The Tetra-aspartate Motif in the Activation Peptide of Human Cationic Trypsinogen Is Essential for Autoactivation Control but Not for Enteropeptidase Recognition*

The activation peptide of vertebrate trypsinogens contains a highly conserved tetra-aspartate sequence (Asp19-22 in humans) preceding the Lys-Ile scissile bond. A large body of research has defined the primary role of this acidic motif as a specific recognition site for enteropeptidase, the physiological activator of trypsinogen. In addition, the acidic stretch was shown to contribute to the suppression of autoactivation. In the present study, we determined the relative importance of these two activation peptide functions in human cationic trypsinogen. Individual Ala replacements of Asp19-22 had minimal or no effect on trypsinogen activation catalyzed by human enteropeptidase. Strikingly, a tetra-Ala19-22 trypsinogen mutant devoid of acidic residues in the activation peptide was still a highly specific substrate for human, but not for bovine, enteropeptidase. In contrast, an intact Asp19-22 motif was critical for autoactivation control. Thus, single Ala mutations of Asp19, Asp20 and Asp21 resulted in 2-3-fold increased autoactivation, whereas the Asp22 → Ala mutant autoactivated at a 66-fold increased rate. These effects were multiplicative in the tri-Ala19-21 and tetra-Ala19-22 mutants. Structural modeling revealed that the conserved hydrophobic S2 subsite of trypsin and the unique Asp218, which forms part of the S3-S4 subsite, participate in distinct inhibitory interactions with the activation peptide. Finally, mutagenesis studies confirmed the significance of the negative charge of Asp218 in autoactivation control. The results demonstrate that in human cationic trypsinogen the Asp19-22 motif per se is not required for enteropeptidase recognition, whereas it is essential for maximal suppression of autoactivation. The evolutionary selection of Asp218, which is absent in the large majority of vertebrate trypsins, provides an additional mechanism of autoactivation control in the human pancreas.

Digestive trypsins are synthesized and secreted by the pancreas as inactive precursors. Physiological activation of trypsinogen takes place in the duodenum, where enteropeptidase (enterokinase) specifically cleaves the Lys 23 -Ile 24 peptide bond (see Refs. 1, 2, and references therein), which corresponds to Lys 15 -Ile 16 in the chymotrypsin-based numbering system (chymo#). 1 The activating cleavage removes a typically 8-amino acid-long activation peptide. In vertebrate trypsinogens, the activation peptide contains a highly conserved tetra-aspartate sequence next to the scissile peptide bond (Fig. 1). Experiments with synthetic peptides and protein substrates indicated that the acidic residues are required for enteropeptidase recognition and cleavage (1)(2)(3)(4)(5)(6)(7). Due to its highly specific extended subsite interactions, in analogy to restriction endonucleases, enteropeptidase became known as a restriction protease, and the Asp-Asp-Asp-Asp-Lys recognition site has been widely utilized as a protein engineering tool for the specific cleavage of fusion proteins. Definitive evidence that the Asp 19 -22 motif of the activation peptide participates in essential subsite interactions with enteropeptidase came from the crystal structure of the bovine enteropeptidase catalytic subunit complexed with an inhibitor analog of the activation peptide, Val-Asp-Asp-Asp-Asp-Lys-chloromethane (8). The structure demonstrated that the P2 and P4 Asp residues (corresponding to Asp 22 and Asp 20 in Fig. 1) formed salt bridges with Lys 99 (chymo#), a unique basic exosite on the catalytic subunit of enteropeptidase. In addition, the P3 Asp (Asp 21 in Fig. 1) was hydrogen-bonded to the hydroxyl group of Tyr 174 (chymo#). The P5 Asp (Asp 19 in Fig. 1) was disordered in the structure. In accordance with structural predictions, mutation of Lys 99 in the catalytic subunit of enteropeptidase abolished trypsinogen activation. Although the crystal structure suggested an important role for at least 3 of the 4 Asp residues in enteropeptidase recognition, a number of studies using various protein substrates or synthetic peptides indicated that a minimal recognition sequence for enteropeptidase consists of a Lys/Arg at P1 and an Asp/Glu at P2, whereas the P3-P5 acidic residues might enhance activity (1)(2)(3)(4)(5)(6)(7). Consistent with the critical role of the P2 Asp was the recent observation that the D22G mutant of human cationic trypsinogen was resistant to activation by bovine enteropeptidase (9).
The inhibitory function of the trypsinogen activation peptide on trypsin-mediated trypsinogen activation (autoactivation) was first demonstrated by chemical modification of the Asp residues in the activation peptide, which greatly enhanced autoactivation (10). Subsequently, tryptic digestion of synthetic model peptides indicated that Asp residues are not favored in the P2-P5 positions (11,12). More recently, biochemical characterization of pancreatitis-associated activation peptide mutations in human cationic trypsinogen confirmed the importance of Asp residues in the activation peptide in autoactivation control. A model peptide with the D22G mutation was cleaved by bovine trypsin at a higher rate compared with the wild type activation peptide (13), and recombinant human trypsinogens carrying the D19A or D22G mutations exhibited markedly increased autoactivation (9). Taken together with previous data, these findings indicated that the tetra-Asp sequence in the mammalian trypsinogen activation peptides has evolved for both efficient inhibition of trypsinogen autoactivation within the pancreas and optimal enteropeptidase recognition in the duodenum. However, to provide further experimental support to this notion, quantitative comparison of enteropeptidase-and trypsin-mediated trypsinogen activation, combined with systematic mutagenesis of the trypsinogen activation peptide, was necessary.
In the present study, the role of the Asp 19 -22 motif in the dual functionality of the activation peptide was investigated by site-directed mutagenesis in human cationic trypsinogen. This human trypsinogen isoform exhibits an unusually high propensity for autoactivation, to the extent that inborn mutations that moderately increase autoactivation cause hereditary pancreatitis (14 -18). Our findings demonstrate that the primary function of the tetra-Asp sequence in the activation peptide of human cationic trypsinogen is suppression of autoactivation. Two inhibitory interactions have been identified: one between Asp 22 of trypsinogen and the conserved hydrophobic S2 subsite of trypsin and another between the unique Asp 218 exosite and Asp 19 -21 of the activation peptide. Together, these interactions can suppress the rate of autoactivation by more than 2 orders of magnitude. In contrast, the Asp 19 -22 sequence is not required for enteropeptidase recognition and confers only a modest catalytic improvement to enteropeptidase-mediated trypsinogen activation in humans.

Materials-N-Benzyloxycarbonyl-Gly-Pro-Arg-p-nitroanilide
was from Sigma, and ultrapure bovine enterokinase was from Biozyme Laboratories (San Diego, CA). Recombinant human proenteropeptidase was from R & D Systems (Minneapolis, MN). Human proenteropeptidase (0.07 mg/ml stock solution; 641 nM concentration) was activated with 50 nM human cationic trypsin in 0.1 M Tris-HCl (pH 8.0), 10 mM CaCl 2 , and 2 mg/ml bovine serum albumin (final concentrations) for 30 min at room temperature and diluted 10-fold to obtain a working stock solution of 64 nM enteropeptidase concentration in 0.1 M Tris-HCl (pH 8.0), 1 mM CaCl 2 , and 2 mg/ml bovine serum albumin. The inclusion of bovine serum albumin was essential for the long term stability of enteropeptidase. Soybean trypsin inhibitor was purchased from Fluka and purified to homogeneity on a bovine trypsin affinity column followed by MonoQ ion exchange chromatography.
Nomenclature-The common names "cationic trypsinogen" and "anionic trypsinogen" are used to denote the two major human trypsinogen isoforms (19). Note that these names reflect only the relative isoelectric points of the proenzymes, and, in fact, all human trypsinogens are anionic. Autoactivation of trypsinogen is a bimolecular self-amplifying process in which trypsin activates trypsinogen to trypsin (trypsin ϩ trypsinogen 3 2trypsin). The term "autoactivation" is used throughout this study to describe trypsin-mediated trypsinogen activation. Amino acid residues in the human trypsinogen sequences are denoted according to their actual position in the native, wild-type preproenzyme, starting with Met 1 . Where indicated by the "chymo#" prefix, the chymotrypsin-based conventional numbering is also used.
Expression and Purification of Recombinant Trypsinogens-Trypsinogen mutants were expressed in the E. coli Rosetta (DE3) strain as cytoplasmic inclusion bodies. Trypsinogen was solubilized in guanidine HCl, renatured in vitro, and purified to homogeneity using ecotin affinity chromatography, as reported previously (16). Trypsinogen preparations were stored on ice, in 50 mM HCl. Concentrations of cationic and anionic trypsinogen solutions were calculated from their UV absorbance at 280 nm, using theoretical extinction coefficients of 36,160, 37,440, 37,320, and 36,040 M Ϫ1 cm Ϫ1 for wild-type cationic trypsinogen, D218Y cationic trypsinogen mutant, wild-type anionic trypsinogen, and Y218D anionic trypsinogen mutant, respectively. The molecular masses of recombinant cationic and anionic trypsinogens were 25,149.3 and 25,077 Da, respectively.
Autoactivation Assays-Trypsinogen in 50 mM HCl was diluted to a 2 M final concentration in the appropriate buffer, and the HCl was neutralized with NaOH, resulting in a 15 mM NaCl concentration. Buffers were used at 0.1 M final concentration, sodium acetate at pH 4.0 and pH 5.0, Na-MES at pH 6.0, Na-HEPES at pH 7.0, and Tris-HCl at pH 8.0. As inert protein, 2 mg/ml bovine serum albumin was included in the autoactivation mixtures. The albumin was extensively dialyzed against distilled water before use. Because commercial albumin preparations slightly inhibit human anionic trypsin, but not cationic trypsin (20), in the autoactivation experiments with anionic trypsinogen, bovine serum albumin was omitted. Autoactivation reactions were initiated by the addition of 10 nM trypsin (final concentration), and reaction mixtures were incubated at 37°C in the presence of 1 mM CaCl 2 (wildtype and mutant cationic trypsinogens) or 10 mM CaCl 2 (wild-type and mutant anionic trypsinogens). These Ca 2ϩ concentrations were previously found optimal for autoactivation of the two trypsinogen isoforms (20). At given times, 2-l aliquots were removed for trypsin activity assay. Trypsin activity was determined using the synthetic chromogenic substrate, N-benzyloxycarbonyl-Gly-Pro-Arg-p-nitroanilide (0.14 mM final concentration) in a 200-l volume. One-minute time courses of p-nitroaniline release were followed at 405 nm in 0.1 M Tris-HCl (pH 8.0), 1 mM CaCl 2 , at room temperature using a Spectramax Plus 384 microplate reader (Molecular Devices).
Calculation of Initial Rates-Progress curve analysis with KINSIM and FITSIM computer programs (21,22) was used to estimate the second-order rate constant of the "trypsin ϩ trypsinogen 3 2trypsin" reaction (see discussion in Ref. 23). Initial rates were then obtained by multiplying the rate constant with the initial concentrations of the reactants (i.e. 10 nM trypsin and 2000 nM trypsinogen), and rates were expressed as nM trypsin generated per min (nM/min). Under the conditions used in this study, degradation of cationic trypsin(ogen) during autoactivation was minimal, and this side reaction was ignored in the calculations. Autoactivation of the less stable anionic trypsinogen was always measured in the presence of 10 mM Ca 2ϩ , which provides sufficient protection against autolysis (20).
Activation of S200A Trypsinogen Mutants-Trypsinogens (2 M concentration) carrying the S200A mutation were activated with 10 nM human cationic trypsin, 50 ng/ml (0.45 nM) bovine enteropeptidase, or 14 ng/ml (0.13 nM) human enteropeptidase at 37°C in 0.1 M Tris-HCl (pH 8.0) and 1 mM CaCl 2 (final concentrations). To eliminate the effect of potential trypsin contamination, the enteropeptidase activation reactions were carried out in the presence of 0.12 M soybean trypsin inhibitor (final concentration), unless indicated otherwise. Aliquots were withdrawn from the activation mixtures, and trypsinogen was precipitated with trichloroacetic acid (10% final concentration). The precipitate was recovered by centrifugation and dissolved in Laemmli sample buffer containing 100 mM dithiothreitol (final concentration), and samples were heat-denatured at 95°C for 5 min. Electrophoretic separation was performed on 13% SDS-polyacrylamide minigels in standard Tris-glycine buffer. For the A 3 D/S200A and A 4 /S200A mutants, the conventional 13% reducing SDS-polyacrylamide gel did not resolve trypsinogen from trypsin; therefore, 21% gels and nonreducing conditions were used for separation. Gels were stained with Brilliant Blue R for 30 min; destained with 30% methanol, 10% acetic acid; and dried. Densitometric quantitation of bands was carried out as described in Ref. 20. Activated trypsin concentrations were determined from the density of the trypsin band, plotted as a function of time, and rates were estimated from linear fits to the initial portions of the reactions.

Zymogen Activation with Human and Bovine
Enteropeptidase-To study the significance of the Asp 19 -22 residues ( Fig. 1) in zymogen activation, individual Ala and Glu replacements were carried out in recombinant human cationic trypsinogen, yielding single Ala mutants D19A, D20A, D21A, and D22A and single Glu mutants D19E, D20E, D21E, and D22E. In addition, tri-Ala replacement of the Asp 19 -21 segment (mutant A 3 D) and tetra-Ala replacement of the entire Asp 19 -22 motif (mutant A 4 ) were also performed. Because of the strong autoactivation of cationic trypsinogen, specific rates of enteropeptidase activation could not be determined with catalytically competent proenzymes. Therefore, the catalytic Ser 200 (chymo# Ser 195 ) was replaced with Ala in wild-type (mutant S200A) and mutant trypsinogens.
Activation of S200A trypsinogen mutants by the enteropeptidase holoenzyme was performed at 37°C in 0.1 M Tris-HCl (pH 8.0), 1 mM CaCl 2 , followed by SDS-PAGE analysis and densitometric quantitation. To inhibit any possible trypsin carryover or contamination from the enteropeptidase preparations, the activation experiments were carried out in the presence of 0.12 M soybean trypsin inhibitor (Kunitz), which abolishes trypsin activity but does not affect enteropeptidase. Kinetic analysis of trypsinogen activation by human enteropeptidase yielded a K m of 1.4 M and a k cat of 35.1 s Ϫ1 (Table I). The K m value is in good agreement with previous reports, whereas the k cat is almost 10-fold higher, probably due to the ultrapure recombinant enteropeptidase preparation or to the different reaction conditions used (see Table I). Subsequently, the activation assays were carried out with 2 M trypsinogen concentration, which corresponds to the approximate K m value of the activation reaction. Unexpectedly, rates of activation of wild-type trypsinogen or mutants D19A, D20A, and D21A were identical when activated with human enteropeptidase (Fig. 1). Activation with bovine enteropeptidase proceeded ϳ4-fold slower (note the different enzyme concentrations used; also see Table I), but no measurable difference was detected between wild-type and D19A, D20A, and D21A trypsinogens. The data indicate that the Asp 19 -Asp 20 -Asp 21 segment does not play any significant role in enteropepti-dase recognition. However, a markedly different picture emerged when mutant D22A was activated with human or bovine enteropeptidase. Bovine enteropeptidase exhibited 24-fold decreased activity for the D22A mutant relative to wild-type trypsinogen, whereas activation by human enteropeptidase was only 1.2-fold slower. Clearly, Asp 22 plays a critical role for recognition by bovine enteropeptidase; however, it contributes only minimally to activation by the cognate human enzyme. The remarkable species specificity observed here underscores the importance of studying physiologically relevant enteropeptidase-trypsinogen pairs to understand the evolutionary processes that shaped zymogen activation.
The results in Fig. 1 strongly argue that, when examined individually, none of the Asp residues within the Asp 19 -22 motif is required for activation by human enteropeptidase and do not explain the strong evolutionary conservation of this sequence. The data also raise the possibility that, in fact, acidic residues within the activation peptide are entirely dispensable for enteropeptidase recognition and cleavage. To address this question, mutant A 3 D, in which only Asp 22 is present as a single acidic residue, and mutant A 4 , which is completely devoid of Asp residues, were activated with enteropeptidase. A number of technical difficulties hindered these experiments. First, removal of the Asp residues from the activation peptide abolished the mobility shift routinely observed on reducing SDS-polyacrylamide gels, and trypsinogen and trypsin migrated at identical positions. To visualize the activation reaction, activated trypsinogens were resolved on 21% gels, under nonreducing conditions. As shown in Fig. 2, in this electrophoresis system, the mobility shift becomes apparent again. Using matrix-assisted laser desorption ionization time-of-flight mass spectrometry, we have confirmed that the gel shift was caused by the removal of the activation peptide (not shown). Second, mutant A 4 was extremely trypsin-sensitive, and even minute amounts of trypsin(ogen) contamination originating from the affinity column or labware could result in significant trypsin-mediated activation. This problem was overcome by inclusion of soybean trypsin inhibitor in the activation reactions, and experiments were carried out with 0.12, 0.5, 1, and 2 M inhibitor concentrations, with identical results. Fig. 2A demonstrates that both human and bovine enteropeptidase activated A 3 D trypsinogen. When compared with activation of wild-type trypsinogen (cf. Fig. 1), the rate of A 3 D trypsinogen activation by human enteropeptidase was identical, whereas activation by bovine enteropeptidase was 2-fold decreased. This observation was consistent with the findings in Fig. 1, indicating that the Asp 19 -21 triplet is not required for recognition by enteropeptidase from either species. Strikingly, A 4 trypsinogen was also activated by human enteropeptidase with a rate that was ϳ2-fold decreased relative to wild-type trypsinogen. In contrast, bovine enteropeptidase did not cleave A 4 trypsinogen to any detectable degree (Fig. 2B). Kinetic analysis of A 4 trypsinogen activation by human enteropeptidase revealed a slightly increased K m (2.1 M) and an ϳ3-fold decreased k cat (11.2 s Ϫ1 ; Table I). Therefore, A 4 trypsinogen is still a specific enteropeptidase substrate, which is cleaved with high catalytic efficiency (k cat /K m ϭ 5.3 ϫ 10 6 M Ϫ1 s Ϫ1 ). Taken together, the results shown in Figs. 1 and 2 clearly demonstrate that the Asp 19 -22 motif per se in human cationic trypsinogen is not essential for recognition and cleavage by human enteropeptidase.
Zymogen Activation by Trypsin (Autoactivation)-Individual Ala mutations in all positions of the Asp 19 -22 motif stimulated autoactivation. At pH 8.0, 1 mM CaCl 2 , and 37°C, mutations D19A and D21A increased the autoactivation rate about 2-fold, whereas mutation D20A enhanced the rate more than 3-fold (Fig.  3). When the combined effect of these three mutations was tested with A 3 D trypsinogen, the autoactivation rate was 13-fold increased, indicating that the inhibitory effects of the Asp residues within the Asp 19 -Asp 20 -Asp 21 triplet are multiplicative. Remarkably, the D22A mutation stimulated autoactivation to such an extent that it was impossible to follow by activity assays. Therefore, trypsin-mediated activation of D22A trypsinogen was analyzed with SDS-PAGE and densitometry using the catalytically deficient D22A/S200A double mutant (Fig. 4). Relative to wildtype trypsinogen, mutant D22A exhibited a drastic increase in trypsin-mediated activation, which was 66-fold higher at pH 8.0. Although data are not shown, activation of the tetra-Ala A 4 / S200A mutant by trypsin was ϳ500-fold accelerated relative to wild-type (S200A) trypsinogen.
The significance of the negative charges within the Asp 19 -22 motif was also assessed by conservative Glu substitutions. Individual replacements of Asp 19 , Asp 20 , and Asp 21 by Glu had minimal, mostly inhibitory effects on autoactivation. Using 2 M trypsinogen and 10 nM trypsin concentrations, at pH 8.0, autoactivation rates of 1.7, 1.8, 0.9, and 0.9 nM/min were determined for wild-type, D19E, D20E, and D21E trypsinogens, respectively (not shown). However, conservative Glu replacement of Asp 22 in mutant D22E resulted in a 5-fold increased autoactivation rate (9.8 nM/min). Similarly, trypsin-mediated activation of the D22E/S200A mutant showed a 5-fold increased rate relative to S200A trypsinogen (Fig. 4). The results not only identify Asp 22 as the critical determinant of autoacti-  vation control but also establish that the entire intact Asp 19 -22 motif is mandatory for maximal suppression of autoactivation in human cationic trypsinogen. 19 -22 Motif of the Trypsinogen Activation Peptide-To visualize the subsite interactions between human cationic trypsin and the trypsinogen activation peptide, we superimposed the structures of human cationic trypsin (24) and bovine enteropeptidase light chain (8), which was crystallized in complex with an inhibitor analog of the activation peptide, Val-Asp-Asp-Asp-Asp-Lys-chloromethane (Fig. 5). The high degree of structural similarity between the two chymotrypsin-like serine proteases allowed the virtual positioning of the Asp 20 -Asp 21 -Asp 22 -Lys 23 sequence in complex with cationic trypsin. With respect to potential interactions of the Asp residues with cationic trypsin, two observations were made. First, the most important determinant of autoactivation, the P2 Asp 22 , faces a hydrophobic groove on trypsin lined by His 63 (chymo# His 57 ), Leu 104 (chymo# Leu 99 ), and Trp 216 (chymo# Trp 215 ). This hydrophobic subsite is well conserved in vertebrate trypsins (25), and it is responsible for the documented hydrophobic P2 preference of trypsin (26). Clearly, a negatively charged Asp residue is unfavored in this environment, whereas the small hydrophobic Ala in the D22A mutant would be readily accommodated. The 5-fold increased autoactivation of the D22E mutant is probably explained by the larger rotamer repertoire of the longer Glu side chain, which might mitigate the conflict between the negative charge and the hydrophobic S2 subsite.

Modeling the Interactions between Cationic Trypsin and the Asp
Second, in our model Asp 218 (chymo# Asp 217 ) on the surface of cationic trypsin came in close proximity to Asp 21 of the activation peptide. In contrast to the conserved hydrophobic S2 subsite, the Asp 218 negative exosite is absent in the vast majority of mammalian trypsins, which typically contain Tyr at this position (see "Discussion"). The structural model in Fig. 5 strongly suggests that the inhibitory action of the activation peptide in human cationic trypsinogen is partly due to an electrostatic clash between the negatively charged Asp 218 and one or more of the Asp residues in the activation peptide, most likely Asp 21 . This hypothesis was particularly attractive, because it further emphasized the highly specialized aspects of the Asp 19 -22 motif in human cationic trypsinogen. To provide functional evidence for this inhibitory interaction, position 218 was subjected to mutagenesis studies.
A Negative Charge at Position 218 Is Required for Inhibition of Autoactivation-First, Asp 218 was replaced with Tyr (mutant D218Y), because human anionic trypsinogen and most other mammalian trypsinogens carry a tyrosine residue at this position. Because protonation of Asp residues might affect the electrostatic repulsion between Asp 218 and Asp 21 , rates of autoactivation were measured over the pH range from 4.0 to 8.0. Autoactivation of wild-type and D218Y trypsinogen was essentially identical at pH 4.0 and pH 5.0; however, in the pH 6.0 -8.0 range, autoactivation of D218Y-trypsinogen was markedly stimulated, whereas wild-type cationic trypsinogen exhibited only a marginal increase (Fig 6, A and B). At the pH 7.0 optimum, the difference in the rates of autoactivation between wild-type and mutant D218Y trypsinogens amounted to 11-fold (32.3 versus 2.9 nM/min). For comparison, the pH dependence of autoactivation was also determined for the single Ala mutants D19A, D20A, and D21A as well as the tri-Ala mutant A 3 D (Fig.  6C). Interestingly, the single Ala mutations caused less pronounced changes (2-3-fold increase at the pH 7.0 optimum) than the D218Y mutation, suggesting that not only Asp 21 but the full Asp 19 -21 sequence is required for the optimal inhibitory interaction with Asp 218 . Consistent with this interpretation, removal of the entire Asp 19 -21 triplet in the A 3 D mutant resulted in an autoactivation profile that was essentially identical to that of the D218Y mutant. Taken together, these results support the predicted electrostatic inhibitory interaction between Asp 218 and the trypsinogen activation peptide.
To characterize further the side chain requirement at position 218 for suppression of autoactivation, Asp 218 was mutated to serine, histidine, or glutamate. These side chains also occur naturally at position 218 (e.g. in bovine cationic trypsin (Ser), in human mesotrypsin (His), or in a snake trypsin (Glu)). As shown in Fig. 7, a marked increase in autoactivation was observed with both the D218S and D218H mutations, indicating that the Ser and His side chains, together with Tyr, are unable to suppress autoactivation of cationic trypsinogen. On the other hand, autoactivation of wild-type and D218E mutant cationic trypsinogens were indistinguishable, demonstrating that a negative charge at position 218 is necessary for inhibition of autoactivation.
Introduction of Asp 218 into Anionic Trypsinogen Suppresses Autoactivation-Human anionic trypsinogen carries a Tyr resi-  Fig. 1). The P1 Lys 23 (red) of the activation peptide interacts electrostatically with the S1 specificity determinant Asp 194 (green; chymo# Asp 189 ). The P2 Asp 22 is facing a hydrophobic groove (yellow) formed by the catalytic His 63 (chymo# His 57 ), Leu 104 (chymo# Leu 99 ), and Trp 216 (chymo# Trp 215 ). The P3 Asp 21 (green) appears to interact with the Asp 218 (chymo# Asp 217 ) surface residue. The catalytic Ser 200 (chymo# Ser 195 ) is shown in orange. The image was rendered using DeepView/Swiss-PdbViewer version 3.7 (available on the World Wide Web at www.expasy.org/spdbv/) (33). due at position 218 and exhibits a pH profile of autoactivation that is similar to the D218Y cationic trypsinogen mutant (Fig. 8; compare with Fig. 6). Because the activation peptide sequence of anionic trypsinogen is identical to that of cationic trypsinogen, it seemed reasonable to assume that replacement of Tyr 218 with Asp should reconstitute the same interaction that operates in cationic trypsinogen. Indeed, the Y218D mutation reduced autoactivation of anionic trypsinogen almost 6-fold at pH 7.0 (5.4 versus 0.9 nM/min; Fig. 8), indicating that the inhibitory repulsion between Asp 218 and the tetra-aspartate tract of the activation peptide has been successfully engineered. DISCUSSION The most surprising finding of this study is that the highly conserved tetra-aspartate sequence in the activation peptide of human cationic trypsinogen is not required for enteropeptidase-mediated activation. Thus, not only were single Ala mutants D19A, D20A, D21A, and D22A all activated normally by human enteropeptidase, but the tetra-Ala replacement mutant A 4 , which is completely devoid of any acidic residues in the activation peptide, was also activated with somewhat reduced but still remarkable efficiency. Therefore, the tenet that human enteropeptidase recognizes its physiological substrate through the Asp 19 -22 motif appears to be wrong. Instead, enteropeptidase recognition seems to be determined by so far uncharacterized distant subsite interactions, in which the heavy chain of enteropeptidase might play a significant role. Interestingly, however, the presence of the P2 Asp 22 in the activation peptide was an almost absolute requirement for  , open symbols). A, autoactivation of 2 M trypsinogen was initiated with 10 nM trypsin (final concentration), and time courses were followed at 37°C in 0.1 M Na-HEPES (pH 7.0) and 10 mM CaCl 2 . B, pH dependence of autoactivation was determined as described under "Experimental Procedures" and in the legend to Fig. 6B. Note that autoactivation experiments with anionic trypsinogen were always performed in 10 mM Ca 2ϩ and in the absence of bovine serum albumin for maximal stability and activity (20). activation by bovine enteropeptidase (see Fig. 1). This observation suggests that the importance of the tetra-Asp motif in enteropeptidase recognition might be species-and isoform-specific. Because only bovine and human cationic trypsinogens have been characterized in detail, it is difficult to generalize the results to other vertebrate trypsinogens. We favor the hypothesis that the diminished significance of Asp 19 -22 in enteropeptidase-mediated activation of human cationic trypsinogen might indicate a recent evolutionary change, and the acidic residues in the activation peptide are necessary for enteropeptidase recognition in the majority of vertebrate trypsinogens.
The critical role of the P2 Asp 22 in activation by bovine enteropeptidase is also in agreement with the available crystallographic data and mutational analysis indicating an essential interaction between Lys 99 (chymo#) of the bovine enteropeptidase catalytic subunit and the P2 Asp (Asp 22 in Fig. 1) of the activation peptide (8). However, the crystal structure also shows that the P4 Asp (Asp 20 ) participates in a salt bridge with Lys 99 (chymo#), and the P3 Asp (Asp 21 ) is hydrogenbonded to the hydroxyl group of Tyr 174 (chymo#). It is not readily apparent why in our study no functional role could be demonstrated for the P3 and P4 Asp residues in trypsinogen activation by bovine or human enteropeptidase. One possible explanation is that the trypsinogen activation peptide interacts differently with the enteropeptidase catalytic subunit (used in the crystallization studies) and the enteropeptidase holoenzyme (used here). Alternatively, the interactions observed in the crystal structure might be redundant and might not translate to better catalytic efficiency.
Our results indicate that the conserved Asp 19 -22 motif has been maintained in human cationic trypsinogen for reasons that are unrelated to enteropeptidase recognition and suggest that autoactivation control is the primary function of this acidic sequence. Indeed, experiments with single and multiple Ala mutants confirmed that each Asp residue plays a role in autoactivation control, and their effects are synergistic in a multiplicative manner. The contribution of Asp 22 is the most significant, as indicated by the 66-fold increased autoactivation of the D22A mutant, whereas mutants D19A, D20A, and D21A exhibited 2-3-fold increased rates of autoactivation. Combination of the D19A, D20A, and D21A mutations in the tri-Ala mutant A 3 D resulted in 13-fold increased autoactivation, whereas the tetra-Ala mutant A 4 was activated by trypsin ϳ500-fold more rapidly than wild-type trypsinogen. Structural modeling revealed two distinct inhibitory interactions between the Asp 19 -22 motif of the activation peptide and cationic trypsin. Asp 22 is oriented toward the conserved hydrophobic S2 subsite of trypsin, formed by His 63 (chymo# His 57 ), Leu 104 (chymo# Leu 99 ), and Trp 216 (chymo# Trp 215 ), resulting in an unfavorable subsite interaction that explains the large effect of the D22A mutation. Furthermore, the unique Asp 218 surface residue, which forms part of the S3-S4 subsite on trypsin, appears to participate in an inhibitory electrostatic interaction with the P3 Asp 21 . Mutagenesis of Asp 218 clearly confirmed that this acidic exosite is essential for inhibition of autoactivation, and removal of the negative charge at position 218 can result in an 11-fold increase. Although modeling suggested that Asp 218 interacts with the P3 Asp 21 , we remain tentative about this interaction, because single Ala mutation of Asp 21 (D21A) stimulated autoactivation only 2-fold, and triple Ala mutation of the Asp 19 -21 sequence was necessary to achieve the same degree of autoactivation stimulation as with the D218Y mutant. To reconcile the structural and functional data, a plausible explanation is that an intact tetra-Asp sequence is required to position Asp 21 for optimal repulsion with Asp 218 . In single Ala mutants D19A, D20A, or D21A, the rearrangement of the remaining Asp residues can maintain a partial inhibitory interaction with Asp 218 , whereas the combined tri-Ala mutant A 3 D exhibits full relief from the Asp 218 -dependent inhibition. An uninterrupted tetra-Asp sequence also appears to be essential for Ca 2ϩ binding to the activation peptide. Although data are not shown, autoactivation of wild-type cationic trypsinogen was stimulated by Ca 2ϩ , whereas no effect was observed with mutants D19A, D20A, D21A, and D22A. It is likely that Ca 2ϩmediated stimulation is important for physiological zymogen activation in vertebrates and represents an additional selective pressure for the evolutionary conservation of the intact tetra-Asp motif in the activation peptide.
The present study also provides a molecular explanation for the previously described unique ability of human cationic trypsinogen to autoactivate at acidic pH (27). This is clearly due to the electrostatic repulsion between Asp 218 and the acidic activation peptide, which becomes prominent above pH 5.0 and suppresses autoactivation. As a result, despite the typical pHdependent stimulation of trypsin activity, autoactivation remains essentially unchanged between pH 5.0 and pH 8.0 (see Fig. 6). In contrast, bovine trypsinogen (28) or human anionic trypsinogen (20) (see also Fig. 8) autoactivate much better at pH 8.0 than at pH 5.0.
In evolutionary terms, the selective advantage of Asp 218 seems to lie in protection against premature trypsinogen autoactivation at neutral or alkaline pH, which prevails in the pancreatic ducts. However, this mechanism of autoactivation control appears to be rare among vertebrate trypsinogens. This position (chymo# 217) can carry a variety of residues (Tyr, Ala, Ser, His, Ile, Asp, and Glu), but there is a strong preference for Tyr, particularly in mammalian trypsinogens (25). Besides human cationic trypsinogen, Asp is found in rat trypsinogen V, but in this minor isoform, the tetra-Asp sequence of the activation peptide is disrupted by an Asn residue, suggesting that the electrostatic repulsion might not be optimal. In the recently released genomic sequence of the beta T-cell receptor locus from the rhesus macaque monkey (Macaca mulatta), the try9 and try13 trypsinogen isoforms contain Asp 218 (GenBank TM entry AC149201). Finally, Glu was identified at this position in a partial trypsinogen cDNA of the snake Bothrops jararaca (GenBank TM entry AF190273) and in a genomic trypsinogen sequence of the green pufferfish Tetraodon nigroviridis (Gen-Bank TM entry CAG00064).
A possible explanation why Asp 218 is conspicuously missing from most vertebrate trypsinogens is that the true selective advantage is not suppression of autoactivation per se but regulation of the pH dependence of autoactivation in such a manner that ensures essentially identical autoactivation over the pH range from pH 5.0 to pH 8.0. This could offer the obvious physiological benefit of enhanced zymogen activation in the duodenum when acidic gastric output lowers the pH transiently. It is noteworthy that trypsinogen activation by enteropeptidase has a similarly broad pH optimum, suggesting a case of convergent evolution to cope with the inhibitory effect of gastric acid. We speculate that in most species enteropeptidase-mediated trypsinogen activation is sufficient to achieve rapid and complete trypsinogen activation in the duodenum, whereas in humans, and possibly in a handful of other species, trypsinogen autoactivation is also required for full zymogen activation.
The same mechanism that ensures efficient zymogen activation in the duodenum might have significant pathological consequences as well. Autoactivation of cationic trypsinogen in the acidic secretory compartment can lead to premature intraacinar trypsinogen activation, even in the absence of cathepsin B activity (27). The situation is further aggravated in hereditary pancreatitis, in which inborn mutations increase the pro-pensity of cationic trypsinogen to autoactivate (14 -18). This notion is also supported by the fact that genetic variants of human anionic trypsinogen, which cannot autoactivate under acidic conditions, have not been found in association with hereditary pancreatitis or other forms of human pancreatitis (29,30).