Adipocyte Lipid-binding Protein Complexed with Arachidonic Acid

Vol. 269,No. 41, ISSUEof October 14, pp. 25339-25347, 1994 Printed in U.S.A.

Tm J o m m OF BIOIGGICAL CHEMISTRY D 1994 by The American Society for Biochemistry and Molecular Biology, Inc.

Adipocyte Lipid-binding Protein Complexed with Arachidonic Acid TITRATION C ~ O R I ~ E T RAND Y X-RAY C R Y S T ~ ~ O G ~ H STUDIES" IC (Received for publication, May 9, 1994, and in revised form, June 28, 1994)

Judith M. LaLonde, Melanie A. Levenson$, Jeremy J. Roe, DavidA. Bernloh&g, and Leonard J. BanaszaM From the Department of Biochemistry, School of Medicine, Uniuersity of Minnesota, Minneapolis, Minnesota 55455 and the $Department of Biochemistry and the $Instituteof Human Genetics, College of Biological Sciences, University of Minnesota, St. Paul, Minnesota-55108

The associationof the adipocyte lipid-binding proteina major role in both the development o f fat cells and in mediand ating theeffects of a variety of hormones which affect adipose oleic acid (cis, 1 8 1 A@)has been examined by titration metabolism. For example, when cultured p r e a d i ~ c are ~ s incalorimentry. In addition, the crystal structure of ALBP cubated with arachidonic acid, they produce large quantities of with bound arachidonic acid has also been obtained. prostaglandins (2) and the process of adipose conversion is of the arachidonic acid-ALBP markedly inhibited (3). The release of arachidonic acid from Crystallographic analysis complex along withthe previously reported oleic acid- phospholipid in adipose tissue has been shown to be associated ALBP structure (Xu, Z., Bernlohr, D. A, and Banaszak, with the signaling pathway of tumor necrosis factor-a (4). FurJ. (1993) J. BioZ. Chern. 268, 7874-7884) provides a thermore, arachidonic acid down-regulates the insulin-dependframework for the molecular examination of protein- ent glucose transporter gene (GLUT4)in 3T3-Lladipocytes by lipid association. Isothermal titration calorimetry re- inhibiting transcription and enhancing mRNA turnover (5). vealed high affinity association of both unsaturated Although there are no reported arachidonic acidspecific fatty acidswith the protein. The calorimetric data binding proteins, it would seem reasonable to consider that the yielded the following thermodynamic parameters for arachidonic acid:Kd= 4.4 pz, n = 0.8, AG = -7370 cdmol, mobilization and utilization of polyunsaturated fatty acids A€€ = -6770 cdmol, and TAS 5 +600 cal/mol. For oleic within the cell makes useof the intracellular fatty acid-binding acid, the ~ e r m o d y n parameters ~c were Kd= 2.4 pz, n proteins. The adipocyte lipjd-binding protein (ALBP)' is a = 0.9, A 0 = -7770 cdmol, A€€= -6050 callmol, and TAS = member of a protein family of intracellular lipid-binding pro+1720 calfmol. The identification of t h e r m ~ ~ a m i c a l f yteins (6). The members of this large family generally have dominating enthalpic factors for both fatty acids are 130-132 amino acids and areexpressed in a wide variety of cell consistentwith the crystallographic studies demon- types possessing active lipid metabolism (6). strating the interaction of the fatty acid carboxylate The three-dimensional structures o f the intracellular lipidwith a combinationof Arg1OS,Arglg, and Tyrlas. binding proteins include 10 anti-parallel P-strands arrangedin The crystallographicrefinement of the. protein- a barrel structure. Within the p-barrel is an enclosed water arachidonate complex wascarried out to 1.6 A with the filled cavity. Two a-helices are located on the surface covering resultant R factor of 0.19. Within the cavity of the crys- part o f what would be one openend of the barrel. The binding talline binding protein,the arachidonate was found in asite for hydrophobicligands is within the cavity, whichis closed hairpin conformation. The conformationof the bound at the other end by the packing of side chains. Despite the ligand is consistent with acceptable torsional angles and hydrophobic nature of the ligands, the cavity is lined with both the four cis double bondsin arachidonate. Theseresults hydrophobic and polar amino acids. The ligand binds with its demonstrate that arachidonate is a ligand for ALBl? polar end buried within the cavity, The dissociat~on constants They provide t h e ~ o d ~and ~ istructural c dataconfor the hydrophobic ligands generallyvary from nanomolar to cerning the physical basis for protein-lipid interaction and suggest that intracellular lipid-binding proteins micromolar. For ALBP, crystaI structures have been determined of commay mediate the biological effects of polyunsaturated plexes with both saturated (palmitate, 16:O and stearate, 18:O) fatty acids in vivo. and monounsaturated (oleate, 18:l)fatty acids (1, 7). The fatty acid bound in the cavity is stabilized by interactions between Arachidonic acid (20:4,all cis is a polyunsaturated the carboxylate of the fatty acid and the side chains of ArglZ6, fatty acid knownto be an important mediator of adipose tissue TyrlZ8,and Arg'06 through an intervening water molecule. The development and physiology. This polyunsaturated lipid is the scheme therefore appears to involve the burying of a single immediate precursor of the short-lived family of eicosanoid hor- negative charge (ligand carboxylate) in a cavity near two posimones which includes the prostaglandins, leukotrienes, and tive charges (protein Arg'06 and ArglZ2"). The postulated role of the adipocyte lipid-bind~g protein as thromboxanes. Arac~donicacid has been i m p l i c a ~ das playing a mediator of intracellular solubilization and transportof fatty * This work was supported by National Institutes of Health Grant acids in adipocytes suggested that ALBP may bind polyunsatuGM13925 (to L. 3. B.) and National Science Foundation Grant rated fattyacids such as arachidonic acid. Furthermore, moduDMB9118658 (to D. B.). The costs of publication of this article were lation of the intracellular concentration of free, unbound defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C.Section 1734 solely to indicate this fact. ' The abbreviations usd are: ALBP, adipocyte lipid-binding protein; TiTb whom correspondence should be addressed: Dept. of Biochemis- AR, arachidonic acid; PA,palmitic acid; HDS, hexadecanesulfonicacid; try, School of Medicine, University of Minnesota, Minneapolis, MN AR.ALBP, PA.ALBP etc, the complexes of fatty acid with ALBP; RMS, 55455. Tel.: 612-626-6597;Fax:612-625-2163. root mean square.

(ALBP)with arachidonic acid (allcis, 204

A&8J1714)

1;.

A53*J1J4)

25339

25340

Interactions Acid-ALBP Arachidonic

arachidonic acid via association with the fatty acid-binding proteins could be expected to have significant effects on t h e metabolic processes affected by such lipids. However, it was clear from the studies of t h e C18 fatty acids with crystalline ALBP that the ligand binding cavity was not long enough in any dimension to accommodate C20 fatty acids unless the ligand protruded beyond the surface of the protein.ofFurthermore, with multiplecis carbon double bonds, the bound ligand had tohave a different conformation from the saturated the or monounsaturated ligands whichhad been studied so far (1,7). Here, we reportthat arachidonate binds to ALBP and describe the binding determinants as studied by titration calorimetry and x-ray crystallography. In addition, we summarize thethermodynamic analysis of oleic acid association with ALBP. The analysis of both the structural and thermodynamic data yields information regarding the physical basis for protein-lipidinteraction and could suggest possible mechanisms for controlling the functionaland developmental effects of arachidonic acid in fat cells by alteringits binding to ALBP.

TABLE I Data collection statistics for ALBP crystals containing arachidonate AR.ALBP

No. of crvstals No. observations No. of reflections VuI >2 No. of unique reflections Average redundancy Average Vu, outer shell % collected possible to 1.70 A % collected possible to 1.60 A

3 4.0 1.60 79,181 70,836 16,949 4.7 5.7 99.9 96

unit and unit cell dimensions ofa = 120.74, b = 37.70, c = 28.57 A, p = 92.34".Data collection statistics aregiven in Table I. A nearly complete data set was obtained from a single crystal with excellent internal agreement between symmetry related reflections. Crystal Structure Refinement-For the model refinement, the combined x-ray crystallographic and energy minimization was carried out using simulated annealing as implemented in the computer program, X-plor (11). Model building and electron density map inspection used the molecular modeling program Chain (12). For model adjustment, EXPERIMENTAL.PROCEDURES both 2 IFI, - IF,I and IF, I - IF,I electron density maps were Materials-Recombinant murine apoALBP was purified following calculated and contoured at 1.0 and 2.5 u levels, respectively. The structure of AR.ALBP was solved using the coordinates of previously reported methods (8). Arachidonic and oleic acids were purapoALBP (7) with waters and side chains having alternate conformachased from NuCheck Prep., Elysian, MN.All other materials were tions removedfrom the coordinate set used to calculate the initial reagent grade or better. Sulfhydryl ntrations-Ligand binding was assessed indirectly by phases. In addition, the starting temperaturefactors for all atoms were set to 15.0 A'. The course of refinement is described in Table 11. Refineusing the sulfhydryl modification assay (9). For each assay, a 20 p~ solution ofALBP wastitrated with 160 p~ 5,5'-dithiobis-(2-nitrobenzoic ment proceeded in 10 macrocycles as higher resolution data was slowly added in shells. The data were partitioned into a working set (90%of acid) in the absence and presence of arachidonate. The production of the thionitrobenzoate anion was monitored spectrophotometricallyusing a the data) and a testset (10%of the data) for calculation of the Rfree(13). Because of the ambiguity in the positioning of arachidonate in the molar extinction coefficient of 13,600 at 412 nm. Previous studies have initial IF, I - IF, I difference map, two models were constructed and demonstrated that the single modifiablesulfhydryl, CYS"~, protected is refined in parallel. Each of the two models contained a unique conforfrom modificationin the presence of a bound lipid (9). Crystallization lkials-Prior to crystallization, the protein was mation for arachidonate. This ambiguity was resolved as higher resoequilibrated in 12.5 mM HEPES, pH 7.5, 1 mM dithiothreitol. The ALBP lution data were added as described under "Results." In the first maccomplex with arachidonate (AR.ALBP) was formed by adding a 2-fold rocycle, 80 steps of rigid body refinement indicated no significant shift molar excess of the ligand. Arachidonic acid was addedin ethanol to a of the molecule in the unit cell. In the second cycle, positional refinement at low resolution and subsequent inspection of the 2 IF, I - IF, I 10 mgiml solution ofALBP so that the ethanol concentration in the protein solution did not exceed1%(v/v).The protein-ligand solution was and I Fo I - IF, I electron density did not indicate any major changes gently stirred overnight at room temperature. Crystallization was in the starting modelfor the protein conformation. After simulated achieved using the hanging drop vapor diffusion method by adding annealing and refinement at 3.0 A resolution, the second arachidonate NaH,PO,&HPO, buffer, pH 6.6-7.2, to a final concentration of 2.1 to model exhibited unusual dihedral angles and bad contacts between neighboring carbon atoms along the aliphatic chain. Based on these 3.0 M. Calorimetric Methods-Fatty acid was solubilized at 25 or 50 p~ in initial results, the second arachidonate model was rebuilt. The first during the process. In the protein model, phosphate buffer (50 mM sodium phosphate, pH 7.0), with a final etha- model, however, behaved well nol concentration of 0.1%. Protein for calorimetry was equilibrated in main chain and side chain adjustments were made at residues L y 8 , SeP3,Ser", Lys", Lys"", and ThrIZ5. phosphate buffer and concentrated t o 1-2 mM. Ethanol was added to the In the next stage of simulated annealinghfinement at2.5-A resoluprotein stock to 0.1% to minimize heats of dilution. Protein was injected in ten 10-pl aliquots into 1.4 ml of either phosphate buffer or ligand to tion, a peptide bond flip occurredat residues 36 and 37. At this stage, 25 measure heats of binding. For determination of thermodynamic param- water moleculeswere added to the model.Solventmoleculeswere eters, heats of dilution of protein were subtracted from heats of binding. added based on the following criteria: spherical appearance of 2 IF, I IF, I and IF, I - IF,I , electron density contoured at 2.5 and 3.0 u All experiments were conducted at 26 "C. levels, and a heteroatom within hydrogeld bonding distance between 2.7 Data were collected using a Microcal OMEGA differential titration and 3.2 A. In the fifth macrocycle at 2.0-A resolution, the 23 factors were calorimeter and analyzed with the accompanying ORIGIN computer software. Based upon prior crystallographic studies (1,7), the stoichi- refined following simulated annealing. 1 Fo I - 1 F, I density appeared ometry of binding ( n ) ,enthalpy of binding ( A H ) , and binding constant both near Met40and CYS"~ aatdistance consistent with oxidation. Both (K,) were determined by fitting the datato a function that modeled one Met4" and CYS"~weremodeled with an oxidized sulfur. In addition class of independent sites (10). Free energy of binding was determined Phe57 was changed to Ala to confirm the position of its side chain. At 1.8-Aresolution, an alternateside chain conformationwas apparent for from the association constant using the equation A G = -RTlnK,,. Entropy of binding was then calculated with the equation A G = AH - T A S . residue Lys". With the addition of data to 1.7 A, the ambiguity in theorientation of All AA values were calculated as the value for oleic acidminus that for arachidonic acid. Five independent titration experiments were carried arachidonate was resolved. The first model, with direction as in palmiout for each ligand. In all cases, the experimental data were analyzed tate, gave better electron density maps and B factors. The ligand conseparately and the fitted parameters averaged to yield the reported formations will be discussed under "Results." An alternate side chain conformation was added for Asp4'. Two conformations for Phe5' have thermodynamic values and statistical errors. X-ray Methods-X-ray diffraction data was collected to 1.6-A resolu- been observed in apoALBP and holoALBP complexes (see Ref 7). In tion using a Siemens multiwire area-detector with CuK, radiation from apoALBP, Phe57is positioned near residues Ala33and Ala75 closingthe a Rigaku RU-200 generator equipped with graphite crystal monochro- proposed portal entry site of the ligand. In holoALBP complexes, a second side chain conformer is observed allowing open access of the mator. Data were collectedwith a crystal to detector distance of 120 mm and with a frame to frame rotation of 0.25". With the generator oper- portal to the ligand. In this study, Phe5' was returned to the model in ating at 45 kV/200 m A , the counting time/frame was 150 s. The the open conformation.In the last three macrocycles ofpositional and €3 AR.ALBP complex crystallized isomorphously with the previously re- factor refinement, additional water molecules were added, the OCCUported high resolution apo- and holoALBP crystals, (1,7).TheAR.ALBP pancy of the ligand was refined, an unknown compound modeled as crystal form belongs to space group C2 with one molecule/asymmetric propionic acid was placed near Glu14 (see Ref. 11, and alternative side

,

Arachidonic Acid-MBP Interactions

25341

TABLE I1 Crystallographic refinement of ALBP.arachidonate complex The following abbreviations are used: rigid, rigid body refinement; poshref, positional and temperature factor refinement; sa, simulated annealing. Ave B, average temperature factor. Method Cycle

Resolution A

No. reflections

10-1.6 204.0 8-3.5 8-3.0 8-2.5 8-2.0 10-1.8 10-1.7 10-1.6 10-1.6 10-1.6

16,731 1,135 1,545 2,521 4,431 8,570 11,933 14,125 16,594 16,594 16,594

Starting

1 2 3 4 5 6 7 8 9 10

Rigid Pos Sa Sa Sa Sa Poshref Podbref Poshref Poshref

No.

Ave B

atoms

15.0 15.0 15.0 15.0 15.0 15.0 13.9 15.6 15.6 16.6 16.6

RMSd

1,025 1,025 1,025 1,025 1,037 1,037 1,072 1,087 1,108 1,118 1,137

0.19 0.40 0.57 0.58 0.73 0.53 0.49 0.29 0.19 0.11

R,JR

factor"

0.348/0.343 0.322/0.306 0.271/0.182 0.353/0.197 0.329/0.232 0.292/0.215 0.274/0.217 0.241/0.200 0.241/0.200 0.227/0.195 0.220/0.188

R,, is defined by Briinger (1993). Ten percent of the F, values are not used for refinement but serve as anindependent check on improvements in the coordinate set as exmessed in the calculated structure factors. Rh.eeis the crystallographic R factor for the reflections omitted from the refinement. chain positions were added for SeP3 and IleE3. The crystals diffract to high resolution (1.6 A), and therefined model of the arachidonate complex produced a high quality structure. The final R factor was 18.8%with an average B factor for protein atoms of 14.8 A'. The final refinement statistics for the AR.ALBP complex are given in Table 111. The model contains 1137 atoms, including 90 solvent molecules. The final R,, and R factor are approximately 1%lower than in the two previously reported structures of ALBP complexes (1). The higher number of solvent molecules and a different modeling of residue Lys"could accountfor this difference.Residues Lys',Lys3',Lys7', Lysloo,Lys105,and LyslZoare disordered. Residues Lys", Asp4', Ser53,and IleE3have two alternate side chain conformations, each present at half occupancy. The coordinates from the crystallographic study have been deposited in the Protein Data Bank.

RESULTS

To determine if arachidonate would bind to ALBP, the cysteine modification assay of Buelt and Bernlohr (9) was first utilized.Previous studies have demonstrated that a bound fatty acid (oleic acid) sterically blocks Cys'" modification by sulfhydryl-reactive reagents. Protection of Cys'17 modification by a fatty acid is therefore taken as an indicator of ligand arachidonate, CYS"~ was association. In the presence of 200 largely protectedfrom 5,5'-dithio-bis(2-nitrobenzoicacid) modification (results notshown). These results were similar to those previously obtained using oleic acid and were consistent with the prior x-ray crystallographic analysis. To accuratelydeterminethe affinity of arachidonate for ALBP, and to compare its binding to thatof oleic acid, isothermal titration microcalorimetry was utilized. Use of this technique permitted simultaneous determinationof stoichiometry, associatioddissociationconstants,andthefree energy, enthalpy, and entropy of binding. The calorimetry data are presented in Fig. 1 and summarized inTable A? The calorimetric analysis of arachidonate binding to ALBP revealed a large negative enthalpicfactor with a rather minor entropic component. The stoichiometry of ligand binding (n = 0.8) was consistent with previous binding and crystallographic data indicating a single ligand-binding site (6-9). The large, dominating enthalpic contribution t o the overall binding energyis consistent with mutagenesis studies which demonstrated that the interaction between the fatty acid carboxylate and the arginine andtyrosine residues are key componentsto the binding process (1,14). Calorimetric analysis of the oleic acid.ALBP complex revealed similar results.They are shown in Fig. 2 and theoverall thermodynamic parameters for both ligands aregiven in Table n! The stoichiometry of binding (n = 0.9) and the relatively large enthalpiccontribution ( A H = -6050 caYmol) to theoverall free energy are again consistent with the conclusion that ionic interactions between the lipid carboxylateand theprotein basic residues are essential for ligand association. The enthalpiccon-

TABLEI11 Final refinement statistics of the crystalline ALBP.arachidonate complex A. x-ray data

AR.ALBP

Resolution limit li Initial R,JR factor Final R,JR factor No. macrocycles No. reflections No. of atoms No. of solvent atoms Average B factor A'

1.6 0.348/0.343 0.220/0.188 10

16,731 1,137 90 16.6

B. Refined model properties RMS deviations

ARALBP

Bonds li Angles' Dihedrals" Impropers"

0.022 1.632 28.263 1.281

tribution for oleate was very similar to that for arachidonate ( A A H = +720 caYmo1). A larger entropic contribution to the binding, relative to thatof arachidonate provide an additional favorable driving force (A(TAS) = +1120 caYmo1) to the equilibrium which leadstothe somewhat greater affinity of the protein for oleic acid than arachidonic acid (AAG = -390 cal/mol). The x-ray studies of the crystalline complex also indicated a single binding site. Experience with other ligands had shown that fatty acids bound to crystalline ALBP have to be studied with caution since electron density for some carbon atoms of the hydrocarbon chain may be disordered within the binding site. Consequently, the x-ray results were first analyzed by so-called difference maps before refinement was initiated. To calculate the difference map, part of which is shown in Fig. 3 A , thestructure factors IF, I (see"Experimental Procedures") were derived from the AR.ALBP crystal and IF, I represents the x-ray data setfrom apoALBF! The difference map between these two crystalline forms clearly showed electron density for the arachidonatefrom the carboxylate headgroup t o carbon 20 on the aliphatic chain. The difference map also indicated that the additional electron density and, consequently, the ligand were contained entirely in the binding cavitywith anapproximate 180"change in direction from head to tail.However, as canbe seen in Fig. 3 A , a break in density occurs shortly after the carboxylate group. This initially suggested two possible interpretations for the position and direction from C, to C,, of arachidonate. The first would proceed from head t o tail as in stearate, oleate, palmi-

25342

Arachidonic Acid-ALBP Interactions Time (mint

Time (m~n)

: 0

m

Y 0

-1 0

I

I

5

10

:I

17

33

25

J

-3.9

-4.5

i

8

t

0

I

I

5

10

Injection Number

Injection Number FIG.1. Isothermal titration calorimetry of arachidonic acid binding to ALBP.Upper graph:this shows the uncorrected calorimetric tracings of the heats associated with arachidonic acid binding to ALBP. Ten 10-pl aliquots of protein (typically 1-2mhl) in phosphate buffer containing 0.1% ethanol were injected into 1.4ml of arachidonic acid (50 p~ in phosphate buffer containing ethanol) and the indicated heats measured as a function of time. Lower graph: the measured heats from each peak were integrated, the heats of dilution from the protein alone subtracted, and the resultant net change was fitted to the equations described under “Experimental Procedures.” The solid line through the points is a least-squares fit to the datacalculated assuming a single site binding model. Results are shown from only one of five similar experiments.

FIG.2. fsothennal titration calorimetry ofoleic acid binding to ALBP. See Fig. I and “Experimental Procedures.’’ Upper graph: this contains the uncorrected data obtained from calorimetric tracings of ALBP titrated into 50 +IM oleic acid in phosphate buffer, pH 7.0. Lower graph: the integrated heats were analyzed as described and fit to the data assuming a single binding site. Results are shown from one of three such experiments.

Fig. 3A, was the best interpretation forboth lFoll- lFo12 difference and 2 IF,I - IFJ electron density maps. It also resulted in the lowest €2 factor. In previous analyses, close contacts between two side chains, Lys” and Met3j, and their symmetry mates occurred along a TABLEIV 2-fold symmetry axis (1). These contacts were explained in Results of ~ o ~ h e r m~a l l o r i ~ ~ t ~ terms of disorder in which an alternate side chain conformation Parameter Arachidonic acid“ Oleic acid” A6 existed. In order to avoid steric interference, this model would require that Lys21 and its symmetry mate display alternate Kd fw) 4.4 t 1.1 2.4 * 1.0 0.9* 0.2 n (mol/mol) 0.8 2 0.2 conformations at any one time. Although an explanation of two AG (cal/mol) -73802130 -7770t580 temporally distinct conformations for LysZ1and its symmetry LLH(cal/mol) -677021360 -6050&1320 mate were proposed, electron density for such a model was TAS (cal/mol) +600 t 1420 +1720 * 1340 BAG (cal/mol) -390 t 710 weak. In the crystal structure of ARpALBP, LysZ1and its sym+720 A M (cal/mol) t 2680 metry mate clearly display electron density for alternate conA1TAS) (cal/mol) +1120=2760 -~ formations. The appearance of an alternate side chain for Lys2’ a Data are tabulated based upon fiveindependent binding isotherms confirms the correctness of the original model of discrete disfor arachidonic acid and three experiments for oleic acid.Each experi- order at thecrystallographic 2-fold symmetry axis in which the ment was performed as described under “Experimental Procedures.” A h values are calculated as thatfor oleic acid minus that for arachi- positioning of the side chains must be synchronized (1). Met35and its symmetry mate in the current model do not donic acid. have close contacts between the sulfur atoms as observed in earlier models (1).It is not known why Met35in the AR-ALBP tate, and hexadecanesulfonic acid, and the second wouId pro- model is different from the earlier studies. There is a difference ceed in the opposite direction. In Fig. 3A, to visualize the al- of 0.8 A in theposition of the sulfuratom of Met3‘ in ARaALBP ternate interpretation imagine connecting the carboxylate to relative to that in PA.ALBP. In addition as observed in other C,, and renumbering the carbon atoms in reverse order. Note ALBP.ligand complexes, unaccounted for electron density apthat this is possible because the difference electron density is peared near residue GIu14 and was modeled as propionic acid (1).Finally, Cys117and Met40again appeared t o be partially shaped almost like the letter “ 0 . We initially felt that themap was not adequate to determine oxidized (1). In spite of a few differences, the crystal structure of which of the two different conformations waspresent. Instead, as noted under “Experimental Procedures,” each orientation of AR.ALBP agrees well with the complexes formedwith stearate, arachidonate was refined separately. Refinement of these two oleate, palmitate, and hexadecanesulfonate (1,7).The average orientations, however, indicated that the first, as depicted in RMS difference to four ALBP.ligand complexes was 0.1 and “~

Arachidonic Acid-ALBP Interactions

25343

A

Frc. 3. Stereodiagrams of electron density for arachidonate boundALBP. to The refined modelofAFt.ALBP is shown in bold lines with the carboxylate oxygens labeledOEl and OE2 and the positionof the olefinic bonds labeledat C5, C8,C11, and C14. A, the electron density map calculated using lFoll- IFJ2 as coefficients prior to refinement is contoured at a 2.0 CT level. This so-called “difference electron density map” is shown with the appropriate model atoms. The map was calculated using is described more fully in the text. B, the final electron density map 2 lFol - IF,I as coefficients and contoured at a 1.0 w level.

0.93 A for main chain and side chain atoms, respectively. RMS differences between AFt-ALBP and apoALBP are somewhat higher at 0.32 and 1.01 A again for main chain and side chain atoms, respectively. Regarding the positioning of side chains, residues LysZ1,A r $ O , Pros8, Leu4*,Ser53,Gld4, and Phe57disRMS differences between AR.ALBP and play thelargest PA.ALBP. The conformational difference at LysZ1and previous structures wasdiscussed above. For residues A r $ O , Leu4*, G ~ u the ~ ~differences , correspond to side chain rotations and displacements not associated with ligand binding. In all the ALBP-ligand complexes examined, Phe5’ has highB factors (40 k ) . The larger RMS differences among different crystal structures and the high temperature factors reflect the dynamic disorder for this residue in the crystalline state and most IikeIy in solution. The final 2 lFol - lFcl electron density map for arachidonate is shown in Fig. 3B. The occupancy for arachidonate is0.9 2nd the averageB factor for atoms in thebound fatty acid is 42 A2. The highest temperature factors occur at carbons 13 to 17 where the corresponding electron density is the weakest. As can be seen in Fig. 4,the bound arachidonate lies completely buried within thecavity of ALBP. It does not protrude slightly

from the surface of the protein as observed with palmitate, stearate, and oleate. Because of its hairpin appearance, several crude dimensions can be attached to its bound conformation. Assuming the nearly closed hairpin conformation resembles an ellipsoid, the overall length of the fatty acid as it extends over the major ellipsoid axis (C, to C14)is 13.3A. The width or minor ellipsoidal axis of the aliphatic chain is 10.8 A. The volume that arachidonate occupies in the cavity is 8%larger than oleate or stearate and 20% larger than thatof palmitate. From head to tail the redirection of the aliphatic chainis approximately 165 ’. The two most abrupt changes in chain direction occur at C,, and C16. The bend from C , to C,, of 1 0 2 O resembles that observed for the same region in the saturated fatty acids palmitate and stearate and the unsaturated fatty acid oleate. The insertion of a cis-double bond in an aliphatic chain introduces a 44 ’ change in direction (15).The conformation of arachidonate bound to crystalline ALBP has a total chain redirection of 165 o closely approaching the theoretical value of 176 for four cis-double bonds. The conformation of arachidonate does not yield a n entirely flatmolecule with its four double bonds lying in the same plane. On the contrary, three of the four O

25344

Arachidonic Acid-ALBP I n t ~ ~ u c ~ ~ o ~ s

A

B

FIG.4. Diagram of arachidonate buriedwithin the cavity of crystalline AR-ALBP.A, ALBP depicted as a ribbon drawing. The bound by black ball and stick model.B, stereodiagram showing the amino acid side chains that form contacts with arachidonate arachidonate is illustrated drawn in unshaded and shaded ball andstick representation, respectively. The C, protein atoms of ALBP are drawn with a thin line.

olefinic bonds (at Cs,C,,, and CJ lie in one plane while the first at C, lies approximately 60 ' from this plane. With a few exceptions, the contacts formed between arachidonate andALBP are homologous with those observed with the other~BP.ligand complexes (1,7). These contacts are listed in Table V. The carboxylate headgroup forms hydrogen bonds and electrostatic interactions withArg126and TyrlZsas well as Arglo6 through an intervening water molecule. This set of ligandprotein interactions is highly conserved in all ALBP-ligand complexes examined (1,7). Because arachidonate i s entirely buried within the cavity some newprotein-ligand contacts along the aliphatic chain are established while some contacts observed with stearate, oleate, and palmitate near the portal (Phe5') are absent. The missing contacts would occur with atoms belonging to residues Metzo, Val32,Phe57,and Lys5,. The hairpin or ellipsoidal conformation

of arachidonate prevents the formation of contacts between the fatty acid and these residues. The total number of polar and non-polar contacts formed with the aliphatic chain is 18 and 15, respectively. For the saturated ligands, the ratio of non-polar t o polar contacts is greater than 1 whereas as for the unsaturated ligands this ratio is less than 1. In the case of arachidonate, the greater number of polar contacts can be accounted for by a set of contacts at C,, and C,, in thealiphatic chain. Residue Ser53has two alternate side chain conformations in AR.ALBP. In one of these conformations the hydroxyl group makes a polar contact with Cz0.A secondnovel contact is with at carbon C,8. This residue has an RMS difference of 0.76 A in comparison with other ALBP.ligand complexes. The pyrrolidine ring of Pro3' points away from C,, in the case of arachidonate and in the opposite direction in thepresence of palmitate acid. This subtle

Arachidonic Acid-UBP Interactions TABLEV Closest contacts between arachidonate and the amino acid side chains of ALBP AR-ALBP

Arachidonate

OEl

Met4'.

Tvr128,

WAT 1. WAT 8

CYS"' Cys"7 Cys"', WAT 10,WAT 3 WAT 4, WAT 10, WAT 3

Asp76

Asp76 Ala1'

Th?', Ala33,Asp76 T h P , Ala33,Ala75 Ala33,Ala75 Ala33 Ala33

Ala33

Cl'

.-.

1 '8

Ul'

c,,

WAT 6, WAT 8, Ser53

difference in thepucker of the pyrrolidine ring prevents a prohibitively close contact (3.17 A? between C,,of arachidonate and Pro3*. As observed in the complexes ofALBP with stearate, oleate, and palmitate, several conserved water molecules form van der Waals contacts withthe aliphatic chainof the bound arachidonate. They are listed in Table VI. Conserved waters 3,4, and10 make the same contacts at C, and C , in arachidonate as observed in palmitate. The distance between Wat 3 and C,andC,, however, is shorter in the case of ?rachidonate= (3.1 A) than observed with bound palmitate (3.4A). This 0.3 A difference is the same order of magnitude as the avFrage RMS difference for all conserved water molecules (0.35 A) found in the cavity of ALBP.ligand complexes. If this shorter distance isexperimentally significant, then it suggests weak hydrogen bonding between this water (Wat 3) and the 7~ bond of the bound arachidonate. &-Double bonds also occur at c,, c,,,and c,,, but no similar water interactions are present. The double bonds at position C,, C,,, and C14lie in the same plane which is oriented approximately 60 from the double bond at C,. New water contacts occur between Wat 6 and Wat 8 at C,, (1). Hence, although arachidonate occupies a larger portion of the cavity, there is little change in the distribution of ordered solvent within the cavity. Only one of the 10 conserved water molecules, Wat 9, observed in otherALBP.ligand complexes (1) is not observed in theAR.&BP structure. Wat 9, if present in AR.ALBP would make a 3.1 A contact withC,, in arachidonate. In previous structures of liganded protein, this waterformed a hydrogen bond to theside chain of S e P . The presence of alternate conformations for Ser53 in the AR.ALBP structure may further weaken this location as a water binding site. One new water molecule is observed in the cavity ofAR.ALBP. This water (2126) forms H-bonds to the guanidium group of ArglZ6 and additionalhydrogen bonds to main chainoxygens of Ala33 and Ala3,. O

DISCUSSION

Crystallographic analysis hasindicated that ALBP binds the polyunsaturated fatty acid, arachidonate, with its carboxylate associated with Arg106,Arg",, and Tyrl''. In the headgroup region, the binding mode for this fatty acid is the sameas that observed for other saturated, and monounsaturated fatty acids (1, 7). However, the aliphatic chain of arachidonate bound to

25345

TABLEVI Waters in close contact with arachidonate ARA atom

ARA c5 ARA c5 ARA C 6 ARA 20 ARA 20

3.68

Water

Distance

w3

3.12 A

w10 w4 W6 W8

3.74 A 3.46 A 3.77 A

A

ALBP adopts a hairpin or U-shaped conformation which prevents itfrom protruding at thesurface of the protein. Furthermore, the arachidonate in thecrystal structure shows that the molecule when bound is not a flat U with all four cis-double bonds lying in the sameplane. Rather, the firstdouble bond of the U is distorted out of plane relative to the other three. Examination of the protein-ligand and water-ligand interactions does not suggest anyspecific set of contacts that stabilize the cis-double bonds in arachidonate.Although Wat 3 forms a weak hydrogenbond to thedouble bond at C,-C,, the position of this water is conserved within other ALBP.ligand complexes. Thus no new protein-ligand interactions at the positions of the double bonds have been identified in the crystal structure of AR.ALBP. The additional number of atoms in arachidonate, stearate, appear to compared to palmitate, oleate, and be accommodated by continuing curvatureinthe bound conformation. Calorimetry has enabled usto observe the energetics of arachidonate andoleate bindingto ALBP. The dissociation constant for the arachidonatebinding reaction is 4.4PM while that for oleic acid is somewhat lower, 2.4 VM. As was shown in Table IV, the binding energies were derived from a major enthalpic component with a smaller entropic contribution. The macroscopically measured enthalpic andentropic factors are netvalues and are comprised of multiple microscopic components, both positive and negative. A quantitative correlation between the thermodynamic dataandthestructuralstudiesis not yet possible. However, a qualitativeinterpretationcan be made and we have summarized these potential correlations in Table VII. The calorimetric measurements indicate that over 90% of the Gibbs free energy for the binding of arachidonic acid to ALBP occurs as a consequence of the enthalpic component. For oleic acid binding, 77%was contributed by the AH. This isconsistent with the calorimetric results reported for fatty acid binding to mutant forms of cellular retinol-binding protein I1 and intestinal fatty acid binding protein where AH contributed 66 and 87%,respectively, to the total free energy (16). Enthalpic contributions areattributable to non-covalent bonding interactions. Clearly visible electrostatic and hydrogen bonding interactions in the crystal structure occur between Arg106,ArglZ6,and T y r l Z 8 in theALBP cavity and the carboxylate of the bound arachidonate or oleate. These are listed as items 1-4 in Table VII. The contribution of these interactions to binding was further demonstrated by mutation eitherof kg1', or QrlZ8or both of these key residues. The resulting ALBP is capable of only minimal fatty acid binding (14). The rather small A A H between arachidonate and oleate (+720 cal/mol) indicates thatfor these two lipids, despite the large change in ligand conformation in the crystalline state, the enthalpic components t o the binding energy are very similar. The second entry in Table VI1 is a favorable enthalpic contribution occurring as a result of the relief of repulsive interactionsbetween the two positively charged arginine side chains. We suspect this to be a contributing factor since no evidence has ever been found for the presence of bound counterions in apoALBP or other similar apoproteins (6). Using Coulombs law, a crude estimate of the unfavorable enthalpic

25346

Arachidonic Acid-ALBP Interactions TABLE VI1 Structural factors related to the thermodynamics of fatty acid binding to ALBP Enthalpic components

Positive factors 1. Coulombic interaction between ligand carboxylate and k g ' @ and 2. Removalof unfavorableCoulombicinteraction between Arg'06and in the apoprotein. 3. Hydrogen bonding between TyrlZ8 andthe ligand carboxylate. 4. Hydrogen bonding between 71 electrons of ligands and side chain atoms. Positive or negative factors 5. Dispersion forces between ligand and protein, or between ligand and preexisting water network. 6. Changes in protein electrostatic interactions as a result of changes in cavity properties uponligand binding. 7. Dispersion forces between atoms of the ligand in the free and bound state. Entropic components

Positive factors 8. Loss of water from the cavity. 9. Relief of solvent ordering as fatty acids are removed from water equilibrium. Positive or negative factors 10. Degrees of freedom lost by ligand. 11. Degrees of freedom lost by protein side chains as a result of ligand binding.

electrostatic interactions in the apoform and the favorable forces in the holo form can be determined. The electrostatic attractive forces with bound arachidonate are about 1.7 times the repulsive interactions in apoALBP. These estimates are based on the crystallographic coordinates using full positive and negative charges for the 2 arginines and the ligand carboxylate, and a +0.5 charge for the hydroxyl of TyrlZ8.The fore-mentioned estimate is possible ifone assumes that no change in the local dielectric constant occurs as a result of ligand binding. If the Coulombic and H-bonding interactions are selected as the primary driving forces in ALBP-fatty acid binding, a dilemma arises. From our studies and those of others, the fatty acid-bindingproteins appear to have affinity only for long chain fatty acids, typically those containing 12 or more carbon atoms. Short chain fatty acids seemto be incapable of binding to these cavity proteins. Hence, while the ionic interactions are an essential component of the binding process, other contacts extending down the length of the lipid must be necessary for high affinity binding. Moreover, the length of the hydrocarbon chain may not be as crucial as are thenumber of molecular contacts. Indeed, the same mutant form of ALBP that was incapable of binding a straightchain fatty acid still demonstrated binding to 12-(9-anthroyloxy)oleic acid, which differedin structureby the addition of the bulky anthracene group (14). Based on the fore-mentioned observations, dispersion forces or as they are frequently referred to, van der Waals' contacts, between the protein and the aliphatic portion of the ligand also provide a contribution to enthalpy that isimportant to binding. These are described as items 5-7 in Table VII. One of them (Table VII,no. 7) would have a negative impact on binding. An unfavorable enthalpic effect could occur as a result of the loss of dispersion forces between atoms within the ligand as the result of a ligand conformational changes upon binding to the protein. At least in the bound state, the arachidonate is clearly in a hairpin conformation. Other workers have studied conformations for arachidonate in thecrystalline (17,191 and free states (20). The conformation of arachidonate in the solid-phase studies does not seem relevant to the present structure. The isolated molecule studies by Rich (20) combined energy minimization and molecular dynamics of arachidonate. Since a protonated form of the fatty acid was used in the calculations, it too may not be comparable to the binding studies described here. We have assumed that thebound formof the fattyacid is ionized. Nonetheless, some of the lowest energy conformers

from this study were U-shaped (see for example, AA73(400),p. 92, Ref. 20). The x-ray crystallographic analysis indicates that at least one bound water molecule (Wat 3 in Tables V and VI) may be making a weak H-bond with the T bond of the bound arachidonate at position c,-c,. Thermodynamically, this H-bond would contribute favorable enthalpy to the binding process and suggests that theconserved water molecules should be considered part of the protein structure insofar as they contribute to the free energy of binding. This may explain why 12-(9-anthroyloxy) oleic acid binds to the R126L, Y128F ALBPmutant. In this case, the loss of enthalpy due to the lack of Coulombic interaction and H-bonding mayhave been compensated by an increased number of van der Waals contacts between the anthroyloxy lipid and the proteidwaternetwork. Table VI1 also lists other intuitive entropic factors associated with the binding energy. They receive potential contributions from multiple sources: protein conformational changes upon binding, restraint in thenumber of degrees of freedom forthe bound ligand, displacement of water from the protein cavity, ordering of water molecules remaining in the cavity to accommodate the ligand, and relief of solvent ordering as ligands are removed fromthe waterflipid equilibrium. Only 8 in Table VI1 is addressed by an examination of the crystal structure of the AR.ALBP or oleate.ALBP complexes.In thisinstance, a loss of water from the cavity is clearly seen by the reduced volumeof the binding cavity and is discussed further below. Protein conformation, accordingto apo versus holo comparisons, shows only verysmall changes, and probably contributes little to the overall entropy changes. The most significant factors appear to pertain to ligand and water order versus disorder. Intuitively, one might expect the change in entropy upon ligand binding to be highly unfavorable ( i e . negative), since the ligand must go from a stateof many degrees of conformational freedom in solution to one of very few in the bound state. However, as noted above, little is known about the conformational state of the fatty acid in solution. Upon ligand binding, the hydration shell of water that exists surrounding the lipid in an aqueous environment is partially replaced by apolar cavity side chain contacts and existing ordered water resulting in a positive entropy due to release of hydration sphere waters. The A(TAS) between oleate and arachidonate binding (+1120 call mol) implies that the increase in free energy associated with oleic acid binding derives primarily from entropic factors.

Interactions Acid-ALBP Arachidonic Approximately 20-25 disordered water molecules are probably displaced from the protein cavity assuming that the volume of a fatty acid is approximately 275 A3 while that of a single water molecule is estimated at 12 A3. This entropic contribution is reasonable only ifthe waters within the cavity are assumed to be more ordered than inbulk solution. The volume that arachidonate occupies within the cavity is only 8%greater than thatfor oleate suggesting that theA(TAS) between oleate and arachidonate binding does not arise from differencesin the numbers of expelled water molecules. Alternate origins for the difference in entropy may be changes in thedegrees of freedom within the lipid itself or in the hydration shell of water surrounding the free uersus bound fatty acid. No experimental data is presently available to address these possibilities. In summary, x-ray crystallographic and calorimetric studies have been used to study a protein-fatty acid complex. While detailed structuraldata of the ALBP.arachidonate and ALBP.oleate complexes is now available, little information is available about the fatty acid conformationin solution. To some extent even qualitative comparisons between the structural and thermodynamic results are difficult. Nonetheless, the enthalpic contribution derived from the structural picture of a carboxylate placed between 2 arginines and 1 tyrosine side chain appears a major factor. The entropic contribution which has been observed calorimetrically is more difficult to explain directly by the x-ray studies. The single most obvious sourceis the loss of degrees of freedom by the bound ligand. Speculation leads also to the commonly used idea that water activity is increased. This would be attributable to two phenomena: ordered water around the unbound fatty acid and the release of water from the cavity binding site. Future experiments will focus on the use of site-directed mutagenesis to alter thenumber of ordered water molecules within the ligand binding cavity.

25347 Acknowledgments-Weare gratefulto associates in the Banaszak and Bernlohr laboratories,especially Christopher Kane,for their helpN discussions. We thank Per Kraulis for the use of the drawing program MOLSCRIFT (18).We are grateful to Ed Hoeffner for his continued help in the use of the computing and x-ray diffraction instrumentation, and we are indebted to the Minnesota Supercomputer Institute for the use of their resources duringthe refinement. REFERENCES 1. LaLonde, J. M., Bernlohr, D. A,, and Banaszak, L. J. (1994) Biochemistry 33, 4885-4895 2. Hyman, B. T.,Stoll, L. L., and Spector,A. A. (1982) Bioch. Biophys. Acta 713, 375-385 3. Serrero, G., Lepak, N. M.,andGoodrich, S. P. (1992) Endocrinology 131, 2545-2551 4. Reid, T.,Ramesha, C. S., and Ringold,G. M. (1991)J. Bid.Chem. 266,1658016586 5. Tebbey, P. W., McGowan, K. M., Stephens, J. M., Buttke, T.M., and Pekala, P. H. (1994) J. B i d . Chem. 269,639-644 6. Banaszak, L., Winter, N., Xu, Z, Bernlohr, D. A., Cowan, S., and Jones, T. A. (1994) Adu. Protein Chem. 46,89-149 7. Xu, Z., Bernlohr, D., and Banaszak, L. J. (1993)J. Biol. Chem. 268.7874-7884 8. Xu, Z., Buelt, M., Banaszak, L., and Bernlohr, D. (1991) J. B i d . Chem. 266, 14367-14370 9. Buelt, M. K., and Bernlohr, D. A. (1990) Biochemistry 29, 7408-7413 10. Wiseman, T.,Williston, S., Brandts, J. E , and Lin, L. (1989)Anal. Biochem. 179,131-137 11. Briinger, A. T.,Kuriyan, J., and Karplus, M. (1987) Science 236,458-460 12. Sack, J. (1988) J. Mol. Gruphics 6, 224-225 13. Briinger, A. (1993) Acta Crystallogx D 49, 24-46 14. Sha, R., Kane, C. D., Xu, 2.. Banaszak, L., and Bernlohr, D. (1993) J. B i d . Chem. 268.7885-7892 15. Mead, J.F., ifin-Slater, R. B., Howton, D. R., and Popjak, G. (1986) Lipids: Chemistry, Biochemistry,and Nutrition, p. 44, Plenum Press, New York 16. Jakoby, M. G., Miller, K. R., Tonner, J. J., Bauman, A., Cheng, L., Li, E., and Cistola, D. P. (1993) Biochemistry 32, 872-878 17. Abrahamsson, S., and Ryderstedt-Nahringbauer,I. (19621Actu Crystallogr:16,

1261 18. Kraulis, P. (1991) J . Applied Crystallogr: 24, 946950 19. Ernst, J.,Sheldrick,W., and Fuhrhop,J-H. (1979)Z.Naturforsch. 34,706711 20. Rich, M. (1993)Biochim. Biophys. Acta 1178,87-96