J. Mol. Biol. (1991) 221, 583-601

Refined Structure of the Hirudin-Thrombin

Complex

Timothy J. Rydelt, Alexander Tulinskyl Department of Chemistry, Michigan East Lansing, MI 48824,

State University U.S.A.

Wolfram Bode and Robert Huber Max-Plan&-Institute fur Biochemie Da033 Martinsried, Germany (Received

11 February

1991; accepted 13 May

1991)

The structure of a recombinant hirudin (variant 2, Lys47) human a-thrombin complex has been refined using restrained least-squares methods to a crystallographic R-factor of 0.173. The hirudin structure consists of an N-terminal domain folded into a globular unit and a long 17.peptide C-terminal in an extended chain conformation. The N-terminal domain binds at the active-site of thrombin where Ilel’ to Tyr3’ penetrates to the catalytic triad. The x-amino group of He1 of hirudin makes a hydrogen bond with OG of Ser195 of t,hrombin, the side-chains of Ilel’ and Tyr3’ occupy the apolar site, Thr2’ is at the entrance to, but. does not enter, the Sl specificity site and Ilel’ to Tyr3’ form a parallel j-strand with Ser214 to Gly219. The latter interaction is antiparallel in all other serine proteinase-protein inhibitfor complexes. The extended C-terminal segment of hirudin, which is abundant, in acidic residues, makes many electrostatic interactions with the fibrinogen binding exosite while the last five residues are in a 3,, helical turn residing in a hydrophobic patch on the thrombin surface. The precision of the complementarity displayed by these two molecules produces numerous interactions, which although independently generally weak, together are responsible for the high degree of affinity and specificity. Although hirudin-thrombin and nPhe-Pro-Arg-chloromethyl ketone-thrombin differ in conformation in the autolysis loop (1~~~145 to Gly150), this is most likely due to different crystal packing interactions and changes in circular dichroism between the two are probably due to the inherent flexibility of the loop. An RGD sequence, which is generally known to be involved in cell surface receptor interactions, occurs in thrombin and is associated with a long solvent channel filled with water molecules leading to the surface from the end of the Sl site. However, the RGD triplet does not appear to be able to interact in concert in a surface binding mode. Ke2/words: hirudin:

thrombin;

blood

1. Introduction Thrombin (EC 3.4.21.5) is a glycoprotein that is representative of a sub-class of serine proteases that play a central role in thrombosis and hemostasis. It is generated in the final events of blood coagulation from prothrombin (Mann, 1987) where it converts fibrinogen into clottable fibrin in the

t Present address: Miami Valley Laboratories. The Procter & (iamble (Jo., PO Box 398707. Cincinnati, OH 45P39-8707, 1’.S.A. $ Author t,o whom all correspondence should be addressed.

coagulation;

inhibition;

exposite

formation of t#hrombi by exhibiting specificity largely attributed to an anion-binding recognition exosite that is distinct from the catalytic site (Fenton, 1981, 1986). Although this exosite is also implicated in thrombin interactions with other substances, heparin binds to a different exosite (Church et al., 1989). Thrombin also activates other coagulation factors such as V, VIII. XTII and protein C (Fenton, 1981) but the processes are modified when thrombin associates with thromhomodulin (Esmon et al., 1986). a-Thrombin consists of two polypeptide chains of 36 (A-chain) and 259 (Bchain) residues linked by a disulfide bridge (Table 1: Butkowski et al., 1977; Thompson et al., 1977; Degen et al., 1983). The crystallographic structure of

584

1’. J. Rydel et al.

Table 1 Summary

of final restrained least-squares paran~eter.~ldeviatio~Ls a?Ld R-factor

statistics

Distances (A): Bond lengths Bond angles Planar l-4 Planes (8): Peptides Aromatic groups (Ihiral volumes (A3): Non-bonded contracts (A): Single torsion Multiple torsion Possible H-bond Thermal parameters (A2) Main-chain bond Main-chain angle Side-chain bond Side-chain angle Torsion angles (deg) Planar Staggered Orthonormal Diffraction pattern a(lF,I) = A+U(sinfI/1-l/6) < llF,l-lE’,ll > = 54 dmin (A) 413 3.39 3.00 2.75 2.56 2.42 2.30

Reflection number

Wol)

< IIE’,F,II ’

K-factor shell

R-factor sphere

3043 3380 3335 3160 3112 2826 2200

34 30 27 25 23 22 21

89 64 56 47 43 39 38

0~150 0136 @I76 0.190 0212 0235 0265

ti150 0143 @151 0157 0163 0168 0,173

human a-thrombin inactivated with PPACK(t) has been determined at 1.9 A (18 = 0.1 nm) resolution (Bode et al., 1989a), where it was shown to possess structural similarity to other trypsin-like serine proteases but with insertion loops centered around Asp60E. Va1149C and Gly186C (Table 1); the 60A-I insertion loop protrudes into the active-site vicinity and, among other things, narrows the substrate cleavage binding cleft. a-Thrombin is susceptible to proteolytic cleavage by trypsin and autolysis at Arg77A (Fenton et al.. 1977). This form has been designated B-thrombin and can undergo further cleavage at Arg67 with the t Abbreviations used: PPACK, n-Phe-Pro-Argchloromethyl ketone; rHV2-K47. recombinant hirudin variant 2-Lys47; Cysls’. prime trailing residue number designates hirudin; PEG, polyethylene glycol; NAG, X-acetylglucosamine: c.d., circular dichroism; BPTI, bovine pancreatic trypsin inhibitor; n.m.r., nuclear magnetic resonance; r.m.s.. root-mean-square.

concomitant loss of the undecapeptide (Boissel et al., 1984). The loop from Lysl45 to Lysl49E (Fig. 1) is also highly sensit’ive to proteolytic cleavages, including a,utolysis, and has often been referred to as the autolysis loop. Thus, in addition to the p-cleavage, y-thrombin has a break at Lys149E. Other forms of thrombin are &-thrombin, resulting from limited proteolysis with elastase and characterized by a single cleavage between Ala149A-Asn149B (Kawabata et al., 1985) and [thrombin. produced by cathepsin G or chymotrypsin with a cleavage at Trp148 (Brezniak et al., 1990). All of the cleaved forms of thrombin exhibit reduced clotting activity. The principal inhibitor of thrombin in blood circulation is antithrombin III (Travis & Salvesen, 1983), however, the most potent natural inhibitor of thrombin is hirudin, a g&residue protein from the medicinal leech Hirudo ,medicinalis European (Markwardt, 1970). By virtue of disulfide links, hirudin consists of a compact N-terminal domain

Refined Structure

NN2-T

qf the Hirudin-Thrombin

CCLRPLTKKKSLED D

I G

R-CODE G D I Y s

K

A s

Complex

T K

E

R

G

K E

L L B-Chain

D

w

I[

P

P

P

N YT

S

p2-IVKGSDAKIGYSPWQVHLTRKPQKLLCGASLISDRWVLTAABCLLKNDLL 16 25 35

45

55

1

60

65

L

R

VRIGKBSRTRYLNIKKIS~LKKIYIBPRYNWRNLDRDIALNKLKKPVAF 85 75

95

105 V NG

------I ASL SDYIEPVCLPl~5RETALQAGYKGRVTGWGNLKKTWTGQPSVLQVVNLPIVE 135 115

AK

145

KG Y DK RPVCKDSTRIRITDNnlCAGKPRGDACKGDSGGPFVMKSPNRWYQNGIVS 185 175 165

195

R WGE-GCDDGKYGIYTEVCRLKKUIQKVIDQFGE-Cm6 225 215

245

235

155

TN 205

Figure 1. Sequence of a-thrombin. Insertions with respect to chymotrypsin indicated as protrusions from linear chain and designated with letters in text; as described in Bode it al.. 1989a).

possessing a two-disulfide, double-loop structure (B, C, D, Fig. 2) preceded by an ordinary disulfide loop (A) and followed by a long 26-residue C-terminal chain. Unlike antithrombin III, however, hirudin is specific for thrombin forms with fibrinogen activity (Fenton et al., 1979) with which, in the case of a-thrombin, it forms a remarkably stable non-covapossessing a lent, stoichiometric, 1 : 1 complex binding constant reportedly as low as 2 x lo-l4 M (Stone & Hofsteenge, 1986)) although catalytically active thrombin is not required for complexation (Stone et al.: 1987). Kinetic and equilibrium studies indicate that hirudin interacts simultaneously with the catalytic site and the fibrogen binding exosite of thrombin (Fenton, 1981; Chang, 1983). Studies with synthetic peptides have shown that the exosite also binds with the C-terminal decapeptide and related fragments of hirudin (Krstenansky & Mao, 1987; Ni et al., 1990) and that it contains a number of lysyl residues of thrombin (Chang, 1989). Whereas natural hirudins are mixtures of variants, recombinant techniques produce homogeneous preparations. The recombinant proteins lack a sulfated Tyr63; however, they only have about tenfold reduced affinity (Dodt et al., 1988; Braun et al., 1988) and possess Ki values in the pica-molar range (Degryse et al., 1989). The hirudin used in the crystallographic structure determination described here of the hirudin-thrombin complex was a recom-

binant form, variant 2 with Lys47 (Harvey et al., 1986). It differs from the most abundant natural iso-inhibitor in eight residues and by the absence of a sulfated tyrosine residue at position 63 (Fig. 2).

Figure 2. The sequence of recombinant hirudin variant 2-lysine 47 (rHV2-K47). Residues indicated outside circles are those of the most abundant natural iso-inhibitor.

T. J. Rydel et al.

586

Two solution structures of recombinant hirudins have been detemined by n.m.r. (Folkers et al., 1989; Haruyama & Wuthrich, 1989). Except for residues Gly31’ to Gly36’, the structure of the N-terminal domain of hirudin (residues 3’ to 48’) was fixed by n.m.r. while the remainder of the C-terminal was completely disordered (16 residues). We report here the highly refined X-ray crystallographic structure of the rHVZK47 human a-thrombin complex at 2.3 A resolution. A preliminary report of the structure determination has already appeared elsewhere (Rydel et al., 1990) as was a lower resolution structure of a variant 1 complex (Grutter et al.. 1990). A comprehensive report of the thrombin structure has been given by Rode et al. (1991). Unlike the n.m.r. solution struttures, the C-terminal of hirudin is structurally ordered in the complex. Tn contrast to predictions (Fenton & Bing, 1986; Johnson et nl., 1989). the N-terminal tripeptide of hirudin interacts with t’hr catalytic site of thrombin and Lys47’ does not occupy the specificity pocket while the interaction of the [:-terminal chain with the fibrinogen exosite of thrombin is confirmed.

2. Materials

and Methods

(‘rystals of t,he rHVSK47 human cc-thrombin complex were grown by vapor diffusion using PEG 4000. MgCl, and acetate buffer at pH 4.5 (Rydel et al.. 1990). The crystals are tetragonal. n = b = 9@39 a, c = 132.97 ,A. space group P-I,,,*) 2 l with 8 molecules per unit cell (protein fraction approx. 39yd). S-ray diffraction intensit) data were collected at, 6°C’ to 2.3 a resolution with an Enraf-Xonius FAST television area drtect,or using a Rigaku rotating anode S-ray source operat’ing at 5% kI%‘. Pertinent statistics regarding the diflraction data are summarized elsewhere (Rydel rt al.. 1990). The measurrments are fairl,v romplete. show excellent internal consistency and are observable out, to the diffraction limit. Patterson methods were used to d&ermine the orietrtation of the thrombin molecule in crystals of t.he nomples (Rossman & Blow. 1962: Huber. 196.5). The rotation search was performed with the thrombin structure of PPACK-thrombin (Bode rt al.. 1989a). The position of the t,hrombin molecule in the unit, cell was fixed by a translation search with programs written by Lattman (1985), modified by J. Deisenhofer & R,. Huber, using intensity data from 8.0 t,o 3.0 A resolution. These calculations established the position of the molecule and the correct enantiomorphic space group as well. The crystallographic R-value (R = XIIF,[-IF,ll/ClF,I) of this solution was 039. The rotational and the translational paramet’ers were refined further with t,he rigid body refinement program TRAREF (Huber & Schneider. 1985), reducing K to 031. The resulting model of the complex was refined using the energy-restraint crystallographic refinement program EREF (Jack & T,evit,t. 1978) with interac+ivr 1978). The graphics interventions using FRO110 (Jonrx cyclic procedure was repeated 5 times and the R-value (*onverged at @193. The refinement was c.ontinued using restrained leastthe program PROFFT squares procedures with (Hendrickson & Konnert, 1980: Finzel. 1987). The latter takes into account certain shortcomings associat’ed with EREF refinement. such as ability t’o restrain thermal

-90

-0

-90 L______.--___-

F

_1__________, I

-180

I

*

I

,

I

I

-90

I

0

I

,

,

90

I

I

1-180

180

PHI

Figure thrombin.

3. Ramachandran (ilycine

residues

4. rl/ angles not, displayed.

of hirudin

factors between at,oms. I)rI)tidr planarity and intrrmolecular contacts, along wit,h an option to refine orcupanties. In addition. since there was sufficient electron density in the last map near NIX of AsnCiOC+to model the 1st NA(: moiety of the oligosaccharide attached to thrombin. it was included into subsequent calrulations. The PROFFT refinement began at) 2.5 a resolut,ion with water st’ructure in the calculation whirh l)rovrd to he illconditioned showing many large positional and OIYUpancy shifts. Therefore. all the solvent was removrcl (R = 0.27): the strurture then refined quirkly and smoothly to K = 0.23. At this stage. water molecules were included again. These were chosen if they appeared in the (2lF,l-lF,l) density and in both the X.0 t,o 2.5 ip and 7.0 to 2.5 A resolut,ion (IF,/-lF,l) difference maps (>25 0). and it they were in the proximity of a hydrogen donor OI acceptor. The inclusion of 76 wat,er molrcules followed b?; 30 cycles of refinement rrduc>ed R tjo 0.20. The resolution was then including

increased to 2.3 A and the refinement c-ontinued. progressively more solvent along with graphical

interventions for model rehuilding purposes, The strut.ture ultimately converged at, K = 0.173 with an average thermal parameter of 35 a2 using 265 wat,er molecules and 21,056 reflertions from 7.0 t,o 2.3 14 resolut,ion. The final target parameters of the refinement and t,heir r.m.s. values are listed in Table 1. These values correspond to 2336 atoms (14 for the NAG of carhohydrak) of thrombin (96?6); 447 of hirudin (89’+,) and the water molerules all of which were defined in terms of 12,459 variahlrs (I.7

observations/parameter). The cl,-angles of 91 O. and only 3 residues are not in conformationally allowrtl regions (Fig. 3): all of t*hesr occur in the A-chain with Serl E and AspI-CL being adjacent to the t,ermini of t,hca chain. The fit of the 1st hrxose of the carbohydrate c.haitl of thromhin to the electron densit’y was good but additional density for t,he remainder of t,hr chain tlicl not develop during the course of the refinement. inclicaating that it, was disordered in the interstit,ial solvent region between molecules in the crystal. From an examination of the R-factor as a function nf scattering angle (Tahlr I) a

Refined Structure of the Him&n-Throw&in

Complex

587

Figure 4. Stereoview of the CA structure of the hirudin-thrombin complex. Hirudin is shown as a bold line; disordered residues are shown as broken lines; N and C-terminal of thrombin and C-terminal of hirudin designated: N-terminal of hirudin in active site A; B and C are fibrinogen binding exosite and autolysis loop, respectively.

co-ordinate error (Luzzati. 1952) of about @20 to 025 L% has been estimated for the structure of the complex?.

3. Results (a) Structure of hirudin of hirudin in the The first 48 residues hirudin-thrombin complex are organized into a compactly folded domain similar to that observed in solution by n.m.r. (Folkers et al., 1989; Haruyama &

Wuthrich, 1989). However, the C-terminal chain of hirudin, completely disordered in solution, is welldefined in the complex and in an extended conformation (Fig. 4: Rydel et al, 1990). The first stretch (Glu49’ to Gly54’) makes numerous electrostatic and polar interactions with thrombin while the second (Asp55 to Gln56’) begins likewise but finishes in a distorted 3,c helical turn that resides in a hydrophobic patch on the thrombin surface. The conformation of the N-terminal domain of hirudin is related to close intramolecular contacts of a three-disulfide core (Fig. 5). The disulfide bonds of CysS’-Cysl4’ and Cysl&Cys28 are nearly perpendicular (distance between disulfide midpoints of 497 A) while Cys16’-Cys28’ and Cys22’-Cys39’ are nearly parallel (disulfide midpoint separation of 5.35 A). Similar close disulfide interactions have also been observed in C3a anaphylatoxin (Huber et al., 1980), in kringle structures of blood coagulation/ fibrinolysis (Tulinsky et al., 1988; Mulichak & Tulinsky, 1990) and in squash-seed trypsin inhibitor (Bode et d., 19893). A list of the sulfur-sulfur distances in the hirudin molecule is presented in Table 2. The net result is that the loop segments B, C and I), which comprise a double loop in hirudin (Fig. 2), fold into three unique loops. This, t The coordinates of the hirudin-thrombin have been desposited in the Protein numbers 1HTC and SHTC!).

Data

complex

Bank

(entry

combined with the close stacking of the other disulfide bond (Cys6’-Cysl4’). produces four threedimensional loops (Fig. 5). A solvent-accessibility calculation (Lee & Richards, 1971) indicates essentially zero accessibility for all the disulfide bonds of hirudin except possibly Cys39’, which, in any case, is very small (18% side-chain, 33% totad). The N-terminal domain possesses short stretches of antiparallel B-structure (Table 3). The j?l strands have two hydrogen bonds while another hydrogen bond precedes the strand (Table 4). The individual ribbons of this element are connected together by a type II’ reverse turn (Crawford et al., 1973; Tl of Table 3). A type II turn (T2) is present in loop C and a longer antiparallel p-finger (/IS of Table 3) is generated in loop D (Figs 4 and 5), which contains four hydrogen bonds (Table 4). Although a reverse turn must join the strands of this finger together (T3 of Table 3), this turn is also disordered in the crystal structure (Figs 4 and 5: Rydel et al.,1990), as it is in solution (Folkers et al., 1989; Haruyama &

Wuthrich, 1989). In addition to the disulfide core, many intramolecular hydrogen bonds also stabilize the N-terminal domain of hirudin (Table 4). These were chosen using distances less than 3.1 A and donorhydrogen-acceptor angles greater than 125” as criteria. In all, 22 hydrogen bonds occur in the domain, of which eight are in turns1 and p-structure. The hydrogen-bond interactions involving Lys47’ appear to be important in extending and terminating this globular domain out to and at residue Pro48’, respectively. Although there are a number of main-chain hydrogen bonds between Cys39’ and Pro48’, which is additionally a conserved stretch of sequence, and the remainder of the

1 Although the hydrogen bond distance is 3.3 A in Tl and T2, these turns are very clear and imply the hydrogen bond.

588

T. J. Rydel et al.

1’

) 6

\ 48’

8’

Figure 5. Stereoview of the folding of the hirudin K-terminal domain. The A to I) loops correspond to those of Fig. I: disulfide bridges and side-chains of Thr4’, Asp5’ and Lys47’ are shown as bold lines; hydrogen bonds and ill-defined residues 32’ to 35’ are shown as broken lines.

domain, the a-amino group of Lys47’ is involved in two apparently crucial hydrogen bonds with residues of the N-terminal pentapeptide of hirudin (Thr4’ and Asp5’: Fig. 5); moreover, Lys47’ N also hydrogen bonds to Asnl2’ 0 (Table 4). Such interactions bring the N and C-termini of the domain in relatively close proximity, increasing the size but maintaining the compactness of the domain. The two proline residues that flank Lys47’ on either side might well help maintain its position. These three residues also initiate a polyproline II helix (Hl: Table 3) like that of collagen (Yonath & Traub, 1969) which terminates at His51’. The folding of the N-terminal domain in the complex is similar to that observed by n.m.r. (Folkers et aE., 1989; Haruyama & Wuthrich, 1989). An optimal superposition of 38 CA atoms with the average of 32 n.m.r. structures? gave an r.m.s. difference of 0.86 A. Although the r.m.s. difference in the side-chain positions increases to about 1.95 A, it is comparable with the r.m.s. deviations from the average n.m.r. structure (Folkers et aE., 1989). The most serious discrepancy with respect to the crystallographic structure occurs in the disulfide positions (r.m.s. = 2.4 A) that are fairly well determined in the complex. Two of the disulfide bonds are displaced by about a half bond length and one bond t We thank Dr G. Marius Clore for providing us with the n.m.r. coordinates of the average solution structure of hirudin prior to their distribution by the Protein Data Bank.

length from the n.m.r. structure, respectively, while Cys16’-Cys28’ has a very different orientation in the complex: x . . x’ = -165, 81, 85. 76. -175” (complex); -60, -169, 75, -168, -40” (n.m.r.). The otherwise close correspondence between the n.m.r. and the crystal structures indicates that the N-terminal domain changes little when it is complexed with thrombin. The C-terminal of hirudin consists of t.wo extended stretches of chain with a bend at Asp55’ (Fig. 4). The first segment is approximately 18 A long. The initial residues are the last three of the polyproline helix while the last three residues have poorly defined electron density. The second segment of the C-terminal is approximately 19 A long and consists of 15 A in extended conformation (Asp55 to Pro60’) followed by a type III 3,, reverse turn (Crawford et al., 1973; Table 4). The extended

Table 2 h’u@~r-sulfur distances (if) Atom 6’ 14 16’ 22’ 28’ 39

6’

in hirudin

14’

lti’

p’

2411 *

4.72 6.43

6-11 856 4.70

* Indicates a disulfide

hand.

28’ 3.74 543 UK* 548

39’ 7.65 8.32 .522 205* 65H

Refined Structure of the Hirudin-Thrombin

589

Complex

Table 3 Secondary Element

structural elements of hirudin Residues

Type

~-strurturr Pl B2

Antiparallel Antiparallel

Helix Hi

Polyproline

Rrvrrsr Tl T2

Cys14’ to (‘~~16’: Asn20’ to Cys22’ Lys27’ t>o Gly31’; Gly36’ to Va140’

II

turns

Type 11

(:lu17’. (:ly18’, SerlS’, AsnXO Glp23’, Lys24’, Gly25’, AsnAG

1

Ser32’. Ssn33’.

Type II’

T3 T4

Type

(In) Hirudin-thrombin

i,nteraction

of the hirudin-thrombin The CA-structure complex is shown in Figure 4 where principal subsites and loops are designated. The docking of the N-terminal domain with thrombin is similar globally to the manner in which other natural serine prot)einase protein inhibitors interface with protease enzymes. Additionally, an important component to the binding in both is in or near the active site region. However, whereas the C-terminal peptide of hirudin is disordered in solution, it is well-defined in the complex where it extends across the surface of the t.hrombin molecule for 40 A terminating

Table 4 bonds of the X-terminal

Donor

(‘ys6’ Thr7’ c:1ux (:Iyio’ (iIn I’ Asnl:’ , lJPU13 11PU I I5’ (+:i,*)’ ._-

N N N N SE” s r h N s

Lysr’i (‘,vs”x’ 1Ie”!l’ l,ruXY (:I1138 c*ys:w V&O’ (:lp42’ (:Iy44’ ‘l’hr4.5’ I&7’ Lys47’ Lvs47’

s s N N N N N N N S s 2;71 S%

Acceptor Lru1.5 (+lnll’ (:lnl I’

0 OEl OEI

(‘vs48’ 0

(h’

Thr45’ (‘ys22’ Thr4’ (‘ysl.?’ Yal40’ (ilnll’ (+ln38’ Sri-9 Ilr”9’ (‘luli’ 1:ys?i ( :1,v2