Portal de Periódicos da CAPES

Crystal Structure Analysis of Warfarin Binding to Human Serum Albumin

2001; Elsevier BV; Volume: 276; Issue: 25 Linguagem: Inglês

10.1074/jbc.m100575200

1083-351X

I. Petitpas, Ananyo A. Bhattacharya, Sue Twine, Malcolm EAST, Stephen Curry,

Pharmacogenetics and Drug Metabolism

Human serum albumin (HSA) is an abundant transport protein found in plasma that binds a wide variety of drugs in two primary binding sites (I and II) and can have a significant impact on their pharmacokinetics. We have determined the crystal structures at 2.5 Å-resolution of HSA-myristate complexed with the R-(+) and S-(−) enantiomers of warfarin, a widely used anticoagulant that binds to the protein with high affinity. The structures confirm that warfarin binds to drug site I (in subdomain IIA) in the presence of fatty acids and reveal the molecular details of the protein-drug interaction. The two enantiomers of warfarin adopt very similar conformations when bound to the protein and make many of the same specific contacts with amino acid side chains at the binding site, thus accounting for the relative lack of stereospecificity of the HSA-warfarin interaction. The conformation of the warfarin binding pocket is significantly altered upon binding of fatty acids, and this can explain the observed enhancement of warfarin binding to HSA at low levels of fatty acid.1h9z1ha2 Human serum albumin (HSA) is an abundant transport protein found in plasma that binds a wide variety of drugs in two primary binding sites (I and II) and can have a significant impact on their pharmacokinetics. We have determined the crystal structures at 2.5 Å-resolution of HSA-myristate complexed with the R-(+) and S-(−) enantiomers of warfarin, a widely used anticoagulant that binds to the protein with high affinity. The structures confirm that warfarin binds to drug site I (in subdomain IIA) in the presence of fatty acids and reveal the molecular details of the protein-drug interaction. The two enantiomers of warfarin adopt very similar conformations when bound to the protein and make many of the same specific contacts with amino acid side chains at the binding site, thus accounting for the relative lack of stereospecificity of the HSA-warfarin interaction. The conformation of the warfarin binding pocket is significantly altered upon binding of fatty acids, and this can explain the observed enhancement of warfarin binding to HSA at low levels of fatty acid.1h9z1ha2 human serum albumin 2,3,5-triiodobenzoic acid Human serum albumin (HSA)1 is the major protein component of blood plasma but is also distributed to the interstitial fluid of the body tissues. The protein binds a number of relatively insoluble endogenous compounds such as unesterified fatty acids, bilirubin, and bile acids and thus facilitates their transport throughout the circulation (1Peters T. All about Albumin: Biochemistry, Genetics, and Medical Applications. Academic Press, San Diego1995Google Scholar, 2Kragh-Hansen U. Dan. Med. Bull. 1990; 37: 57-84PubMed Google Scholar). HSA is also capable of binding a wide variety of drugs (1Peters T. All about Albumin: Biochemistry, Genetics, and Medical Applications. Academic Press, San Diego1995Google Scholar, 2Kragh-Hansen U. Dan. Med. Bull. 1990; 37: 57-84PubMed Google Scholar, 3Carter D.C. Ho J.X. Adv. Protein Chem. 1994; 45: 152-203Google Scholar), and much of the interest in this abundant protein derives from its effects on drug delivery. Drug binding to plasma proteins such as HSA can be an important determinant of pharmacokinetics, restricting the unbound concentration and affecting distribution and elimination. In some cases the major fraction of the administered drug is sequestered by HSA; this is particularly true of warfarin, a widely used anticoagulant, which is 99% bound to the protein under normal therapeutic conditions and consequently has a small volume of distribution and low clearance (4Holford N.H.G. Benet L.Z. Katzung B.G. Basic and Clinical Pharmacology, 7th Ed. Appleton and Lange, Stamford, CT1998: 34-49Google Scholar). HSA has a limited number of binding sites for endogenous and exogenous ligands, so that drug binding to the protein may be affected by a variety of factors. Although the effects on pharmacokinetics of drug-drug competition for the same sites on HSA are generally held to be of little clinical import (4Holford N.H.G. Benet L.Z. Katzung B.G. Basic and Clinical Pharmacology, 7th Ed. Appleton and Lange, Stamford, CT1998: 34-49Google Scholar, 5Lindup W.E. Bridges J.W. Chasseaud L.F. Gibson G.G. Progress in Drug Metabolism. 10. Taylor & Francis, London1987: 141-185Google Scholar), physiological or diseased states that cause variations in the plasma levels of albumin or its primary endogenous ligands can influence drug binding and may require dosages to be closely monitored. Added to this, genetic polymorphisms in HSA can also alter drug binding and may further complicate the clinical picture (6Petersen C.E. Ha C.E. Harohalli K. Park D.S. Bhagavan N.V. Chem. Biol. Interact. 2000; 124: 161-172Crossref PubMed Scopus (35) Google Scholar, 7Wilding G. Blumberg E.S. Vesell E.S. Science. 1977; 195: 991-994Crossref PubMed Scopus (16) Google Scholar). To understand the molecular basis of these effects, structural information is required to fully delineate the binding sites for drugs and endogenous ligands. Such information will also be invaluable to efforts to exploit the carrier properties of HSA in the development of novel therapeutic reagents for drug targeting (8Beljaars L. Molema G. Schuppan D. Geerts A. De Bleser P.J. Weert B. Meijer D.K. Poelstra K. J. Biol. Chem. 2000; 275: 12743-12751Abstract Full Text Full Text PDF PubMed Scopus (129) Google Scholar) or oxygen transport (9Tsuchida E. Komatsu T. Hamamatsu K. Matsukawa Y. Tajima A. Yoshizu A. Izumi Y. Kobayashi K. Bioconjugate Chem. 2000; 11: 46-50Crossref PubMed Scopus (51) Google Scholar). HSA, a 585-residue protein, is monomeric but contains three structurally similar α-helical domains (I–III); each domain can be divided into subdomains A and B, which contain six and four α-helices, respectively (3Carter D.C. Ho J.X. Adv. Protein Chem. 1994; 45: 152-203Google Scholar, 10He X.M. Carter D.C. Nature. 1992; 358: 209-215Crossref PubMed Scopus (3566) Google Scholar). Structural studies have mapped the locations of the fatty acid binding sites (11Curry S. Mandelkow H. Brick P. Franks N. Nat. Struct. Biol. 1998; 5: 827-835Crossref PubMed Scopus (1210) Google Scholar, 12Bhattacharya A.A. Grüne T. Curry S. J. Mol. Biol. 2000; 303: 721-732Crossref PubMed Scopus (775) Google Scholar) and the primary drug binding sites on the protein (10He X.M. Carter D.C. Nature. 1992; 358: 209-215Crossref PubMed Scopus (3566) Google Scholar). The fatty acid binding sites are distributed throughout the protein and involve all six subdomains; by contrast many drugs bind to one of the two primary binding sites on the protein, known as Sudlow's sites I and II (13Sudlow G. Birkett D.J. Wade D.N. Mol. Pharmacol. 1975; 11: 824-832PubMed Google Scholar). Although examples of drugs binding elsewhere on the protein have been documented (13Sudlow G. Birkett D.J. Wade D.N. Mol. Pharmacol. 1975; 11: 824-832PubMed Google Scholar, 14Sjoholm I. Ekman B. Kober A. Ljungstedt-Pahlman I. Seiving B. Sjodin T. Mol. Pharmacol. 1979; 16: 767-777PubMed Google Scholar, 15Bhattacharya A.A. Curry S. Franks N.P. J. Biol. Chem. 2000; 275: 38731-38738Abstract Full Text Full Text PDF PubMed Scopus (501) Google Scholar), most work has focused on the primary drug sites. These investigations have largely employed competition binding methods to investigate the selectivity of the primary drug binding sites. Drug site I, where warfarin binds, has been characterized as a conformationally adaptable region with up to three subcompartments (16Fehske K.J. Schläfer U. Wollert U. Müller W.E. Mol. Pharmacol. 1981; 21: 387-393Google Scholar, 17Kragh-Hansen U. Mol. Pharmacol. 1988; 34: 160-171PubMed Google Scholar, 18Yamasaki K. Maruyama T. Kragh-Hansen U. Otagiri M. Biochim. Biophys. Acta. 1996; 1295: 147-157Crossref PubMed Scopus (242) Google Scholar). Crystallographic studies located drug sites I and II in subdomains IIA and IIIA, respectively, and provided the first detailed description of the geometry and chemistry of the binding environment (10He X.M. Carter D.C. Nature. 1992; 358: 209-215Crossref PubMed Scopus (3566) Google Scholar). Several reports have analyzed structural aspects of the binding of an aspirin analogue, 2,3,5-triiodobenzoic acid (TIB), to defatted HSA (10He X.M. Carter D.C. Nature. 1992; 358: 209-215Crossref PubMed Scopus (3566) Google Scholar), to the HSA-myristate complex (11Curry S. Mandelkow H. Brick P. Franks N. Nat. Struct. Biol. 1998; 5: 827-835Crossref PubMed Scopus (1210) Google Scholar), or to equine serum albumin (19Ho J.X. Holowachuk E.W. Norton E.J. Twigg P.D. Carter D.C. Eur. J. Biochem. 1993; 215: 205-212Crossref PubMed Scopus (100) Google Scholar). However, these studies were done at fairly modest resolutions (3.0–4.0 Å), and in any case it is not clear whether TIB, which carries three iodine atoms, is properly representative of typical HSA drug ligands. The binding of warfarin to HSA and the interactions of the drug with other HSA ligands have been investigated for many years. Warfarin (3-(α-acetonylbenzyl)-4-hydroxycoumarin) binds to drug site I with high affinity (Kd ≈ 3 μm) (13Sudlow G. Birkett D.J. Wade D.N. Mol. Pharmacol. 1975; 11: 824-832PubMed Google Scholar, 20Diana F.J. Veronich K. Kapoor A.L. J. Pharm. Sci. 1989; 78: 195-199Abstract Full Text PDF PubMed Scopus (61) Google Scholar,21Loun B. Hage D.S. Anal. Chem. 1994; 66: 3814-3822Crossref PubMed Scopus (255) Google Scholar). A crystallographic study in 1992 confirmed subdomain IIA as the binding locus for a single molecule of warfarin but was unable to give specific details on the binding interactions because of the limited resolution of the data (10He X.M. Carter D.C. Nature. 1992; 358: 209-215Crossref PubMed Scopus (3566) Google Scholar). Warfarin shares this binding site with a range of other drugs (including phenylbutazone, tolbutamide, and indomethacin) and thus competes with them for binding to HSA (13Sudlow G. Birkett D.J. Wade D.N. Mol. Pharmacol. 1975; 11: 824-832PubMed Google Scholar, 16Fehske K.J. Schläfer U. Wollert U. Müller W.E. Mol. Pharmacol. 1981; 21: 387-393Google Scholar,17Kragh-Hansen U. Mol. Pharmacol. 1988; 34: 160-171PubMed Google Scholar). Other studies have shown that low levels of fatty acids (22Vorum H. Honoré B. J. Pharm. Pharmacol. 1996; 48: 870-875Crossref PubMed Scopus (88) Google Scholar, 23Birkett D.J. Myers S.P. Sudlow G. Mol. Pharmacol. 1977; 13: 987-992PubMed Google Scholar) or elevated pH (24Wilting J. van der Giesen W.F. Janssen L.H. Weideman M.M. Otagiri M. Perrin J.H. J. Biol. Chem. 1980; 255: 3032-3037Abstract Full Text PDF PubMed Google Scholar, 25Wanwimolruk S. Birkett D.J. Biochim. Biophys. Acta. 1982; 709: 247-255Crossref PubMed Scopus (55) Google Scholar) may enhance the affinity of HSA for warfarin by up to 3-fold, but the molecular mechanisms of these effects are not well understood. To elucidate the molecular details of the interaction of warfarin with HSA, we report here the determination of the crystal structures of HSA-myristate complexed with both theR-(+) and S-(−) enantiomers of the drug. Recombinant HSA, produced in yeast and supplied by Delta Biotechnology Ltd. (Nottingham, UK), was purified by gel filtration, complexed with excess myristic acid, and crystallized as described previously (11Curry S. Mandelkow H. Brick P. Franks N. Nat. Struct. Biol. 1998; 5: 827-835Crossref PubMed Scopus (1210) Google Scholar, 12Bhattacharya A.A. Grüne T. Curry S. J. Mol. Biol. 2000; 303: 721-732Crossref PubMed Scopus (775) Google Scholar). Racemic warfarin, purchased from Sigma-Aldrich (A-2250) was separated into R-(+) and S-(−) enantiomers using established protocols (26West B.D. Preis S. Schroeder C.H. Link K.P. J. Am. Chem. Soc. 1961; 83: 2676-2679Crossref Scopus (151) Google Scholar). Crystals of HSA-myristate were harvested into solutions containing 37% (w/v) polyethylene glycol 3350, 50 mm potassium phosphate, pH 7.0. Warfarin was dissolved at 20 mm in harvest buffer at pH 12.0 with continuous stirring for 24 h; the pH was then restored to 7.0 with concentrated HCl. Crystals were soaked first in harvest buffer containing 10 mm R-(+) or S-(−) warfarin enantiomer for 15 h and then in buffer containing the same enantiomer (20 mm) for at least 24 h. X-ray diffraction data were collected at room temperature at station X-11 of the European Molecular Biology Laboratory Outstation at DESY, Hamburg, Germany (see Table I). The data were indexed and measured with MOSFLM. 2A. Leslie, personal communication. For both enantiomers, crystals of the HSA-myristate-warfarin complex were isomorphous with the HSA-myristate complex (11Curry S. Mandelkow H. Brick P. Franks N. Nat. Struct. Biol. 1998; 5: 827-835Crossref PubMed Scopus (1210) Google Scholar, 12Bhattacharya A.A. Grüne T. Curry S. J. Mol. Biol. 2000; 303: 721-732Crossref PubMed Scopus (775) Google Scholar). The protein model for this structure, stripped of its ligands, was used for initial phasing of the x-ray data. The model, split into its six subdomains, was initially refined using a rigid body protocol in CNS (27Brünger A.T. Adams P.D. Clore G.M. DeLano W.L. Gros P. Grosse-Kunstleve R.W. Jiang J.S. Kuszewski J. Nilges M. Pannu N.S. Read R.J. Rice L.M. Simonson T. Warren G.L. Acta Crystallogr. Sect. D Biol. Crystallogr. 1998; 54: 905-921Crossref PubMed Scopus (17028) Google Scholar) and then subjected to cycles of positional and B-factor refinement before calculation of initial Fo −Fc and 3Fo − 2Fc maps. These maps were used to guide the positioning of the fatty acid and warfarin ligands and bound waters and to make manual adjustments to the protein prior to further cycles of refinement. Details of the refinement statistics are summarized in Table I. Coordinates have been deposited with the Protein Data Bank (identification codes are given in Table I). Figures depicting the structure were prepared using BOBSCRIPT (28Esnouf R. J. Mol. Graph. 1997; 15: 133-138Google Scholar) and RASTER3D (29Merrit E.A. Bacon D.J. Methods Enzymol. 1997; 277: 505-524Crossref PubMed Scopus (3881) Google Scholar).Table IData collection and model refinement statisticsR-(+)S-(−)Data collectionHSA-myristate-warfarin complexa(Å)187.45187.54b(Å)38.8438.77c (Å)95.8895.80β (°)105.59105.38Resolution range (Å)30–2.530–2.5Independent reflections2309023058Multiplicity 1-aValues for the outermost resolution shell are given in parentheses.1.9 (2.0)2.1 (2.1)Completeness (%)98.4 (98.4)98.2 (98.2)I/ςI4.4 (1.8)4.9 (1.8)Rmerge(%) 1-bRmerge = 100 × ΣhΣj‖Ihj −Ih‖/ΣhΣj Ihjwhere Ih is the weighted mean intensity of the symmetry-related reflections Ihj.6.5 (31.4)6.0 (37.2)Model RefinementProtein Data Bank identification1h9z1ha2Number of nonhydrogen atoms47504750Number of water molecules2823Rmodel(%) 1-cRmodel = 100 × Σhkl‖Fobs −Fcalc‖/Σhkl Fobs whereFobs and Fcalc are the observed and calculated structure factors, respectively.20.420.0Rfree(%) 1-dRfree is theRmodel calculated using a randomly selected 5% sample of reflection data omitted from the refinement.25.424.8R.m.s. 1-eR.m.s., root mean square. deviation from ideal bond lengths (Å)0.0070.007R.m.s. deviation from ideal angles (°)1.21.2R.m.s. deviation in B-factors main/side chain (Å2)1.4/2.21.4/2.2Average B-factor (Å2)59.463.61-a Values for the outermost resolution shell are given in parentheses.1-b Rmerge = 100 × ΣhΣj‖Ihj −Ih‖/ΣhΣj Ihjwhere Ih is the weighted mean intensity of the symmetry-related reflections Ihj.1-c Rmodel = 100 × Σhkl‖Fobs −Fcalc‖/Σhkl Fobs whereFobs and Fcalc are the observed and calculated structure factors, respectively.1-d Rfree is theRmodel calculated using a randomly selected 5% sample of reflection data omitted from the refinement.1-e R.m.s., root mean square. Open table in a new tab HSA-myristate complexes were prepared and crystallized as described previously (11Curry S. Mandelkow H. Brick P. Franks N. Nat. Struct. Biol. 1998; 5: 827-835Crossref PubMed Scopus (1210) Google Scholar, 12Bhattacharya A.A. Grüne T. Curry S. J. Mol. Biol. 2000; 303: 721-732Crossref PubMed Scopus (775) Google Scholar, 30Carter D.C. Chang B. Ho J.X. Keeling K. Krishnasami Z. Eur. J. Biochem. 1994; 226: 1049-1052Crossref PubMed Scopus (234) Google Scholar). Over a period of at least 48 h the crystals were soaked in increasing concentrations of the warfarin R-(+) and S-(−) enantiomers (Fig. 1) up to a final concentration of 20 mm (see “Experimental Procedures”). The soaked crystals were isomorphous to crystals of HSA-myristate, and the structures were solved by molecular replacement. Difference electron density maps showed clear density for a single warfarin molecule in subdomain IIA in each case (Fig.2). The density was consistent with the open form of the warfarin molecule (31Valente E.J. Lingafelter E.C. Porter W.R. Trager W.F. J. Med. Chem. 1977; 20: 1489-1493Crossref PubMed Scopus (97) Google Scholar); features corresponding to the coumarin, benzyl, and acetonyl moieties of the drug were evident and gave a clear indication of the overall orientation of the drug in the pocket. For both R-(+) and S-(−) warfarin there nevertheless seemed to be a possible ambiguity in the orientation of the coumarin moiety. However, although the density appeared to allow the possibility of flipping the coumarin group by 180° about the C2-C13 bond (Figs. 1 and 2), attempts to refine the drug in the alternative orientation resulted in higher B-factors for the ligand, a poorer fit to 3Fo − 2Fc electron density maps, and significantly more residual difference density (data not shown). The density is thus consistent with a single dominant conformation of the bound drug for each of the enantiomers. Models forR-(+) and S-(−) warfarin bound to HSA-myristate were refined to 2.5 Å and have Rfree values of 25.4 and 24.8%, respectively, and good stereochemistry (TableI).Figure 2Experimental F o − F c difference electron density map for (A) R-(+) and (B) S-(−) warfarin in subdomain IIA, where F o values are the observed HSA-myristate-warfarin amplitudes, and F c values are the calculated amplitudes from the protein model from the HSA-myristate structure (11Curry S. Mandelkow H. Brick P. Franks N. Nat. Struct. Biol. 1998; 5: 827-835Crossref PubMed Scopus (1210) Google Scholar); the phases were derived from the same model. Oxygen atoms in the warfarin molecule are shown indark gray.View Large Image Figure ViewerDownload Hi-res image Download (PPT) The overall structures of the HSA-myristate-warfarin complexes are very similar to those reported for HSA-myristate and other HSA-fatty acid complexes (11Curry S. Mandelkow H. Brick P. Franks N. Nat. Struct. Biol. 1998; 5: 827-835Crossref PubMed Scopus (1210) Google Scholar, 12Bhattacharya A.A. Grüne T. Curry S. J. Mol. Biol. 2000; 303: 721-732Crossref PubMed Scopus (775) Google Scholar); there are only minor conformational changes in a number of side chains located at the drug binding site. As expected, warfarin binds within subdomain IIA in Sudlow's drug site I (Fig.3 A) (10He X.M. Carter D.C. Nature. 1992; 358: 209-215Crossref PubMed Scopus (3566) Google Scholar, 13Sudlow G. Birkett D.J. Wade D.N. Mol. Pharmacol. 1975; 11: 824-832PubMed Google Scholar, 32Bos O.J. Remijn J.P. Fischer M.J. Wilting J. Janssen L.H. Biochem. Pharmacol. 1988; 37: 3905-3909Crossref PubMed Scopus (81) Google Scholar) and displaces the myristate molecule that binds to this domain (in fatty acid site 7, FA7) with relatively low affinity (12Bhattacharya A.A. Grüne T. Curry S. J. Mol. Biol. 2000; 303: 721-732Crossref PubMed Scopus (775) Google Scholar). In addition to the drug bound in IIA, there is a fragment of difference density adjacent to the myristate bound to subdomain IB at the same location observed for triiodobenzoic acid (11Curry S. Mandelkow H. Brick P. Franks N. Nat. Struct. Biol. 1998; 5: 827-835Crossref PubMed Scopus (1210) Google Scholar). The density would be large enough to encompass the coumarin moiety of warfarin and may represent a secondary binding site for the drug, but there is no density for the other portions of the drug, and a second ligand molecule has therefore not been incorporated into the model. The binding pocket is formed by the packing of all six helices of subdomain IIA (Fig. 3 C). The R-(+) and S-(−) enantiomers bind in the pocket in almost identical conformations, both in the open configuration (31Valente E.J. Lingafelter E.C. Porter W.R. Trager W.F. J. Med. Chem. 1977; 20: 1489-1493Crossref PubMed Scopus (97) Google Scholar). The coumarin and benzyl moieties of theR-(+) and S-(−) forms are nearly perfectly superimposable, presumably because stabilization of the open conformation allows these groups to rotate about the C2-C13 bond. The main difference in the drug conformations occurs in the acetonyl group, which branches from the chiral carbon and is located at the mouth of the pocket. However, at 2.5 Å this difference is barely detectable in our electron density maps. The finding that the enantiomers bind in essentially the same way to HSA is consistent with the observation that they have similar binding affinities for the protein. At 25 °C, close to the temperature of our diffraction data collection, theR-(+) and S-(−) enantiomers have dissociation constants of 3.8 and 2.9 μm, respectively (21Loun B. Hage D.S. Anal. Chem. 1994; 66: 3814-3822Crossref PubMed Scopus (255) Google Scholar), although the difference is slightly more pronounced at lower temperatures (20Diana F.J. Veronich K. Kapoor A.L. J. Pharm. Sci. 1989; 78: 195-199Abstract Full Text PDF PubMed Scopus (61) Google Scholar,21Loun B. Hage D.S. Anal. Chem. 1994; 66: 3814-3822Crossref PubMed Scopus (255) Google Scholar). Given the ability of HSA to accommodate a wide range of different drug compounds in site I, it is perhaps surprising to find that warfarin fits quite snugly into the binding pocket (Fig. 3 C). The binding site has two sub-chambers that accommodate different portions of the warfarin molecule. The interaction between drug and protein appears to be dominated by hydrophobic contacts, but there are also a number of specific electrostatic interactions. The benzyl moiety binds in a sub-pocket formed by Phe-211, Trp-214, Leu-219, and Leu-238 with additional aliphatic contacts from Arg-218 and His-242 (Fig.3 B). Direct contact between the benzyl ring and the side chain of Trp-214 explains why modification of this residue reduces warfarin binding to HSA (33Fehske K.J. Muller W.E. Wollert U. Velden L.M. Mol. Pharmacol. 1979; 16: 778-789PubMed Google Scholar). The proximity of the benzyl ring to Arg-218 also accounts for the reduction in binding affinity associated with mutations at this position (6Petersen C.E. Ha C.E. Harohalli K. Park D.S. Bhagavan N.V. Chem. Biol. Interact. 2000; 124: 161-172Crossref PubMed Scopus (35) Google Scholar). The coumarin moiety fits into the main chamber furthest from the entrance, which is the same portion of site I that is occupied by TIB (11Curry S. Mandelkow H. Brick P. Franks N. Nat. Struct. Biol. 1998; 5: 827-835Crossref PubMed Scopus (1210) Google Scholar). Fig. 3 B shows that this chamber has an additional side pocket (delimited by Leu-219, Arg-222, Phe-223, Leu-234, Ile-264, Leu-257, and Ile-290) that is not occupied by warfarin but may accommodate hydrophobic portions of other site I drugs. The coumarin group makes primarily hydrophobic contacts with the surrounding side chains (TablesII andIII and Fig. 3 C). The roof and floor of the pocket are delimited by Ala-291 and Leu-238, respectively; the back end of the coumarin ring contacts Ile-260, Ile-264, Ile-290, and the aliphatic portions of Arg-257 and Ser-287 and lies close to but does not directly contact the fatty acid in the adjacent binding site (FA2). According to the view in Fig. 3 C, there are further hydrophobic contacts on the right flank from Val-241 and on the left from the aliphatic portion of Arg-222. In addition, two of the three oxygen atoms contribute to electrostatic interactions. The O4 atom of the hydroxyl group makes hydrogen bonds to the side chain of His-242 (2.9 Å) and to a bound water molecule (2.8 Å). On the other side of the drug, although the O1 atom faces a gap in the wall of the binding pocket and has no specific interaction with the protein, the O2 atom is positioned 3.7 Å from Nε of Arg-222. Closer to the pocket entrance the O3 of the acetonyl group lies 3.3 Å from NH2 of Arg-222. The combination of electrostatic interactions made by O2, O3, and O4 probably helps to fix the orientation of the bound warfarin, but complementarity between the shape of the drug and the pocket is clearly also important for binding. There is no significant difference in the electrostatic contacts made by the two enantiomers.Table IIPotential electrostatic interactions between warfarin and HSAWarfarin atomAmino acid contactDistance of closest approachR-(+)S-(−)ÅO2Arg-222 (Nɛ)3.63.9O3Arg-222 (NH2)3.43.2O4His-242 (Nɛ2)2.92.9O4Water 162.82.8 Open table in a new tab Table IIIVan der Waals interactions between warfarin and HSAWarfarin moietyAmino acidDistance of closest approachR-(+)S-(−)ÅCoumarinArg-2223.63.9Phe-2235.35.6Leu-2383.53.5Val-2414.94.6Arg-2574.04.2Ile-2603.73.8Ile-2644.34.2Ser-2873.83.8Ile-2904.14.0Ala-2913.73.8BenzylPhe-2113.83.9Trp-2143.83.6Ala-2153.64.1Arg-2184.44.1Leu-2193.93.8AcetonylArg-2223.43.2 Open table in a new tab Drug binding to HSA has been extensively investigated over the past 30 years. Much of this work has been performed using binding and competition assays to map out the number and selectivity of binding sites on the protein (13Sudlow G. Birkett D.J. Wade D.N. Mol. Pharmacol. 1975; 11: 824-832PubMed Google Scholar, 16Fehske K.J. Schläfer U. Wollert U. Müller W.E. Mol. Pharmacol. 1981; 21: 387-393Google Scholar, 17Kragh-Hansen U. Mol. Pharmacol. 1988; 34: 160-171PubMed Google Scholar, 18Yamasaki K. Maruyama T. Kragh-Hansen U. Otagiri M. Biochim. Biophys. Acta. 1996; 1295: 147-157Crossref PubMed Scopus (242) Google Scholar, 34Sudlow G. Birkett D.J. Wade D.N. Mol. Pharmacol. 1976; 12: 1052-1061PubMed Google Scholar), although more recently, mutagenesis techniques have been applied to the study of drug interactions (6Petersen C.E. Ha C.E. Harohalli K. Park D.S. Bhagavan N.V. Chem. Biol. Interact. 2000; 124: 161-172Crossref PubMed Scopus (35) Google Scholar, 35Dockal M. Chang M. Carter D.C. Ruker F. Protein Sci. 2000; 9: 1455-1465Crossref PubMed Scopus (125) Google Scholar). Although initial crystallographic studies at moderate resolution have focused mainly on the binding of TIB, a drug analogue, to HSA (10He X.M. Carter D.C. Nature. 1992; 358: 209-215Crossref PubMed Scopus (3566) Google Scholar, 11Curry S. Mandelkow H. Brick P. Franks N. Nat. Struct. Biol. 1998; 5: 827-835Crossref PubMed Scopus (1210) Google Scholar), there has been a lack of high resolution structural information with which to interpret the amassed biochemical and biophysical data on HSA-drug interactions. The structures reported in this paper provide the first high resolution view of what might reasonably be termed a “classic” site I drug bound to HSA. Although many different drugs bind to site I, they are generally bulky heterocyclic compounds with a negative charge located toward the center of the molecule (34Sudlow G. Birkett D.J. Wade D.N. Mol. Pharmacol. 1976; 12: 1052-1061PubMed Google Scholar). Warfarin exemplifies this type of compound, and the structure of the HSA-myristate-warfarin complex shows how this broad selectivity is achieved. The coumarin and benzyl moieties are accommodated in separate but adjacent chambers of what has previously been described as a “sock-shaped” pocket (10He X.M. Carter D.C. Nature. 1992; 358: 209-215Crossref PubMed Scopus (3566) Google Scholar). The polar features of the drug are all found at the basic mouth of the pocket, with the exception of the hydroxyl oxygen (O4), which interacts specifically with the side chain of His-242 on one wall of the pocket (Fig. 3 C). Both enantiomers of warfarin bind in essentially the same way to the protein. This accounts for the poor stereoselectivity of albumin and suggests that the prospects for developing it as a reagent for chiral separations of warfarin are unfavorable. Conversely, the structure clearly reveals the details of the HSA-warfarin interaction and provides a structural basis for the possible design of warfarin derivatives with altered HSA binding properties. The binding of several warfarin derivatives to HSA has already been characterized; acenocoumarin binds more loosely than warfarin, and phenprocoumon binds more tightly (16Fehske K.J. Schläfer U. Wollert U. Müller W.E. Mol. Pharmacol. 1981; 21: 387-393Google Scholar, 22Vorum H. Honoré B. J. Pharm. Pharmacol. 1996; 48: 870-875Crossref PubMed Scopus (88) Google Scholar, 34Sudlow G. Birkett D.J. Wade D.N. Mol. Pharmacol. 1976; 12: 1052-1061PubMed Google Scholar). The weaker association of acenocoumarin probably occurs because the addition of an NO2 group to the benzyl ring sterically hinders its accommodation in the hydrophobic compartment formed primarily by Phe-211, Trp-214, Leu-219, and Leu-238. By contrast the tighter binding of phenprocoumon, in which the CH2-CO-CH3 acetonyl group found on warfarin is replaced by a propyl moiety (C3H7), may be due to the formation of hydrophobic contacts between this propyl group and the side chains of Trp-214 and Arg-218 (Fig. 3 C). Comparison with the structure of HSA-myristate-TIB (11Curry S. Mandelkow H. Brick P. Franks N. Nat. Struct. Biol. 1998; 5: 827-835Crossref PubMed Scopus (1210) Google Scholar) reveals differences in the ways that TIB and warfarin bind to site I. The heterocyclic coumarin moiety of warfarin binds in the same location as the benzyl ring of TIB. Although the planes of the aromatic rings are inclined at about 20° to each other, they make many of the same interactions with the walls of this part of the pocket, and both drugs can also from specific hydrogen bonds with His-242. The presence of bulky iodine atoms projecting from the TIB benzyl group make it too wide to fit into the sub-pocket that accommodates the benzyl ring of warfarin. Our structural results offer a plausible explanation for the observation that binding of fatty acids to HSA can enhance the affinity of the protein for warfarin and some other site I ligands (22Vorum H. Honoré B. J. Pharm. Pharmacol. 1996; 48: 870-875Crossref PubMed Scopus (88) Google Scholar, 23Birkett D.J. Myers S.P. Sudlow G. Mol. Pharmacol. 1977; 13: 987-992PubMed Google Scholar, 34Sudlow G. Birkett D.J. Wade D.N. Mol. Pharmacol. 1976; 12: 1052-1061PubMed Google Scholar). The presence of up to 3 mol of long chain fatty acids per mol of HSA can increase the affinity for warfarin by almost a factor of three (22Vorum H. Honoré B. J. Pharm. Pharmacol. 1996; 48: 870-875Crossref PubMed Scopus (88) Google Scholar, 23Birkett D.J. Myers S.P. Sudlow G. Mol. Pharmacol. 1977; 13: 987-992PubMed Google Scholar). Fatty acid binding to HSA induces a substantial conformational change in the protein involving rotations of domains I and III relative to domain II (11Curry S. Mandelkow H. Brick P. Franks N. Nat. Struct. Biol. 1998; 5: 827-835Crossref PubMed Scopus (1210) Google Scholar, 12Bhattacharya A.A. Grüne T. Curry S. J. Mol. Biol. 2000; 303: 721-732Crossref PubMed Scopus (775) Google Scholar, 36Curry S. Brick P. Franks N.P. Biochim. Biophys. Acta. 1999; 1441: 131-140Crossref PubMed Scopus (466) Google Scholar). One of the principal driving forces behind this conformational change is binding of a fatty acid molecule to the binding site (FA2) that spans the interface between subdomains IA and IIA and lies close to the warfarin binding site. In the absence of fatty acid the side chain of Tyr-150 from the linker connecting helices h2 and h3 of subdomain IB projects into the warfarin binding site (Fig.3 D). However, occupation of site FA2 by myristate displaces domain IB, rotating helices h2 and h3 and pulling the side chain of Tyr-150 out of the drug pocket to allow it to hydrogen-bond with the carboxylate group of the fatty acid (FA2). In concert with the movement of Tyr-150 the side chain of Arg-257 rotates to hydrogen-bond to the same fatty acid carboxylate. The changes induced by fatty acid binding, removal of a hydroxyl group from the drug pocket and repositioning of the aliphatic portion of Arg-257, thus render the binding environment for the coumarin moiety more hydrophobic. This appears to be the most likely explanation for the increased binding affinity of HSA for warfarin. Curiously, the binding of phenprocoumon is only marginally (∼20%) enhanced under similar conditions (22Vorum H. Honoré B. J. Pharm. Pharmacol. 1996; 48: 870-875Crossref PubMed Scopus (88) Google Scholar). This suggests that minor conformational changes induced by fatty acid binding in the vicinity of the mouth of the pocket may also affect drug binding. The enhancement of warfarin binding by fatty acids is maximal with 3–4 mol of fatty acid bound per mol of HSA (22Vorum H. Honoré B. J. Pharm. Pharmacol. 1996; 48: 870-875Crossref PubMed Scopus (88) Google Scholar, 23Birkett D.J. Myers S.P. Sudlow G. Mol. Pharmacol. 1977; 13: 987-992PubMed Google Scholar). At higher fatty acid concentrations the affinity for warfarin falls off, an observation that is probably attributable to direct competition between the drug and fatty acid for binding to the pocket in IIA. This interpretation is supported by the observation that fatty acid binds to this site (designated FA7) in a manner that is suggestive of a low affinity interaction (primarily because of the absence of specific interactions between the protein and the fatty acid carboxylate group) (12Bhattacharya A.A. Grüne T. Curry S. J. Mol. Biol. 2000; 303: 721-732Crossref PubMed Scopus (775) Google Scholar). Thus we would only expect to find displacement of the drug by high levels of fatty acid. Over the pH range 6–9 HSA undergoes a transition from the neutral to the basic form that is linked to an increase in the affinity of the protein for a number of site I ligands, including warfarin (which binds about 3-fold more tightly at alkaline pH) (24Wilting J. van der Giesen W.F. Janssen L.H. Weideman M.M. Otagiri M. Perrin J.H. J. Biol. Chem. 1980; 255: 3032-3037Abstract Full Text PDF PubMed Google Scholar). The precise nature of the structural changes associated with the neutral-basic transition is incompletely characterized. The transition appears largely to involve changes in domains I and II, although there is some involvement of domain III (32Bos O.J. Remijn J.P. Fischer M.J. Wilting J. Janssen L.H. Biochem. Pharmacol. 1988; 37: 3905-3909Crossref PubMed Scopus (81) Google Scholar). It remains unclear whether there is any link between the structural changes induced by shifts in pH and those induced by fatty acid binding, although the observation that fatty acid binding lowers the pH of the midpoint of the neutral-basic transition suggests a possible parallel between the two effects (25Wanwimolruk S. Birkett D.J. Biochim. Biophys. Acta. 1982; 709: 247-255Crossref PubMed Scopus (55) Google Scholar). In view of the paucity of data on this point, attempts to explain the pH effect on warfarin binding to HSA from the crystal structure are problematic. Our results show that the only group in the binding site that is likely to be titratable in the range pH 6–9 is the side chain of His-242, which makes a hydrogen bond to the drug. However, raising the pH would deprotonate the histidine, which would be expected to reduce the strength of its interaction with warfarin, contrary to the results of binding experiments. It thus appears that other effects of pH upon the protein are responsible for tighter binding at alkaline pH, and further work will be required to determine what these are. Recently, methods to study structural aspects of drug binding using engineered fragments of HSA have been developed in an effort to make the analysis more tractable (35Dockal M. Chang M. Carter D.C. Ruker F. Protein Sci. 2000; 9: 1455-1465Crossref PubMed Scopus (125) Google Scholar, 37Mao H. Gunasekera A.H. Fesik S.W. Protein Expression Purif. 2000; 20: 492-499Crossref PubMed Scopus (23) Google Scholar). Although such approaches are likely to be very valuable, particularly because they use fragments small enough to be analyzed structurally by nuclear magnetic resonance techniques, truncations of the protein may give rise to inadvertent structural perturbations of the drug binding sites. This study and an earlier structural analysis of HSA complexed with the general anesthetics propofol and halothane (15Bhattacharya A.A. Curry S. Franks N.P. J. Biol. Chem. 2000; 275: 38731-38738Abstract Full Text Full Text PDF PubMed Scopus (501) Google Scholar) demonstrate the feasibility of high resolution crystallographic studies of the binding of drugs to the intact protein and show that the structural information obtained provides a richly detailed view of the binding site with which to interpret the biochemical and biophysical data already accumulated on drug interactions with the protein. Further work on other site I and site II drugs will allow us to build up a much more complete picture of drug interactions with HSA and will provide a structural basis for a more rational approach to drug design either to exploit or avoid the impact of HSA on drug delivery.