J. Mol. Biol. (1990) 212, 541-552

Solution Structure of the Kringle 4 Domain from Human Plasminogen by ‘H Nuclear Magnetic Resonance Spectroscopy and Distance Geometry K. A. Atkinson and R. J. P. Williams Inorganic Chemistry Laboratory, University of Oxford South Parks Road, Oxford OX1 3&R, U.K. (Received 21 June 1989; accepted 29 November 1989) Kringle 4 is an autonomous structural and folding domain within the proenzyme plasminogen. Homologous domains are found throughout the blood clotting and fibrinolytic proteins. In this paper, we present the almost complete assignment of the lH nuclear magnetic resonance (n.m.r.) spectrum of the kringle 4 domain of human plasminogen. A detailed structural analysis has been completed. The sequential pattern of nuclear Overhauser enhancements indicated little regular secondary structure but rather a series of turns and loops connecting /J-strands. A small stretch of antiparallel p-sheet was identified between the residues 61 to 63 and 71 to 73 and the close proximity of other strands was determined from two-dimensional nuclear Overhauser enhancement spectra. Slowly exchanging amide (NH) resonances were found to be associated with residues of the /?-sheet and neighbouring strands that support the hydrophobic core of the domain, A total of 526 interproton distance constraints and two hydrogen bonds were specified as input to the distance geometry program DISGEO. Tertiary structures were produced that were consistent with the n.m.r. data. The structures were compared with that of our earlier model fragment 1 determined based on n.m.r. studies and with that of prothrombin crystallographically.

1. Introduction

These studies showed the kringle domain to contain little regular secondary structure but rather to consist of a loose assembly of loops and turns whose conformation is defined by the disulphide bridges, a number of hydrogen bonds and a small piece of antiparallel P-sheet on which the hydrophobic core of the lysine-binding site is based. Other workers have made further assignments by comparison of human plasminogen kringle 4 with bovine, porcine and chicken homologues (Ramesh et al., 1986;

Plasminogen (Mr x 90,000) is the physiological precursor to the fibrinolytic serine protease sequence contains five plasmin. Its primary “kringle” domains prior to the protease domain (Sottrup-Jensen et al., 1978). These homologous domains are also found in a number of other proteins of the blood clotting and fibrinolytic systems (prothrombin, tissue plasminogen activator, urokinase, factor XII, apolipoprotein (a); see Blake et al., 1987). They are found to be autonomous structural and folding domains and are believed to mediate the binding of these proteins to other proteins (Patthy et al., 1984), e.g. plasminogen to fibrin (Wiman & Wall&n, 1977). The primary sequence of the kringle domain is characterized by a set of three disulphide bridges. A number of kringle domains have been found to bind lysine and its analogues. The kringle 4 of human plasminogen (N, z 10,000) domain contains such a binding site (Ryan, 1987). In previous nuclear magnetic resonance (n.m.r.t) studies, the resonances of a number of residues have been assigned (Trexler et al., 1983, 1985; De Marco et al., 1985; Mabbutt & Williams, 1988) and a model of the kringle 4 domain was constructed largely on the basis of side-chain interactions (Esnouf et al., 1985). 0022~2836/90/070541-12 $03.00/O

TAbbreviations used: n.m.r, nuclear magnetic resonance spectroscopy; nOe, nuclear Overhauser enhancement; p.p.m., parts per million; COSY, 2-dimensional correlated spectroscopy; NOESY: S-dimensional nOe spectroscopy; RELAY, S-dimensional relayed coherence transfer spectroscopy; DOUBLERELAY, 2-dimensional double-relaye’d coherence transfer spectroscopy; TOCSY, 2-dimensional total correlation spectroscopy; r.m.s.d., root-meansquare distance; nOe connectivities are denoted by d,,, where A and B indicate the relevant protons (see also legend to Fig. 3); interatomic distances are denoted by d*-a. Standard 3-letter abbreviations are used to name amino acids in the text; single-letter codes are used in Figures and Tables for clarity. The residue numbering system used throughout is that of Trexler et al. (1983). Pu’ote the deletion at position 59. 541

0 1990 Academic PrtsssLimited

542

R. A. Atkinson

and R. J. P. Williams

Petros et al., 1988) and using photo-chemically induced dynamic nuclear polarization ‘H n.m.r. spectroscopy (De Marco et al.; 1986, 1989). Nuclear magnetic resonance studies of a number of other kringle domains provided evidence of structural similarity between the domains (Trexler et al., 1985; Motta et al., 1987; Thewes et al., 1987; Oswald et al., 1989). The structure of only one kringle domain (that contained in prothrombin fragment 1) has been determined crystallographically (Olsson et al., 1982; Park & Tulinsky, 1986; Harlos et al., 1987) and refined to 0.28 nm (Tulinsky et al., 1988). The binding sites of kringles 1 and 4 of plasminogen have been modelled using these co-ordinates (Motta et al., 1987; Ramesh et al.; 1987). The outline structure of kringle 4 from plasminogen (Esnouf et al., 1985) has been compared to the unrefined crystal 1 (Williams et structure of prothrombin fragment al.: 1986). We are now able to present a considerably more detailed comparison. The development of two-dimensional n.m.r. techniques (Ernst et aZ., 1987) and their application to proteins (Wiithrich, 1986, 1989) has been well documented elsewhere. Two approaches to the sequential assignment of the ‘H n.m.r. spectra of proteins have arisen. The first is based on the identification of spin systems in experiments that reveal J-correlated interactions (Neuhaus et al., 1985) followed by placement of the spin systems in sequence t’hrough detection of sequential nuclear Overhauser enhancements (noes) (Wiithrich, 1986; Chazin et al., 1988; Chazin & Wright,, 1988). The second approach relies on the initial identificat,ion of only a few spin systems (e.g. those of aromatic residues). Sequential noes are used to identify runs of “main-chain” (or “backbone”) resonances whose position in the sequence is indicated by the resonances of the known residues (Englander & Wand, 1.987; Saudek et al., 1989a,b). The side-chain spin systems are then traced and c.hecked against the known sequence. This latter approach has been applied here to a system in which some of the residues have been assigned in previous studies in the absence of ligands (Mabbutt & Williams, 1988). In this study, spectra were recorded in the presence of the ligand 4-(aminomethyl)benzoic acid. We shall relate the results presented here to those obtained from kringle 4 in t’he absence of ligand at a later date. nOe data provide a measure of interproton distances. The detection of a la,rge number of noes between assigned proton resonances allows the determination of tertia,ry structure through the use of distance geometry algorithms (Have1 et al., 1983; Braun & G6, 1985; Wagner et al., 1987). These methods have been used successfully to determine the solution structure of a number of small proteins in the absence of crystallographic data (Williamson et al., 1985; Cooke et al., 1987; Saudek et al., 1989aJ). In this paper, we describe the use of the program DISGEO (Have1 & Wiit’hrich, 1984) to calculat,e the tertiary st’ructure of the kringle 4 domain from human plasminogen.

2. Materials (a) Preparation

and Methods of kringle

4 dopmain

The kringle 4 domain of human plasminogen was isolated by autodigestion of plasmin to produce the heavy chain (Lawrence, 1987), followed by digest,ion with elastase and gel filtration (Sottrup-Jensen et al., 1978). (b) Z&ear

,magnetie

resonance

experiments

The n.m.r. experiments were performed using samples containing 4 x 10m3 x-kringle 4 dissolved in ‘HZ0 or ‘Hz0 containing up to 30% (v/v) “H,O, with @l ;Mand 10 x 10e3 iv-4-(aminomnethyl)chloride sodium benzoic acid. Samples of solutions of sodium deuteroxide and deuterium chloride were used to adjust the pH of the samples to pH 67 (meter reading uncorrected for isotope effect). Dioxan was used as an inbernal reference set at 3.75 p.p.m. relative to trimethylsilpl-propanesulphonic acid. All spectra were recorded at 37°C and 500 MHz on a GE/Xcolet spect’rometer equipped with an Oxford Instruments superconducting magnet and a ‘H 5 mm probe. All experiments (COSY (Aue et al.; 1976), RELAY and DOUBLERELAY (Eich et al.; 1982; Wagner, 1983), TOCSY (Braunschweiler & Ernst, 1983; Bax & Davis. 1985), POESY (Jeener; et al., 1979; Anil-Kumar et al.. 1980), PreTOCSY COSY a.nd PreTOCSY NOESY (Otting & Wiithrich, 1987)) were carried out in the phase-sensitive mode using the method of States et al. (1982). with the transmitter offset on the ‘HO’H resonance and w&h CYCLOPS phase cycling (Hoult & Richards. 1975). The ‘HOIH resonance was suppressed by irradiation during the relaxation time (typically I.0 s) and durjng the mixing time, except for the PreTOCSY POESY experiment when irradiation was applied only in t’he last l/3 of the mixing time. NOESY spectra were recorded with mixing times of 80. 120, 160, 190 and 240 ms and the PreTOCSY NOESY spectra with a mixture time of 160 ms, all with a lOgb random variation of the mixing time. RELAY, DOUBLER,ELAY a.nd TOCSY spectra were recorded with total mixing times of 30; 60 and 45 ms, respectively. Typically, 512 t, increments were collected for J-correlated spectra and 256 increments for the NOESY experiments, with 64 transients/increment. The data, were multiplied by t’rapezoidal and double exponential functions in t, and either similar fun&ions in t, or a shifted cosine bell prior to zero filling and Fourier transformation. The final digital resolution for spect,ra recorded in ‘HZ0 was 3.91 Hz/pt, il. each dimension and 2.93 Hz/pt for spectra recorded in ‘H,O. All spectra shown (Figs 1 and 2) are unsymmetrized with the Fl dimension horizontal.

(b) Amide exchange Slow XH exchange was measured by freeze-drying the sample of kringle 4 containing 4-(aminomethyl)benzoic acid in ‘H,O and dissolving it in ‘H,O. The first spectrum was acquired 12 min after dissolution. One-dimensional spectra and short COSY experiments (8 transients, 256 increments) were collected over 24 h. Those residues that gave a crosspeak in the “fingerprint” region (NH-CaH) of the final COSY spectrum are indicated by t,he tallest ba,rs in Fig. 3. Shorter bars indicate residues whose NH resonance exchanged faster. Data for other labile resonances are not shown in Fig. 3.

Solution

Structure

of Plasminogen

(c) Distance geometry calculations Distance geometry calculations were performed on a DEC microVAX II computer using the program DISGEO (Have1 & Wiithrich, 1985), differing from the described method only in that each substructure was taken to completion, without discarding those that converge less well. Sequential interproton distances (i.e. those involving only the NH, C”H and CPH protons of adjacent residues) were obtained by classification of crosspeaks in the POESY and PreTOCSY NOESY spectra acquired with mixing times of 160 ms as strong (distance