Role of the occluding loop in cathepsin B activity.

Within the lysosomal cysteine protease family, cathepsin B is unique due to its ability to act both as an endopeptidase and a peptidyldipeptidase. This latter capacity to remove C-terminal dipeptides has been attributed to the presence of a 20-residue insertion, termed the occluding loop, that blocks the primed terminus of the active site cleft. Variants of human procathepsin B, where all or part of this element was deleted, were expressed in the yeast Pichia pastoris. A mutant, where the 12 central residues of the occluding loop were deleted, autoprocessed, albeit more slowly than the wild type proenzyme, to yield a mature form of the enzyme with endopeptidase activity comparable with the wild-type cathepsin B, but totally lacking exopeptidase activity. This deletion mutant showed a 40-fold higher affinity for the inhibitor cystatin C, suggesting that the occluding loop normally restricts access of this inhibitor to the active site. In addition, the binding affinity of the cathepsin B propeptide, which is a potent inhibitor of this enzyme, was 50-fold increased, consistent with the finding that the loop reorients on activation of the proenzyme. These results suggest that the endopeptidase activity of cathepsin B is an evolutionary remnant since, as a consequence of its membership in the papain family, the propeptide must be able to bind unobstructed through the full length of the active site cleft.

The lysosomal cysteine protease cathepsin B is unique within the papain superfamily in that it acts both as an endopeptidase and an exopeptidase. Thus in addition to making internal cleavages (1), it also removes C-terminal dipeptide units from the substrate (peptidyldipeptidase activity) (2). The x-ray crystal structure of cathepsin B suggests the molecular basis for this dual character. Relative to other papain-like proteases, cathepsin B contains an extra structural element termed the "occluding loop" (3), which blocks off one end of the substrate binding cleft (see Fig. 1a). As proposed by Schechter and Berger (4), the protease active site cleft is considered to consist of a series of subsites, each accommodating an amino acid residue of the peptide substrate. Those subsites binding the residues C-terminal to the peptide bond undergoing cleavage are termed the primed sites (S 1 Ј to S 3 Ј ) and accept the P 1 Ј to P 3 Ј substrate residues. The subsites accepting the N-terminal side of the substrate, the P 1 to P 3 residues, are termed the unprimed sites (S 1 to S 3 , with the numbering in each case starting with the residues bordering the cleavage site). The structure of rat cathepsin B containing the covalently bound inhibitor Z-Arg-Ser(OBz)-chloromethylketone (where Z represents benzyloxycarbonyl and Bz represents benzoyl) (5) maps out the unprimed substrate binding subsites and indicates that the occluding loop plays a defining role in primed site specificity. The strategic position of two histidine residues in this loop (His 110 and His 111 ) 1 suggests that they can act as acceptors for the negatively charged carboxylate of the P 2 Ј residue C terminus. The structure of a human cathepsin B complex with the E-64 2 analogue CA030 (the ethyl ester of epoxysuccinyl-Ile-Pro-OH) (6) shows this can indeed be the case. The free carboxyl group on the proline residue in the dipeptide moiety of this inhibitor interacts with these two histidines suggesting that Ile and Pro occupy the P 1 Ј and P 2 Ј subsites, respectively, as would be expected for substrate binding in the exopeptidase mode.
While the occluding loop appears to play an important role in the binding of substrates that are cleaved by the exopeptidase activity of cathepsin B, it would be expected to interfere with the binding of extended peptides. It would also be expected to obstruct the binding of protein protease inhibitors such as the cystatins, which in the case of the papain/stefin B complex have been shown to form extensive interactions with the protease primed sites (7). Since the occluding loop represents a separate and relatively poorly ordered region of cathepsin B (5), we reasoned that it could be deleted without compromising the structure of the remaining enzyme.
We have previously demonstrated the efficient recombinant expression of rat cathepsin B as an ␣-factor fusion construct in the yeast Saccharomyces cerevisiae (8). This system proved much less effective for the expression of the human enzyme (9); however, we have recently shown that high levels of human cathepsin B can be produced using similar constructs in the methanotropic yeast Pichia pastoris (10). We have used this system to produce forms of human cathepsin B lacking the occluding loop and thus allowing us to evaluate its role in cathepsin B function.  (5)) and papain (Protein Data Band code 9pap (11)) formed the basis for structural modelling. The Kollman all atom force field and Powell minimizer were used with the program SYBYL 6.1 (Tripos Assoc. Inc.). The non-bonded cutoff was set at 8 Å and the dielectric constant to 80. The segment Cys 108 to Cys 119 in the cathepsin B occluding loop was deleted, and the remaining ends were joined through a peptide linkage (mutant M1, see Fig. 1, a and c). Energy minimizations were performed allowing Ser 104 to Asp 124 and residues within 8 Å to move, while the rest of the molecule remained fixed. The minimizations were judged to be complete when the root mean square of the gradients was smaller than 0.1 kcal/mol Å.
Construction of Human Cathepsin B Deletion Mutants-A cDNA construct, consisting of wild-type human procathepsin B as a fusion with the preproregion of yeast ␣-factor, had been prepared previously in pVT105 (a derivative of the multiple purpose yeast shuttle vector pVT100U (12), where the ADH promoter had been deleted). The consensus sequence for oligosaccharide substitution in the mature protein had been removed by the substitution S115A (9). XhoI and NotI restriction sites were then introduced into the C-terminal region of the ␣-factor proregion and the 3Ј-untranslated region of procathepsin B, respectively, to allow subcloning into the P. pastoris expression vector pPIC9 (Invitrogen). Using single-stranded DNA from the construct in pVT105, two deletion mutants of human procathepsin B were constructed by loop-out mutagenesis. The oligonucleotide 5Ј-CCG TAC TCC ATC CCT CCC ACG GGG GAG GGA GAT ACC-3Ј was used to remove Cys 108 to Cys 119 , yielding the mutant M1 (see Fig. 1c). The oligonucleotide 5Ј-GTA GGG TGC AGA CCG TAC GAA GGT GTT CAA AAG TGT AGC AAA ATC TGT G-3Ј was used to delete residues Ser 104 to Pro 126 and add four amino acids from the papain sequence (Glu 89 -Gly 90 -Val 91 -Gln 92 ) as a linker, giving the mutant M2 (Fig. 1d). Sequencing of the complete procathepsin B cDNA inserts following mutagenesis confirmed the expected sequences. Both constructs were digested with XhoI and NotI, and the procathepsin B fragment was subcloned into pPIC9.
Expression of Mutants in P. pastoris-For integration into the Pichia genome, the pPIC9 based constructs were linearized by cleavage with BglII and purified. The P. pastoris host strain GS115 (Invitrogen) was then transformed with the linearized constructs by electroporation. Positive transformants were grown for 2 days in medium containing glycerol as the carbon source followed by growth in the presence of methanol for a further 2 days to induce expression of recombinant protein. Protein secreted into the culture supernatants was analyzed by SDS-PAGE and Western blotting using sheep anti-human cathepsin B (13).
Purification of M1 and M2 Mutant Proteins-After concentration, supernatants were dialyzed against 50 mM sodium acetate, 1 mM EDTA (pH 5.0) to allow autoprocessing of proenzyme mutants. The mature mutant M1 was reversibly inactivated by addition of methylmethanethiosulfonate (1 mM) and purified on an FPLC system (Pharmacia) using an SP-Sepharose fast flow column. The enzyme eluted with 150 mM NaCl. Mutant M2, which was not able to autoprocess, was purified in its proenzyme form using an SP-Sepharose fast flow column under the conditions described above. The proenzyme form of mutant M1 was purified by gel filtration chromatography using a Sephacryl 200HR column equilibrated in 25 mM Tris, 150 mM NaCl (pH 7.5).
Active Site Titration-The concentration of active wild-type enzyme was determined by titration using the cysteine protease inhibitor E-64 (Sigma) (14). The mutant M1 proved to be poorly inhibited by E-64, so enzyme concentrations were estimated by titration with recombinant human cystatin C prepared as described previously (15).
The proteolytic activity of cathepsin B mutants against protein substrates was performed using azocasein (18) in 50 mM sodium acetate (pH 5.2), containing 200 mM NaCl, 1 mM EDTA, 2 mM DTT. After incubation at 30°C for 30 min with 0.5% azocasein, proteins were trichloroacetic acid-precipitated, and A 366 of the supernatants containing liberated peptides was measured.
pH Activity Profiles-k cat /K m values were measured at 0.1 pH intervals over the range of pH The reaction buffers were 50 mM sodium citrate (pH 3.0 -6.0), 50 mM sodium phosphate (pH 5.8 -7.8), and 25 mM sodium borate (pH 7.7-8.4). All buffers contained 200 mM NaCl, 1 mM EDTA, and 2 mM DTT. The pH for each reaction was checked immediately after measurement of the initial rate.
Thermostability Measurements-Cathepsin B (0.7 nM) in 50 mM sodium acetate buffer (pH 5.2) containing 200 mM NaCl, 1 mM EDTA was incubated at 20 or 30°C for various times. The remaining enzymatic activity was measured using the same buffer system containing 2 mM DTT, 1% acetonitrile, and 5 M Z-Phe-Arg-MCA.
Interaction of Cystatin C with M1 Mutant and Wild-type Cathepsin B-Enzymes and human cystatin C were incubated either in 50 mM sodium acetate buffer (pH 5.2), containing 200 mM NaCl, 1 mM EDTA, 2 mM DTT, and 1% acetonitrile, or in 50 mM sodium phosphate (pH 6.0), containing 1 mM EDTA, 200 mM NaCl, 2 mM DTT, and 3% DMSO, for 5 min at room temperature. The residual enzyme activity was measured using 5 M Z-Phe-Arg-MCA. Values of K i were calculated by non-linear regression analysis using the general equation describing tight binding inhibition. Precise cystatin C concentrations were determined by titration with papain (19).
Inhibition by the Cathepsin B Propeptide-The rat cathepsin B propeptide (PCB1, residues 1-56) was prepared as described previously (20). Inhibition experiments were performed either in 50 mM sodium acetate (pH 5.2), 200 mM NaCl, 1 mM EDTA, 2 mM DTT, 1% acetonitrile or in 50 mM sodium phosphate (pH 6.0), 100 mM NaCl, 0.2 mM EDTA, 2 mM DTT, 0.025% DMSO using 5 M Z-Phe-Arg-MCA as substrate. Under these conditions, progress curves for the inhibition of the mutant M1 and wild-type cathepsin B by PCB1 at pH 5.2 and 6.0 followed typical slow-binding kinetics as defined by the equation, where P is the product formed, v i and v s are the initial and steady-state velocities, respectively, t is the reaction time, and k obs is the apparent second order rate constant for inhibition. Nonlinear regression analysis using the program Enzfitter (Elsevier-Biosoft, Cambridge, U. K.) gave the individual parameters (v i , v s and k obs ) for each progress curve. For each data set, the enzyme-inhibitor dissociation constant (K i ) values were obtained from the relationship v i /v s Ϫ 1 ϭ [I]/K i (21). The plot of k obs versus [I] remained linear over the range of inhibitor concentrations studied (2-20 nM), giving an estimate of k on and confirming that inhibition by PCB1 occurred by a one-step process (20). The dissociation rate constant k off was calculated from the relationship K i ϭ k off /k on .
SDS-PAGE Analysis-Proteins were treated with 10 M of E-64. When necessary, they were deglycosylated with endoglycosidase H. After precipitation with methanol (4 volumes), deglycosylation was performed in 25 mM sodium citrate buffer (pH 5.0) by incubation with 1 unit of endoglycosidase H (Boehringer Mannheim) for 1 h at 37°C. Proteins were analyzed by gel electrophoresis (12%) in the presence of SDS under reducing conditions.

Modelling of Cathepsin B Occluding Loop Mutants-
The possibility that the cathepsin B occluding loop could be deleted without causing significant perturbation of the remaining protein was investigated by molecular modelling. Inspection of the x-ray crystal structure of cathepsin B suggested that the disulfide-bonded segment, Cys 108 to Cys 119 , could be deleted and a peptide bond formed between Pro 107 and Thr 120 (Fig. 1c). Energy minimization of the resulting structure, the mutant M1 (Fig. 1a), showed that, on the formation of this bond, only a few residues on the surface of the molecule would be displaced, suggesting that deletion of this segment would not seriously affect the overall structure of the molecule.
Relative to papain, which lacks the occluding loop, the mutant M1 still included eight residues of this structural element.
For the complete deletion, the segment 104 -126 was removed, and the resulting gap closed using a connecting tetrapeptide derived from the homologous region of papain (Fig. 1d). Superimposition of the modified cathepsin B structure (the mutant M2) with that of papain suggested that these changes could be accommodated without radical restructuring of the protein.
Energy minimization of this more drastically modified form was not attempted.
Expression and Purification of the Procathepsin B Mutants-Deletion of part or all of the occluding loop was accomplished by loop-out mutagenesis. Both mutant proteins were expressed and secreted at levels comparable with the wild-type human enzyme (approximately 20 mg/liter of culture medium) and as previously observed for rat procathepsin B (10). The mutant M1 precursor was purified by gel filtration, and the proenzyme form of the mutant M2 was purified by ion exchange chromatography. About 10 mg of pure proenzymes were obtained in each case (Fig. 2). The recombinant cathepsin B proenzymes expressed in P. pastoris were heterogeneous due to modification of the N-linked oligosaccharide moiety on the proregion and migrated with apparent molecular masses ranging from 40 to 52 kDa with a dominant band at the leading edge of the smear (Fig. 2, lanes 1 and 5). Following enzymatic deglycosylation with endoglycosidase H, each cathepsin B form resolved as a single band of the expected size of 36.5 kDa (Fig. 2, lanes  2 and 6), migrating slightly faster than the deglycosylated wild-type proenzyme. The far UV circular dichroism spectra of the mutant proenzymes were similar to that of the wild-type proenzyme, suggesting correct folding (data not shown).
In Vitro Processing of the M1 and M2 Mutants-Wild-type procathepsin B is readily processed upon activation at acid pH (9), leading to the generation of the mature, active, single chain form of the enzyme (Fig. 2, lane 4). Processing of the M1 precursor to yield a proteolytically active enzyme proceeded at a much slower rate than that observed for the wild-type enzyme (Fig. 2, lane 3). Five days at pH 5.0 were required to totally activate the mutant M1, whereas the wild type enzyme was completely processed following overnight dialysis at 4°C. Processing was accelerated somewhat under lower pH conditions. Inclusion of the cysteine protease inhibitor, E-64, during dialysis prevented maturation of the precursor, suggesting that activation occurred as the result of autoprocessing. N-terminal sequencing demonstrated a three residue extension relative to  the fully processed lysosomal form. Attempts to process the mutant M1 using extrinsically added proteases (pepsin, papain, and proteinase K) were unsuccessful as the protein was either totally resistant or rapidly degraded (data not shown).
The mature mutant M1 was purified by ion-exchange chromatography on an SP-Sepharose column. From 1 liter of culture medium, 4 mg of pure mature M1 could be obtained, which migrated as a single band with a slightly higher mobility than wild-type cathepsin B (Fig. 2, lane 3). The mutant M2 did not autoprocess at acidic pH, even following lengthy incubation at 37°C. Processing could be attained in the presence of exogenous wild-type cathepsin B but not with other extrinsically added proteases such as pepsin, papain, or proteinase K. This conversion was shown to be both time-and dose-dependent as in the case of rat procathepsin B (C29S) (8). Further characterization of the mature mutant M2 was not carried out due to the inability to separate it from the wild-type cathepsin B used for processing.
Thermal Inactivation of the Mature M1 Mutant-The mutant M1 was found to be much more sensitive to thermal inactivation than the wild-type enzyme (Fig. 3). Loss of activity followed simple first order kinetics. At 30°C, inactivation of the mutant M1 occurred with a t 1/2 of 11 min compared with 110 min for the wild-type enzyme. A similar differential effect was seen at 20°C (t 1/2 of 28 and 280 min, respectively).
Endo-and Exopeptidase Activities of the M1 Mutant-Kinetic analysis was carried out with synthetic and protein substrates. Kinetic parameters, determined using the amidomethylcoumaryl substrate, Z-Phe-Arg-AMC, showed that removal of the external part of the loop had no significant effect on the k cat /K m value at pH 5.2 (Table I) although both K m and k cat showed a 5-fold decrease for the mutant M1 compared with the wild-type enzyme. At pH 6.0, k cat /K m was about 1.6-fold lower for the mutant M1 (Table I), consistent with a decrease in k cat at this pH. Little change in K m was observed between pH 5.2 and 6.0.
Activity of the mutant M1 was also tested against the protein substrate, azocasein (Table II). The deletion mutant was able to cleave this substrate with a specific activity similar to that of the wild-type enzyme. This activity was completely abolished in the presence of the cysteine protease inhibitor E-64. These estimates of the proteolytic activity of the mutant M1 are probably underestimates since, as shown in Fig. 3, there were substantial activity losses during the incubation times used in these assays. The peptidyldipeptidase activity of the mutant M1 was measured using the quenched fluorescence substrate dansyl-Phe-Arg-Phe(NO 2 )-Leu. Under the experimental conditions used, no exopeptidase activity was detectable for the mutant M1 in acetate or citrate buffers at pH 5.0, while native cathepsin B showed activity (k cat /K m of 432,000 M Ϫ1 s Ϫ1 ) comparable with that determined previously by Pohl et al. (17).
pH Activity Profiles Using the Substrate Z-Phe-Arg-MCA-As shown previously, wild-type cathepsin B exhibits a complex pH-activity profile that can best be fitted to a model involving five dissociation events, three in the ascending limb and two in the descending limb (22). Deletion of the occluding loop had a dramatic effect on the pH-activity profile (Fig. 4). Maximal activity was observed at pH 5, while the high levels of activity found for wild-type cathepsin B at pH 7-8 were not found with the mutant M1. Below pH 4.5, the activity dropped very rapidly. This loss of activity was in part the result of the reduced stability of the mutant M1 at acidic pH.
Inhibition of the M1 Mutant by Human Cystatin C-Since it had been proposed that the occluding loop of cathepsin B would obstruct the binding of cystatins to the active site (7), inhibition of the mutant M1 by human cystatin C was studied. At pH 6.0, a K i value of 0.17 nM for the wild-type cathepsin B/cystatin C interaction was found, consistent with values obtained by others (23,24) (Table III). However, the mutant M1 exhibited a 40-fold higher affinity for the inhibitor. The same differential effect was observed at pH 5.2. However, the affinity of both forms of the enzyme for the inhibitor was lower than at the higher pH.   (25,26) indicates the occluding loop adopts a different conformation in the proenzyme so that the propeptide is able to bind through the entire active site cleft. Such a reorientation of the occluding loop must occur for the free propeptide to bind to the mature enzyme. In agreement with this proposal, the affinity of the free cathepsin B propeptide for the mutant M1 was found to be 54-fold higher than for the wild-type enzyme at pH 6.0 (Table IV). A similar differential effect was seen at pH 5.2. In agreement with the proposal that the occluding loop must be reoriented during propeptide binding, the on rate is increased in the case of the mutant M1, whereas the off rate is essentially unchanged.

DISCUSSION
The occluding loop of cathepsin B represents a unique structural element relative to the other members of the papain superfamily. The three-dimensional structures of cathepsin B (3) and of a cathepsin B complex with an inhibitor occupying the S 1 Ј and S 2 Ј subsites (6) provide clues with respect to the exopeptidase activity. However, the relationship of the loop to endopeptidase activity is not clear. In order to study the role of the occluding loop, we have expressed a mutant of cathepsin B (M2) engineered to mimic the local structure of the other members of the papain superfamily. Although the folding of the precursor of this cathepsin B mutant did not appear to be altered significantly, as shown by circular dichroism spectroscopy (data not shown), this mutant did not autoprocess. In contrast, a more subtle deletion (M1) did allow processing to occur, albeit at a reduced rate relative to the wild-type precursor. Since the activation of wild type and mutant M1 procathepsin B proceeds by an autoactivation mechanism, the inability of the mutant M2 to undergo processing may be related to a stronger binding of the propeptide to the active site region. The slower autoprocessing of the mutant M1 and its higher affinity for the free propeptide support this proposal.
Studies of the mature mutant M1 provide insight into the functional role of the loop. Its partial truncation reduced the stability of the molecule indicating that this structural element contributes to the stability of the protein. In fact, the three dimensional structure of cathepsin B shows that, while the loop is attached to the left domain of the enzyme, residues 108 -119 interact with the right domain, suggesting that this region of the protein may be stabilized by interdomain interactions. In this regard, it was shown previously that interdomain perturbation in papain affects enzyme stability (27).
The three-dimensional structure also suggested an explanation of the unique ability of cathepsin B to act as an exopeptidase (3,6). Indeed, while the mutant M1 was proteolytically active, it showed no detectable exopeptidase activity, validating the previously proposed role of the loop residues His 110 and His 111 as "acceptors" for the negative charge on the substrate P terminus. On the other hand, the K m for the endopeptidase substrate Z-Phe-Arg-MCA decreased five-fold, showing that the peptide is able to bind to the active site in a more efficient manner. Similar endoprotease activity against a protein substrate for the wild type and M1 was demonstrated although the value for the mutant M1 is likely to have been an underestimate due to its reduced stability.
High molecular weight inhibitors also provide a method to probe active site accessibility. The recently determined threedimensional structures of rat (25) and human (26) procathepsin B reveal that the occluding loop is a mobile element. In the precursor molecule, the loop adopts an alternate conformation to that occupied in the mature enzyme, no longer impeding access to the primed side of the active site cleft. Indeed, the free cathepsin B propeptide must trigger a similar conformational transition on binding to the mature enzyme as it is a potent, specific inhibitor of cathepsin B (20) implying that it has the ability to displace the occluding loop. This finding also implies that shortening the loop should increase the binding affinity of the free propeptide to cathepsin B, a prediction confirmed on removal of 12 residues of the loop in the mutant M1 which enhanced the affinity of the free propeptide by more than 50-fold over that for the wild-type enzyme. As expected from the mechanism of the interaction, the on rate is increased while the off rate remains essentially unchanged.
It was pointed out previously that the conformation of the occluding loop in mature cathepsin B is incompatible with cystatin binding (3), as reflected by the high K i value for formation of the cystatin C/cathepsin B complex relative to that of cysteine proteases lacking the occluding loop (28). This is a similar situation to that described above for the binding of the propeptide and suggests that a significant alteration of the loop position occurs when the inhibitor binds to the cathepsin B active site. Indeed, partial deletion of the loop, as in the mutant M1, increases the affinity of cathepsin B for cystatin C by over 40-fold relative to the wild-type enzyme, a value similar to that found for propeptide inhibition. The K i value for the mutant M1 approaches that reported for cathepsin L (23), a cysteine protease lacking the occluding loop. All of these kinetic and structural observations are consistent with the predictions from molecular modelling of increased access to the active site on deletion of the loop.
Although the influence of the loop on the active site accessibility is clear, the occluding loop also has an influence on the hydrolytic activity of cathepsin B. For instance, the reduction of k cat by over 6-fold at pH 5.2 for the mutant M1 shows that the gain of binding efficiency (lower K m ) is offset by a reduction in substrate turnover. Also, modification of the loop lowers the   optimal pH of activity by over 2.5 pH units although the exact meaning of this effect remains obscure. The lack of symmetry of the pH-activity profile of wild-type cathepsin B has been attributed to the proximity of the loop and the protonation state of residues His 110 and His 111 (22). Shortening of the loop with the consequent removal of these two residues promotes a more symmetrical profile. Taken together, these data show the dramatic effect of the loop on the mechanism of activity. Besides its demonstrated role in exopeptidase activity, the occluding loop of the cathepsin B clearly limits access of macromolecules to the active site. It is interesting to speculate on the evolution of cathepsin B from an ancestral papain-like cysteine protease (29). The addition of the occluding loop to facilitate peptidyldipeptidase activity had to remain compatible with the propeptide binding pattern and activation mechanism common to this family of enzymes, as demonstrated by the structures of procathepsin B and procathepsin L (25,30). We propose, therefore, that the endoprotease activity retained by cathepsin B is an evolutionary remnant and that the main role of cathepsin B in its natural environment, the lysosome, is to act as an exopeptidase. The residual endopeptidase activity, while not essential intracellularly, still allows the enzyme to play an extracellular role as an endopeptidase that is poorly inhibited by cystatins (31). This may have important pathological consequences.