J. Mol. Biol. (1997) 266, 601±609

JMB MS 2517 [10/2/97]

The 2.6 AÊ Structure of Antithrombin Indicates aConformational Change at the Heparin Binding Site

Richard Skinner1*, Jan-Pieter Abrahams2, James C. Whisstock1

Arthur M. Lesk1, Robin W. Carrell1 and Mark R. Wardell1

1Department of HaematologyUniversity of Cambridge, MRCCentre, Hills Road, CambridgeCB2 2QH, UK2MRC Laboratory of MolecularBiology, MRC Centre, HillsRoad, Cambridge CB2 2QHUK

Present address: M. R. Wardell,Biochemistry and Molecular BiophUniversity in St. Louis, School of MEuclid Avenue, St. Louis, MO 6311

0022±2836/97/080601±09 $25.00/0/mb

The crystal structure of a dimeric form of intact antithrombin has beensolved to 2.6 AÊ , representing the highest-resolution structure of an active,inhibitory serpin to date. The crystals were grown under microgravityconditions on Space Shuttle mission STS-67. The overall con®dence in thestructure, determined earlier from lower resolution data, is increased andnew insights into the structure-function relationship are gained. Clearand continuous electron density is present for the reactive centre loopregion P12 to P14 inserting into the top of the A-b-sheet. Areas of theextended amino terminus, unique to antithrombin and important in thebinding of the glycosaminoglycan heparin, can now be traced furtherthan in the earlier structures. As in the earlier studies, the crystals containone active and one latent molecule per asymmetric unit. Better de®nitionof the electron density surrounding the D-helix and of the residues impli-cated in the binding of the heparin pentasaccharide (Arg47, Lys114,Lys125, Arg129) provides an insight into the change of af®nity of bindingthat accompanies the change in conformation. In particular, the observedhydrogen bonding of these residues to the body of the molecule in thelatent form explains the mechanism for the release of newly formedantithrombin±protease complexes into the circulation for catabolicremoval.

# 1997 Academic Press Limited

Keywords: antithrombin; 2.6 AÊ structure; X-ray crystallography; heparin-pentasaccharide binding; dimer

*Corresponding author

Introduction

Antithrombin, a serine protease inhibitor (serpin)and the major inhibitor of blood coagulation, isone of only a few molecules that requires the bind-ing of the polysulphated glycosaminoglycan hepar-in for full activation (Olson & BjoÈrk, 1992). Theminimal high-af®nity binding pentasaccharide se-quence from heparin has been elucidated (Lindahlet al., 1980; Casu et al., 1981; Thunberg et al., 1982;Choay et al., 1983) and the approximate bindingsite for this sequence on the antithrombin moleculehas been mapped biochemically (for a review, seeCarrell et al., 1995). This binding site is de®ned bybasic residues of the D-helix, the amino-terminaltip of the A-helix and the ¯exible amino-terminusof the molecule, extended relative to other serpins.

Department ofysics, Washington

edicine, 660 South0, USA

960798

The site is known to undergo a conformationalchange upon binding of the heparin pentasacchar-ide (Einarsson & Andersson, 1977; Villanueva &Danishefsky, 1977; Nordenman et al., 1978;Nordenman & BjoÈrk, 1978; Olson et al., 1981;Gettins, 1987) and a further conformational changeupon docking of antithrombin to its target protease(Danielsson & BjoÈrk, 1983; Olson & Shore, 1986).

There is good evidence that the formation of thecomplex with the protease is accompanied by achange in the conformation of the reactive centreloop of the serpin from an exposed stressed (S)state to a relaxed (R) state in which the loop is in-corporated into the A-b-sheet of the molecule. ThisS! R change occurs also on reactive loop cleavageof serpins or even when loop incorporation is in-duced, as occurs here during crystallisation, to givethe latent conformation. All of these R confor-mations, complexed, reactive loop cleaved or latentforms, possess a signi®cantly lower af®nity for theheparin pentasaccharide compared to the circulat-ing inhibitory conformation (BjoÈrk & Fish, 1982;

# 1997 Academic Press Limited

602 The 2.6 AÊ Structure of Antithrombin

JMB MS 2517 [10/2/97]

BjoÈrk et al., 1992; Carrell et al., 1991). Here we pre-sent a 2.6 AÊ structure of dimeric antithrombin con-sisting of one molecule in the latent conformationand one molecule in the active conformation(Wardell et al., 1993) that shows regions of theamino-terminus previously unseen in lower-resol-ution structures (Schreuder et al., 1994; Carrell et al.,1994). In particular, the side-chain positions in thepentasaccharide binding site are better de®ned,and a comparison of the active and latent confor-mations provide a structural explanation for theloss of heparin af®nity that occurs in the transitionfrom the S to R conformations.

Results and Discussion

Refinement

The solution of a 2.6 AÊ structure of dimeric anti-thrombin has enabled several regions of the mol-ecule, critical for its function, to be de®nedsuf®ciently for meaningful conclusions to bedrawn about the structure-function relationships ofthe molecule. The ®nal model is summarised andthe stereochemical statistics of the model areshown in Table 1. The Ramachandran plot showsthat 81.3% of all residues adopt the most favouredf and c main-chain torsion angles, while eightnon-glycine residues fall outside the allowed re-gions. These residues are all found in the most dis-ordered regions of the structure and are likely toadopt multiple conformations within the crystal.The average temperature factor is 54.7 AÊ 2 for all

Table 1. Summary of merging, re®nement and stereochemic

Two molecules per asymmetric unit. One latent molecule (L) and ocontains residues 1 to 29, 43 to 432; the L molecule contains residueacetylglucosamine residue per molecule; 79 water molecules.

ParameterSpace group

Unit cell dimensions a � 61.4MultiplicityRmerge (%)

F/sResolution range (AÊ )

Total number of re¯ectionsCompleteness (%)

Number of atoms per asymmetric unit:ProteinWater

CarbohydrateTotal

Stereochemistry (r.m.s. deviation from ideal geometry)Bond lengths (AÊ )Bond angles (�)

Trigonal planes (AÊ )General planes (AÊ )Torsion angles (�)

B-value correlations (AÊ 2)

Resolution (AÊ ) 5.00± 4.57± 4.03± 3.67± 3.41±Bin 4.57 4.03 3.67 3.41 3.22

R-factor 0.15 0.16 0.18 0.21 0.24

Free R-factor 0.22 0.28 0.27 0.29 0.34

main-chain atoms and 65.2 AÊ 2 for all side-chainatoms. The loops ¯anking the D-helix (residues 106to 116 and 130 to 138) have weak electron densityin both the I and the L-molecules, their tempera-ture factors are high and they proved to be someof the most dif®cult parts of the molecule to build.This was also the case for regions of the extendedamino-terminus as will be discussed below. Resi-dues 1 to12 and 378 to 382 in the I-molecule and 7to 12 and 405 in the L-molecule were added com-pared to the those present in the 3.0 AÊ structure(Carrell et al., 1994). In addition, N-acetylglucosa-mine residues could be built onto Asn155 in bothmolecules, and 79 water molecules were includedin the ®nal structure.

The complete reactive centre loop of antithrombinin the active conformation is de®ned by the ad-ditional presence of residues 378 to 382 in themodel. The loop is partially inserted from residueP10 to P14 (notation according to Schechter &Berger, 1967) in the A-b-sheet, where it is held inplace by a network of seven hydrogen bonds. Theside-chain of the P13 residue (Glu381) is hydrogenbonded to the side-chain oxygen atom of Tyr220 ins3A, further anchoring the loop. The new structureshows that the gap between s3A and s5A beneaththe partially inserted reactive centre loop is ®lledby an ordered water molecule.

Heparin pentasaccharide binding site

The binding site on antithrombin for the heparinpentasaccharide is one of the functionally import-

al statistics

ne active molecule (I); 819 residues in total: the I molecules 7 to 12, 14 to 26, 43 to 395, 405 to 432; one N-

ValueP21

1 AÊ , b � 98.31 AÊ , c � 90.41 AÊ , a � g � 90�, b � 103.32�1.84.8

29.1730.0 to 2.6

24,64674 (53.9 between 2.8 AÊ and 2.62 AÊ )

65297928

6636

0.092.160.050.011

20.05310.834

3.22± 3.06± 2.93± 2.82± 2.72± Overall3.06 2.93 2.82 2.72 2.64 (%)

0.26 0.27 0.29 0.33 0.37 21.7

0.31 0.31 0.37 0.41 0.63 29.9

The 2.6 AÊ Structure of Antithrombin 603

JMB MS 2517 [10/2/97]

ant regions of the molecule. The higher-resolutiondata has led to an improvement of the electrondensity for the key residues involved in the bind-ing site (in particular, Arg47, Lys125 and Arg129).This binding site on antithrombin has been eluci-dated elsewhere using a combination of evidencefrom biochemistry, site-directed mutagenesis andthe study of natural antithrombin variants as wellas variant synthetic pentasaccharides (for a review,see Carrell et al.,1995). Recent recombinant studies(Kridel et al, 1996) have indicated the additional in-volvement of Lys114 in the broader heparin bind-ing site, and we demonstrate here its closeproximity and likely involvement in the core pen-tasaccharide binding site. This core binding sitehas been narrowed down to the basic residues onthe D-helix, the amino-terminal end of the A-helixand the surrounding structural features, in particu-lar residues from the amino terminus. This regionis well de®ned in the 2.6 AÊ structure and allows adetailed structural comparison between the twoconformations of antithrombin, which are knownto differ markedly in their pentasaccharide bindingaf®nities.

The D-helix seems to have relative freedom tomove, as it is ¯anked by two highly ¯exible loops.The ¯exibility of these loops is re¯ected in theirhigh temperature factors and by the fact that theirconformation differs between the active and latentmolecules. In the active conformation, which has ahigher af®nity for the pentasaccharide, the D-helixis slightly kinked and the side-chains of the keypentasaccharide binding residues Arg47 (A-helix),Lys114 (loop at base of D-helix), Lys125 andArg129 (D-helix) are solvated as their terminal Natoms are not stabilised by any intra-molecular hy-drogen bonds (Figure 1(a)). By comparison, the D-helix in the latent conformation is unkinked, longerby half a turn and the side-chains of three of thepentasaccharide binding residues are hydrogenbonded to other regions of the molecule: Arg47 tothe backbone of Leu112, Lys125 to the backbone ofIle7 and Arg129 to the side-chain of Asn45(Figure 1(b)). In addition to this, the side-chain ofLys114 points away from the pentasaccharidebinding site in the latent conformation. In thelatent molecule the electron density for Lys125 iswell de®ned (Figure 2(a)) and clearly shows theorientation of the side-chain that is hydrogen-bonded to Ile7. As seen in Figure 2(b), the side-chain of Lys125 in the active molecule points intofree solution and makes no stabilising hydrogenbond. The electron density for the Nz and Ce atomsof the side-chain is disordered, which, given thatthe residue is in free solution, is to be expected.Also, in the active conformation, Ne of Arg129 hy-drogen bonds to Glu414. This stabilises the confor-mation of the side-chain and holds it in anorientation where the amino groups are more ac-cessible to solvent and thereby are more readilyavailable to form ionic interactions with the sul-phate groups of the pentasaccharide.

Analysis of the pentasaccharide binding site hasshown that a conformational rearrangement result-ing in a signi®cant redistribution of positive chargetakes place upon adoption of the lower-af®nitylatent conformation. As a result, this positivecharge is largely ``hidden'' from the surface of themolecule. The effect of the structural rearrange-ment upon adoption of the latent conformation isto redistribute and dissipate the positive chargedensity of this region (Figure 3(a) and (b)), thusrendering it less capable of making the initial bind-ing interactions with the highly negatively chargedpentasaccharide sequence. This loss of initial elec-trostatic attraction must have a profound effect onthe binding of the pentasaccharide, as it has beendemonstrated that this binding is predominantlyionic in nature (Nordenman & BjoÈrk, 1981).

The detailed conformational change in the heparinbinding site ®ts satisfyingly with the observedchange with natural mutations of antithrombinthat result in thrombosis. Substitutions of Arg47 toCys, Ser or His and of Arg129 to Glu all result in adecreased af®nity for the heparin pentasaccharideand hence to a predisposition to thrombosis (Koideet al., 1984; Owen et al., 1987; Borg et al., 1988;Gandrille et al., 1990). Similarly, the chemicalblocking of Lys125 or its recombinant substitutionresults in a decreased heparin af®nity (Petersonet al., 1987; Fan et al., 1994), as does the recombi-nant substitution of Lys114 (Kridel et al., 1996). The2.6 AÊ structure suggests how this primary bindingsite on the D and A-helices interacts with theamino and carboxy-termini of the molecule. In par-ticular, the structure helps explain a previouslyperplexing observation with another natural mu-tant of antithrombin associated with thrombosis:Ile7! Asn (Brennan et al., 1988). This substitutioncreates an extra glycosylation site, and the associ-ated loss of heparin af®nity had been explained bya steric blocking of the adjacent binding site on theD-helix. However, although only a proportion ofthe Ile7! Asn variant becomes glycosylated at thenew site, all the variant antithrombin had a de-creased af®nity for heparin. The close associationof Ile7 to the heparin binding site, and hence thelikely consequences of any perturbation associatedwith its substitution, is indicated by the disulphidebond between Cys8 and Cys128 and the hydrogenbond between Lys125 and Ile7.

The study of natural mutants of antithrombin hasalso suggested a functional interaction of the pri-mary pentasaccharide binding site with the C-terminal portion of the molecule. There are nowknown to be a series of natural mutants in the C-terminal portion of antithrombin from 402 to 429and particularly in s1C and s4B, that result in mol-ecular instability and in a signi®cantly decreasedaf®nity for heparin (Mille et al, 1994). The hydro-gen bonding of Arg129 on the D-helix to Glu414 ins4B we report here indicates that direct interactioncan occur between the heparin binding site and theC-terminal portion of the molecule. However, this

Figure 1. A representation of the heparin pentasaccharide binding site showing the D-helix (red), the A-helix (green)and the side-chains of the pentasaccharide binding residues (blue). (a) In the active conformation the D-helix iskinked and all of the pentasaccharide binding residues are oriented towards the binding site. (b) In the latent confor-mation, the D-helix is straight and all of the pentasaccharide binding residues are either hydrogen-bonding to otherresidues or pointing away from the binding site or both. The ribbon diagram was produced by MOLSCRIPT(Kraulis, 1991).

604 The 2.6 AÊ Structure of Antithrombin

JMB MS 2517 [10/2/97]

does not, in itself, explain why perturbation of thisC-terminal domains results in a loss of af®nity.

Most satisfying is the clear explanation the struc-ture provides for the loss of heparin binding af®-nity that occurs when antithrombin undergoes theS to R conformational transition. Physiologically,antithrombin primarily functions as an anticoagu-lant in the microcirculation where it is bound andactivated by the heparan side-chains that line thecapillary surface. An essential feature of the func-tion of antithrombin is that the formation of thecomplex with a coagulation protease should be fol-lowed by its release into the circulation with sub-

sequent uptake and catabolism by the liver. Themechanism involved in the release of the anti-thrombin±protease complex is evident from thecomparison of the primary heparin binding site inthe two conformations present in the dimeric struc-ture. The change from the S to the R conformation,as occurs on formation of the complex with theprotease, is seen in the structure to result in a reor-ientation of the exposed arginine and lysine resi-dues at the primary binding site. This reorientationis such that the side-chains of the binding residuescan hydrogen bond to the body of the moleculeand hence allow the release of the complex from

Figure 2. 2Fo ÿ Fc map (contoured at 1.0s) of Lys125 and the surrounding region. (a) In the latent molecule the elec-tron density for the side-chain of Lys125 is well de®ned and the hydrogen bond between this side-chain and the car-boxyl oxygen atom of Ile7 can be clearly seen. (b) In the active molecule the electron density is less well de®ned forthe end portion of the side-chain and no hydrogen bonding can be observed.

JMB MS 2517 [10/2/97]

The 2.6 AÊ Structure of Antithrombin 605

Figure 3. (a) Worm representation of native antithrombin (left) showing the key heparin binding residues (purple) inthe D-helix region and the corresponding electostatic surface potential map (right; red, negative potential; blue, posi-tive potential) with the D-helix region circled. (b) Worm representation of latent antithrombin (left) showing the keyheparin binding residues (purple) in the D-helix region and the corresponding electrostatic surface potential map(right) with the D-helix region circled.

JMB MS 2517 [10/2/97]

The 2.6 AÊ Structure of Antithrombin 607

JMB MS 2517 [10/2/97]

the pentasaccharide domain that binds them to thecapillary heparans.

Materials and Methods

Crystallisation

For each new batch of crystallisation trials, antithrombinwas freshly prepared as described (McKay, 1981) andstored at 4�C. Before crystallisation, the protein wasbrought to a concentration of 20 mg/ml in 20 mM Tris-HCl (pH 8.0). A modi®cation of the sitting-droplet meth-od was used, where the precipitant solution was soakedinto an absorbant circular wick surrounding the sittingdroplet well. The trays were then ¯own on Space Shuttlemission STS-67 as part of the Protein Crystallisation atMicrogravity (PCAM) program run by NASA. The dro-plets were sealed from the precipitant in the wells withan elastomer membrane until the crystallisation traysreached microgravity. Before leaving microgravity thesamples were again sealed with the elastomer mem-brane. Each of the droplets contained 20 ml of protein sol-ution and 20 ml of precipitant. The duration of the ¯ightwas 14 days and on landing, the crystal trays were col-lected and examined for crystal growth. Crystals wereobtained in conditions spanning 18 to 20% (w/v) poly-ethylene glycol (PEG) 4000, 50 mM Na/KPO4, 0.05%(w/v) NaN3 (pH 6.7).

Data collection

Data were collected on an MAR image plate scanner(MAR Research, Hamburg, Germany) using a synchro-tron X-ray source (l � 1.488 AÊ , station 7.2 at the Dares-bury Rutherford Appleton Laboratory, Cheshire, UK).The crystal was maintained at a temperature of 100 Kthroughout data collection. Immediately prior to freez-ing, the crystal was immersed in a cryoprotectant (25%(v/v) methylpentane diol (MPD), 19% (w/v) PEG 4000,50 mM Na/KPO4, 0.05% (w/v) NaN3 (pH 6.7) for ®veseconds and then mounted on the goniometer head inthe N2 stream. Because of time constraints, only 100� ofdata were collected in angular increments of 1� around arandom axis in equally spaced wedges. The images wereintegrated using the MOSFLM package (Leslie, 1992)and were scaled and merged over the resolution range26.9 AÊ to 2.6 AÊ using SCALA and AGROVATA, respect-ively, both part of the CCP4 suite of programs (CCP4,1994). We attribute the shrinkage of the unit cell alongthe a-axis by 13%, compared to crystals described byWardell et al., (1993), to the transfer into the cryoprotec-tant and the subsequent freezing of the crystals. The re-sults of these steps are summarised in Table 1.

Molecular replacement and refinement

Molecular replacement was performed using the coordi-nates of the existing 3.0 AÊ structure of antithrombin(Carrell et al., 1994) as a search model with the rotationaland translation functions being determined by the pro-gram AMoRe (Navaza, 1994) within the resolution range30.0 AÊ to 3.0 AÊ , giving a ®nal solution with an R-factorof 37.5%. As with the antithrombin crystals from whichthe 3.0 AÊ structure was solved (Wardell et al., 1993), theasymmetric unit of these crystals contained a dimer con-sisting of one active molecule (I-molecule) and one latentmolecule (L-molecule) where residues 387 to 391 of thereactive site loop of the I-molecule replace residues 400

to 404 as the ®rst strand of the C-b-sheet in the L-mol-ecule.

In a rigid-body re®nement, using TNT (Tronrud et al.,1987), each of the secondary structural elements and in-terconnecting loops was treated as an individual rigidbody. The resulting model was used to calculate sA-weighted (Read, 1986), 2Fo ÿ Fc electron density maps aswell as Fo ÿ Fc difference maps. Manual rebuilding andaddition of new residues was carried out using the pro-gram O (Jones et al., 1991).

A randomly chosen 3.3% (682 re¯ections) of all re¯ec-tions was removed as a test set for the calculation of thefree R-factor (BruÈ nger, 1992), and the rebuilt model wasre®ned by conjugate direction and conventional least-squares methods using TNT. In the latter stages of re-®nement, water molecules were added on the basis ofthe difference electron density, proximity to ordered pro-tein atoms and potential hydrogen-bonding pattern,further reducing the free R-factor. From 448 potentialwater molecules examined, 79 were accepted.

Comparison and analysis of the pentasaccharidebinding region of the active and latent molecules

All backbone and Ca superimpositions were performedusing the program Pinq (Lesk, 1977, 1986). The results ofthese superimpositions were then visually inspectedusing the programs Quanta 4.1 (Molecular Simulations,Inc) and GRASP (Graphical Representation and Analysisof Surface Properties; Nicholls, 1992).

Acknowledgements

This work was supported by the Wellcome Trust, theMedical Research Council and the British Heart Foun-dation. The authors thank the Station Masters at station7.2 of the Daresbury Synchrotron Facility for their assist-ance during data collection. The atomic coordinates andstructure factors of the structure have been deposited atthe Brookhaven Protein Data Bank and have been as-signed the codes 2ANT and R2ANTSF respectively.

References

BjoÈrk, I. & Fish, W. W. (1982). Production in vitro andproperties of a modi®ed form of bovine antithrom-bin, cleaved at the active site by thrombin. J. Biol.Chem. 257, 9487±9493.

BjoÈrk, I., YlinenjaÈrvi, K., Olson, S. T. & Bock, P. E.(1992). Conversion of antithrombin from an inhibi-tor of thrombin to a substrate with reducedheparin af®nity and enhanced conformational stab-ility by the binding of a tetradecapeptide corre-sponding to the P1 to P14 region of the putativereaction bond loop of the inhibitor. J. Biol. Chem.267, 1976±1982.

Borg, J-Y., Owen, M. C., Soria, C., Soria, J., Caen, J. &Carrell, R. W. (1988). Arginine 47 is a prime heparinbinding site in antithrombin. A new variant RouenII, 47 Arg to Ser. J. Clin. Invest. 81, 1292±1296.

Brennan, S. O., Borg, J-Y., George, P. M., Soria, C., Soria,J., Caen, J. & Carrell, R. W. (1988). New carbo-hydrate site in mutant antithrombin (7 Ile-Asn)with decreased heparin af®nity. FEBS Letters, 237,118±122.

608 The 2.6 AÊ Structure of Antithrombin

JMB MS 2517 [10/2/97]

BruÈ nger, A. T. (1992). Free R value: a novel statisticalquantity for assessing the accuracy of crystalstructures. Nature, 355, 472±475.

Carrell, R. W., Evans, D. Ll. & Stein, P. E. (1991). Mobilereactive centre of serpins and the control ofthrombosis. Nature, 364, 737.

Carrell, R. W., Stein, P. E., Fermi, G. & Wardell, M. R.(1994). Biological implications of a 3 AÊ structure ofdimeric antithrombin. Structure, 2, 257±270.

Carrell, R. W., Skinner, R., Wardell, M. R. & Whisstock,J. C. (1995). Antithrombin and heparin. Mol. Med.Today, 1, 226±231.

Casu, B., Oreste, P., Torri, G., Zopetti, G., Choay, J.,Lormeau, J. C., Petitou, M. & SinayÈ , P. (1981). Thestructure of heparin oligosaccharide fragments withhigh anti-(factor Xa) activity containing the minimalantithrombin III binding sequence. Biochem. J. 197,599±609.

CCP4. (1994). Programs for protein crystallography. ActaCrystallog. sect. D, 50, 760±763.

Choay, J., Petitou, M., Lormeau, J. C., SinayÈ , P., Casu,B. & Gatti, G. (1983). Structure-activity relationshipin heparin: a synthetic pentasaccharide with highaf®nity for antithrombin III eliciting high anti-factorXa activity. Biochem. Biophys. Res. Commun. 116,492±499.

Danielsson, AÊ & BjoÈrk, I. (1983). Properties of anti-thrombin-thrombin complex formed in the presenceand in the absence of heparin. Biochem. J. 213, 345±353.

Einarsson, R. & Andersson, L. O. (1977). Binding ofheparin to human antithrombin III as studied bymeasurements of tryptophan ¯uorescence. Biochim.Biophys. Acta, 490, 104±111.

Fan, B., Turko, I. V. & Gettins, P. G. W. (1994). Lysine-heparin interactions in antithrombin. Properties ofK125M and K290M, K294M, K297M variants. Bio-chemistry, 33, 14156±14161.

Gandrille, S., Aiach, M., Lane, D. A., Vidaud, D.,Mohlo-Sabatier, P., Caso, R., de Moerloose, P.,Fiessinger, J-N. & Clauser, E. (1990). Important roleof arginine 129 in heparin-binding site of anti-thrombin III. Identi®cation of a novel mutationarginine 129 to glutamine. J. Biol. Chem. 265, 18997±19001.

Gettins, P. (1987). Antithrombin III and its interactionwith heparin. Comparison of the human, bovineand procine proteins by 1H NMR spectroscopy. Bio-chemistry, 26, 1391±1398.

Jones, T. A., Zou, J-Y., Cowan, S. W. & Kjeldgaard, M.(1991). Improved methods for building proteinmodels in electron density maps and the location oferrors in these models. Acta Crystallog. sect. A, 47,110±119.

Koide, T., Odani, S., Takahashi, K., Ono, T. &Sakuragawa, N. (1984). Antithrombin III Toyama:replacement of arginine-47 by cysteine in hereditaryabnormal antithrombin III that lacks heparin-bind-ing ability. Proc. Natl Acad. Sci. USA, 81, 289±293.

Kraulis, P. J. (1991). MOLSCRIPT: a program to produceboth detailed and schematic plots of proteinstructures. J. Appl. Crystallog. 24, 946±950.

Kridel, S. J., Chan, W. W. & Knauer, D. J. (1996).Requirement of lysine residues outside theproposed pentasaccharide binding region for highaf®nity heparin binding and activation of humanantithrombin III. J. Biol. Chem. 34, 20935±20941.

Lesk, A. M. (1977). Macromolecular marionettes. Comp.Biol. Med. 7, 113±129.

Lesk, A. M. (1986). Integrated access to sequence andstructural data. In Biosequences: Perspectives and UserServices in Europe (Sallone, C., ed.), pp. 23±28, EEC,Bruxelles.

Leslie, A. G. W. (1992). Recent changes to the MOSFLMpackage for processing ®lm and image plate data.CCP4 and ESF-EACMB Newsletter of Protein Crystal-lography, vol. 26, Daresbury Laboratory, UK.

Lindahl, U., BaÈckstroÈm, G., Thunberg, L. & Leder, I. G.(1980). Evidence for a 3-O-sulfated D-glucosamineresidue in the antithrombin-binding sequence ofheparin. Proc. Natl Acad. Sci. USA, 77, 6551±6555.

McKay, E. J. (1981). A simple two-step procedure for theisolation of antithrombin III from biological ¯uids.Thromb. Res. 21, 375±382.

Mille, B., Watton, J., Barrowcliffe, T. W., Mani, J-C. &Lane, D. A. (1994). Role of N- and C-terminalamino-acids in antithrombin binding topentasaccharide. J. Biol. Chem. 269, 29435±29443.

Navaza, J. (1992). AMoRe: an automated package formolecular replacement. Acta Crystallog. sect. A, 50,157±163.

Nicholls, A. (1992) . GRASP: Graphical Representation andAnalysis of Surface Properties, Columbia University,New York.

Nordenman, B. & BjoÈrk, I. (1978). Binding of low-af®nityand high-af®nity heparin to antihrombin. Ultra-violet difference spectroscopy and circular dichro-ism studies. Biochemistry, 17, 3339±3344.

Nordenman, B. & BjoÈrk, I. (1981). In¯uence of ionicstrength on pH on the interaction between high-af®-nity heparin and antithrombin. Biochim. Biophys.Acta, 672, 227±238.

Nordenman, B., Danielsson, AÊ & BjoÈrk, I. (1978). Thebinding of low-af®nity and high-af®nity heparin toantithrombin. Fluorescence studies. Eur. J. Biochem,90, 1±6.

Olson, S. J. & Shore, J. D. (1986). Transient kinetics ofheparin-catalyzed protease inactivation by antith-rombin III. The reaction step limiting heparin turn-over in thrombin neutralization. J. Biol. Chem, 261,13151±13159.

Olson, S. J. & Bjork, I. (1992). Regulation of thrombinby antithrombin and heparin cofactor II. InThrombin: Structure and Function (Berliner, L. J. ,ed.), pp. 159±217, Plenum Publishing Corpor-ation, New York.

Olson, S. T., Srinivasan, K. R., BjoÈ rk, I. & Shore, J.(1981). Binding of high af®nity heparin to anti-thrombin III. Stopped ¯ow kinetic studies of thebinding interaction. J. Biol. Chem. 256, 11073±11079.

Owen, M. C., Borg, J-Y, Soria, C., Soria, J., Caen, J. &Carrell, R. W. (1987). Heparin binding defect in anew antithrombin III variant: Rouen, 47 Arg to His.Blood, 69, 1275±1279.

Peterson, C. B., Noyes, C. M., Pecon, J. M., Church,F. C. & Blackburn, M. N. (1987). Identi®cation of alysyl residue in antithrombin which is essential forheparin binding. J. Biol. Chem. 262, 8061±8065.

Read, R. J. (1986). Improved Fourier coef®cients formaps using phases from partial structures witherrors. Acta Crystallog. sect. A, 42, 140±149.

Schechter, I. & Berger, A. (1967). On the size of activesites in proteases. 1 Papain. Biochem. Biophys. Res.Commun. 27, 157±162.

Schreuder, H. A., de Boer, B., Dijkema, R., Mulders,J., Theunissen, H. J. M., Grootenhuis, P. D. J. &Hol, W. G. J. (1994). The intact and cleaved

The 2.6 AÊ Structure of Antithrombin 609

JMB MS 2517 [10/2/97]

human antithrombin III complex as a model forserpin±proteinase interactions. Nature Struct. Biol.1, 48±55.

Thunberg, L., BaÈckstroÈm, G. & Lindahl, U. (1982).Further characterization of the antithrombin-bind-ing sequence in heparin. Carbohydr. Res. 100, 393±410.

Tronrud, D. E., Ten Eyck, L. F. & Matthews, B. W.(1987). An ef®cient general-purpose least-squaresre®nement program for macromolecular structures.Acta Crystallog. sect. A, 42, 489±501.

Villanueva, G. B. & Danishefsky, I. (1977). Evidence fora heparin-induced conformational change on antith-rombin III. Biochem. Biophys. Res. Commun. 74, 803±809.

Wardell, M. R., Abrahams, J-P., Bruce, D., Skinner, R. &Leslie, A. G. W. (1993). Crystallisation and prelimi-nary X-ray diffraction analysis of two confor-mations of intact antithrombin. J. Mol. Biol. 234,1253±1258.

Edited by R. Huber

(Received 18 July 1996; received in revised form 12 November 1996; accepted 12 November 1996)