The Three-dimensional Structure of Recombinant Leech-derived Tryptase Inhibitor in Complex with Trypsin

The x-ray crystal structure of recombinant leech-derived tryptase inhibitor (rLDTI) has been solved to a resolution of 1.9 Å in complex with porcine trypsin. The nonclassical Kazal-type inhibitor exhibits the same overall architecture as that observed in solution and in rhodniin. The complex reveals structural aspects of the mast cell proteinase tryptase. The conformation of the binding region of rLDTI suggests that tryptase has a restricted active site cleft. The basic amino terminus of rLDTI, apparently flexible from previous NMR measurements, approaches the 148-loop of trypsin. This loop has an acidic equivalent in tryptase, suggesting that the basic amino terminus could make favorable electrostatic interactions with the tryptase molecule. A series of rLDTI variants constructed to probe this hypothesis confirmed that the amino-terminal Lys-Lys sequence plays a role in inhibition of human lung tryptase but not of trypsin or chymotrypsin. The location of such an acidic surface patch is in accordance with the known low molecular weight inhibitors of tryptase.

The x-ray crystal structure of recombinant leech-derived tryptase inhibitor (rLDTI) has been solved to a resolution of 1.9 Å in complex with porcine trypsin. The nonclassical Kazal-type inhibitor exhibits the same overall architecture as that observed in solution and in rhodniin. The complex reveals structural aspects of the mast cell proteinase tryptase. The conformation of the binding region of rLDTI suggests that tryptase has a restricted active site cleft. The basic amino terminus of rLDTI, apparently flexible from previous NMR measurements, approaches the 148-loop of trypsin. This loop has an acidic equivalent in tryptase, suggesting that the basic amino terminus could make favorable electrostatic interactions with the tryptase molecule. A series of rLDTI variants constructed to probe this hypothesis confirmed that the amino-terminal Lys-Lys sequence plays a role in inhibition of human lung tryptase but not of trypsin or chymotrypsin. The location of such an acidic surface patch is in accordance with the known low molecular weight inhibitors of tryptase.
Tryptase is the major protein component of mast cell granules (1). In contrast to many other trypsin-like serine proteinases, tryptase is present in the granules in an active form. Human mast cell tryptase is catalytically active as a tetramer, stabilized by heparin proteoglycans from the mast cells that are stored and secreted together with the protease. Human tryptase is unique inasmuch as no endogenous inhibitors have yet been detected for this enzyme, although rat tryptase has been shown to be inhibited by trypstatin (2), a fragment of the inter-␣-trypsin inhibitor (3). Regulation of human tryptase activity is thought to be mediated by dissociation of the tetramer, a process that may be accelerated by competition for heparin between the protease and other heparin-binding proteins. Human mast cell tryptase has been isolated from lung (4,5), skin (6), pituitary (7), and a mast cell leukemia cell line (8). Two cDNAs have been cloned from human lung (HLT␣ and HLT␤) (9,10) and three from human skin (HST1, HST2, and HST3) (11), where HLT␤ and HST2 possess the same sequence. Recent data suggest that different isoforms may exhibit differences in their specificities (12).
The physiological roles of tryptase are as yet unclear, although in vitro studies have implicated its function in neuropeptide turnover, coagulation, and the catabolism of extracellular matrix proteins (reviewed in Ref. 1). The central role of mast cells in allergic and inflammatory reactions suggests that tryptase may be involved in the pathogenesis of disorders such as asthma, interstitial lung diseases, psoriasis, arthritis, gingivitis, and peridontitis. Only a few synthetic inhibitors of tryptase have been reported so far (13)(14)(15). Knowledge of the tryptase structure would be of great benefit for the design of pharmacologically active lead compounds in the development of drugs for the treatment of these diseases. Attempts have been made to model the structure of human mast cell tryptase (16) and the corresponding mouse enzyme (17) by homology to other serine proteinases. These studies indicate the location of putative heparin binding sites and suggest that the limited activity of human tryptase may be due to a restricted active site cleft. Attempts to crystallize tryptase have been unsuccessful to date, presumably due to the requirement of (generally heterogeneous) glycosaminoglycans for stability.
We have recently identified and characterized leech-derived tryptase inhibitor (LDTI), 1 a protein-type inhibitor of human mast cell tryptase from the medicinal leech Hirudo medicinalis, (18) and have expressed it in Saccharomyces cerevisiae (19). LDTI inhibits tryptase with a K i of 1.4 nM and, in addition, trypsin and chymotrypsin in the nanomolar range. Together with the plasmin inhibitor bdellin B-3 from H. medicinalis (20) and the thrombin inhibitor rhodniin from Rhodnius prolixus (21), LDTI belongs to a small group of nonclassical Kazal-type inhibitors related to, for example, the Japanese quail ovomucoid inhibitor (22) but exhibiting a distinctive disulfide pattern (Table I). We have recently solved the structure of rLDTI in solution (23) and that of rhodniin in complex with thrombin (24).
In this paper, we describe the crystal structure of rLDTI in complex with porcine trypsin, solved to a resolution of 1.9 Å. The compact molecule binds to the active site of trypsin in a canonical (i.e. substrate-like) manner. Its distinctive disulfide pattern results in a disk-shaped molecule that suggests a restricted active site cleft of tryptase. The deletion of one turn of ␣-helix allows the amino-terminal residues to follow a path along the surface of the proteinase novel for Kazal-type inhibitors. Based on the structure, we have used site-directed mutagenesis to establish a specific role of the amino-terminal basic residues of LDTI in its interaction with human lung tryptase. The implications of these results for the structure and inhibition of tryptase are discussed.
Oligonucleotides were purchased from Pharmacia. The Escherichia coli-S. cerevisiae shuttle and expression vector pVT102U/a and the yeast strain S-78 were kindly provided by T. Vernet (Biotechnology Research Institute, Montreal, Canada) and by C.-W. Chi and Y.-S. Zhang (Shanghai Institute of Biochemistry, China) (25,26). The other E. coli strains and vector constructions harboring rLDTI are described in Ref. 19.
Crystallization and Structure Solution-An excess of rLDTI was added to porcine trypsin (Sigma) in 20 mM MOPS, 20 mM NaCl, pH7, and the resulting complex was concentrated to 10 mg/ml. Ellipsoidal crystals of the complex were obtained using vapor diffusion with 10% polyethylene glycol 6000, 2.3 M phosphate, pH 8. The crystals belong to the tetragonal space group P4 3 2 1 2 with cell constants a ϭ b ϭ 63.4 Å, c ϭ 131.2 Å, ␣ ϭ ␤ ϭ ␥ ϭ 90°and contain one complex in the asymmetric unit. Data to 1.9-Å resolution were collected on an image plate (MAR Research, Hamburg) and evaluated using Mosflm (27) from the CCP4 package. A total of 139,563 reflections (21,466 independent reflections, 98.6% complete to 1.9 Å) were measured with an overall R sym of 0.082. The structure was solved using AMoRe (28), with the trypsin component of the porcine trypsin-mung bean inhibitor complex (29) as starting model. Successive rounds of model building using the program O (30) and refinement using the program X-PLOR (31) parameterized according to Ref. 32 allowed the positioning of all 223 trypsin residues, 31 of LDTI's 46 residues, 149 water molecules, and 1 calcium ion. The final model has a crystallographic R-factor of 0.197 for all reflections to 1.9 Å, and consists of 1996 active atoms, with root mean square devia-tions of 0.008 Å in bond lengths and 1.805°in bond angles and an average temperature factor of 21 Å 2 .
Cassette Mutagenesis and Expression of the rLDTI Variants-Standard techniques of molecular cloning were performed according to Ref. 33, with DNA sequencing and yeast genetic methods according to Ref. 34. The deletion variants rLDTI-var1 and rLDTI-var2 and the substitution variant rLDTI-var3 were constructed by cassette mutagenesis using the cloning vector pRM 5.1.5 (19). Briefly, after cleavage of vector DNA with XbaI/BglII, the small fragment coding for the fusion linker and the N terminus of rLDTI was substituted by appropriate hybridized oligonucleotides (cassettes). The oligonucleotide sequences for the XbaI/ BglII fragments were 5Ј For expression in S. cerevisiae, the mutated rLDTI genes were isolated by XbaI/HindIII cleavage and ligated into yeast shuttle vector pVT102U/␣. The resulting expression vectors pRM 20.2.1, pRM 21.2.1, and pRM 22.2.1 were used to transform S. cerevisiae S-78 (35). Standard yeast expression experiments were performed as described in Ref. 19. Procedures for the mutagenesis, expression, and purification of the thrombin-directed mutants rLDTI-var1, rLDTI-var2, rLDTI-var3, rLDTI-var4, and rLDTI-var5 are described in the accompanying paper (36).
SDS-polyacrylamide gel electrophoresis of proteins was performed with 15-25% polyacrylamide gels (37). The gels were self-prepared and run in either a conventional apparatus or the PhastSystem ® (Pharmacia). Isoelectric focussing was performed with the PhastSystem ® using the calibration kit, pH 3.5-9.3, from Pharmacia.
Selected fractions from the cation exchange chromatography, usually 2-3 nmol of protein, were analyzed by reversed phase HPLC as detailed previously (18). Partial N-terminal sequence analysis was performed with the gas phase sequencer 473A (Applied Biosystems GmbH, Weiterstadt, Germany) following the instructions of the manufacturer.
The protein concentration was determined using the Pierce BCA* protein assay with bovine serum albumin as standard protein (38) or by measuring the absorbance at 280 nm and using theoretical values for aromatic residues and cystines according to Ref. Mass Spectrometry-Reverse phase HPLC-purified protein fractions were infused into an atmospheric pressure ionization source fitted to a tandem quadrupole instrument, API III (Sciex, Thornhill, Ontario, Canada). The sample solutions were delivered to the sprayer by a syringe infusion pump (model 22, Harvard Apparatus, Inc., South Natick, MA). The liquid flow rate was set at 5 ml/min for sample introduction. The instrument m/z scale was calibrated with the ammonium adduct ions of polypropylene gycol. The average molecular mass values of the protein were calculated from the m/z peaks in the charge distribution profiles of the multiple charged ions (40,41). Theoretical masses for the variants were calculated assuming three disulfide bridges with the GCG DNA/protein analysis software (42).
Enzymatic Measurements-The concentrations of the biologically ac-TABLE I Sequences of Kazal-type inhibitors, aligned according to cysteines BDB3, bdellin B3; RHOD1 and RHOD2, first and second domains of rhodniin, respectively; JPQ, third ovomucoid domain from the Japanese quail. The asterisk denotes the position of the P 1 residue; secondary structure is denoted by a (␣-helix) and b (␤-sheet). Residues mutated in this work are in boldface type.

LAAVSVDCSEYPKPAC-PKDYRPVCGSDNKTYSNKCNFCNAVVESNGTLTLNHFGKC
tive LDTI variants were determined by titration with trypsin, assuming an equimolar interaction between each inhibitor and the enzyme. Bovine pancreatic trypsin was standardized by active site titration using p-nitrophenyl pЈ-guanidinobenzoate (43). Equilibrium dissociation constants (K i ) were determined essentially as suggested by Bieth (44). Briefly, a constant concentration of bovine pancreatic trypsin, chymotrypsin, or tryptase isolated from human lung tissue (in the presence of 50 g/ml heparin to stabilize the protease) was incubated with increasing concentrations of each inhibitor; the time necessary to reach equilibration of the enzyme-inhibitor complex was determined in preliminary experiments. Subsequently, the residual enzyme activities were measured by following the hydrolysis of suitable substrates. Apparent K i values (K i (app) ) were calculated by fitting the steady state velocities to the equation for tight binding inhibitors (45) using nonlinear regression analysis as follows, where V i and V 0 are the velocities in the presence and absence of an inhibitor, and E t and I t the total concentrations of enzyme and inhibitor, respectively. Since the LDTI variants bind to only two of the four catalytic subunits of the tryptase tetramer in the range of concentrations used, Equation 1 was modified by subtracting V r (i.e. the remaining velocity at high concentrations of inhibitor) from V i and V 0 (18).
Equilibrium dissociation constants (K i ) for the complexes of the LDTI variants with chymotrypsin were obtained from the K i app values after correction for competition between the inhibitors and the substrate; even at high substrate concentrations competition was not detectable during the measurements of residual enzyme activities of tryptase and trypsin.
K i values for the inhibition of tryptase and factor Xa by the low molecular weight inhibitors TPDCA and DX9065a (Fig. 4) were calculated according to Dixon (46) from data measured at pH 7.6 and 25°C using the chromogenic substrate Tos-Gly-Pro-Arg-pNA.

Crystal Structure of LDTI in Complex with Porcine Trypsin-
Apart from a few surface-located side chains, the trypsin moiety is almost completely defined by electron density. With the exception of the side chain of Tyr 217 , which swings out to accommodate the amino-terminal residues of LDTI ( Fig. 1; see below), no significant conformational changes are seen in the trypsin component compared with other porcine trypsin structures (29,(47)(48)(49). The LDTI moiety is well defined in the vicinity of the proteinase, but it is characterized by elevated temperature factors and disrupted density further away from trypsin. In particular, amino acid residues Lys 1 -Lys 2 , Gly 15 -Arg 19 , Ser 33 -Ser 36 , and the C-terminal residues Pro 41 -Asn 46 are defined by either weak or no electron density. In contrast, the NMR solution structure (23) is well defined for residues Cys 4 -Cys 40 , while the terminal peptides Lys 1 -Val 3 and Pro 41 -Asn 46 are mobile.
LDTI secondary structure consists of a short central ␣-helix (Ser 24 -Asn 30 ) and a small antiparallel ␤-sheet (residues Val 13 -Gly 15 , Thr 20 -Tyr 21 , and Ile 34 -Glu 37 ) characteristic for the clas-sical Kazal-type inhibitors (Fig. 2). Due to a large deletion, however, the helix is one turn shorter than that in Japanese quail ovomucoid third domain (22) or either domain of rhodniin (24) (Fig. 2). The third ␤-strand is in part poorly defined, suggesting a degree of flexibility in the crystal. Residues P3-P3Ј (Cys 6 -Lys 11 ) exhibit the highly conserved "canonical" or substrate-like conformation of the small serine proteinase inhibitors (50). As a result of its "nonclassical" disulfide pattern, the active site wedge of LDTI is rather narrow, as was also observed for rhodniin (24). In contrast to either of the two domains of rhodniin, however, the first disulfide bridge (Cys 4 -Cys 29 ) possesses a right handed spiral conformation that causes the amino terminus to exit the active site cleft to the "south," rather than to the "west" (Fig. 1). Such a conformation of the disulfide is sterically hindered in rhodniin and Japanese quail ovomucoid third domain, since it would result in a clash of the amino-terminal residues with those of the longer ␣-helix.
Although not well defined, there is sufficient density to show that the two amino-terminal lysine residues reach out toward the trypsin loop containing residue 148 (referred to as the "148-loop"). Structure-based sequence alignment of mast cell tryptases with trypsin (16) reveals this loop to be acidic in the tryptases (see Fig. 3), making it conceivable that the basic amino terminus of LDTI makes an electrostatic contribution to its interaction with tryptase. To test this hypothesis, a series of N-terminally truncated and charge-deleted variants were constructed using cassette mutagenesis. Fig. 3). The binding of LDTI (green) to trypsin (orange) causes the side chain of Tyr 217 to flip outwards. This is a result of the conformation of the disulfide Cys 4I -Cys 29I ("I" suffix used to distinguish inhibitor residue numbers from those of trypsin). This figure was prepared using the program O (30). Expression of Variants-The yeast expression system gave yields of the variants of the order of 12 mg/liter, whose purity and identity were analyzed using SDS-polyacrylamide gel electrophoresis, isoelectric focusing, and reverse phase HPLC. Correct processing of the ␣ mating type leader fusion protein was verified by partial N-terminal sequencing in each case. Mass spectroscopy of the variants yielded major masses for LDTI-var1 of 4609.3 Da (calculated mass, 4609.35 Da), for LDTI-var2 of 4480.8 Da (calculated mass, 4481.18 Da), and two masses in a 1:1 ratio for LDTI-var3 of 4736.9/4719.4 Da (calculated mass, 4737.45 Da). The lower mass peak observed for LDTI-var3 was attributed to partial formation of pyroglutamate. Variants rLDTI-var1 and rLDTI-var2 both showed an additional minor mass with a reduced value of 114.0 Ϯ 1 Da compared with the major peak, interpreted as the truncation of the C-terminal Asn 46 by endogenous yeast proteinases.

FIG. 2. Superposition of C-␣ atoms of the Kazal-type inhibitors LDTI (thick lines), Japanese quail ovomucoid third domain (22) (medium lines), and rhodniin's first domain (24) (thin lines). The
Inhibition of Tryptase by the Variants-Equilibrium dissociation constants were measured for each variant with trypsin, tryptase and chymotrypsin (Table II). Sequential deletion of the first and second N-terminal lysines (variants 1 and 2, respectively) had a marked effect on anti-tryptase activity but only a negligible effect on the interaction with trypsin or chymotrypsin. This was also the case for the substitution of these residues by the uncharged polar residue glutamine (variant 3), substantiating the positive charge requirement. A series of variants designed to allow thrombin inhibition (see accompanying paper (36)) showed decreased tryptase inhibition. Mutation of four residues in the reactive site (rLDTI-var3) resulted in a modest decrease in tryptase inhibition, as did variant rLDTI-var2, C-terminally elongated by an acidic peptide. Tryptase inhibition was effectively abolished in the remaining C-terminal variants rLDTI-var1, rLDTI-var4, and rLDTI-var5. Some of these effects were also mirrored in the inhibition of trypsin and chymotrypsin.
Having established the existence of a negatively charged patch on the tryptase surface to the south of the active site cleft, it was decided to test the inhibition of tryptase by bisamidine inhibitors (Fig. 4). TPDCA inhibits tryptase with a K i of 0.85 M, whereas no inhibition could be measured for DX9065a.

DISCUSSION
The heparin-associated tetrameric tryptase has so far defied attempts at crystallization. In this paper, we present structural and mutagenesis data for rLDTI, a high affinity inhibitor of human tryptase. Using the structure of LDTI as a template, we are able to address aspects of the tryptase structure. Compared with the "classical" Kazal-type inhibitors (e.g. the Japanese quail ovomucoid inhibitor (22)), rLDTI has a narrow disk-like proteinase-binding region. This is a result of its characteristic disulfide pattern, which it shares with rhodniin and bdellin B-3. The structure of the rhodniin-thrombin complex (24) revealed this narrow binding region to be essential for the fit to thrombin's restricted active site cleft, bounded to the "north" by its characteristic 60-insertion loop (51).
According to the structural alignment proposed by Johnson and Barton (16), tryptase does not possess such a 60-insertion loop. The active site cleft should, however, be occluded to the west by a pronounced 9-residue insertion with respect to trypsin residues 173-176 (Fig. 3) immediately after the intermediate helix. The southerly exit of LDTI's amino-terminal residues from trypsin's active site cleft suggests that LDTI is favorably adapted to avoid this restriction in tryptase. The southerly route is made possible by the conformation of disulfide Cys 4 -Cys 29 , in turn facilitated by the deletion of the last turn in the central ␣-helix with respect to the ovomucoid inhibitors or rhodniin (Fig. 2). This disulfide may be in equilibrium with other conformers in the free inhibitor; a specific conformation was not observed in the NMR structure due to a lack of nuclear Overhauser effects constraining the amino terminus (23). An indication that the right handed spiral conformation might be well populated in the free inhibitor, and hence of functional significance, comes from the fact that the side chain of trypsin Tyr 217 swings out to accommodate the amino-terminal residues (Fig. 1). In all other published structures of porcine trypsin (29,(47)(48)(49), this side chain makes favorable stacking interactions with the aromatic side chain of Tyr 172 .
The southerly route of the amino-terminal residues would lead to a juxtaposition of the basic Lys 1 -Lys 2 amino terminus with the acidic 148-loop of human tryptases (Fig. 3). Progressive deletion of these residues results in a clear deterioration in  inhibitory potency of LDTI for tryptase (Table II), as does their mutation to the neutral Gln 1 -Gln 2 variant. That these mutations had hardly any effect on the inhibition of trypsin and chymotrypsin points strongly to a specific role of electrostatic exosite interactions in the inhibition of tryptase by LDTI.
Such an electrostatic interaction is consistent with the known low molecular weight inhibitors of tryptase (13)(14)(15). The most potent inhibitors consist of two aromatic amidino functions linked via a suitable spacer (Fig. 4). Clearly, one amidino function would occupy the primary specificity pocket of tryptase. It is quite conceivable that the second amidino moiety could make electrostatic interactions with the acidic 148-loop, similar to the proposed interaction of LDTI's amino terminus, with the edge of the aromatic group nestling against the 173insertion loop to the west. The crystal structure of trypsin in complex with bis-benzamidine (1,1Ј,3Ј,1Љ)-terphenyl-3,3Љ-dicarbamidinium (TPDCA) (52) shows just such an interaction between the distal amidino group and the side chain of Asn 143 . This could explain the inhibition of tryptase by TPDCA, 2,7bis-(4-amidinobenzylidene)-cycloheptan-1-one (BABCH), and bis-(5-amidino-2-benzimidazolyl)methane (BABIM) despite their differences in size, length, and functional groups.
Coagulation factor Xa is also inhibited by dibasic inhibitors, and indeed most of the compounds shown in Fig. 4 inhibit both tryptase and factor Xa. In factor Xa, however, the receptor for the distal basic moiety is found near the S4 pocket, i.e. toward the northwest of the active site (53,54), which would be inaccessible in tryptase due to the 173-insertion loop. We hypothesize that these inhibitors must adapt to suit the respective active sites. Comparison of the factor Xa:tryptase inhibition ratios suggests that DX9065a and BABCH prefer a "factor Xa conformation" in solution, while BABIM and TPDCA adopt a "tryptase conformation." The factor Xa-specific inhibitor DX9065a fits ideally to the active site of factor Xa (53, 54); its rigid structure would preclude binding to the south, which is in agreement with its total lack of tryptase inhibitory activity. The particularly low K i observed for BABIM (13) may be further enhanced by its increased basicity.
Our results suggest a tentative low resolution model for the lining of the active site cleft of the human tryptase monomer (Fig. 3). LDTI makes use of two dominant features of tryptase: the restriction of the active site to the west and an acidic surface patch to the south. Sequences of tryptase show a further insertion of four predominantly hydrophobic residues between trypsin residues Tyr 39 and His 40 (16) at the eastern edge of the active site cleft (Fig. 3). This loop may restrict the active site further, which could explain the marginal decrease in tryptase inhibition by rLDTI-var3. Such a conclusion should be considered circumspectly in view of the altered properties of this variant with respect to trypsin and chymotrypsin. In contrast to the C-terminal acidic extension variants rLDTI-var1, -var4, and -var5, rLDTI-var2 shows only a mild diminution of anti-tryptase activity, comparable with that of the N-terminal deletion variant rLDTI-var2. The reason for this cannot be explained by our model of the tryptase monomer; it may reflect interactions with other subunits of the tetrameric tryptase. Similarly, the model presented here does not explain the observed 50% maximal inhibition achieved by LDTI for the cleavage of small substrates by tryptase; presumably, the binding of LDTI to one monomer of tryptase precludes access of a further inhibitor molecule to the neighboring active site in the tetramer.
Although our data do not allow conclusions about the quarternary organization of tryptase or the role played by heparin to be drawn, the model presented here serves as a first step in understanding the structural features of this intriguing enzyme and its inhibition.