Probing the Binding Sites of Antibiotic Drugs Doxorubicin and N-(trifluoroacetyl) Doxorubicin with Human and Bovine Serum Albumins

We located the binding sites of doxorubicin (DOX) and N-(trifluoroacetyl) doxorubicin (FDOX) with bovine serum albumin (BSA) and human serum albumins (HSA) at physiological conditions, using constant protein concentration and various drug contents. FTIR, CD and fluorescence spectroscopic methods as well as molecular modeling were used to analyse drug binding sites, the binding constant and the effect of drug complexation on BSA and HSA stability and conformations. Structural analysis showed that doxorubicin and N-(trifluoroacetyl) doxorubicin bind strongly to BSA and HSA via hydrophilic and hydrophobic contacts with overall binding constants of K DOX-BSA = 7.8 (±0.7)×103 M−1, K FDOX-BSA = 4.8 (±0.5)×103 M−1 and K DOX-HSA = 1.1 (±0.3)×104 M−1, K FDOX-HSA = 8.3 (±0.6)×103 M−1. The number of bound drug molecules per protein is 1.5 (DOX-BSA), 1.3 (FDOX-BSA) 1.5 (DOX-HSA), 0.9 (FDOX-HSA) in these drug-protein complexes. Docking studies showed the participation of several amino acids in drug-protein complexation, which stabilized by H-bonding systems. The order of drug-protein binding is DOX-HSA > FDOX-HSA > DOX-BSA > FDOX>BSA. Drug complexation alters protein conformation by a major reduction of α-helix from 63% (free BSA) to 47–44% (drug-complex) and 57% (free HSA) to 51–40% (drug-complex) inducing a partial protein destabilization. Doxorubicin and its derivative can be transported by BSA and HSA in vitro.

Introduction Doxorubicin ( Fig. 1) is an effective chemotherapeutic agent for the treatment of breast cancer, malignant lymphomas, soft tissue sarcoma and various neoplastic diseases. The injury to nontargeted tissues often complicates cancer treatment by limiting doxorubicin dosage and diminishing the quality of patients' life during and after doxorubicin treatment [1]. Furthermore, the use of doxorubicin has been limited by a dose-related and irreversible cardiotoxicity as well as by the emergence of drug resistance [1]. Nano-particle delivery systems of doxorubicin are a promising approach to increase its antineoplastic efficacy and lower its sideeffects by site-specific drug delivery via active targeting mechanism [2,3]. Several delivery systems have been tested for doxorubicin transport using various nanoparticles by chemical conjugation or physical encapsulation [4][5][6][7][8][9][10]. Polymer conjugates such as PEGalbumins have been recently used for distribution of doxorubicin in vivo [11]. Serum albumins are also used for drug delivery in vitro [12].
Serum albumins are the major soluble protein constituents of the circulatory system and have many physiological functions [13]. The most important property of this group of proteins is that they serve as transporters for a variety of organic and inorganic compounds including metal ions. BSA (Fig. 1) has been one of the most extensively studied of this group of proteins, particularly because of its structural homology with human serum albumin (HSA). The BSA molecule is made up of three homologous domains (I, II, III) which are divided into nine loops (L1-L9) by 17 disulfide bonds. The loops in each domain are made up of a sequence of large-small-large loops forming a triplet. Each domain in turn is the product of two subdomains (IA, IB, etc.). Xcrystallographic data [14] show that the albumin structure is predominantly a-helical with the remaining polypeptide occurring in turns and extended or flexible regions between subdomains with no b-sheets. BSA (Fig. 1) has two tryptophan residues that possess intrinsic fluorescence [15]. Trp-134 in the first domain and Trp-212 in the second domain. Trp-212 is located within a hydrophobic binding pocket of the protein and Trp-134 is located on the surface of the molecule. HSA (Fig. 1) is a globular protein composed of three structurally similar domains (I, II and III), each containing two subdomains (A and B) and stabilized by 17 disulphide bridges [15,16]. Aromatic and heterocyclic ligands were found to bind within two hydrophobic pockets in subdomains IIA and IIIA, namely site I and site II [15][16]. Seven binding sites for fatty acids are localized in subdomains IB, IIIA, IIIB and on the subdomain interfaces [15]. While there are marked similarities between BSA and HSA in their compositions (Fig. 1), HSA has only one tryptophan residue Trp-214, while BSA contains two tryptophan residues Trp-212 and Trp-134 that can be used as fluorophores. Fluorescence quenching is considered as a useful method for measuring binding affinities. Fluorescence quenching is the decrease of the quantum yield of fluorescence from a fluorophore induced by a variety of molecular interactions with quencher molecule [17,18]. Therefore, using the quenching of the intrinsic tryptophan fluorescence of BSA (Trp-212 and Trp-134) and HSA (Trp-214) as a probing tool allows us to study the interactions of doxorubicin and its derivative with serum proteins in an attempt to characterize the nature of drug-protein complexation.
In this report, we present spectroscopic analysis and docking studies of the interaction of doxorubicin and N-(trifluoroacetyl) doxorubicin with HSA and BSA in aqueous solution at physiological conditions, using constant protein concentration and various drug contents. Structural information regarding drug binding site and the effect of drug-protein complexation on the stability and conformation of BSA and HSA is also reported here.

Materials
HSA and BSA fraction V were purchased from Sigma Chemical Company (St-Louis, MO) and used as supplied. Doxorubicin hydrochloride was generously provided by Pharmacia/Farmitalia Carlos Erba, Italy and N-(trifluoroacetyl) doxorubicin was synthesized according to the published methods [19,20]. Other chemicals were of reagent grades and used as supplied.

Preparation of Stock Solutions
Protein (BSA or HSA) was dissolved in aqueous solution (40 mg/ml or 0.5 mM) containing 10 mM Tris-HCl buffers (pH 7.4). The protein concentration was determined spectrophotometrically using the extinction coefficient of 36 500 M 21 cm 21 at 280 nm [21]. A drug solution of 1 mM was prepared in 10 mM Tris-HCl and diluted to various concentrations in Tris-HCl (pH 7.4).

FTIR Spectroscopic Measurements
Infrared spectra were recorded on a FTIR spectrometer (Impact 420 model), equipped with deuterated triglycine sulphate (DTGS) detector and KBr beam splitter, using AgBr windows. Solution of drug was added dropwise to the protein solution with constant stirring to ensure the formation of homogeneous solution and to reach the target drug concentrations of 0.125, 0.25 and 0.5 mM with a final protein concentration of 0.25 mM. Spectra were collected after 2 h incubation of BSA or HSA with drug solution at room temperature, using hydrated films. Interferograms were accumulated over the spectral range 4000-600 cm 21 with a nominal resolution of 1 cm 21 and 150 scans. The difference spectra [(protein solution + drug solution) -(protein solution)] were generated using water combination mode around 2300 cm 21 , as standard [22]. When producing difference spectra, this band was adjusted to the baseline level, in order to normalize difference spectra.

Analysis of Protein Conformation
Analysis of the secondary structure of BSA and HSA and their drug complexes was carried out on the basis of the procedure previously reported [23]. The protein secondary structure is determined from the shape of the amide I band, located around 1650-1660 cm 21 . The FTIR spectra were smoothed and their baselines were corrected automatically using Grams AI software. Thus the root-mean square (rms) noise of every spectrum was calculated. By means of the second derivative in the spectral region 1700-1600 cm 21 six major peaks for BSA, HSA and complexes were resolved. The above spectral region was deconvoluted by the curve-fitting method with the Levenberg-Marquadt algorithm and the peaks corresponding to a-helix (1660-1654 cm 21 ), b-sheet (1637-1614 cm 21 ), turn (1678-1670 cm 21 ), random coil (1648-1638 cm 21 ) and b-antiparallel (1691-1680 cm 21 ) were adjusted and the areas were measured with the Gaussian function. The areas of all the component bands assigned to a given conformation were then summed up and divided by the total area [24,25]. The curve-fitting analysis was performed using the GRAMS/AI Version 7.01 software of the Galactic Industries Corporation.

Circular Dichroism
CD Spectra of BSA, HSA and their drug complexes were recorded with a Jasco J-720 spectropolarimeter. For measurements in the far-UV region (178-260 nm), a quartz cell with a path length of 0.01 cm was used in nitrogen atmosphere. Protein concentration was kept constant (12.5 mM), while varying drug concentrations (0.125, 0.25 and 0.5 mM). An accumulation of three scans with a scan speed of 50 nm per minute was performed and data were collected for each nm from 260 to 180 nm. Sample temperature was maintained at 25uC using a Neslab RTE-111 circulating water bath connected to the water-jacketed quartz cuvettes. Spectra were corrected for buffer signal and conversion to the Mol CD (De) was performed with the Jasco Standard Analysis software. The protein secondary structure was calculated using CDSSTR, which calculates the different assignments of secondary structures by comparison with CD spectra, measured from different proteins for which high quality X-ray diffraction data are available [26,27]. The program CDSSTR is provided in CDPro software package which is available at the website: http:// lamar.colostate.edu/˜sreeram/CDPro.

Fluorescence Spectroscopy
Fluorimetric experiments were carried out on a Varian Cary Eclipse. Solutions containing drug 1 to 100 mM in Tris-HCl (pH = 7.4) were prepared at room temperature (2461uC). Solutions of HSA and BSA containing 10 mM in 10 mM Tris-HCl (pH = 7.4) were also prepared at 2461uC. The fluorescence spectra were recorded at l exc = 280 nm and l em from 287 to 500 nm. The intensity at 347 nm (tryptophan) was used to calculate the binding constant (K) according to previous literature reports [28][29][30][31][32][33].
On the assumption that there are (n) substantive binding sites for quencher (Q) on protein (B), the quenching reaction can be shown as following.
The binding constant (K A ), can be calculated as: The fluorescence intensity is proportional to the protein concentration as follows: Results from fluorescence measurements can be used to estimate the binding constant of drug-protein complex. From eq 4: The accessible fluorophore fraction (f) can be calculated by modified Stern-Volmer equation.
Where F 0 is initial fluorescence intensity and F is fluorescence intensities in the presence of quenching agent (or interacting molecule). K is the Stern-Volmer quenching constant, [Q] is the molar concentration of quencher and f is the fraction of accessible fluorophore to a polar quencher, which indicates the fractional fluorescence contribution of the total emission for an interaction with a hydrophobic quencher [17,18]. The plot of F 0/ (F 0 -F) vs 1/ [Q] yields f 21 as the intercept on y axis and (f K) 21 as the slope. Thus, the ratio of the ordinate and the slope gives K.

Molecular Modeling
The structure of BSA was predicted by automated homology modelling using SWISS-MODEL Workspace [34] from the amino acid sequence NP-851335 [35,36]. The structure of free HSA (PDB id: 1AO6, chain A) obtained by X-ray crystallography was used as a template [36]. These two proteins share 78.1% of sequence identity, which is sufficient to obtain reliable sequence alignment [37]. Images of the structures were generated using Pymol (DeLano Scientific, Palo Alto, CA, USA). RMSD between model and template proteins was 0.20 Å for positions of backbone atoms, as calculated with DeepView/Swiss-PdbViewer 4.0.1 (Fig. 1). The quality of the predicted BSA structure was found to be similar to the structure of free HSA used here as a template, using structure and model assessment tools of SWISS-MODEL workspace.
The docking studies were performed with ArgusLab 4.0.1 software (Mark A. Thompson, Planaria Software LLC, Seattle, Wa, http://www.arguslab.com). The structure of BSA was obtained from the above method and three dimensional structure of drug was generated from PM3 semi-empirical calculations, using Chem3D Ultra 6.0. A blind docking approach was taken as the whole protein was selected as a potential binding site. The docking runs were performed on the ArgusDock docking engine using high precision with a maximum of 200 candidate poses. The conformations were ranked using the Ascore scoring function, which estimates the free binding energy. Upon location of the potential binding sites, the docked complex conformations were optimized using a steepest decent algorithm until convergence, within 40 iterations. Amino acid residues within a distance of 3.5 Å relative to drug were considered involved in the complexation.

FTIR Spectra of Drug Complexes with BSA and HSA
The drug interactions with BSA and HSA were characterized by infrared spectroscopy and its derivative methods. The spectral shifting and intensity variations of protein amide I band at 1656-1655 cm 21 (mainly C = O stretch) and amide II band at 1547-1543 cm 21 (C-N stretching coupled with N-H bending modes) [24,33,38] were monitored upon drug interaction. The difference spectra [(protein solution + drug solution) -(protein solution)] were obtained, in order to monitor the intensity variations of these vibrations and the results are shown in Figures 2 and 3. Similarly, the infrared self-deconvolution with second derivative resolution enhancement and curve-fitting procedures [23] were used to determine the protein secondary structures in the presence of drug ( Figure 4 and Table 1).
At low drug concentration (0.125 mM), decrease of intensity was observed for the protein amide I at 1658-1656 and amide II at 1544-1543 cm 21  . The observed spectra shifting amide I and amide II bands are due to drug binding to protein C-O and C-N groups, while the decrease in the intensity of the amide I band in the spectra of the drug-protein complexes suggests a major reduction of protein a-helical structure at high drug concentrations [39].

CD Spectra
The conformational changes observed from infrared results for BSA and HSA and their drug complexes are consistent with CD spectroscopic analysis shown in Table 2. The CD results show that free BSA has a high a-helix content 59%, b-sheet 14%, turn 11% and random coil 16% (Table 2), consistent with the literature report [41]. The free HSA contains a-helix 54%, b-sheet 16%, turn 15% and random coil 15% (Table 2). Upon drug complexation, major reduction of a-helix was observed from 59% in free BSA to 53-50% in drug-BSA and from 54% to 50-45% in the drug-HSA complexes ( Table 2). The decrease in ahelix was accompanied by an increase in the b-sheet, turn and random coil structures ( Table 2). The major reduction of the ahelix with an increase in the b-sheet, turn and random structures are consistent with the infrared results, indicating a partial protein destabilization (Tables 1 and 2). However, the drugs induced more perturbations of HSA structure than that of BSA conformation (Table 1) consistent with more stable drug-HSA complexes than those of drug-BSA adducts, which will be discussed in fluorescence spectroscopy.

Fluorescence Spectra and Stability of Drug Complexes with BSA and HSA
HSA contains a single polypeptide of 585 amino acids with only one tryptophan (Trp-214) located in subdomain II A. BSA contains two tryptophan residues Trp-134 and Trp-212 located in the first and second domains of protein hydrophobic regions. Tryptophan emission dominates both HSA and BSA fluorescence spectra in the UV region. The decrease of fluorescence intensity of HSA and BSA has been monitored at 347 nm for drug-protein systems (Fig. 5A-D show representative results for each system). The plot of F 0/ (F 0 -F) vs 1/[drug] (Fig. 5A'-D' show representative plots for drug-protein complexes). Assuming that the observed changes in fluorescence come from the interaction between drug and HSA or BSA, the quenching constant can be taken as the  (Fig. 5A'-D'). The association constants calculated for the drug complexes suggest strong affinity for drug-protein binding, compared to the other ligand-protein adducts [42,43]. However, HSA forms more stable complexes than BSA due to major hydrophilic and hydrophobic drug-HSA interactions.
In order to verify the presence of static or dynamic quenching in drug-protein complexes we have plotted F 0 /F against Q and the results are show in Fig. 6. The plot of F 0 /F versus Q is linear for drug-BSA and drug-HSA adducts indicating that the quenching is mainly static in these drug-protein complexes [44]. The K q was estimated according to the Stern-Volmer equation: where F 0 and F are the fluorescence intensities in the absence and presence of quencher, [Q] is the quencher concentration and K D is the Stern-Volmer quenching constant (K q ), which can be written as K D = k q t 0 ; where k q is the bimolecular quenching rate constant and t 0 is the lifetime of the fluorophore in the absence of quencher, 5.9 ns for BSA and 5.6 ns for HSA [17]. The quenching constants (K q ) are 3.

Docking Studies
Our spectroscopic results were complemented with docking experiments in which doxorubicin and N-(trifluoroacetyl) doxorubicin were docked to BSA and HSA to determine the preferred binding sites for DOX and FDOX. The docking results shown in Fig. 8 and Table 3 Table 3). It is evident that several amino acids with hydrophobic and hydrophilic characters are in contact with DOX and FDOX in these BSA and HSA complexes ( Fig. 8 and Table 3). Molecular modeling presented in Table 3 showed major differences in the binding sites of DOX and FDOX with BSA and HSA (different amino acids are involved in drug-protein complexation). Docking results showed stronger drug-protein complex formed with HSA than BSA which is consistent with our spectroscopic results K DOX-HSA = 1.

Conclusion
Based on our spectroscopic and docking results, doxorubicin and N-(trifluoroacetyl) doxorubicin bind BSA and HSA via hydrophilic and hydrophobic contacts with more stable complexes formed with HSA than BSA. The drug-protein binding involves several amino acids residues, which stabilized by H-bonding network. Drug interaction alters protein secondary structure for both BSA and HSA causing a partial protein destabilization. It is clearly shown that HSA and BSA can transport DOX and FDOX in vitro. Table 3. Amino acid residues involved in drug-BSA and drug-HSA complexes with the free binding energy for the best selected docking positions.