of analytical grade and were used as received. Apparatus. Absorption spectra were obtained on a Hitachi U-2000 double-beamed spectrophotometer. Fluorescence steadystate spectra were recorded on a SPEX FluoroMax-3 fluorimeter from Jobin-Yvon HORIBA. Fluorescence decay times were obtained by adapting the time-correlated single-photon-counting (TCSPC) technique using a time-resolved spectrometer from IBH equipped with a data station hub, TBX-04 photon detection module, Nanoled (295 or 334 nm), and a 5000 M fluorescence monochromator, all from IBH. Dielectrical constants () for the solvents studied at 294 ( 1 K were measured using either cyclohexane (CH) ( ) 2.023 at 293 K) or methanol (MeOH) ( ) 33.62 at 293 K) as reference calibration standards on a BI-870 liquid dielectric constant meter from Brookhaven Instruments. Quartz cuvettes (1.0 cm, 3.0 mL Suprasil from Hellma) were used in all spectroscopic experiments. 1H NMR experiments were performed on an AC-250 spectrometer operating at 250 MHz from Bruker. NMR Studies. Aliquots (17.5 mM to 3.31 M) of acetic-d4 containing a constant concentration of warfarin (34 mM) were added to an NMR test tube (500 µL) containing 34 mM of warfarin dissolved in chloroform-d. The chemical shifts, δ (ppm), for the different protons studied were recorded at 298 K. FID-spectra were processed in the software MestReC (v. 4.1.1.0) for Windows, and the resulting peaks from the observed proton resonances in the acetic acid-d4 titration experiment were integrated using the software PeakFit (v. 4.12) from Seasolve, United States, with the Voigt amplitude G/L algorithm and the smoothing function filter of Savitsky-Golay. All estimations of proton resonance peak areas yielded R2 > 0.95.
J. Phys. Chem. B, Vol. 111, No. 35, 2007 10521 Absorption Spectroscopy. Solutions of warfarin (14 µM) dissolved in heptane (HEP), chloroform (CHCl3), ethyl acetate (EtOAc), acetic acid (AcOH), tetrahydrofuran (THF), dichloromethane (DCM), 2-propanol (2PrOH), ethanol (EtOH), acetonitrile (ACN), mixtures of ACN:water (99.9:0.1 and 99.5:0.5 v/v), and mixtures of ACN:NaOH (aq) (99.9:0.1 and 99.5:0.5 v/v, from a 1 M NaOH (aq) stock solution) were scanned between 200 and 400 nm. Absorption spectra for each of the solvents or solvent mixtures studied were obtained at 294 ( 1 K. Fluorescence Spectroscopy. Typically, all the collected fluorescence steady-state emission and excitation spectra were obtained for 14 µM solutions of warfarin in CHCl3, AcOH, 2PrOH, EtOH, or ACN. These spectra were corrected for wavelength-dependent optical components at 294 ( 1 K with the excitation and emission monochromator spectral band widths at 1 and 2 nm, unless otherwise stated. Steady State. Initially, the dependency of solvent polarity on the fluorescence emission was investigated in different solvents using either 300 nm (CHCl3, AcOH, 2PrOH, EtOH, and ACN) or 305 nm (HEP, EtOAc, THF, and DCM) as the wavelength of excitation. Absorption was kept below 0.2 to prevent self-absorption of the emission and inner filter effects. At the excitation wavelength 305 nm, the spectral band widths were placed at 2 and 4 nm. The calculation of fluorescence quantum yields, ΦF, was performed using anthracene (14 µM) as a standard reference compound at the excitation wavelengths 300 or 340 nm in CH (ΦF ) 0.36).29 The ΦF for warfarin in the different solvents were calculated from a triplicate of experiments according to eq 1.30
∫0∞ IW(V˜ )dV˜ Φ W ) ΦR ∞ ∫0 IR(V˜ )dV˜
(
1 - 10-A
R
1 - 10-A
W
)( ) nW nR
2
(1)
where ΦR is the quantum yield of the reference standard used and the integrals ∫ IW(V˜ )dV˜ and ∫ IR(V˜ )dV˜ represent the emission band areas between 320 and 500 nm for warfarin and the standard reference compound, respectively. AW and AR are the absorbances at the wavelengths of excitation studied. Finally, nW and nR are the refractive indices for warfarin and the standard reference. To investigate the wavelength dependency of ΦF, fluorescence excitation spectra were recorded and the excitation monochromator was typically scanned from 250 nm up to 15 nm below the wavelength set as the static emission wavelength. Steady-State Titration Studies. Aliquots of AcOH (stock solution of 1.75 M in ACN or 17.5 M in CHCl3) were added to ACN (V0 ) 3.0 mL) or CHCl3 (V0 ) 2.0 mL) solutions of warfarin at 278, 288, and 298 K where the constant temperature was maintained by a thermostat (Gebru¨der Haake Gmbh, Germany). Fluorescence emission spectra were recorded both before and after sequential addition of AcOH. After each experiment, the maximum intensities at 407 nm in ACN and 355 nm in CHCl3 were corrected for the dilution effect upon the addition of AcOH. The excitation and emission monochromator slit widths were set to 2 and 4 nm. In a different series of titration experiments, different volumes of CHCl3 or THF (0.3, 0.6, 1.0, and 5.0% (v/v)) or glycerol (GLC) (0.1, 0.5, 1.0, 5.0, 10, and 15% (v/v)) were sequentially added to solutions of warfarin (3.0 mL) dissolved in ACN. All titration experiments were performed using an excitation wavelength of 305 nm. The fluorescence emission spectra were obtained after
10522 J. Phys. Chem. B, Vol. 111, No. 35, 2007 addition of solvent, rigorous stirring, and incubation for a minimum of 2.0 min. Apparent quenching constants were calculated by linear regression using the software package GraphPad Prism (version 4.00, GraphPad Software, United States). Time Resolved. The time-resolved measurements, with resolution down to sub-nanosecond time regime, was performed with time-correlated single-photon counting (TCSPC) technique. Fluorescence decay times of warfarin in ACN were measured at different emission wavelengths (345, 360, 380, 389, 400, 408, 420, and 450 nm) set by a single grating monochromator with spectral band widths of 8 or 16 nm. The HSA-warfarin binding studies were performed in PBS buffer using 10 µM of warfarin and 0.0065-0.65 mg mL-1 of HSA, using 420 nm as the fluorescence emission wavelength set by a single grating monochromator with spectral band widths of 12 or 16 nm. Apparent dissociation constants (KD) were determined from the isotherms derived from changes in the amplitude (Ai) of the state studied using one site binding model. The excitation source was either an IBH NanoLed (IBH) producing ∼295 nm (warfarin solvent experiments) or 334 nm (HSA-warfarin binding experiments) excitation pulses at 1.0 MHz repetition rates. The photon-counting rate was always lower than 2% of the excitation source repetition rate to avoid photon pile-up effects. A typical instrumental pulse using an IBH TBX-04 detector at a time calibration of 13 ps per channel and a scattering sample of LUDOX was about 600 ps at full width half-maximum. To remove the additional excitation pulse from the NanoLed at around 430 nm, a cutoff filter (UG-11) was kept in front of the excitation source.31 Before every sample measurement, the instrument response function was measured using LUDOX solution (excitation and emission wavelengths 295 and 334 nm, respectively). The fluorescence decays were collected over 4096 channels, and fluorescence lifetimes were calculated using either a double or a triple exponential method within the software package Analysis Software (version 6.1.51, IBH). Typically, curve-fits were accepted when χ2 e 1.2. Computational Methods. All the density functional theory (DFT)-B3LYP32,33 calculations were performed using the software package Gaussian (v. 03, revision B.02 Gaussian, Inc., Pittsburgh PA). The geometry of the studied isomers of warfarin were optimized using the B3LYP/6-31G(d,p) basis set. All geometrical parameters were varied until convergence was obtained, as indicated by the changes in the total molecular energy being smaller than a threshold of 1 × 10-6 Hartree. Dipole moments were calculated in vacuo and in water as well as the Gibbs free solvation energies, ∆GSolv, (water, polarizable continuum model, PCM with ) 78.39) using the B3LYP/6311G(2df,p) basis set). Finally, energies and probabilities (oscillator strengths, f) of eight singlet and three triplet excited electronic transition states were predicted using the optimized structures and the B3LYP/6-31G(d,p) basis set in vacuo. The current methodology employed is well-known to overestimate excitation energies by 3-5 kcal mol-1 (∼0.2 eV) for non-charge-transfer excitations. For pure charge-transfer (CT) excitations, the error introduced by the methodology is approximately twice this amount (0.3-0.4 eV).34 The predicted UV spectrum of warfarin was approximated using a Gaussian distribution of the absorption probabilities within the software SWizard (revision 4.2, S. I. Gorelsky, SWizard program, http://www.sg-chem.net). In these simulations, the UV/vis print window was set to 5000-50 000 cm-1, the width at half-height was set to 3000 cm-1 the absorption
Karlsson et al.
Figure 1. Absorption spectra of warfarin (14 µM) in solvents with different polarities: CHCl3 (black), AcOH (brown), 2PrOH (green), EtOH (blue), and ACN (red).
bandwidth of 0.35 eV, and the energy/frequency scaling coefficient was set to 1.0. Results and Discussion The structural dynamics of warfarin and the resultant ensemble of isomeric forms that can be adopted complicate our understanding of the function of warfarin in clinical contexts such as bioavailability, for example, its binding to transport proteins and receptors. Surprisingly, despite the structural diversity available to warfarin, the structural variation has not generally been taken into account when examining its binding to protein targets. Moreover, the fluorescence spectroscopic behavior of this substance in its various isomeric forms as a function of environment has not been clearly established. The UV absorption behavior of warfarin was investigated in a series of organic solvents with differences in polarity and hydrogen bonding ability, Figure 1. From these studies, a major absorption was observed between 240 and 360 nm with two distinct absorption bands centered at 280 and 310 nm. Altering the polarity of the media by changing solvent resulted in minor wavelength shifts ((1 nm) in the obtained absorption spectra. Examination of the structural composition of the spectra afforded additional information. Notably, transitions from less polar solvents, for example, CHCl3 or AcOH, to the more polar 2PrOH, EtOH, or ACN resulted in spectral broadening. An intensity increase in the absorption at 330-350 nm, induced by the polar solvents, was found to be most evident in the case of EtOH. This effect is attributed to the presence of significant hydrogen bonding between the solute and the solvent molecules, which may provide an explanation for the broadening of the spectrum previously observed for other coumarin derivatives in EtOH.35 The complex absorption pattern of the spectrum indicates that different isomers of warfarin exist simultaneously in solution. NMR studies confirmed the presence of different isomeric distributions in the various solvents studied (see below). Theoretical studies were then initiated to clarify the role and contribution of different isomers to the absorption spectrum. Nine previously reported isomeric structures of warfarin were modeled and energy minimized using the B3LYP density function. This approach was used to calculate conformational energies, ∆GSolv, and dipole moments in vacuo and in water as well as to predict the first eight singlet and three triplet electronic excitation energies and probabilities (Table S2, Figures S2S3). The predicted UV spectrum of warfarin in vacuo was normalized and compared to the experimentally obtained spectrum in EtOH, Figure 2, which was found to have an
Spectrophysics of Warfarin Implications
Figure 2. Calculated B3LYP singlet absorbance spectrum for the structures of warfarin studied (O) compared with the experimentally determined spectrum in EtOH (blue).
Figure 3. Absorbance spectra of warfarin in different solvents: 1 mM (red) and 5 mM (blue) of NaOH (aq) in ACN (black), containing 17 µM of warfarin.
absorption shoulder evident at ∼330-355 nm. 2PrOH and ACN induced similar shoulders in this wavelength range, as seen in Figure 1. From these predictions, it was observed that isomers 1(-) and 4(-), see Chart 1, were found to absorb in the region of the absorption shoulder. In addition, it was predicted that 1(H) or the hemiacetal isomers, 2 and 3, of warfarin absorbed at shorter wavelengths in the region e290 nm. Collectively, the B3LYP predictions of the absorption energies of warfarin indicated that the isomers 1, 1(-), 4, and 4(-), in the absence of intramolecular interactions, absorb at significantly lower energies compared to the absorption of 1(H), 2, and 3. To test the B3LYP prediction of the position of the 1(-) and 4(-) S0fS1 absorption band, spectra were acquired using warfarin in ACN and solutions of warfarin in ACN containing 1 or 5 mM NaOH (aq), Figure 3. As proposed, additions of NaOH (aq) favored the formation of 1(-) and 4(-). This could be clearly observed in the UV spectrum as one broad S0fS1 absorption band evolving as the concentration of NaOH increased with a maximum at 320 nm. Control experiments were performed, adding the same volumes of pure water, with, however, only very modest changes in the UV spectrum (data not shown). An assignment of the absorption behavior of the isomers of warfarin in organic solvents suggests that the S0fS1 band for each of 1, 1(-), and 4(-) lies predominantly in the region 320-360 nm, and the S0fS1 absorption band for cyclic isomers lies in the region 250-305 nm. Interestingly, the distinctly different spectra of the cyclic, open, and open deprotonated forms of warfarin allow for the ready identification of the presence of these forms in a sample with the help of simple UV-absorption spectra.
J. Phys. Chem. B, Vol. 111, No. 35, 2007 10523
Figure 4. Fluorescence emission spectra (λexc ) 300 nm) of 14 µM of warfarin in A, CHCl3 (black); B, AcOH (brown); C, 2PrOH (green); D, EtOH (blue); and E, ACN (red). Solvent background is subtracted.
Initially, a series of fluorescence excitation and emission spectra were recorded in the same range of solvents studied by UV spectroscopy. The spectral emission band shape was independent of excitation wavelength upon S0fS1 absorption, confirming the purity of the sample of warfarin used in these experiments. From the emission fluorescence spectra, Figure 4, it was found that in the polar solvents studied, two fluorescence bands could be observed. The existence of these bands was most evident in ACN, where the maximum of the first band was ≈360 nm. The second band was broad, featureless, and highly red-shifted, with its maximum at 407 nm, and occupied the same region as the fluorescence emission generated in nonpolar solvents such as CHCl3 and AcOH. The presence of the additional fluorescence band in the emission spectra in the polar solvents 2PrOH and EtOH was not as evident as the one observed in ACN. Nonetheless, the significant red-shift in the emission made it reasonable to assume that it also exists in these solvents. Fluorescence excitation spectra were obtained to characterize the nature of the fluorescing populations in the different organic solvents studied. Observing a static emission wavelength (at ∼the emission wavelength maximum) and varying the wavelength of excitation will typically yield information on the wavelength dependency on the quantum yield, ΦF, of a fluorophore. Spectra recorded in CHCl3 and AcOH both showed a complete overlap with the previously obtained absorption spectra, Figure 5a, which again confirm the sample purity. However, in the polar solvents 2PrOH, EtOH, and ACN, the excitation spectrum is broad and red-shifted in comparison to the observed absorption spectrum, Figure 5b-d. This phenomenon has previously been reported for a bis-coumarin derivative where the authors explained this abnormal fluorescence behavior as resulting from the existence of rotational isomers in solution.8 A closer examination of the excitation spectra in the polar solvents studied showed that spectra show a fluorescence enhancement at red-side and subsequently became broader in the region where absorption of 1(-) and 4(-) earlier was predicted by B3LYP calculations. Typically, the broadest excitation spectrum was observed in ACN where two fluorescence bands were previously observed in the emission spectrum. Accordingly, we ascribe this fluorescence enhancement phenomenon to the presence of 1(-) and 4(-) in solution. Estimations of ΦF were obtained using either 300 nm or 340 nm as the wavelengths of excitation, Table 1. Notably, excitation at 340 nm resulted in an evident increase in ΦF in the polar solvents studied, compared to excitation at 300 nm, which again is a consequence of a longer wavelength absorption because of
10524 J. Phys. Chem. B, Vol. 111, No. 35, 2007
Karlsson et al.
Figure 5. Normalized and overlapped absorbance (abs) and excitation (exc) spectra in (a) CHCl3 (λem ) 355 nm, abs: solid black line, exc: dashed black line) and AcOH (λem ) 350 nm, abs: solid brown line, exc: dashed brown line) and (b) 2PrOH (λem ) 387 nm, abs: solid green line, exc: dashed green line), (c) EtOH (λem ) 385 nm, abs: solid blue line, exc: dashed blue line), and (d) ACN (λem ) 350 nm, abs: solid red line and exc: dashed red line; λem ) 407 nm, exc: dashed purple line).
TABLE 1: Quantum Yield, ΦF, Determinationsa ΦF (×
a
10-2)
solvents
λexc ) 300 nm
λexc ) 340 nm
CHCl3 AcOH 2PrOH EtOH ACN
0.24 ( 0.01 0.55 ( 0.01 1.6 ( 0.01 2.7 ( 0.01 0.78 ( 0.01
5.9 ( 0.02 3.8 ( 0.01 9.1 ( 0.21
Values are presented as mean ( standard error.
TABLE 2: Time-Resolved Fluorescence Lifetimes and Their Relative Contributions to the Emission Spectrum at λexc) 295 nm relative amplitudeb at different λem (nm)
lifetime (ns)a solvent
τ1
τ2
CHCl3
