Bovine Proenteropeptidase Is Activated by Trypsin, and the Specificity of Enteropeptidase Depends on the Heavy Chain*

Enteropeptidase, also known as enterokinase, initiates the activation of pancreatic hydrolases by cleaving and activating trypsinogen. Enteropeptidase is synthesized as a single-chain protein, whereas purified enteropeptidase contains a ≈47-kDa serine protease domain (light chain) and a disulfide-linked ≈120-kDa heavy chain. The heavy chain contains an amino-terminal membrane-spanning segment and several repeated structural motifs of unknown function. To study the role of heavy chain motifs in substrate recognition, secreted variants of recombinant bovine proenteropeptidase were constructed by replacing the transmembrane domain with a signal peptide. Secreted variants containing both the heavy chain (minus the transmembrane domain) and the catalytic light chain (pro-HL-BEK (where BEK is bovine enteropeptidase)) or only the catalytic domain (pro-L-BEK) were expressed in baby hamster kidney cells and purified. Single-chain pro-HL-BEK and pro-L-BEK were zymogens with extremely low catalytic activity, and both were activated readily by trypsin cleavage. Trypsinogen was activated efficiently by purified enteropeptidase from bovine intestine (K m = 5.6 μm and k cat = 4.0 s−1) and by HL-BEK (K m = 5.6 μm and k cat = 2.2 s−1), but not by L-BEK (K m = 133 μm and k cat = 0.1 s−1); HL-BEK cleaved trypsinogen at pH 5.6 with 520-fold greater catalytic efficiency than did L-BEK. Qualitatively similar results were obtained at pH 8.4. In contrast to this striking difference in trypsinogen recognition, the small synthetic substrate Gly-Asp-Asp-Asp-Asp-Lys-β-naphthylamide was cleaved with similar kinetic parameters by both HL-BEK (K m = 0.27 mm and k cat = 0.07 s−1) and L-BEK (K m = 0.60 mm and k cat = 0.06 s−1). The presence of the heavy chain also influenced the rate of reaction with protease inhibitors. Bovine pancreatic trypsin inhibitor preferred HL-BEK (initial K i = 99 nm and final K i * = 1.8 nm) over L-BEK (K i = 698 nm andK i * = 6.2 nm). Soybean trypsin inhibitor exhibited a reciprocal pattern, inhibiting L-BEK (K i * = 1.6 nm), but not HL-BEK. These kinetic data indicate that the enteropeptidase heavy chain has little influence on the recognition of small peptides, but strongly influences macromolecular substrate recognition and inhibitor specificity.

Enteropeptidase, originally named enterokinase when it was discovered by Pavlov (1), is a membrane-bound serine protease of the duodenal mucosa that cleaves trypsinogen to generate active trypsin. In almost all vertebrate species, a short trypsinogen activation peptide is released that terminates with the sequence Asp-Asp-Asp-Asp-Lys (2). Following activation, trypsin cleaves and activates other zymogens in pancreatic secretions, including chymotrypsinogen, proelastase, procarboxypeptidases, and some prolipases. Thus, enteropeptidase initiates a simple two-step enzymatic cascade that activates digestive hydrolases within the lumen of the gut. The biological importance of this pathway is demonstrated by the severe intestinal malabsorption and diarrhea that is caused by congenital enteropeptidase deficiency (3,4).
Bovine enteropeptidase is synthesized as a single-chain precursor of 1035 amino acid residues (5) that appears to require proteolytic activation, suggesting that enteropeptidase may not be the "first" protease of the digestive hydrolase cascade. Active enteropeptidase has been cleaved after Arg-800 to produce a disulfide-linked heterodimer with an amino-terminal Ϸ120-kDa heavy chain and a Ϸ47-kDa light chain; ϳ40% of the apparent mass of these polypeptides is due to glycosylation (6,7). The enteropeptidase heavy chain consists of an aminoterminal membrane-spanning domain, a mucin-like domain, two repeats found in complement serine proteases C1r and C1s, a MAM domain (so named for similar motifs first identified in the metalloprotease meprin, the Xenopus laevis neuronal protein A5, and protein-tyrosine phosphatase Mu), and a macrophage scavenger receptor cysteine-rich repeat (reviewed in Ref. 8). The light chain is a typical chymotrypsin-like serine protease. The activation cleavage site between the heavy and light chains has the sequence Val-Ser-Pro-Lys2Ile, which might be recognized by trypsin or other trypsin-like proteases. The identity of the endogenous proenteropeptidase activator is unknown. If trypsin were responsible in vivo, this would raise the logical question of how such a closed trypsin-enteropeptidase cycle could be initiated (reviewed in Ref. 8).
The determinants of enteropeptidase substrate specificity are not understood fully and may not be confined to the catalytic serine protease domain. The enteropeptidase light chain has been isolated in active form by partial reduction and alkylation of purified bovine intestinal enteropeptidase (9), with an average of Ϸ3 alkylated cysteine residues/molecule. A similar protein has been made by expression of recombinant enteropeptidase light chain (10), which is predicted to have at least 1 unpaired cysteine. These light chain preparations had markedly reduced ability to activate trypsinogen, but normal activity toward Gly-Asp-Asp-Asp-Asp-Lys-␤-naphthylamide. Similar selective defects in activity toward trypsinogen have been achieved by heating (6,11) or acetylation (12) of two-chain enteropeptidase. Although the structural cause of the impaired trypsinogen activation is uncertain, these observations suggest that the catalytic center and at least one secondary binding site on enteropeptidase cooperate to recognize trypsinogen.
To determine the influence of the enteropeptidase heavy chain on substrate recognition, secreted variants of recombinant proenteropeptidase were prepared, with or without the heavy chain. Proenteropeptidase was shown to be a zymogen with extremely low intrinsic protease activity, and it was activated efficiently by trypsin. The presence of the heavy chain slightly inhibited proenteropeptidase activation and had almost no effect on the recognition of the small peptide substrate Gly-Asp-Asp-Asp-Asp-Lys-␤-naphthylamide. However, the heavy chain had profound effects on the recognition of macromolecular substrates (trypsinogen) and inhibitors (pancreatic trypsin inhibitor and soybean trypsin inhibitor).
Construction of Expression Vectors-BEK cDNA clone A8 (5) was used as the template for expression vector construction. To generate the heavy chain-containing HL-BEK construct, an XhoI site was created at nucleotide 701 (before the first low density lipoprotein receptor-like domain) by PCR with primers 5Ј-gat atc aca gga aac tag tct cac tcga gtg ccc acc tg-3Ј (SpeI and XhoI sites are underlined) and 5Ј-cgt ctt cct cct ggt cca gtt g-3Ј (complement of nucleotides 1350 -1371 of BEK). The PCR product was digested with SpeI and BstXI; BEK clone A8 was digested with SpeI and partially digested with BstXI and then ligated to the PCR fragment to generate pBlue-HL. The desired fragment of pBlue-HL was isolated by digestion with XhoI and HindIII and then ligated to SalI/HindIII-digested His 6 vector pET-28b(ϩ) (Novagen, Madison, WI) to generate plasmid pET-28bHL. Human prothrombin signal peptide sequence was amplified by PCR using plasmid p⌬PT (13) as template and primers 5Ј-gga att ccc cat gga agc ttg aat tcc cag gag ctg aca c-3Ј (adjacent NcoI and HindIII sites underlined) and 5Ј-gat cga tct cat gac act tgt ggc ggt gaa tg-3Ј (BspHI site underlined). The product was digested with NcoI and BspHI and ligated to NcoIlinearized pET-28bHL to generate plasmid pET-28bHL/Nco, which contains the prothrombin signal peptide, the His 6 tag, and HL-BEK cDNA. This insert was removed by digestion with HindIII, made blunt, and ligated to SmaI-digested pNUT (14) to produce the expression plasmid pNUTHL. Plasmid pNUTL, encoding L-BEK lacking the enteropeptidase heavy chain, was constructed similarly except that an XhoI site was introduced at nucleotide 2453 in the first step by PCR with primers 5Ј-aat agc tct aga ctc gag agt aac tac aaa tca tgt ggg-3Ј (adjacent XbaI and XhoI sites underlined) and 5Ј-aaa ggc att aca ata aat ag-3Ј (complement of nucleotides 2998 -3017 of BEK). The PCR product was digested with XbaI and NsiI and ligated in place of the corresponding fragment of the parent template to generate plasmid pBlue-L. For all plasmids, the segments derived by PCR were sequenced to confirm the accuracy of the construction.
Expression and Purification of Recombinant BEK-Baby hamster kidney cells were transfected with plasmid pNUTHL or pNUTL by a calcium phosphate precipitation method and selected with methotrexate (400 M). Resistant colonies were pooled and propagated in Dulbecco's modified Eagle's medium with 10% fetal calf serum and ZnCl 2 (50 M). The cells were then washed three times with phosphate-buffered saline and maintained in serum-free Dulbecco's modified Eagle's me-dium with CaCl 2 (0.1 mg/ml). Cells were grown in 1-liter roller bottles, and conditioned medium was collected after 3 days and stored at Ϫ20°C. Medium was applied to a Q Fast Flow column (4 ml; Pharmacia Biotech Inc.) equilibrated with TB, pH 7.6 (20 mM Tris-HCl, pH 7.6, and 20 mM NaCl). The column was washed with 300 ml of TBS, pH 7.6, and eluted with a 50-ml linear NaCl gradient (20 -500 mM) in Tris-HCl, pH 7.6. Proenteropeptidase was detected after activation with trypsin by cleavage of GD 4 K-NA as described below. Pro-HL-BEK eluted at Ϸ260 mM NaCl; pro-L-BEK did not bind to the Q Fast Flow column and was recovered in the flow-through fractions. Active fractions were pooled; dialyzed against TB, pH 7.9; and applied to a Ni 2ϩ -nitrilotriacetic acid-agarose column (5 ml; QIAGEN Inc.). After washing with 200 ml of TB, pH 7.9, either recombinant pro-HL-BEK or pro-L-BEK was eluted with a 50-ml linear gradient for which the limit buffer was TB, pH 7.9, containing 60 mM imidazole and 500 mM NaCl. The eluted protein was dialyzed against TB, pH 7.6, and applied to a Mono Q column (2 ml; Pharmacia Biotech Inc.). After washing with 30 ml of TB, pH 7.6, enteropeptidase was eluted with a 30-ml linear NaCl gradient (20 -500 mM). Purified recombinant pro-HL-BEK or pro-L-BEK was dialyzed against TB, pH 7.6, and stored at Ϫ70°C.
Activation of Proenteropeptidase by Trypsin-Proenteropeptidase (50 nM) in 20 mM Tris-HCl, pH 7.6, 150 mM NaCl, and 5 mM CaCl 2 was warmed to 37°C, and trypsin was added to a final concentration of 1 or 5 nM. Samples were removed at intervals, and the reaction was terminated either by adding sample loading buffer (2% SDS and 2% ␤mercaptoethanol) for analysis by SDS-PAGE (15) and silver staining (16) or by adding ovomucoid to a final concentration of 50 nM for assay of activated enteropeptidase with GD 4 K-NA (17). The amino-terminal amino acid sequences of activated HL-BEK and L-BEK were determined after SDS-PAGE and electroblotting onto a polyvinylidene difluoride membrane as described previously (18).
To produce activated enteropeptidase free of trypsin, recombinant proenteropeptidase (500 nM, 5 ml) was incubated with immobilized trypsin-Sepharose 4B (100 l; Worthington) for 2 h at room temperature. The extent of the reaction was monitored by SDS-PAGE. Activated recombinant enteropeptidase was filtered and concentrated by ultrafiltration (Centricon-30, Amicon, Inc.). The concentration of activated enteropeptidase was determined by active-site titration with fluorescein mono-p-guanidinobenzoate (Sigma) using trypsin titrated with p-nitrophenyl p-guanidinobenzoate as a reference standard (19,20).
Enzyme Kinetics-Kinetic parameters for trypsinogen activation were determined at pH 5.6 to minimize the autoactivation of trypsinogen as described previously (6). Assays (1 ml) contained trypsinogen (0 -40 M) in 50 mM sodium citrate, pH 5.6, at room temperature. The reaction was initiated by adding enteropeptidase (1 nM). At different times, 100-l samples were removed and added to a microplate containing 100 l of 20 mM Tris-HCl, pH 8.4, 150 mM NaCl, and 500 M S-2765, and absorbance at 405 nm was recorded as a function of time. The concentration of trypsin generated was determined by comparing the rate of S-2765 hydrolysis with that of a standard of pure trypsin for which the concentration of active sites was determined by titration with p-nitrophenyl pЈ-guanidinobenzoate (19,20).
Kinetic parameters for trypsinogen activation were also determined at pH 8.4. To prevent autoactivation of trypsinogen, ovomucoid was included to sequester the trypsin product (21). Assays (1 ml) contained 0 -40 M trypsinogen, 25 mM Tris-HCl, pH 8.4, 10 mM CaCl 2 , and 40 M ovomucoid at 37°C. The reaction was initiated by adding enteropeptidase (0.3 nM). At different time intervals, 100-l samples were removed and added to 20 l of 0.2 M HCl to achieve a final pH of 2.0. This condition inactivates enteropeptidase and dissociates the trypsin-ovomucoid complex. To quantitate the free trypsin, an equal volume of 500 M S-2765 in 20 mM Tris-HCl, pH 8.4, and 150 mM NaCl was added, and absorbance at 405 nm was recorded as a function of time. This concentration of S-2765 was confirmed to prevent reassociation of trypsin and ovomucoid at pH 8.4, permitting accurate measurement of the trypsin as described under "Enzyme Kinetics." Kinetic parameters for cleavage of the synthetic peptide substrate GD 4 K-NA were determined as described previously (17). Assays (60 l) contained 0 -1 mM GD 4 K-NA, 25 mM Tris-HCl, pH 8.4, and 10 mM CaCl 2 at 37°C. The reaction was initiated by adding enteropeptidase (15 nM). At selected times, 10-l samples were removed and added to 3.5 l of 2 M HCl. Free ␤-naphthylamine concentration was determined as described previously (17).
Values for K m and k cat were obtained by directly fitting to the Michaelis-Menten equation by nonlinear least-squares regression. Under all assay conditions, the consumption of substrate (trypsinogen or GD 4 K-NA) was Ͻ15% of the total.
at room temperature (21°C) and contained 20 mM Tris-HCl, pH 8.0, 150 mM NaCl, 20 mM CaCl 2 , 0.1% polyethylene glycol 8000, 500 M S-2366, and either 0 -500 nM BPTI or STI. The reaction was initiated by adding enteropeptidase (0.2 nM), and absorbance at 405 nm was recorded (two points/min) over 30 min. The data were fitted by nonlinear leastsquares regression (KaleidaGraph, Synergy Software, Reading, PA) to Equation 1 (22), where [P] is the concentration of p-nitroanilide product at time t, v s is reaction velocity at steady state, v o is the initial reaction velocity, and k obs is the apparent rate constant for the transition from v o to v s . Reaction mechanisms associated with relatively slow enzyme inhibition include the following (Mechanisms A-C).
In Mechanism A, enzyme and inhibitor associate slowly and therefore approach equilibrium slowly. In Mechanism B, enzyme and inhibitor rapidly and reversibly form an initial "loose" complex (EI), which isomerizes slowly to the final complex (EI*). Mechanism C involves the slow isomerization of enzyme to a form (E*) that binds inhibitor. These mechanisms can be distinguished by the relationship of k obs to inhibitor concentration ([I]) (Equations 2-4),  (22). For both mechanisms, the plot is linear. For Mechanism A, the line passes through the origin. For Values for k 2 or k 4 were calculated from the values of k obs , v s , and v o that were determined by fitting to Equation 1. As indicated, values for k 1 (Mechanism A) or k 3 and K i (Mechanism B) were calculated by fitting appropriate parameters to Equation 2 or 3 by nonlinear least-squares regression. For Mechanism A,

Expression and Purification of Recombinant Bovine
Enteropeptidase-To facilitate the production of recombinant enteropeptidase, the membrane-spanning domain near the amino terminus was replaced with the human prothrombin signal peptide, allowing secretion into conditioned medium. A His 6 tag was included after the signal peptidase cleavage site, enabling purification by chromatography on Ni 2ϩ -nitrilotriacetic acid resin. Two forms of recombinant enteropeptidase were constructed (Fig. 1) to test the contribution of the heavy chain to macromolecular recognition. HL-BEK was truncated just before the first low density lipoprotein receptor-like domain and contains the remainder of the heavy chain and the light chain. L-BEK was truncated after the last cysteine residue of the macrophage scavenger receptor-like domain and contains mainly the light chain attached to a short carboxyl-terminal segment of the heavy chain. This preserves the disulfide bond predicted to link Cys-788 of the heavy chain to Cys-912 of the light chain, based on the organization of chymotrypsinogen and other homologous serine proteases that have disulfide linkages between the protease domain and the activation peptide (Fig.  1). The masses calculated from amino acid composition are 101 kDa for pro-HL-BEK and 36 kDa for pro-L-BEK.
Stably transfected baby hamster kidney cells secreted recombinant pro-HL-BEK or pro-L-BEK to a final concentration of Ϸ2 mg/liter in conditioned medium, and both proteins were readily purified to homogeneity. Upon gel electrophoresis under reducing conditions (Fig. 2) Ϸ150 and Ϸ60 kDa, respectively. Digestion with N-glycanase reduced the apparent masses to 130 kDa (pro-HL-BEK) and 40 kDa (pro-L-BEK) (data not shown), indicating that most of the difference between the calculated and observed masses is due to extensive N-linked glycosylation. Neither pro-HL-BEK nor pro-L-BEK (50 nM) cleaved detectable quantities of the synthetic substrate GD 4 K-NA (400 M) during a 1-h incubation, indicating that single-chain proenteropeptidase is a zymogen.
Activation of Proenteropeptidase by Trypsin-The zymogens pro-HL-BEK and pro-L-BEK were activated readily by trypsin. The time course of pro-HL-BEK activation by 2 nM trypsin (Fig.  3) indicates that generation of enteropeptidase activity correlated with the appearance of a Ϸ43-kDa fragment; no autocatalytic cleavage or activation was observed in the absence of trypsin. Trypsin (1 nM) cleaved and activated pro-L-BEK Ϸ3fold more rapidly than pro-HL-BEK (Fig. 4). The rate of pro-HL-BEK cleavage by trypsin was slightly faster in the presence of 5 mM calcium chloride than in 5 mM EDTA (data not shown), but the dependence on calcium ion concentration was not studied in detail.
Trypsin-activated HL-BEK had an apparent mass of 140 kDa under nonreducing conditions (Fig. 2, fifth lane), similar to that of uncleaved pro-HL-BEK (fourth lane), and contained two disulfide-linked fragments of 133 and 43 kDa (second and third lanes). Compared with the unreduced trypsin-digested sample (Fig. 2, fifth lane), the 133-kDa heavy chain (third lane) stained relatively faintly, suggesting that the heavy chain may be sensitive to degradation by trypsin. The apparent mass of activated L-BEK was also similar to that of uncleaved pro-L-BEK (Fig. 2, eighth and ninth lanes), and upon reduction, L-BEK contained a 43-kDa fragment similar to that of HL-BEK (third lane). The predicted amino-terminal activation fragment of L-BEK therefore appears to be disulfide-linked to the 43-kDa fragment under nonreducing conditions. Amino acid sequencing of the 43-kDa fragment from either HL-BEK or L-BEK gave the sequence Ile-Val-Gly-Gly-Ser-Asp-Ser-Arg-Glu-Gly, indicating that cleavage occurred at the site predicted from the cDNA sequence, after Lys-800 of full-length BEK (5). No amino-terminal sequence could be obtained for pro-L-BEK, pro-HL-BEK, or the 133-kDa chain of HL-BEK, suggesting that the amino terminus of the heavy chain is blocked in these preparations.

Kinetics of Substrate Cleavage by Enteropeptidase Variants-Pro-HL-BEK
and pro-L-BEK were cleaved with immobilized trypsin to avoid trypsin contamination of the active forms, and the concentrations of HL-BEK, L-BEK, and purified intestinal BEK were determined by active-site titration. All three preparations had similar kinetic constants for cleavage of GD 4 K-NA (Table I), and the values for K m and k cat were comparable to those reported for human enteropeptidase: K m ϭ The kinetics of trypsinogen activation were analyzed under two conditions. At pH 5.6, interference from the autoactivation of trypsinogen is minimal, although enteropeptidase has reduced activity (23,24). Under these conditions, BEK and HL-BEK had similar K m and k cat values, which are comparable to those reported previously for BEK: K m ϭ 17 M and k cat ϭ 6.3 s Ϫ1 (7) and K m ϭ 7 M and k cat ϭ 1.48 s Ϫ1 (6). In contrast, L-BEK cleaved trypsinogen relatively slowly; the K m was increased 24-fold, and the k cat was reduced 20 -40-fold relative to HL-BEK and BEK (Table I). Qualitatively similar results were obtained at pH 8.4, which is optimal for enteropeptidase. Under these conditions, the autoactivation of trypsinogen can be significant, but interference from this cause was prevented by inclusion of ovomucoid to inactivate the trypsin product (21). The K m for trypsinogen activation by L-BEK was increased 18-fold, and the k cat was decreased 2.5-3-fold.
Reaction of Enteropeptidase Variants with Inhibitors-The inhibition of enteropeptidase by BPTI was studied in the presence of the competing substrate S-2366, which conveniently slows the rate of inhibition and allows continuous monitoring of the remaining active protease. The values of K m for S-2366 hydrolysis determined by an initial rate method were 1.55 mM for BEK, 2.09 mM for HL-BEK, and 2.42 mM for L-BEK.
BPTI was shown to be a reversible slow binding inhibitor of enteropeptidase (Fig. 5). Addition of HL-BEK to a reaction containing substrate and BPTI caused a slow decline in the rate of substrate hydrolysis that achieved a steady state by Ϸ60 min (Fig. 5, curve 2). The same steady-state rate was approached by dilution of the inactive preformed enteropeptidase-BPTI complex into a similar reaction (Fig. 5, curve 3).
A slow binding inhibition pattern may be due to one of (at least) three mechanisms as discussed under "Experimental Procedures" (22). For Mechanism A, the establishment of equilibrium between E and a single EI complex is simply slow; for Mechanism B, an initial low affinity enzyme-inhibitor complex forms rapidly and then isomerizes slowly to a more stable complex; and for Mechanism C, the free enzyme slowly isomerizes to a form that then binds inhibitor. These mechanisms were distinguished by the dependence of the apparent rate constant (k obs ) on the concentration of inhibitor. For example, progress curves for the reaction of HL-BEK and BPTI were determined at several concentrations of BPTI (Fig. 6A), and values for v s , v o , and k obs were obtained by fitting to Equation 1. The plot of k obs versus [BPTI] is hyperbolic (Fig. 6B). When linearized by double-reciprocal transformation, the plot of Equation 3 has non-zero intercepts (Fig. 6C). These results are consistent with a slow binding inhibitor of Mechanism B.
In contrast to these results with BPTI, the reaction of L-BEK with STI did not exhibit evidence of an initial low affinity complex. The kinetics of inhibition were slow (Fig. 7A), but the plot of k obs versus [STI] is linear (Fig. 7B), and the corresponding double-reciprocal plot passes through the origin (Fig. 7C).
These data are consistent with a slow binding inhibitor of Mechanism A.
Kinetic parameters determined for the reactions of enteropeptidase variants with BPTI and STI are summarized in Table II. The data with BPTI were consistent with a slow binding inhibitor of Mechanism B. For the reaction of BPTI with L-BEK, the final K i * was 3-fold higher and the initial K i was 7-fold higher than for the corresponding reaction with HL-BEK. A reciprocal pattern was observed for reactions with STI: BEK and HL-BEK were completely resistant to STI, whereas L-BEK was inhibited more rapidly and more effectively by STI than by BPTI. Ovomucoid did not inhibit either HL-BEK or L-BEK (data not shown), confirming the suitability of ovomucoid to inactivate the trypsin product during kinetic studies of trypsinogen activation by these recombinant enteropeptidase variants.

DISCUSSION
The activation of trypsinogen is a key step in the activation of other digestive hydrolases within the lumen of the gut, and efficient catalysis of this reaction depends on enteropeptidase. Almost all vertebrate trypsinogens are activated by proteolytic cleavage of a Lys-Ile bond in an amino-terminal peptide that contains the sequence Asp-Asp-Asp-Asp-Lys-Ile (2). Molecular modeling of enteropeptidase suggests that specific basic residues on the surface of the catalytic subunit (light chain) interact directly with the acidic residues of trypsinogen activation peptides (25). Such interactions may account for the recognition of small peptide substrates, but probably are not sufficient to explain the recognition of trypsinogen. The isolated light  chain has been prepared by partial reduction of purified bovine enteropeptidase (9) or by expression of recombinant light chain (10). Both preparations had normal activity toward small synthetic peptides, but had dramatically reduced activity toward trypsinogen. Therefore, the recognition of small substrates requires only the light chain, whereas efficient cleavage of trypsinogen may also depend on the heavy chain. Similar selective defects in trypsinogen recognition were produced in two-chain enteropeptidase by heating (6,11) or by acetylation (12). This behavior suggests that the catalytic center and at least one secondary substrate-binding site (exosite) cooperate to recognize trypsinogen, and exosites sensitive to these treatments could be located on the heavy chain or the light chain.
To investigate the determinants of enteropeptidase substrate specificity, variants were constructed that contain or lack most of the heavy chain. Proteolytic removal of the transmembrane segment appears to have little effect on trypsinogen activation by bovine (9) or porcine (25) enteropeptidase; therefore, the putative membrane-spanning domain was replaced with a signal peptide to enable purification of secreted recombinant proteins from conditioned medium (Fig. 1).
Special attention was paid to the structure of the construct that encodes pro-L-BEK. Several chymotrypsin-like serine proteases have an "extra" disulfide bond that covalently links the activation peptide to the protease domain, and alignment of enteropeptidase with a subfamily of serine proteases suggests that the last cysteine of the heavy chain (Cys-788) is linked to the light chain (Cys-912, or Cys-122 in chymotrypsin numbering) (5). To preserve this predicted disulfide bond and to avoid the generation of an unpaired or abnormally paired cysteine, the construct retained the carboxyl-terminal 17 amino acids of the heavy chain that correspond to the chymotrypsin activation peptide. Covalent association of the short activation peptide and the catalytic domain was confirmed by demonstrating that under nonreducing conditions, pro-L-BEK and trypsin-activated L-BEK have similar electrophoretic mobility, whereas after reduction, the apparent mass of L-BEK (45 kDa) is substantially smaller than that of pro-L-BEK (60 kDa) (Fig. 2).
Both pro-HL-BEK and pro-L-BEK were purified as singlechain proteins and were found to have little (if any) catalytic activity. In particular, pro-HL-BEK had no more than 0.002fold the activity of two-chain HL-BEK toward GD 4 K-NA (Fig.  3). The predicted activation cleavage site of proenteropeptidase occurs in the sequence VSPK 800 IVGG, which has the appearance of a good site for cleavage by trypsin (5). In fact, trypsin readily activated both pro-HL-BEK and pro-L-BEK (Fig. 4) by cleaving them after Lys-800. Pro-HL-BEK was activated Ϸ3fold more slowly than L-BEK, indicating that the heavy chain has a modest inhibitory effect on the recognition of proenteropeptidase by trypsin.
The concentrations of proenteropeptidase tested (0 to Ϸ1 M) were not sufficient to allow accurate determination of the K m for cleavage by trypsin, which must be Ͼ1 M (data not shown). However, trypsin cleaved pro-HL-BEK at a rate of 1.0 s Ϫ1 and pro-L-BEK at a rate of 2.0 s Ϫ1 (each at 50 nM), and these turnover numbers are similar to those of other good trypsin substrates. Thus, proenteropeptidase is made as a single-chain zymogen, and it is activated efficiently by trypsin; enteropeptidase, in turn, activates trypsinogen, begging the question of how such a closed cycle could be initiated in vivo. Additional study is required to determine whether trypsinogen or proenteropeptidase is sufficiently "leaky" for this purpose, or whether another responsible protease may be present in duodenal mucosa or pancreatic secretions.
Although deletion of the heavy chain had only a modest effect on proenteropeptidase activation, it had a dramatic and selective effect on enteropeptidase substrate specificity. BEK, HL-BEK, and L-BEK had similar kinetic parameters for cleavage of GD 4 K-NA, indicating that the heavy chain does not strongly influence the recognition of small substrates (Table I). With the physiological substrate trypsinogen, BEK and HL-BEK also had similar values for K m (5.6 M), and the values for k cat differed Ͻ2-fold (Table I), indicating that the transmembrane and alternative exon domains have little effect on trypsinogen activation. This is consistent with the observation that proteolytic release of enteropeptidase from membranes did not impair trypsinogen cleavage (26). In contrast to these small effects, deletion of the entire heavy chain markedly inhibited the cleavage of trypsinogen (Table I). Compared with HL-BEK, the catalytic efficiency (k cat /K m ) of L-BEK was decreased Ϸ520-fold at pH 5.6 and Ϸ47-fold at pH 8.4. Therefore, heavy chain motifs between the first low density lipoprotein receptor-like domain and the macrophage scavenger receptor-like domain (Fig. 1) strongly influence the recognition of trypsinogen. The defect in trypsinogen activation is not explained by changes in the S2-S5 subsites of L-BEK since GD 4 K-NA cleavage is preserved. These results suggest the enteropeptidase heavy chain contains an exosite that interacts directly with trypsinogen.
In addition to promoting trypsinogen activation, the heavy chain also affects the reactions of enteropeptidase with macromolecular protease inhibitors. BEK purified from intestine was reported to be inhibited by BPTI with a dissociation constant of Ϸ5 nM (7), and recombinant HL-BEK also was found to be inhibited by BPTI with a final K i * of 1.8 nM (Table II). In addition, the kinetics were typical of slow binding inhibition (Fig. 5). Detailed analysis (Fig. 6) was consistent with a mechanism in which a low affinity complex was formed rapidly with a K i of 99 nM, and this complex slowly isomerized to the final high affinity complex with a forward rate constant (k 3 ) of 0.88 min Ϫ1 . Deletion of the heavy chain had little effect on k 3 , modestly increased the final K i * to 6.2 nM, and increased the initial K i Ϸ7-fold to Ϸ700 nM (Table II). An opposite pattern was observed for STI. BEK is completely resistant to STI (7), and HL-BEK also was not inhibited by STI. However, deletion of the heavy chain rendered L-BEK susceptible to slow inhibition by STI with a K i * of 1.6 nM (Table II). In addition, the kinetics of the L-BEK reaction with STI did not indicate the existence of an initial low affinity complex.
Isolated enteropeptidase light chain prepared by partial reduction and alkylation (9) and a recombinant form with an unpaired cysteine residue (10) also are inhibited by STI. Thus, deletion of the heavy chain impairs the inhibition of enteropeptidase by BPTI, but enables inhibition by STI. Unlike typical enteropeptidase substrates, BPTI and STI do not have acidic amino acids adjacent to the site that interacts with the catalytic center, and for this reason, the enteropeptidase heavy chain may be a particularly important determinant of reaction rates with these and other inhibitors.
The kinetic effects of deleting the enteropeptidase heavy chain clearly implicate it in the recognition of macromolecular substrates and inhibitors. The activation of trypsinogen is enhanced by the heavy chain, and its influence is especially significant under acid conditions (pH 4 -6) that commonly occur in the duodenum. The resistance of enteropeptidase to most protease inhibitors also appears to depend, in part, on the heavy chain. The unusual properties of the enteropeptidasetrypsinogen interaction appear analogous to those of certain blood coagulation factors for which the noncatalytic domains of serine proteases contribute to macromolecular recognition. These systems provide detailed insight into how serine proteases and their substrates are modified during evolution to perform specific, regulated functions by exploiting structural features of both catalytic and noncatalytic domains.