298
Biochemistry 1993, 32, 298-309
NMR Studies of the Dynamics of the Multidomain Protein Urokinase-type Plasminogen Activator? Ursula K. Nowak,t Xiang Li,i Andrew J. Teuten,i Richard A. G. Smith,$ and Christopher M. Dobson'J Oxford Centre for Molecular Sciences and Inorganic Chemistry Laboratory, University of Oxford, South Parks Road, Oxford OX1 3QR, U.K., and SmithKline Beecham Pharmaceuticals, Yew Tree Bottom Road, Great Burgh, Epsom, Surrey KT18 SXQ, U.K. Received July 28, 1992; Revised Manuscript Received October 26, I992
ABSTRACT: u-PA (urokinase-type plasminogen activator or urokinase) has been studied under a variety of solution conditions by 1-D and 2-D N M R spectroscopy. Very high quality spectra could be obtained from the recombinant protein despite the high molecular mass (46 kDa) by appropriate choice of solution conditions; mildly acidic pH and low ionic strength were found to be optimal. Comparison of spectra of u-PA with spectra of the isolated kringle and protease domains, the EGF-kringle pair, and a synthetic peptide from the kringle-protease linker region, enabled sequential assignments in the u-PA spectrum to be made for kringle resonances, and domain-specific assignments for many others. Simulations of line shapes in both 1-D and 2-D spectra enabled effective correlation times for the different domains, both isolated and in the intact protein, to be determined. These have permitted a model of the u-PA dynamics to be put forward involving extensive, but not unrestricted, motion between the different domains.
u-PA1 (urokinase-type plasminogen activator, urokinase) is a member of the family of multidomainfibrinolytic proteins and, together with t-PA (tissue plasminogen activator), is one of the major activators of plasminogen. Although the biological role of t-PA is almost solely concerned with the dissolution of blood clots, u-PA has also been found to be important in a number of other processes. Its involvement in processes of tissue degradation has suggested a role for u-PA in the invasiveness of tumors and the process of metastasis (Dano et al., 1985). u-PA has been investigated as a new prognostic marker in cancer, as its level in primary cancers is related to the metastatic potential (Reilly et al., 1991). u-PA is synthesized in the kidney and can be isolated from plasma, urine, and certain cancer cell cultures. It is secreted as a 54-kDa single-chain glycoprotein (scu-PA or prourokinase) which shows some properties of a protease zymogen, such as the lack of reactivity with protease inhibitors (Gurewich et al., 1984) and peptide-chloromethylketone derivatives (Lijnen et al., 1987). It does, however, possess some, albeit low, intrinsic catalytic activity (Gurewich et al., 1988; Lijnen et al., 1990). scu-PA is activated to the fully active two-chain form by cleavage between the Lys-158 and Ile-159 bond by plasmin (Holmes et al., 1985);the two chains remain connected by a disulfidebridge. Recombinant scu-PA has been expressed in Escherichia coli (Winkler et al., 1985; Winkler et al., 1986; Holmes et al., 1985) and has a molecular mass of 46 kDa, reduced from that of the native protein by the absence of carbohydrate; u-PA of urinary origin has a glycosylation site at Asn-302 (Steffens et al., 1982). Recombinant u-PA has t This is a contribution from the Oxford Centre for Molecular Sciences, which is supported by the Science and Engineering Research Council and the Medical Research Council of the United Kingdom. U.K.N. was supported by the EuroEan Science Exchange Programme funded by the Royal Society and the Osterreichischen Akademie der Wissenschaften. * Author to whom correspondence should be addressed. University of Oxford. 5 SmithKline Beecham Pharmaceuticals. 1 Abbreviations: NMR, nuclear magnetic resonance; DQF COSY, double-quantum-filtered correlated spectroscopy; TOCSY, total correlated spectroscopy; u-PA, urokinase-type plasminogen activator; EGF, epidermal growth factor; 1-D, one dimensional; 2-D, two dimensional.
0006-296019310432-298504.0010
been shown to have similar functional properties to u-PA isolated from human urine (Giinzler et al., 1984). Three different domainscan be distinguished in the sequence of u-PA: the N-terminal epidermal growth factor like (EGFlike) domain, the kringle domain, and the C-terminal proteolytic domain. These domains have significant homology with domains found in other proteins and are probably related in evolution (Gilbert, 1978; Patthy, 1985). The EGF-like domain (residues 5-45,5 kDa) is related to human epidermal growth factor and binds to the u-PA receptor (Appella et al., 1987; Blasi et al., 1988). The functional role of the kringle domain (residues 46-135,lO kDa) remains unclear, although is has been shown that the kringle domain of u-PA binds heparin and might also bind to cell-surface polyanions (Stephens et al., 1992). It does not exhibit the fibrin binding specificity found in other homologous domains, such as kringle 1 and 4 in plasminogen and kringle 2 in t-PA. The third domain is a trypsin-likeserine proteasedomain (residues 13641 1, 33 kDa), which contains the "catalytic triad" (His-204, Asp-255, and Ser-356) found in all serine proteases (Steffens et al., 1982). It is connected to the kringle domain by a 16residue long linker peptide and can be isolated following cleavage at Lys-135 during limited proteolysis by plasmin. The resulting low molecular weight u-PA (LMW u-PA) is fully active and is also present in body fluids. Activation of scu-PA to two-chain u-PA (tcu-PA) by cleavage at Lys-158 leads to a new amino terminus which might form an ion pair with Asp-355 by intruding into the proteolytic domain in the manner found for other serine proteases (de Munk & Rijken, 1990). At the present time little is known about the physical and structural properties of intact u-PA. The protein has a high content of @-sheetand @-turnsecondary structure, as seen in CD studies (Mangel et al., 1991). In common with other multidomain proteins, crystallization conditions have proved elusive, and there is no crystal structure presently available. Investigation by small-angleneutron scattering has indicated a highly asymmetric molecule with a radius of gyration of 3 1 A and a maximum dimension of 90 A (Mangel et al., 1991). Although u-PA is large in respect to its feasibility for NMR 0 1993 American Chemical Societv
Dynamics of u-PA
Biochemistry, Vol. 32, No. 1, 1993 299
Inactivated human urinary u-PA was purified on an FPLC system using a gel filtration column (SephacrylS-100 HiLoad column). A 0.1 M solution of NH4HCO3, pH 8.4, was used as the eluting buffer to make direct lyophilization possible. Fractions containing u-PA were pooled after their purity had been confirmed by SDS gel electrophoresis and then lyophilized. The linker peptide from Lys- 136, the cleavage site between the kringle and the proteasedomain, to Lys-158, the activation site of the protease domain, was synthesized on an automated peptide synthesizer using a p-alkoxybenzylalcohol resin and standard Fmoc protocols. The peptide was cleaved from the resin using 90% TFA/5% H20/2.5% 1,2-ethanediol/2.5% anisole and was purified by reverse-phase HPLC. All protein samples were dissolved in D2O or H2O at pH 4.5 and purified from small molecule impurities and salts by five cycles of centrifugation at 4 OC (reducing the volume from 2.0 mL to about 150 pL in each step; 4000g) using Centricon-3 or Centricon-10 concentration cells with a 3- or 10-kDa cut-off membrane, respectively (Amicon,Gloucester, U.K.). The pH was adjusted by adding dilute DCl and NaOD (2% and 0.2%), and pH values quoted are uncorrected for isotope effects. Sample concentrationswere between 0.6 and 2.2 mM. NMR Spectroscopy. NMR spectra were recorded on a Bruker AM600 spectrometer with a proton resonant frequency of 600.13 MHz at the Oxford Centre for Molecular Sciences. Spectra were collected at 29 or 35 OC. 1-D spectra were recorded using 4K or 8K data points over a spectral width of 7812.5 or 8928.5 Hz, collecting 300-1600 scans. Resolution enhancement was achieved using the Lorentzian-Gaussian transformation (GB 0.15, LB -10) prior to zero-filling to 8K. DQF COSY (Rance et al., 1983) and TOCSY spectra (Braunschweiler & Ernst, 1983; Davis & Bax, 1985) were acquired over 2K data points and 5 12-800 t I increments in the absorption mode with time-proportionalphase incrementation (TPPI) for quadrature detection in the tl dimension. Water saturation was achieved by low-power irradiation during MATERIALS AND METHODS the relaxation delay introduced between scans. A total of 80, Protein Samples. Recombinant two-chain u-PA, low 96 or 112 transients were collected for each tl increment. molecular weight u-PA, and the isolated kringle domain were TOCSY spectra were acquired in reverse mode using a provided by Griinenthal GmbH, Aachen, Germany. These MLEV17 pulse sequence (Bax & Davis, 1985) for mixing were derived from single-chain u-PA expressed in E. coli, the (mixing time, 47 ms). 2-D spectra were processed on a Sun twechain protein and fragmentsbeing produced by proteolytic 4/110 workstation using the Felix program provided by Dr. cleavage. The recombinant EGF-kringle domain was provided D. R. Hare (Felix version 1.1, Hare Research Inc.). The data by Dr. D. J. Ballance, Delta Biotechnology Ltd., Nottingham, set was resolution enhanced by double-exponential and U.K., and was secreted from the yeast Saccharomyces trapezoidal multiplicationin t 2 and a combination of doublecereuisiae. Human high molecular weight u-PA of urinary exponential, shifted sine-bell-squared and trapezoidal mulorigin was a gift of the Japan Chemical Research Co. Ltd., tiplication in tl prior to zero-filling in both dimensions. After Japan. zero-filling, the digital resolution was 3.84.2 Hz/pt in both Human urinary and recombinant two-chain u-PA and low dimensions. All spectra were referenced to 1,Cdioxaneat 3.74 molecular weight u-PA were prepared in a 0.1 M Tris, 0.15 PPm. M NaC1, and 20% (v/v) glycerol containing buffer at pH 7.4 Simulations. Spectral simulations for both 1-D and 2-D at a concentration of 10 mg/mL and inactivated with a threefold molar excess of GGACK (L-Glu-L-Gly-L-Arg-chloro- NMR spectra were performed on a Sun 4 workstation. Simulations of 2-D COSY spectra were carried out using a methylketone,Cambridge Bioscience, Cambridge, U.K.) (20 program provided by Dr. C. Redfield which assumes weak mM in ethanol) at 25 OC until the activity against S-2444 was coupling for neighboring protons and applies experimental reduced to less than 0.5% of that of the untreated protein. parameters (line width, coupling, and resolution enhancement Inactivation was carried out to reduce the possibility of functions) to in-phase and antiphase peaks. Tryptophan cross autolysis in the NMR experiments during data acquisition, peaks were simulated with the known values of coupling although this is unlikely to be fast at the pH values used in constants (3J45= 8.03, 3J56 7.08, 3J67 = 8.34, 4J46 1.12, the experiment. The samples were buffer exchanged into 0.1 4J47 = 0.95, sJ47 = 0.81) (Redfield, 1984) and varying line M NH4HC03 at pH 8.5 on Sephadex G-25M (PD-10, Pharmacia) gel filtration columns at 4 OC and lyophilized. width values. The same resolution enhancement functions
studies, important information has been gained from preliminary NMR studies on this protein (Oswald et al., 1989; Bogusky et a1.,1989). Comparison of spectra of the intact protein with those of the isolated domains provided evidence for close structural similarity between the isolated domains and the corresponding units of the intact protein. Furthermore, these studies showed that the domainshave a substantial degree of structural independence in terms of both their dynamical behavior and their folding properties. Indeed, it is to these properties that the feasibility of NMR studies is attributed (Oswald et al., 1989). As with other multidomain proteins, one approach to structural analysis has been to isolate individual domains and to investigate their structures. The kringle domain of u-PA which can be isolated by extended plasminolysis (Bogusky et al., 1989; Mazar et al., 1992) has been studied by NMR (Li et al., 1992), and its secondary structure elements have been determined. These include three antiparallel @-sheets,two helices, and three tight @-turns. One of the helices is analogous to that reported for t-PA kringle 2 (Byeon et al., 1991; de Vos et al., 1992), but the other one is so far unique to the u-PA kringle. The overall tertiary fold has at present been determined only in outline but is similar to the crystal structures of prothrombin kringle 1 (Park & Tulinsky, 1986; Tulinsky et al., 1988), plasminogen kringle 4 (Mulichak et al., 1991; Wu et al., 1991), and especially t-PA kringle 2 (de Vos et al., 1992; Byeon & LlinBs, 1991). In this paper we extend considerably previous NMR studies of u-PA and its component domains and investigate a peptide synthesized to correspond to the linker region between the kringle and the protease domain (residues 136-158). Much improved NMR spectra were achieved by the use of recombinant protein and careful optimization of conditions, and these have enabled us to analyze 1-D and 2-D spectra in some detail. In particular, the nature of domain mobility within u-PA has been investigated by simulating 1-D and 2-D spectra to obtain information about the overall motional correlation times for the domains, and these data have enabled us to put forward a dynamical model for this multidomain protein.
300 Biochemistry, Vol. 32, No. 1, 1993
Nowak et al.
were used for the simulations and for processing the NMR data sets (see above). Positive and negative contours of the cross peaks are plotted at exponentially increasing levels. In order to simulatea tryptophan cross peak, a first estimate of the resonance line width was obtained by determination of the line width of a resolved tryptophan resonance in the 1-D spectrum. A best fit of an experimental 1-D peak to a simulated in-phase signal was made by visual comparison, and then 2-D cross peaksof the same line width were simulated. By adjusting the contour level, the contour profile of the simulated cross peak was compared with the experimental cross peak for a given line width until the two were closely matched. This was then repeated for other cross peaks for the same contour level but different line width values. In addition to visual comparison of the simulated and the experimental cross peaks, the peak heights of the cross p:aks were compared. These were measured within the Felix routine, and the four absolute intensities of the maxima and minima of a crosspeak were averaged. The half-height line width (Av1p) for a Lorentzian line is
1 Avl12= -= L c [ 5 J ( 0 ) r T 2 2077
+ 9J(w) + 6J(2w)]
transverse relaxation of an A3 spin system can be described by using Redfield relaxation matrix theory (Redfield, 1965) or by solving a group of first-order differential equations by forming a set of "normal mode variables" (Werbelow & Marshall, 1973; Werbelow & Grant, 1977). For the latter case, the equations have the form
-d2= R-q where q = ( q l , q ~ , q 3is) ~a set of three variables related to transverse coherences, and at least one of them, e.g., 41, corresponds to the observable transverse magnetization I,. The matrix elements R,, (m,n = 1,3) are each a linear combination of spectral density functions Jijk/(XW) which can be approximated as:
+
where T2 is the transverserelaxation time, and the summation is over all pairwise contributionsto the specificspin resonance. Using the formalism of Lipari and Szabo (1982) to account for internal motions, the spectral density function J(w) can be written as
where T~ is the overall rotational correlation time, 7-l = rC-l Tin-1, and Tin is the correlation time for internal motions. S2 is a generalized order parameter, r is the interproton distance, and the other symbols have their conventional meaning. Provided that Tin
