Crystallographic Analysis Reveals Common Modes of
`Binding of Medium and Long-chain Fatty Acids to
`Human Serum Albumin
`
`AnanyoA.Bhattacharya,TimGru¨neandStephenCurry*
`
`Biophysics Section, Blackett
`Laboratory, Imperial College of
`Science, Technology and
`Medicine, London SW7 2BW
`UK
`
`Human serum albumin (HSA) is an abundant plasma protein that is
`responsible for the transport of fatty acids. HSA also binds and perturbs
`the pharmacokinetics of a wide range of drug compounds. Binding
`studies have revealed significant
`interactions between fatty acid and
`drug-binding sites on albumin but high-resolution structural information
`on ligand binding to the protein has been lacking. We report here a crys-
`tallographic study of five HSA-fatty acid complexes formed using satu-
`rated medium-chain and long-chain fatty acids (C10:0, C12:0, C14:0,
`C16:0 and C18:0). A total of seven binding sites that are occupied by all
`medium-chain and long-chain fatty acids have been identified, although
`medium-chain fatty acids are found to bind at additional sites on the pro-
`tein, yielding a total of 11 distinct binding locations. Comparison of the
`different complexes reveals key similarities and significant differences in
`the modes of binding, and serves to rationalise much of the biochemical
`data on fatty acid interactions with albumin. The two principal drug-
`binding sites, in sub-domains IIA and IIIA, are observed to be occupied
`by fatty acids and one of them (in IIIA) appears to coincide with a high-
`affinity long-chain fatty acid binding site.
`
`# 2000 Academic Press
`
`*Corresponding author
`
`Keywords: crystal structure; fatty acid binding; lipid; human serum
`albumin
`
`Introduction
`
`Long-chain fatty acids are required for the syn-
`thesis of membrane lipids, hormones and second
`messengers, and serve as an important source of
`metabolic energy. Fatty acids are stored as triacyl-
`glycerols in adipose tissue and released into the
`circulation where their low aqueous solubilities
`(typically < 1 mM) are overcome by serum albumin,
`an abundant plasma protein. Human serum albu-
`min (HSA) greatly enhances the transport capacity
`of plasma, since it is present at around 0.6 mM
`and can carry at least six molecules of fatty acid.
`Under normal physiological conditions, HSA car-
`ries around 0.1-2 mol of fatty acid per mol protein
`(Fredricksonetal.,1958).
`
`HSA is capable of binding an extraordinarily
`broad range of drugs, and much of the clinical and
`pharmaceutical interest in the protein derives from
`itseffectsondrugpharmacokinetics(Robertson&
`Brodersen,1991;Duboisetal.,1993;Jakobyetal.,
`´,1996;Itohetal.,1997;
`1995;Vorum&Honore
`Demant&Friche,1998;Mollaetal.,1998).Pioneer-
`ing work by Sudlow found two primary drug-
`binding sites on the protein, named sites I and II
`(Sudlowetal.,1975,1976).Manysubsequentstu-
`dies have shown that the presence of fatty acids
`has unpredictable effects on drug binding, and
`both co-operative and competitive interactions
`havebeenobserved(Vallner,1977;Birkettetal.,
`1978;Wanwimolruketal.,1983;Ivarsen&
`Brodersen,1989;Brodersenetal.,1990;Vorum&
`Honore´,1996;Curryetal.,1998).
`The binding of fatty acids to serum albumin has
`beenstudiedforover40years(Carter&Ho,1994;
`Peters,1995)butourunderstandingoftheseinter-
`actions is far from complete. Although it is now
`well established that the protein has multiple fatty
`acidbindingsitesofvaryingaffinities(Kragh-
`Hansen,1990;Carter&Ho,1994;Peters,1995),the
`
`

`

`722
`
`Fatty Acid Binding to HSA
`
`precise number of binding sites is not known; a
`current consensus seems to be that there are two or
`three high-affinity sites on the protein and at least
`threefurthersitesofloweraffinity(Carter&Ho,
`1994;Peters,1995).Avarietyofbiochemicaland
`biophysical methods have been used to investigate
`the nature and locations of fatty acid binding sites
`and many significant insights have accrued. The
`mixed electrostatic and hydrophobic nature of the
`binding
`interaction
`has
`been
`characterised
`(Reynoldsetal.,1968;Parksetal.,1983;Cistola
`etal.,1987b)andithasbeenshownthatatseveral
`binding sites the carboxylate head-group of the
`bound fatty acid is more rigidly anchored than the
`methylenetail(Hamiltonetal.,1984).Moreover,
`the domain locations of some sites have been
`mapped(Reedetal.,1975;Hamiltonetal.,1991),
`and there are clear indications that the primary
`sites for medium-chain and long-chain fatty acids
`aredistinct(Means&Bender,1975;Soltys&Hsia,
`1978).However,progressinunderstandingfatty
`acid binding has been hampered, due largely to
`the complexity of the multiple binding interactions
`but also to the absence of structural information.
`Almost all of the methods used to identify binding
`sites have resorted to some form of modification of
`theprotein(Reedetal.,1975;Shaklaietal.,1984;
`Hamiltonetal.,1991)ortheligand(Sklaretal.,
`1977;Berdeetal.,1979;Reed,1986),whichinevita-
`bly complicated interpretation of the results. Only
`recently
`have
`structural
`and mutational
`approaches been applied to the analysis of ligand
`binding to HSA.
`The first
`crystallographic analyses of HSA
`revealed that the protein, a 585 amino acid residue
`monomer, contains three homologous a-helical
`domains(I-III)(He&Carter,1992;Carter&Ho,
`1994).Thedomainseachcontaintenhelicesand
`are divided into six-helix and four-helix sub-
`domains (A and B); the first four helices of A and
`B form similar anti-parallel a-helix bundles. These
`investigations found that drug-binding sites I and
`II on HSA are located in sub-domains IIA and
`IIIA,respectively(He&Carter,1992;Hoetal.,
`1993).Wesubsequentlyreportedthecrystalstruc-
`ture of HSA complexed with a medium-chain fatty
`acid(myristicacid,C14:0)(Curryetal.,1998,1999),
`which brought to light the precise locations of six
`fatty acid binding sites. This study also uncovered
`a previously undetected binding site for drug com-
`pounds,
`immediately adjacent to a bound fatty
`
`acidmoleculeinsub-domainIA(Curryetal.,
`1998).
`found in appreciable
`However, C14:0 is not
`quantitiescirculatinginplasma(Saifer&Goldman,
`1961)exceptundercertainconditionsofdiseaseor
`clinicaltreatment(Bach&Babayan,1982;Babayan,
`1987;Bougne `resetal.,1989).Therefore,toassess
`the generality of our findings from the structure of
`the HSA-C14:0 complex, we have extended our
`analysis and present the crystal structures of four
`new HSA-fatty acid complexes. These include com-
`plexes with the saturated fatty acids C10:0, C12:0,
`C16:0 and C18:0 (capric, lauric, palmitic and stearic
`acid,
`respectively;
`in the Cm:n nomenclature
`adopted for this work, m gives the number of
`carbon atoms in the methylene tail and n is the
`number of double bonds). We have re-refined
`the structure of the HSA-C14:0 complex using a
`more complete high-resolution data set than that
`available previously.
`
`Results
`
`Structure determination
`
`HSA-fatty acid complexes were prepared by
`incubating the protein with a large molar excess of
`fattyacid(seeMaterialsandMethods;Table1).
`Five different complexes were prepared in the
`course of the present work using C10:0, C12:0,
`C14:0, C16:0 and C18:0 fatty acids. The complexes
`all crystallized isomorphously with HSA-C14:0
`(Curryetal.,1998)andthestructuresweresolved
`by molecular replacement. Difference electron den-
`sity maps showed clear density for multiple bound
`fatty acids in each case. At most sites (exceptions
`are noted below) the shape of the density, which
`had a significant broadening at one end, clearly
`indicated the position of the carboxylate moiety
`and thus defined the orientation of
`the bound
`lipid. This usually placed the carboxylate group of
`the fatty acid adjacent to an amino acid side-chain
`with which it could make a salt-bridge or hydro-
`gen bond. Models for each HSA-fatty acid complex
`were constructed and refined to resolutions of
`2.44-2.70 A˚ ; the models have Rfree values in the
`range 25.8-28.8 % and reasonable stereochemistry
`(Table2).Whileitwasnotpossibleattheseresol-
`utions to assign the dihedral angles within the
`methylene tails of the bound fatty acids precisely,
`the overall shapes of the tails were well-defined.
`Average B-factors for the different models are rela-
`tively high, ranging from 52-61 A˚ 2; sub-domain
`
`Table 1. Summary of co-crystallisation conditions
`
`Process
`
`C10:0
`
`C12:0
`
`C14:0
`
`C16:0
`
`C18:0
`
`Fatty acid (mM)
`HSA:FA molar ratio
`Fatty acid (mM)
`PEG 3350 (%, w/v)
`PEG 3350 (%, w/v)
`
`Complexing
`Complexing
`Washing
`Crystallisation
`Harvest
`
`10
`>40
`5.0
`28
`35
`
`5
`30
`0.5
`28
`35
`
`2.4
`12
`0.1
`28
`35
`
`2.5
`20
`0.1
`28
`35
`
`0.8
`>20
`0.05
`31
`34
`
`MPI EXHIBIT 1060 PAGE 2
`
`MPI EXHIBIT 1060 PAGE 2
`
`

`

`Fatty Acid Binding to HSA
`
`723
`
`Table 2. Data collection and model refinement statistics
`
`C10:0
`
`C12:0
`
`C14:0
`
`C16:0
`
`C18:0
`
`HSA-fatty acid complex
`
`A. Data collection
`Unit cell dimensions
`a (A˚ )
`b (A˚ )
`c (A˚ )
`b(deg.)
`Sourcea
`Resoution range (A˚ )
`Independent reflections
`Multiplicity
`Completeness(%) b
`I/sI
`Rmerge (%)c
`
`191.84
`39.02
`95.88
`105.12
`
`188.50
`186.78
`38.90
`39.19
`95.77
`95.34
`104.63
`105.15
`D-9.6H-X31D-9.5H-X31D-9.6
`13.0-2.5
`37.5-2.5
`12.0-2.45
`23210
`22984
`23657
`1.9
`3.2
`1.5
`97.5(100)92.7(83.6)96.2(92.4)92.9(87.8)93.7(96.8)
`5.3(2.5)
`9.5(3.1)
`3.7(3.0)
`7.0(27.3)3.8(22.6)5.7(22.4)4.6(17.8)9.0(24.8)
`
`190.09
`38.75
`95.87
`104.96
`
`12.0-2.44
`24122
`2.7
`
`9.4(4.0)
`
`1e7h
`4554
`29
`
`0.006
`1.21
`1.7/2.7
`
`189.60
`38.84
`95.98
`105.49
`
`40.0-2.7
`17689
`2.5
`
`4.5(2.4)
`
`1e7i
`4639
`26
`
`0.007
`1.16
`1.7/2.6
`
`B. Model refinement
`RCSB PBD ID
`Number of non-hydrogen atoms
`Number of water molecules
`Rmodel (%)d
`Rfree (%)e
`r.m.s deviation from ideal bond lengths (A˚ )
`r.m.s deviation from ideal angles (deg.)
`r.m.s. deviation in B-factors main/side-chain (A˚ 2)
`
`1e7f
`1e7e
`4478
`4561
`26
`31
`22.022.522.521.420.4
`27.127.628.827.225.9
`0.006
`0.007
`1.16
`1.23
`1.7/2.6
`1.7/2.6
`
`1e7g
`4598
`16
`
`0.006
`1.10
`1.7/2.7
`
`a D, Sychrotron Radiation Source, Daresbury, UK; H, DESY, Hamburg, Germany.
`b Values for the outermost resolution shell are given in parentheses.
`c Rmerge (cid:136) 100 (cid:6)h(cid:6)jjIhj (cid:255) Ihj/(cid:6)h(cid:6)jIhj, where Ih is the weighted mean intensity of the symmetry-related reflections Ihj.
`d Rmodel (cid:136) 100 (cid:6)hkljFobs (cid:255) Fcalcj/(cid:6)hklFobs, where Fobs and Fcalc are the observed and calculated structure factors, respectively.
`e Rfree is the Rmodel calculated using a randomly selected 5 % sample of reflection data omitted from the refinement.
`
`IIIB consistently exhibits higher
`B-factors (65-80 A˚ 2).
`
`than average
`
`ium-chain fatty acids (C10:0 to C14:0), the data
`indicate the presence of further sites (see below).
`
`Overview of fatty acid binding
`
`For the four new HSA-fatty acid complexes in
`this study, the protein conformations are very simi-
`lar to the structure of HSA-C14:0, which has been
`describedindetailelsewhere(Curryetal.,1998,
`1999).TheHSA-fattyacidcomplexesalldisplay
`the conformational change from the defatted HSA
`structure(He&Carter,1992;Sugioetal.,1999)that
`wasobservedpreviouslyforHSA-C14:0(Curry
`etal.,1998).Thisconformationalchangeinvolves
`substantial rotations of domains I and III relative
`to the central domain II, and appears to be primar-
`ily driven by fatty acid binding to the site located
`at the juncture of sub-domains IA and IIA.
`Our original study on the HSA-C14:0 complex
`positively identified five binding sites for the fatty
`acid. These consist of a single site in sub-domain
`IB, one at the interface between sub-domains IA
`and IIA, two sites in IIIA (Sudlow’s drug-binding
`siteII)andafifthinIIIB(Figure1).AsixthC14:0-
`binding site was suggested, on the basis of frag-
`mentary electron density, to lie at the interface
`between sub-domains IIA and IIB. The present
`results show that all six sites identified for C14:0
`are occupied by all the fatty acids in our study
`(C10:0,C12:0,C16:0andC18:0)(Figure1).In
`addition, a seventh site has been identified within
`the drug-binding cavity of
`sub-domain IIA
`(Sudlow’s drug-binding site I). Moreover, for med-
`
`Common binding sites for medium and
`long-chain fatty acids
`
`In the discussion of fatty acid binding sites, we
`have adopted and extended the site-numbering
`schemethatwasestablishedforHSA-C14:0(Curry
`etal.,1998).
`
`Site 1
`
`This site lies in a D-shaped cavity in the centre
`ofthefour-helixbundleofsub-domainIB
`(Figure2).Incomparisontotheothermajorfatty
`acid binding pockets on HSA, this site is relatively
`open and accessible to solvent. There is clear
`electron density for all
`the medium-chain fatty
`acids (C10:0-C14:0) in their respective complexes
`but the density becomes progressively weaker and
`more broken for longer fatty acids. In the HSA-
`C18:0 complex, the density for the bound fatty acid
`is especially weak at this site and the side-chain of
`Tyr138, which lines the binding cavity and contacts
`the ligand, also appears disordered. In the absence
`of
`fatty acid, Tyr138 stacks with Tyr161 and
`occludes the binding pocket but upon fatty acid
`binding at this site, the tyrosine side-chains rotate
`through about 90 (cid:14) so that their phenyl rings hold
`thelipidinahydrophobicclamp(Curryetal.,
`1999)(Figure2).ThedisorderingofTyr138inthe
`HSA-C18:0 structure is an indication that this site
`
`MPI EXHIBIT 1060 PAGE 3
`
`MPI EXHIBIT 1060 PAGE 3
`
`

`

`724
`
`Fatty Acid Binding to HSA
`
`Figure 1. Structures of HSA complexed with five different fatty acids. The protein secondary structure is shown
`schematically and the domains are colour-coded as follows: I, red; II, green, III, blue. The A and B sub-domains are
`depicted in dark and light shades, respectively. This colour scheme is maintained throughout. Bound fatty acids are
`shown in a space-filling representation and coloured by atom type (carbon, grey; oxygen, red). Where two fatty acid
`molecules bind in close proximity, one of them is shown in a darker shade of grey. All Figures were prepared using
`Bobscript(Esnouf,1997)andRaster3D(Merrit&Bacon,1997).
`
`the
`is not well occupied. The observation that
`weakening of the ligand density at site 1 is pro-
`gressive with increasing chain length for saturated
`fatty acids suggests that the presumed increase in
`binding affinity at this site as the ligand becomes
`longerandmorehydrophobic(Ashbrooketal.,
`1975)doesnotfullycompensatefortheirlower
`aqueous solubilities. Although C18:0 may bind to
`this site at least as tightly as C16:0, its solubility
`may be too low to allow full occupancy; such ‘‘cut-
`off’’ effects have been observed for the protein
`binding of other series of aliphatic compounds
`(Franks&Lieb,1985;Curryetal.,1990).
`All the fatty acids bind at site 1 in the same
`orientation with the carboxylate group hydrogen-
`bonded to Arg117 and to a water molecule that is
`also coordinated by the side-chain hydroxyl group
`of Tyr161 and the carbonyl oxygen atom of
`Leu182. For the longer-chain saturated fatty acids,
`the tail curls around the inside surface of the cavity
`so that the tip of the hydrophobic tail gradually
`approaches His146 at the lower end of the cavity
`opening(Figure2).Thissuggestsanexplanation
`for
`the finding that C16:0
`and C18:0 but
`not C8:0 can block the covalent modification of
`the equivalent His
`in bovine serum albumin
`(BSA)byN-dansylaziridine(Brown&Shockley,
`1982).
`
`Site 2
`
`Located between sub-domains IA and IIA, site 2
`is one of the most enclosed fatty acid binding sites
`on HSA, since even the carboxylate moiety of the
`ligandislargelyshieldedfromsolvent(Figure3(a)).
`Binding of fatty acids at this site is implicated in
`driving the conformational change observed with
`the uptake of ligand because this site is dislocated
`intotwohalf-sitesindefattedHSA(Curryetal.,
`1998,1999);theformationofacontiguouspocket
`to accommodate the fatty acid requires rotation of
`domain I relative to domain II and appears to be
`stabilized by ligand binding. In all cases, the car-
`boxylate head-groups of fatty acids bound to site 2
`are anchored in sub-domain IIA by hydrogen bond
`interactions to the side-chains of Tyr150, Arg257
`andSer287(Figure3(a)).Themethylenetails
`extend linearly within the narrow hydrophobic
`cavity formed by the realignment of IA and IIA;
`the longest fatty acid tails extend furthest into this
`cavity. There is good electron density for all the
`fatty acids (C10:0-C18:0) at this site, indicating that
`even fatty acids with just ten carbon atoms in the
`methylene tail are long enough to pin the two half-
`sites together as a binding pocket. Curiously, we
`have been unable to co-crystallise HSA with C8:0
`in a form that is isomorphous with crystals of the
`other HSA-fatty acid complexes. Simple modelling
`
`MPI EXHIBIT 1060 PAGE 4
`
`MPI EXHIBIT 1060 PAGE 4
`
`

`

`Fatty Acid Binding to HSA
`
`725
`
`Figure 2. Fatty acid binding to site 1 in sub-domain
`IB.TheproteiniscolouredasinFigure1.Thefatty
`acids are coloured using a rainbow scheme: C10:0, red;
`C12:0, orange; C14:0, yellow; C16:0, green; C18:0, blue
`(this
`colour
`scheme
`is maintained in subsequent
`Figures). Selected amino acid side-chains and a bound
`water molecule are shown coloured by atom type.
`
`experiments suggest that this may be because C8:0
`is too short for its methylene tail to make the stabi-
`lising contacts with sub-domain IA that are
`required for the conformational change.
`In the case of C12:0, the electron density map
`showed clear evidence for the presence of a second
`fatty acid molecule binding in the upper part of
`the pocket within IA (designated 20), its tail over-
`
`lapping in an anti-parallel fashion with the first
`lipidmoleculeinsite2(Figure3(b)).Examination
`of a map for the HSA-C14:0 complex calculated
`from a new high-resolution data set (resolution
`2.5A˚ ,Table2)confirmedthatasecondC14:0mol-
`ecule was binding at this site, in a configuration
`essentially identical with that observed for the
`shorter C12:0 fatty acid. The carboxylate head
`group for the second molecule, which is not visible
`in either the HSA-C12:0 or HSA-C14:0 electron
`density maps, presumably extends into solvent
`from the upper surface of sub-domain IA and is
`disordered. In the HSA-C10:0 map, there is only
`very weak, broken density to suggest the presence
`of a second fatty acid molecule in site 2. The bind-
`ing of a second molecule of C10:0 may be much
`weaker because the methylene tails are too short to
`allow the
`overlapping hydrophobic
`contacts
`observed for C12:0 and C14:0. For fatty acids long-
`er than C14:0, there is no density at all for a second
`molecule, perhaps because binding of
`the first
`long-chain fatty acid at this site leaves too little
`room for a second molecule to bind with appreci-
`able affinity.
`
`Sites 3 and 4
`
`Sub-domain IIIA was previously observed to
`contain two molecules of C14:0, bound at sites 3
`and4(Curryetal.,1998).Themoleculesbound
`approximately at right-angles to one another with
`theirmethylenetailsincontact(Figures1and
`4(a)). All the fatty acids included in this study
`exhibited the same pattern of binding within sub-
`domain IIIA, sites 3 and 4 both being occupied.
`Within site 3, the head-group is fixed in the
`same position for all
`fatty acids, by hydrogen
`bonding to Ser342 and Arg348 from IIB, and
`Arg485 from IIIA. As the fatty acid tail
`length
`increases, the tail is forced into a U-bend configur-
`ation, since the longest dimension of the pocket is
`able to accommodate only about 12-14 methylene
`
`Figure 3. Fatty acid binding to
`site 2 in sub-domains IA and IIA.
`(a) Superposition of medium-chain
`and long-chain fatty acids.
`(b)
`Simulated annealing Fo (cid:255) Fc omit
`map contoured at 2.75s, showing
`difference
`electron
`density
`for
`C12:0 in site 2. A second molecule
`is bound in the upper part of the
`site, designated 20.
`
`MPI EXHIBIT 1060 PAGE 5
`
`MPI EXHIBIT 1060 PAGE 5
`
`

`

`726
`
`Fatty Acid Binding to HSA
`
`Figure 4. Fatty acid binding to
`sites 3 and 4 in sub-domains IIIA.
`(a) Superposition of medium-chain
`and long-chain fatty acids.
`(b)
`Model of C18:0 bound in two con-
`figurations in site 4. The configur-
`ation that
`is inverted relative to
`that observed for shorter fatty acids
`is depicted in with dark-grey car-
`bon atoms. This orientation of the
`ligand allows hydrogen bonding of
`the carboxylate group to the amide
`nitrogen atom of S419 and the side-
`chain hydroxyl group of T422.
`
`cases, a single fatty acid molecule binds in an
`extendedlinearconformation(Figure5).Thecar-
`boxylate head-group of the bound fatty acid invari-
`ably interacts with the side-chain of Lys525, aided
`in most cases by Tyr401, while the methylene tail
`extends into the tunnel. As with site 4, the largest
`fatty acid used here (C18:0) is long enough to fill
`the channel completely; indeed, the tip of the tail
`of this fatty acid projects through the far side of
`the channel and is exposed to solvent. However,
`unlike site 4, there is no indication of inversion of
`the fatty acid ligand. Superposition of the HSA-
`fatty acid complex structures for different fatty
`acids suggests that the carboxylate head-groups
`are not as uniformly anchored as for other sites
`(especially sites 2-4); in part, this may be due to the
`inherent flexibility of Lys525.
`
`Site 6
`
`On the basis of fragmentary electron density, our
`original analysis of HSA-C14:0 complexes tenta-
`tively identified a sixth fatty acid binding site in a
`
`groups. The breadth of the pocket allows variation
`in
`the positions
`of
`the
`hydrophobic
`tails
`(Figure4(a)).
`In site 4, which is longer and narrower than site
`3, the conformations of the bound fatty acids are
`similar to one another because of the constriction
`of the hydrophobic tunnel that accommodates the
`fatty acid methylene tails at this site. The carboxy-
`late head-groups all lie in more or less the same
`positions and are hydrogen-bonded by Arg410,
`Tyr411 and Ser489, which lie on the exterior sur-
`face on one side of sub-domain IIIA. As the chain
`length of the bound fatty acid increases, the car-
`boxylate head-group remains fixed and the methyl-
`ene tail protrudes further into the hydrophobic
`tunnel that runs through sub-domain IIIA. The
`longer fatty acids bind in an extended configur-
`ation that follows the shape of the tunnel. For
`C18:0 fatty acid, the methylene tail traverses the
`entire width of the sub-domain, the methyl tips
`emerging into solvent from the opposite surface.
`The electron density for
`this tip is not
`fully
`accounted for by the methyl group, suggesting that
`C18 fatty acids may also bind to site 4 in an
`inverted configuration. A second configuration for
`the C18:0 in site 4 has been incorporated into the
`model, which places the fatty acid carboxyl group
`within hydrogen bonding distance of the amide
`nitrogen atom of Ser419 and the
`side-chain
`hydroxylgroupofThr422(Figure4(b)).Intrigu-
`ingly, the possibility that C18 fatty acids may bind
`in two opposing orientations in site 4 is supported
`by the finding that octadecanedioic acid, which has
`a carboxylate group at either end, binds to sub-
`domainIIIAofBSA(Tonsgard&Meredith,1991).
`Our results suggest that only fatty acid species
`with at least 18 carbon atoms may span the dis-
`tance between the polar ends of the pocket.
`
`Site 5
`
`This site is formed by a hydrophobic channel
`that spans the width of sub-domain IIIB. In all
`
`Figure 5. Superposition of fatty acids bound to site 5
`in sub-domains IIIB. Amino acid side-chains for Y401
`and K525 are shown coloured by atom type.
`
`MPI EXHIBIT 1060 PAGE 6
`
`MPI EXHIBIT 1060 PAGE 6
`
`

`

`FattyAcidBindingtoHSA
`
`727
`
`shallow trench on the surface of the protein at the
`interfacebetweensub-domainsIIAandIIB(Curry
`etal.,1998,1999).Theresultsofthepresentstudy
`with fatty acids of varying chain lengths indicates
`that this site is occupied by both medium and
`long-chain fatty acids. The site differs most signifi-
`cantly from sites 1-5 in that there does not appear
`to be a cluster of well-placed amino acid side-
`chains that might co-ordinate the binding of the
`fattyacidcarboxylatemoiety(Figure6).Theligand
`density is generally stronger on the IIB side of the
`binding site and, in some cases, shows the broad-
`ening associated with a carboxylate group, while
`on the IIA side there is no indication in any map of
`such broadening. The fatty acids modelled at this
`site have thus been built with their carboxylate
`groups interacting with IIB. This is consistent with
`evidence that C16:0 activated with Woodward’s
`reagent K can interact with Lys349 in sub-domain
`IIIB of BSA, which corresponds to Lys351 in HSA
`(Reed,1986).Nonetheless,itshouldbestressed
`that the carboxylate groups of the different fatty
`acids adopt a variety of positions and there is no
`conserved electrostatic interaction. Side-chains that
`may be involved, at
`least
`transiently,
`include
`Arg209, Lys351 and Ser480. In contrast, the middle
`portion of the methylene tail appears to be well
`anchored, probably because
`salt-bridges
`from
`Arg209 to both Asp324 and Glu354 forms a side-
`chain ‘‘strap’’
`that helps to hold it
`in position
`(Figure6).
`The binding of C10:0 differs from that of the
`other
`fatty acids
`in that
`two molecules are
`observed to bind in site 6 in what appears to be a
`linear tail-to-tail configuration. The first C10:0 mol-
`ecule binds in a conformation similar to that of the
`
`other fatty acids, with its carboxylate head-group
`towards IIB;
`the second binds in the opposite
`orientation in the portion of the pocket (designated
`60) that is occupied by the methylene tails of the
`longer fatty acids. This arrangement places the tips
`of the methylene tails of the two C10:0 molecules
`in the centre of the site where the middle portions
`of
`the tails of
`the longer fatty acids are well
`anchored. There is no electron density for the car-
`boxylate moiety for the second C10:0 molecule
`bound in site 6, and the head-group has been
`omitted from the model. Additional unexplained
`density is associated with the side-chain of Lys212
`nearby, a feature not observed for complexes with
`long-chain fatty acids;
`this suggests a possible
`interaction between this residue and the ligand car-
`boxylate group but there is a break between the
`density adjacent to Lys212 and that for the methyl-
`ene tail for the bound C10:0 and at present it is not
`clear how the two might convincingly be united.
`The general absence of clear ligands for carboxy-
`late groups of the fatty acids at this rather open
`site suggests that the binding affinity may be rela-
`tively low. Consistent with this notion, in crystals
`of HSA-C14:0 soaked with the general anaesthestic
`halothane, the fatty acid in site 6 was displaced by
`the drug, whereas sites 1-5 retained their bound
`lipids(Bhattacharyaetal.,2000).
`
`Site 7
`
`A seventh binding site was observed for all fatty
`acids within the hydrophobic cavity of sub-domain
`IIA(Figure7).Althoughthepositionofthesiteis
`analogous to the binding cavity in IIIA (which con-
`tains sites 3 and 4), the site itself is considerably
`smaller. Previously we did not report C14:0 bind-
`
`Figure 6. Superposition of fatty acids bound to site 6
`in domain II. Observe the variability in the positions of
`the carboyxlate head-groups of the fatty acids. The salt-
`bridges from R209 to D324 and E354 are shown as
`dotted lines. The portion of the pocket that binds a
`second molecule of C10:0 (shown with carbon atoms
`coloured red) is labelled Site 60.
`
`Figure 7. Superposition of fatty acids bound to site 7
`in sub-domain IIA. The carboxylate groups of the fatty
`acids were not modelled due to lack of electron density.
`
`MPI EXHIBIT 1060 PAGE 7
`
`MPI EXHIBIT 1060 PAGE 7
`
`

`

`728
`
`Fatty Acid Binding to HSA
`
`ing within IIA, since the original map for this
`ligand contained only a very short crescent of den-
`sity(Curryetal.,1998).However,densityofavery
`similar shape was observed for all other fatty acids
`in the same position and was not observed in
`maps of defatted HSA (data not shown). In all
`cases, it is truncated and lacking the characteristic
`broadening at one end that was used elsewhere to
`locate the carboxylate head-group of the ligand.
`Accordingly, the bound fatty acids have been mod-
`elled conservatively with just their methylene tails.
`Although we cannot be definitive about side-chain
`interactions with the carboxylate head-group, a
`number of basic
`residues,
`including Lys199,
`Arg218, Arg222, His242 and Arg257 (which makes
`a salt-bridge to the carboxylate group of the fatty
`acid bound in site 2), are close enough to be
`involved. The fatty acids bind in a curved con-
`figuration, the tail being co-planar with aromatic
`drugsthatbindatthissite(Curryetal.,1998).Pre-
`vious data show that C14:0 may be displaced from
`site 7 in the presence of 0.5 mM tri-iodobenzoic
`acid and suggest that the site has relatively low
`affinityforfattyacids(Curryetal.,1998),whichis
`consistent with the lack of a fixed carboxylate
`interaction. However, other evidence (see below)
`suggests that site 7 in sub-domain IIA may be a
`primary binding site for shorter-chain fatty acids.
`
`Additional binding sites for medium-chain
`fatty acids
`
`The shortest fatty acid used in this study, C10:0,
`binds at two additional sites on the protein that
`were not initially observed for any of the longer
`fatty acids (sites 8 and 9). Both sites occur within
`the large crevice between domains I and III.
`
`Sites 8 and 9
`
`A single molecule of C10:0 is observed to bind
`close to the base of the inter-domain crevice and is
`actually tucked into the juncture between domains
`IIandIII(site8;Figures1and8(a)).Themethylene
`tail of the C10:0 molecule bound to site 6 helps to
`form the hydrophobic end of the site 8 pocket,
`indicating a possible cooperative interaction. At the
`other end of the cavity, there is an opening ringed
`by polar residues (Lys195, Lys199, Arg218, Asp451
`and Ser454). However, the carboxylate group of
`the bound fatty acid does not appear to have a
`basic amino acid ligand, though there is a hydro-
`gen bond to the side-chain of Ser454.
`Located further up the inter-domain crevice, site
`9 provides a rather open binding environment so
`that only one flank of the bound C10:0, which
`faces towards the bottom of the crevice, is in full
`contactwiththeprotein(Figure8(b)).Onthetop
`flank of the ligand (towards the solvent) there are
`contacts from the side-chain of Lys190 and from a
`bridging interaction between Glu187 of domain I
`and Lys432 of domain III, which is similar to the
`side-chain strap that occurs in site 6 (see above).
`
`Figure 8. Additional binding sites for C10:0 fatty acid.
`(a) Site 8, which occurs at
`the base of
`the crevice
`between sub-domains IA-IB-IIA on one side and IIB-
`IIIA-IIIBontheother(seeFigure1).(b)Site9,which
`lies in an upper region of the crevice. A salt-bridge
`between D187 and K432 across the top of the crevice
`helps to hold the ligand in place.
`
`This interaction is not observed in any other com-
`plex because the residues involved are too far
`apart; it occurs in HSA-C10:0 because of small but
`significant movements of domains I and III, which
`cause a 1-2 A˚ narrowing of the crevice between
`them. In the configuration modelled for the bound
`fatty acid, the carboxylate head-group of the C10:0
`is salt-bridged to the side-chain of Lys436; how-
`ever, both ends of the binding pocket have distri-
`butions of polar side-chains and the shape of the
`electron density does not exclude the possibility
`that the orientation of the fatty acid is the opposite
`ofthatshowninFigure8(b).
`
`MPI EXHIBIT 1060 PAGE 8
`
`MPI EXHIBIT 1060 PAGE 8
`
`

`

`Fatty Acid Binding to HSA
`
`Discussion
`
`HSA serves to greatly amplify the capacity of
`plasma for
`transporting fatty acids. However,
`because the protein binds a wide variety of drugs,
`it can have a serious impact on drug pharmacoki-
`netics. While previous studies have established
`that fatty acids and drugs may interact both coop-
`eratively and competitively, the structural basis for
`these effects is not well understood. Here, we have
`focused on the interactions of the protein with its
`primary class of ligand and examined how HSA
`binds a variety of fatty acid molecules. The study
`includes one of the most physiologically important
`fatty acid ligands known to bind HSA in vivo: pal-
`miticacid(C16:0)(Saifer&Goldman,1961).The
`structures of the five HSA-fatty acid complexes
`permit an extensive and highly detailed survey of
`the interactions of medium and long-chain fatty
`acids with the protein. The results show that there
`is a common set of seven binding sites for fatty
`acids ranging in chain length from ten to 18 carbon
`atoms and reveal how the variable geometry of the
`pockets can accommodate fatty acids of different
`lengths. Further, our findings have identified four
`additional binding locations on the protein that
`can be accessed by medium chain-length fatty
`acids (C10:0-C14:0).
`The binding sites are distributed across the three
`homologous domains of HSA, though each domain
`accommodates its complement of fatty acids in a
`differentway,asnotedpreviously(Hamiltonetal.,
`1991;Curryetal.,1998)(Figure1).Althoughthere
`are clear variations in the modes of binding of
`fatty acids at different sites, some common themes
`emerge. For sites 1-5, the carboxylate moieties of
`the fatty acids are anchored by the same electro-
`static interactions to basic and polar side-chains,
`irrespectiveofchainlength(Figures2-7).Att