Conformation of One- and Two-chain High Molecular Weight Urokinase Analyzed by Small-angle Neutron Scattering and Vacuum Ultraviolet Circular Dichroism*

The structures of one- and two-chain high molecular weight human urokinase were analyzed by small-angle neutron scattering and vacuum ultraviolet circular di- chroism. Both one- and two-chain high molecular yeight urokinases exhibited a radiu5 of gyration of 3 1 A and a maximum dimension of 90 A. Neither parameter was affected by the presence of lysine sufficient to saturate all the lysine-binding sites in human plasminogen. These physical parameters are consistent with the sedimentation coefficient of high molecular weight urokinase and indicate that both proteins are highly asymmetric. Neither protein contained much 0-helix or parallel @-sheet. Most of the secondary struc- ture was in the form of antiparallel &sheet and 8-turns, very similar to the secondary structure of plas- minogen. The macroscopic kinetic constants, K , and k,,,, for the hydrolysis of (<Glu-Gly-Arg-NH)2-rhoda- mine by two-chain high molecular weight urokinase and low molecular weight urokinase which lacks the epidermal growth factor and kringle domains were similar. These structural and kinetic data are consist- ent with the domains in both forms of urokinase being independent structural and functional units. HBS (Melhado et al., 1982). The synthesis and purification of the fluorogenic substrate (<Glu-Gly-Arg-NH)2-rhoda- mine will be described elsewhere. Kinetic experiments were performed in HBS in which the concentration of two-chain HMW urokinase or LMW urokinase was 1.55 nM and the (<Glu-Gly-Arg-NH)2-rhodamine concentration was varied from 100 to 800 PM. An excitation wavelength of 492 nm and an emission wavelength of 523 nm were used with a hand pass of 5 nm.

Human urokinase is a highly specific serine protease that activates the zymogen plasminogen, present in all body fluids, to the potent serine protease plasmin. Plasminogen activation is involved in a wide variety of normal and abnormal extracellular physiological processes that require proteolytic activity (for a review, see Dan0 et al., 1985). This enzyme system is highly regulated to be triggered only where and when plasmin is needed. There are two major plasminogen activators. Urokinase is involved in such physiological processes as cell migration and tissue remodeling whereas the other major plasminogen activator, tissue plasminogen activator, is more involved in thrombolysis.
There are multiple forms of human urokinase: one-chain HMW' (high molecular weight) urokinase, two-chain HMW * This work was supported by the Office of Health and Environmental Research of the United States Department of Energy. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. t To whom correspondence should be addressed. § Present address: Institute of Biophysics, Academia Sinica, Beijing, China.
As are all the proteins involved in plasminogen activation, urokinase is composed of domains. A cysteine-rich aminoterminal domain from amino acids Ile5 to Lys4' shares sequences with human epidermal growth factor. It is by way of this domain that both one-and two-chain HMW urokinase bind to a specific receptor on the cell surface (Vassalli et al., 1985;Stoppelli et al., 1985;Appella et al., 1987;Roldan et al., 1990;Behrendt et al., 1990;Mangel, 1990). Sequences from amino acids Cysso t o L Y S '~~ are similar to sequences in kringle 5 of human plasminogen (Gunzler et al., 1982b). Kringles are triple-loop structures of about 80 amino acids characteristically constrained by three disulfide bridges (Magnusson et al., 1976). There are five kringles at the amino-terminal end of plasminogen (Sottrup-Jensen et al., 1978) and two in tissue plasminogen activator (Pennica et al., 1983). The kringles have a high degree of sequence similarity and are autonomous structural and folding domains (Trexler and Patthy, 1983;Patthy, 1985). The protease domain at the carboxyl terminus of human urokinase contains sequences similar to those in serine proteases (Strassburger et al., 1983). Although its sequence and the observation that it cleaves its physiological substrate plasminogen at an Arg-Val bond indicate it is trypsin-like, its disulfide pattern resembles that of chymotrypsin.
There are indications that the domains in two-chain HMW urokinase are independent structural units. In two-dimensional NMR studies, the (two-dimensional) NOESY (above diagonal) and COSY (below diagonal) spectra of two-chain HMW urokinase in D20 were compared with the spectra of fragments of urokinase containing the EGF-kringle domain, the kringle domain, and the protease domain (Oswald et al., 1989;Bogusky et al., 1989). The intact protein showed an unusually well-resolved spectrum for such a large molecule. The resonance line widths were not greatly increased from those of the isolated domains. This was specifically interpreted as being a consequence of substantial independent motion between the EGF-kringle domains and the protease domain which overcomes the broadening effect expected from the slow overall tumbling rate of the intact molecule. Temperature studies indicated that the structures and stabilities of isolated domains are similar to those in the intact protein and specifically suggested there is a high degree of independent motion between the kringle and protease domains.
Recently, we showed there are two major conformations in human plasminogen, a closed conformation in which there is domain interaction among the five kringles and the protease domain and an open conformation in which domain interaction is abolished (Mangel et al., 1990;Ramakrishnan et al., 1991). The closFd form of native plasminogen has a radius of gyration of 39 A. Upon conversion to an open form by occupation of a weak lysine-binding site or by removal of the first 77 amino acids at the pHz terminus, the radius of gyration changes to 56 or 51 A, respectively. Since the secondary structure of the closed form of plasminogen did not change upon conversion from the closed to the open form, we interpreted this as indicating the structure of the closed form of plasminogen is formed by domain interaction and that upon conversion to the open form this interaction is abolished. Here we ask whether there are dramatic differences in conformation between the one-and two-chain forms of HMW urokinase. We employ small-angle scattering because it can be used to study directly and precisely the conformation of proteins in solution. Analysis of the low angle part of the scattered intensity yields the radius of gyration and the maximum dimension of a particle without using any assumptions (Guinier and Fournet, 1955). We also employ vacuum ultraviolet circular dichroism (CD), to estimate the secondary structure content of one-and two-chain HMW urokinase. T o determine whether the EGF and kringle domains influence the specific activity of the proteinase domain, we characterize the interaction of two-chain HMW urokinase and LMW urokinase with a new fluorogenic substrate (<Glu-Gly-Arg-NH)2-rhodamine.

MATERIALS AND METHODS
buffered saline (HBS) which contained 0.01 M Hepes, p H 7.2, 0.137 All small-angle scattering experiments were performed in Hepes-M NaC1, 2.68 mM KC1,0.91 mM CaC12, and 0.49 mM MgC12. One-and two-chain HMW urokinase and LMW urokinase were a gift from Collaborative Research, Inc. The concentrations of one-and twochain HMW urokinase were determined from the absorbance at 280 nm using & = 15.3. Prior to each experiment the urokinases were fractionated on a Sephacryl S-200 column in HBS to remove possible aggregates and then concentrated by filtration under nitrogen pressure with an Amicon PM-10 ultrafiltration membrane. Concentrated protein solutions were dialyzed for 15 h, with one change of buffer, against a 50-fold volume of HBS made up in D,O. After each experiment, the urokinases were analyzed by NaDodS04-polyacrylamide gel electrophoresis under reducing conditions (Laemmli, 1972) which indicated they had not changed (data not shown).
Small-angle neutron scattering measurements were done on the H9B spectrometer (Schneider and Schoenborn, 1984) in the High Flux Beam Reactor at Brookhaven National Laboratory. The mean incident wavelength varied, depending upon the experiment, from 4.5 to 7.5 A. The protein concentration in samples was around 10 mg/ ml. Samples and their corresponding buffers were loaded into cylindrical quartz cells with a 1-mm path length. The transmissions were 0.91-0.92. All measurements were performed a t 4 "C.
In each experiment, the scattering was measured from the sample, a cell containing the buffer which had a transmission nearly identical to that of the sample, an empty cell, and a blocked beam. Scattered neutrons were measured on an He3 area detector. To obtain the scattered intensity from the macromolecule, we radially averaged the scattering and corrected for background and buffer scattering. The scattering was then normalized for beam intensity which was monitored using a low efficiency fission detector during the measurement, thickness, transmission, and protein Concentration.
The radius of gyration and forward scatter were estimated from a least-squares fit to the linear region of Guinier plots of the data, (ln[Z(k)] versus k'), using the assumption that at low angles Z(k) = i(0) exp(-k2Rg2/3). The range for the linear region was k = 0.020-0.045 A".
Vacuum ultraviolet circular dichroism spectra were obtained using the vacuum spectrometer U9B of the National Synchrotron Light Source at Brookhaven National Laboratory (Sutherland et al., 1980(Sutherland et al., , 1982. Experiments were performed at room temperature in a quartz cell with a 12.5-fim path. The protein concentration was about 5 mg/ ml in HBS. trations of two-chain HMW urokinase and LMW urokinase were For measurement of the macroscopic kinetic constants, the concendetermined by active site titrations with 3 PM fluorescein-mono-pguanidinobenzoate in HBS (Melhado et al., 1982). The synthesis and purification of the fluorogenic substrate (<Glu-Gly-Arg-NH)2-rhodamine will be described elsewhere. Kinetic experiments were performed in HBS in which the concentration of two-chain HMW urokinase or LMW urokinase was 1.55 nM and the (<Glu-Gly-Arg-NH)2-rhodamine concentration was varied from 100 to 800 PM. An excitation wavelength of 492 nm and an emission wavelength of 523 nm were used with a hand pass of 5 nm.

RESULTS AND DISCUSSION
Small-angle neutron scattering data from one-and two-chain HMW urokinase in D20 are shown in Fig. 1 in a plot of I ( k ) uersus k where I ( k ) is the scattered intensity as a function of k = 4asinO/X, 26 is the scattering angle, and X is the mean wavelength. The scattered intensity was normalized for incident neutron flux, thickness, transmission, and protein concentration. 4usin8/X, 28 is the scattering angle, and X is the mean wavelength. Open circle is one-chain HMW urokinase; closed circle is twochain HMW urokinase. and Two-chain HMW Urokinase For small-angle neutron scattering the intensity at very small angles obeys the Guinier relationship (Guinier and Fournet, 1955). Guinier plots, ln[Z(K)] uersus k2, of the low-angle scattering data in Fig. 1 from one-and two-chain HMW urokinase in D20 are shown in Fig. 2. The radius of gyration Rg is most accurately measured in DzO from the slope of a Guinier plot. The radius of gyration of one-chain HMW yrokinase was 29.9 A and of two-chain HMW urokinase 32 A, Table I. Similar experiments were performed in the presence of 0.05 M 6aminohexanoic acid (data not shown). ThF radius of gyration of one-chain HMW uTokinase was 31.0 A and of two-chain HMW urokinase 30.7 A, Table I.
The maximum dimensions of one-and two-chain HMW urokinase were estimated from the scattering data in Fig. 1 using an indirect Fourier transformation described by Moore (1980) to calculate the length distribution function P( r ) . This function describes the distribution of distances between pairs of points in a particle. P( r)dr is the probability that two points in a particle are at a distance between r and r + dr. The scattering data were expanded in a finite series with basis functions that have the property that P ( r ) = 0 for r > d, , , , where dm,, is a parameter corresponding to the assumed maximum dimension of the particle. If dm,, is too small, the value of the reduced x* for the expansion will be very large. As the value of dm,, is increased, the reduced x* drops until it reaches a lower limiting value. The smallest value of dm,, for which x2 reaches its limiting value is the smallest maximum dimension of the particle consistent with the data. The variation of x2 with dm,, is shown in Fig. 3, and from this data the maximum dimensio? for both one-and two-chain HMW urokinase was about 90 A, Table I. The vacuum ultraviolet circular dichroism spectra of oneand two-chain urokinase were very similar, Fig. 4. There was one positive peak at 186 nm and one negative peak at 198 nm. The spectra were obtained down to 178 nm because although spectra measured down to 200 nm can be used to determine accurately the amount of a-helix in a protein, they cannot be used to solve for other structures (Siege1 et al., 1980). The data were analyzed by a method that uses inverse CD spectra for each of the five major secondary structures of proteins (Compton and Johnson, 1986). The method predicts the percent of each secondary structure by forming the dot product of the corresponding inverse CD spectrum, expressed as a vector, with the CD spectrum of the protein digitized in the same way. Oneand two-chain HMW urokinase contained little or no a-helix or parallel @-sheet. Most of the secondary structure was in the form of antiparallel @-sheet and @-turns, Table 11. The secondary structure of HMW urokinases was remarkably similar to that of native plasminogen which contains little or no a-helix or parallel @-sheet, 36% antiparallel P-sheet and 28% @-turns, " Mangel et al. (1990).  Table I1 (Mangel et al., 1990). This was not unexpected since the kringle and protease domains in HMW urokinases are similar in sequence to the five kringles and protease domain in plasminogen (Patthy, 1985). Under identical experimental conditions we obtained the CD spectra of one-and two-chain HMW urokinase in the presence of 0.05 M 6-aminohexanoic acid, a concentration sufficient to saturate the weak lysine binding sites in plasminogen (Mangel et al., 1990). The spectra and the percent of each secondary structure were identical to those obtained in the absence of 6-aminohexanoic acid (data not shown).
There are several indications that our results on the secondary structure of one-and two-chain HMW urokinase are reliable. The method of analysis places no constraint on the sum of structures, and the percent of a structure is permitted to be negative. Thus, sums of structure between 90 and 110% and the absence of large, negative percent of structure are indicative of a successful analysis (Compton and Johnson, 1986). The sum of structures was 102% for both one-and twochain HMW urokinases. The one negative percent was for parallel P-sheet, and this was only 3%. As additional controls, we obtained the CD spectrum of chymotrypsin and analyzed it in an identical way (Mangel et al., 1990). The results compare favorably to the secondary structure of chymotrypsin determined by x-ray crystallography, Table 11. Also, the presence of 6-aminohexanoic acid had no effect on the secondary structure of chymotrypsin (Mangel et al., 1990).
Our data indicate that one-and two-chain HMW urokinase are highly asymmetric particles. If the partial specific volumes are 0.754 ml/g and the molecular weights are 46,375, calculated from the amino acid composition, then the R, of both particles would be 18.6 A if they were spheres. Experiment!lly, the R, values were 30 A. For a sphere with an R, Of 18.6 A, the diameter or maximum dimension would be 48 A. The maximum dimensions were 90 A, based upon low angle neutron scattering data. Our data are consistent with the experimentally determined sedimentation coefficient for HMW urokinase which is 3.3 when corrected for temperature and in water (Barlow, 1976). Based upon our R, value, we calculate that the sedimentation coefficient should be 3.6 (Kumosinski and Pessen, 1982). An important conclusion from these experiments is that the R,, maximum dimension and secondary structure of onechain HMW urokinase remain similar upon conversion to the two-chain HMW form. Thus, conversion from an inactive form to a form that can activate plasminogen is not accompanied by a dramatic change in conformation. Plasminogen undergoes a massive change in conformation accompanied by an increase in flexibility upon conversion from the closed to the open form (Mangel et al., 1990;Ramakrishnan et al., 1991). The K,,, for activation of the closed form of plasminogen by two-chain HMW urokinase is much higher than the in vivo concentration of plasminogen, and the K, decreases a t least 10-fold upon conversion to the open form . We interpreted the decrease in K,,, to be the result of the increase in flexibility upon conversion to the open form which could facilitate the binding of two-chain HMW urokinase. The change in conformation that activates HMW urokinase upon cleavage of the one-chain form to the two-chain form must be much more subtle than the conformational differences between the two forms of plasminogen.
The domains in both forms of HMW urokinase appear to function independently. The EGF domain mediates the binding of both forms of HMW urokinase to a cell surface receptor (Appella et al., 1987). The function of the kringle in HMW urokinase is unknown. The kringles in plasminogen contain lysine-binding sites that mediate the binding of plasminogen to fibrin (Thorsen, 1975;Lucas et al., 1983a, 198313;Suenson and Thorsen, 1982;Bok and Mangel, 1985), the extracellular matrix (Knudsen et al., 1986), and the cell surface (Plow et al., 1986). They are also involved in determining the conformation of the closed form of plasminogen by domain interaction (Mangel et al., 1990;Ramakrishnan et al., 1991). The data presented here indicated that 6-aminohexanoic acid at a concentration sufficient to saturate the lysine-binding sites in plasminogen has no effect on either the radii of gyration or secondary structures of one-or two-chain HMW urokinases. There is evidence for a weak benzamidine-binding site in the kringle of both forms of HMW urokinase (Winkler et al., 1985;Winkler and Glaber, 1986). In plasminogen there is also a weak benzamidine-binding site in kringle 5 (Holleman et al., 1975;Vardi and Patthy, 1981).
If the protease domain is an independent functional unit then the specific activities of two-chain HMW urokinase and LMW should be the same even though LMW urokinase is missing the EGF and kringle domains. We measured the macroscopic kinetic constants for the hydrolysis of the fluorogenic substrate (<Glu-Gly-Arg-NH)z-rhodamine, Table 111. The Michaelis constants were identical within experimental error as were the catalytic rate constants. The latter were calculated using as enzyme concentrations the concentrations of active sites based upon active site titrations with fluorescein-mono-p-guanidinobenzoate (Melhado et al., 1982). The

Conformation of
One-and Two-chain HMW Urokinase macroscopic kinetic constants were determined in Hepesbuffered saline and in the absence of added salt because the activity of urokinase is sensitive to ionic strength. Physiological ionic strength increased the Michaelis constants and decreased the catalytic rate constants for both two-chain HMW urokinase and LMW urokinase, Table 111. That the macroscopic kinetic constants were the same indicate the EGF and kringle domains do not affect the catalytic behavior of the protease domain in the hydrolysis of small, organic, amide substrates. This is consistent with data from Oswald et al. (1989) and from Bogusky et al. (1989) in which a comparison of the transition temperatures of the kringle domain in fragments and in intact two-chain HMW urokinase indicates that any interaction between the folded kringle and protease domain has no significant influence on the thermal stability of the kringle domain.
Even though the domains in urokinase are independent structural and functional units, the activation of plasminogen by urokinase can be regulated to occur at specific times and places. First, conversion from the inactive one-chain form to the two-chain form can be regulated. Second, both one-and two-chain HMW urokinase can be localized by binding to a specific receptor on the cell surface (Blasi, 1988), and they can be released upon conversion to LMW urokinase. Third, one-chain HMW urokinase is insensitive to attack by protease inhibitors, and the half-life of two-chain HMW urokinase can be regulated by the local concentration of its two major inhibitors, PAI-1 and PA1 -2. Fourth, plasminogen activation by two-chain HMW urokinase can be regulated by the form of its substrate plasminogen (Mangel et al., 1990;Ramakrishnan et al., 1991). The Michaelis constant for activation of the closed form of plasminogen by two-chain HMW urokinase is much higher than the in uiuo concentration of plasminogen, and upon conversion to the open form the K , decreases 10fold . Fifth, the binding of plasminogen to higher ordered structures such as fibrin (Bok and Mangel, 1985) and the extracellular matrix (Knudsen et al., 1986) increases its local concentration so that it is preferentially activated by urokinase over plasminogen in solution.