Plasticity of Extended Subsites Facilitates Divergent Substrate Recognition by Kex2 and Furin*

Yeast Kex2 and human furin are subtilisin-related proprotein convertases that function in the late secretory pathway and exhibit similar though distinguishable patterns of substrate recognition. Although both enzymes prefer Arg at P1 and basic residues at P2, the two differ in recognition of P4 and P6 residues. To probe P4 and P6 recognition by Kex2p, furin-like substitutions were made in the putative S4 and S6 subsites of Kex2. T252D and Q283E mutations were introduced to increase the preference for Arg at P4 and P6, respectively. Glu255 was replaced with Ile to limit recognition of P4 Arg. The effects of putative S4 and S6 mutations were determined by examining the cleavage by purified mutant enzymes of a series of fluorogenic substrates with systematic changes in P4 and/or P6. Whereas wild Kex2 exhibited little preference type for Arg at P6, the T252D mutant and T252D/Q283E double mutant exhibited clear interactions with P6 Arg. Moreover, the T252D and T252D/Q283E substitutions altered the influence of the P6 residue on P4 recognition. We infer that cross-talk between S4 and S6, not seen in furin, allows wild type and mutant forms of Kex2 to adapt their subsites for altered modes of recognition. This apparent plasticity may allow the subsites to rearrange their local environment to interact with different substrates in a productive manner. E255I-Kex2 exhibited significantly decreased recognition of P4 Arg in a tetrapeptide substrate with Lys at P1, although the general pattern of selectivity for aliphatic residues at P4 remained unchanged.

The subtilisin superfamily includes a subfamily of related processing proteases, the proprotein convertases that function in the late secretory pathway of diverse eukaryotic organisms including Saccharomyces cerevisiae, Caenorhabditis elegans, Drosophila melanogaster, and mammals (1)(2)(3)(4). Unlike the degradative subtilisins, which display a broad substrate specificity for hydrophobic residues (5), the proprotein convertases are post-translational modifying enzymes that process secretory proteins in a sequence-specific manner. In general, these pro-teases cleave C-terminal to clusters of basic residues, but their exact sequence specificity differs among the members of this family, even though they are Ն45% identical within their subtilisin-related domains. Similarities and differences in substrate recognition were illustrated by the enzymatic characterization of two members of this family, the S. cerevisiae protease, Kex2, and the human homologue, furin.
A detailed understanding of substrate recognition by Kex2 and furin has emerged from extensive analysis of the purified secreted, soluble enzymes using model peptide substrates. Based on these studies, the consensus cleavage site for Kex2 was determined to be (Ali/Arg)-Xaa-(Lys/Arg)-Arg2 (where Ali indicates an aliphatic amino acid), with the principal determinants being a basic residue at P 2 and Arg at P 1 (6 -9). 1 A conservative substitution of Lys for Arg at P 1 reduced k cat /K m of Kex2 1000-fold (10) and resulted in a change in the ratelimiting step from deacylation to acylation (10 -12). Kex2 exhibits a less stringent preference at P 4 , with a dual specificity for either a basic or an aliphatic residue (10). Furin also exhibits a strict requirement for Arg at P 1 , but, unlike Kex2, it has reduced selectivity for P 2 and increased dependence on P 4 recognition (13)(14)(15). For example, substitution of Ala for Arg at P 4 resulted in a 2500-fold decrease in k cat /K m (14).
Kex2 and furin also differ in P 6 recognition. Furin exhibited a 10-fold preference for Arg versus Ala at P 6 (14). Furthermore, the presence of a basic residue at P 6 can partially compensate for the lack of Arg at P 4 (14, 16 -18). An examination of physiological Kex2 substrates does not indicate any obvious P 6 selectivity, and in experiments with peptide substrates, Kex2 exhibits only a 2-fold preference for Arg at P 6 (14). This difference in P 6 recognition was also observed in interactions with derivatives of eglin-c that had been engineered to be potent inhibitors of Kex2 and furin (19). Kex2 exhibited only a slightly higher (ϳ3-fold) affinity for Arg (as opposed to Gly) at P 6 in an eglin-c variant having Arg at P 1 and P 4 (19). However, this same substitution of Arg for Gly at P 6 had a striking qualitative effect on the interaction of the inhibitor with furin, in that it caused the eglin-c variant to be cleaved. This result implies that the mode of P 6 recognition also is fundamentally different between Kex2 and furin, suggesting that, unlike Kex2, furin has a well defined S 6 subsite.
Crystal structures of subtilisin-inhibitor complexes, such as that of subtilisin bound to Streptomyces subtilisin inhibitor, illustrate that the principal residues in subtilisin BPNЈ that contact P 4 are Tyr 104 and Ile 107 (20). Based on these structural studies, several groups have mutated the S 4 pocket in an attempt to alter the P 4 substrate specificity of subtilisin (21)(22)(23)(24)(25)(26). Wells and co-workers (27,28) found that substitutions of Asp for Tyr 104 in subtilisin BPNЈ increased cleavage of substrates containing a P 4 Arg, but the resulting mutant protease did not discriminate between Arg and Phe at this position. In another study, acidic residues in furin predicted to interact with P 4 were mutated, and the mutant furin enzymes were co-transfected with a furin substrate, pro-von Willebrand factor. Substitution of Val for Asp 233 in furin, at a position equivalent to Tyr 104 in subtilisin, resulted in an enzyme that cleaved pro-von Willebrand factor with Ala at P 4 better than the wild type (WT) substrate (29).
In this work, the differences in substrate recognition by the S 4 and S 6 subsites of Kex2 and furin were explored by mutagenesis. Residues predicted to contribute to the specificity of P 4 and P 6 binding and that were different in Kex2 and furin were mutated in the yeast enzyme, and the substrate specificity of the mutants was analyzed. Substitutions in Kex2 were chosen prior to the availability of crystallographic data for Kex2 or furin and thus were based on examination of threedimensional structures of subtilisins and the amino acid sequences and structural models of Kex2 and furin (20, 30 -33). One group of mutations was generated with the goal of making Kex2 specificity more furin-like, by increasing selectivity for basic residues at P 4 and for Arg at P 6 . In addition, the model for the dual specificity of the S 4 subsite was tested by making a Kex2 mutant that was predicted to exhibit reduced recognition for basic residues at P 4 while retaining selectivity for aliphatic P 4 residues. Recently, the x-ray crystal structures were solved of the Kex2 catalytic domain complexed with tripeptidyl and tetrapeptidyl boronic acid inhibitors and of the furin catalytic domain complexed with a tetrapeptidyl chloromethylketone (34 -36). Through the comparison of the P 4 -S 4 interactions in Kex2 and furin, the structures have allowed us to interpret the results of these mutagenesis experiments with greater clarity. The biochemical data presented here will be discussed in light of the crystallographic structures.
Materials-DNA restriction enzymes, T4 DNA ligase, and oligonucleotides were from Invitrogen, and Pfu turbo polymerase was from Stratagene. Peptide substrates Boc-LKR2MCA and Pyr-RTKR2MCA were from Bachem, and all other peptide substrates were synthesized as described previously (7,10,14). General laboratory reagents were from Sigma and Fisher.
Site-directed Mutagenesis-All of the mutations were made by overlap extension (37). The following primers were used to make point mutations: (i) T252D, GGTGATATTACTGACGAAGATGA (sense) and TCATCTTCGTCCGTAATATCACC (antisense); (ii) E255I, CGGAA-GATRTRGCTGCTAGCTTGATTTA (sense) and TAAATCAAGCTAG-CAGCYAYATCTTCCG (antisense); (iii) Q283E, GGAAGACATTTA-GAAGGCCCTAG (sense) and CTAGGCCCTTCTAAATGTCTTCC (antisense); and (iv) V289A, GTGACCTGGCCAAAAAGGC (sense) and GCCTTTTTGGCCAGGTCAC (antisense). The template for all single mutations was pAL7, (38) a pRS314-based vector encoding the fulllength KEX2 gene with an additional XhoI site downstream of the P-domain (39), and its expression was regulated by its WT promoter. Q283E-Kex2 served as a PCR template for creation of the T252D/ Q283E-Kex2 double mutant. PCR products were subcloned into pAL7as HindIII to BglII fragments, and the incorporation of each mutation was confirmed by DNA sequencing (University of Michigan DNA sequencing core).
Expression and Purification of Mutants-The substituted Kex2 DNAs were recombined with a linearized expression vector for the production of secreted, soluble Kex2 mutants. The general method was described in Ref. 40. Briefly, the expression vector pAL10 was a deriv-ative of the secreted, soluble Kex2 expression plasmid, pG5KEX2⌬613 (9), in which a XhoI site was inserted in place of internal KEX2 sequences from a point 370 nucleotides downstream from the start codon to a point 1788 nucleotides downstream from the start codon, just 3Ј to sequences encoding the P-domain. pAL7 vectors encoding the Kex2 mutants were linearized with BamHI and co-transformed into CBO17 with XhoI-digested pAL10. Transformants containing recombinant plasmids and thus encoding mutant-secreted, soluble Kex2 were selected on synthetic dextrose complete-Ura plates. Individual colonies were grown overnight in synthetic dextrose complete-Ura liquid medium and then inoculated into 1040 expression medium (41). After growth at 30°C for 24 h, the medium was checked for activity. Equal amounts of medium and substrate solution (140 M BocQRR2MCA, 400 mM BisTris, 2 mM CaCl 2 ) were mixed in wells of a 96-well plate and release of the fluorogenic reporter was determined using a Molecular Devices fmax fluorescence plate reader. The cell cultures secreting active enzyme were reinoculated into fresh medium and incubated for 24 h at 30°C. The enzymes were purified as described (9). The purified proteins were active site-titrated as described (7,9).
Kinetic Characterization of Wild Type and Mutant Enzymes-Pseudo first order and saturation measurements were carried out at 37°C in 0.2 M Bis Tris, 1 mM CaCl 2 , 0.1% Triton X-100 as described (7,10,14).
Error Analysis-The error for all experiments is listed as S.D. in the form of the percentage of deviation of the average value for each data point. These values were calculated using Microsoft Excel. 4 and P 6 -To increase recognition of basic side chains at P 4 and P 6 , putative S 4 and S 6 residues were selected by comparison of Kex2 and furin sequence alignments and model structures as well as on results of previous mutagenesis experiments (24,28,29). Again, these residues were chosen prior to the availability of any crystallographic data. Substitution of Asp for Thr 252 in Kex2, a position equivalent to S 4 residue Tyr 104 in subtilisin BPNЈ and Asp 233 in furin, was originally introduced to increase recognition of basic versus aliphatic residues at P 4 (Figs. 1 and 2). Because substitutions at this position in both subtilisin and furin exhibited significant alterations in their P 4 specificities, this residue was also considered a good candidate for tuning the P 4 specificity of Kex2 (24,28,29). The S 6 subsite was more difficult to model because the degradative subtilisins described to date do not have a distinct binding pocket for P 6 , but Seizen et al. (32) tentatively assigned an insertion loop, with respect to subtilisin, in furin to form the S 6 subsite. Only very recently, the crystallographic data for Kex2 clarified the structure of this insertion (see "Discussion"). Within this region, furin has Asp at amino acid 264, equivalent to Gln 283 in Kex2 (Figs. 1 and 2). Gln 283 was mutated to a Glu to mimic the charge at that position in furin while minimizing the change in geometry in the binding site. Glu at this position was expected to be well tolerated as it is also found in PC1/3 (33). The T252D/Q283E double mutant was constructed to determine whether this would result in a Kex2 mutant with furin-like specificity at P 4 and P 6 . Finally, to decrease recognition of basic residues at P 4 while maintaining interactions with aliphatic side chains, Ile was substituted for Glu 255 , a potential site of interaction with basic residues equivalent to Ile 107 in subtilisin and Glu 236 in furin.

Mutation of Kex2 to Alter Recognition of Basic Residues at P
Effects of the T252D and Q283E Substitutions on Specificity for Basic and Aliphatic Amino Acids at P 4 in the Context of Tetrapeptide Substrates-Although the majority of known FIG. 1. Comparison of Kex2 and furin sequences in the putative S 4 and S 6 subsites. After comparison of charged residues between Kex2 and furin, Thr 252 and Gln 283 , which were basic for furin and were implicated in extended subsite recognition, were chosen for mutagenesis. Another substitution was made for conserved residue between Kex2 and furin, Glu 255 . physiological Kex2 substrates have an aliphatic residue at P 4 , purified Kex2 can also cleave substrates with Arg at P 4 . In fact, using the model substrates Ac-␤YKR2MCA 2 and Ac-RYKR2MCA, Kex2 exhibited a 3-fold preference for the P 4 Arg substrate (see Table I) (10). Further investigation indicated that the positive charge of the guanidinium group of Arg, and not the aliphatic portion, was the critical determinant for the S 4 recognition of Arg at P 4 , suggesting that Kex2 binds aliphatic and basic residues using different binding modes (10). Such dual specificity at P 4 is not observed with furin, which has a clear preference for Arg at this position (14). Thus, increasing the net negative charge in the S 4 subsite of Kex2 would be expected to disfavor the binding of aliphatic residues and enhance binding of basic ones. Indeed, T252D-Kex2 exhibited k cat /K m values with the model substrates Ac-␤YKR2MCA and Ac-RYKR2MCA that indicated a 5-fold preference for Arg over Nle at P 4 (see Table I). Unexpectedly, the Kex2 mutant with a putative S 6 mutation, Q283E-Kex2, also exhibited an increased, 4-fold, preference for Arg over Nle, whereas the double mutant T252D/Q283E-Kex2 displayed similar specificity constants to wild type with these substrates. Thus, the very slightly enhanced recognition of Arg over Nle at P 4 in tetrapeptides with Arg P 1 was observed with both of the putative S 4 and S 6 mutant enzymes but not the double mutant.
Previously, it was demonstrated that P 4 recognition becomes more important when Lys is substituted for Arg at P 1 (10). Substitution of Lys for Arg at this position resulted in a change of the rate-limiting step from hydrolysis of the acyl enzyme intermediate (deacylation) to its formation (acylation) (10,12). Comparison of k cat /K m values for Ac␤YKK2MCA and AcRYKK2MCA showed an increasing preference for P 4 Arg versus Nle in the order WT Ͻ T252D-Kex2 Ͻ Q283E-Kex2 Ͻ T252D/Q283E-Kex2 (see Table I). T252D/Q283E-Kex2 exhibited a 6-fold preference for Arg over Nle at P 4 in the context of Lys at P 1 . Although the effects of the T252D and Q283E substitutions were not additive, each mutation contributed toward the increased specificity of the double mutant for P 4 Arg versus Nle. Relative to WT-Kex2, T252D-Kex2 exhibited a 7-fold higher k cat /K m for AcRYKK2MCA and a 3-fold higher k cat /K m for Ac␤YKK2MCA (see Table I). Q283E-Kex2 exhibited a 2.5-fold reduction k cat /K m for Ac␤YKK2MCA relative to WT-Kex2 (see Table I). These results suggested that both an increased acylation rate with the P 4 Arg substrate and a decreased acylation rate with the P 4 Nle substrate contribute to the enhanced discrimination of Arg versus Nle by T252D/Q283E-Kex2.
Backbone Contacts at P 5 and P 6 Affect P 4 Recognition by WT but Not Mutant Forms of Kex2-Most kinetic analyses of Kex2 specificity have been performed with substrates lacking a P 6 residue. To evaluate the contribution of the S 6 -P 6 interaction toward the processing of hexapeptide substrates, the proteolysis of a series of hexapeptide substrates was analyzed using pseudo first order kinetics (Fig. 3). However, in addition to specific interactions, P 5 and P 6 residues in hexapeptide substrates could conceivably provide nonspecific backbone contacts that could reduce the relative importance of P 4 binding. This possibility was tested for WT and mutant enzymes by comparing the k cat /K m ratio for a pair of hexapeptide substrates with Ala at P 5 and P 6 (AcAARYKR-MCA and AcAAAYKR2MCA) to the k cat /K m ratio for the analogous tetrapeptide substrates (AcRYKR2MCA and AcAYKR2MCA; Fig. 4, Table I). In the case of WT Kex2, the specificity for Arg versus Ala at P 4 decreased from 28-fold in the tetrapeptide context to 3.4-fold in the hexapeptide context. Moreover, the presence of nonspecific contacts at P 5 and P 6 resulted in a 3.7-fold increase in k cat /K m of WT Kex2 for AcAAAYKR2MCA as compared with AcAYKR2MCA, but no such increase was observed with any of the mutants (Table I). Furthermore, the mutants were markedly more specific than WT Kex2 for Arg versus Ala at P 4 in the P 5 , P 6 Ala substrates. T252D-Kex2 displayed a 6.6-fold and Q283E-Kex2 displayed a 13-fold preference for P 4 Arg in the context of the hexapeptides. Even more strikingly, T252D/ Q283E-Kex2 exhibited 32-fold higher k cat /K m values with Arg versus Ala at P 4 with both the tetrapeptide and hexapeptide substrates (Table I and Fig. 4). Unlike wild type, the additional potential nonspecific contacts at P 5 and P 6 did not result in a decrease in P 4 specificity. T252D/Q283E-Kex2 displayed a 10fold higher specificity than did WT-Kex2 for P 4 Arg in the context of the nonspecifically extended substrates.
T252D Affects P 6 Recognition-Although the presence of Ala at P 5 and P 6 did not affect the P 4 specificity of the Kex2 mutants, this fact did not rule out P 6 recognition by these enzymes. A preference for Arg at P 6 was exhibited to some degree by both WT and the mutant Kex2 proteins (Table I and Fig. 5). In terms of k cat /K m for cleavage of AcRAAYKR2MCA and AcAAAYKR2MCA, both WT and Q283E-Kex2 exhibited relatively modest preferences for Arg (4-and 2.4-fold, respectively). Surprisingly, however, the T252D mutation in the pu- 2 The abbreviations used are: WT, wild type; Nle or ␤, norleucine; , norvaline; , cyclohexylalanine; MCA, 7-amino-4-methylcoumarin. tative S 4 subsite of Kex2 had a dramatic effect on P 6 recognition, with T252D-Kex2 exhibiting a 14-fold higher k cat /K m for Arg than for Ala at P 6 . This enhanced discrimination was retained in the T252D/Q283E-Kex2 enzyme, which was 15 times more specific for Arg at P 6 . Thus, the T252D mutation unexpectedly affected both P 4 and P 6 recognition and had a larger impact on S 6 -P 6 interactions than on S 4 -P 4 interactions.
In the case of furin, the P 6 side chain can clearly make a significant contribution to substrate recognition (14,16,42). A possible manifestation of this P 6 recognition was the observation that purified, soluble furin was inhibited by high concentrations (Ն5 M) of hexapeptide, but not tetrapeptide, substrates (14). 3 In contrast, high concentrations of hexapeptide substrates did not inhibit WT Kex2 (14). Because T252D/ Q283E-Kex2 showed improved P 6 recognition relative to WT Kex2, saturation kinetics were performed using AcRAKYKR2MCA. However, the T252D/Q283E-Kex2 exhibited saturation kinetics indicating that substrate inhibition did not occur (Fig. 6). In the cleavage of AcRAKYKR2MCA, T252D/Q283E-Kex2 had a lower k cat than WT Kex2 (40 s Ϫ1 for T252D/Q283E-Kex2 and 200 -250 s Ϫ1 for WT) but also exhibited a lower K m (0.2 M T252D/Q283E-Kex2 and 0.8 M for WT) (14).
Cross-talk between S 4 and S 6 Subsites of Kex2-The k cat /K m ratios for cleavage of AcRARYKR2MCA and AcRAAYKR2MCA were compared with k cat /K m values for cleavage of AcAARYKR2MCA and AcAAAYKR2MCA to reveal whether the presence of a favorable residue (i.e. Arg) at P 6 altered the P 4 specificity profile of WT or mutant forms of Kex2 (Fig. 7). Whereas P 4 specificity of WT and Q283E-Kex2 was only slightly decreased when Arg was present at P 6 , specificity for Arg at P 4 was substantially reduced with Arg at P 6 in the case of T252D-Kex2 and T252D/Q283E-Kex2 (Table I and Fig. 7). The relative preference of T252D-Kex2 for P 4 Arg versus Ala decreased from 6-fold with Ala at P 6 to 1.2-fold with Arg at P 6 . The effect of P 6 on P 4 specificity was even more pronounced with T252D/Q283E-Kex2. In the context of P 6 Ala, the double mutant had a 32-fold preference for P 4 Arg versus Ala. The inclusion of Arg at P 6 decreased P 4 specificity nearly 10-fold, resulting in only a 3.5-fold preference for P 4 Arg.
The E255I Substitution Reduces Recognition of Basic Residues at P 4 -The effects of the E255I, predicted to lessen the recognition of basic residues at P 4 by removing an acidic residue from the S 4 pocket, were clearest in the context of a P 1 Lys residue. E255I-Kex2 showed a marked decrease in preferential processing of a substrate with a P 4 Arg relative to one with a P 4 Ala when Lys was present at P 1 (Table II and Fig. 8). Whereas WT Kex2 exhibits a ϳ100-fold higher k cat /K m for AcRYKK2MCA than for AcAYKK2MCA, E255I-Kex2 exhibited only a 6-fold preference for the P 4 Arg substrate (Table II and Fig. 8). This was the result of a 4-fold decreased k cat /K m for cleavage of AcRYKK2MCA by the mutant enzyme as compared with the WT combined with a 4-fold increase in k cat /K m for cleavage of AcAYKK2MCA by the mutant enzyme as compared with the WT.
In contrast, E255I-Kex2 exhibited little or no decrease in the recognition of aliphatic residues at P 4 in the context of Lys at P 1 . Relative to WT Kex2, E255I-Kex2 exhibited 2-4-fold higher k cat /K m values for cleavage of P 1 Lys substrates having an aliphatic residue, Ala, Nle, Val, or cyclohexylalanine, at P 4 . E255I-Kex2 cleaved Ac␤YKK2MCA, with norleucine at P 4 , with a k cat /K m increased ϳ2-fold relative to the WT enzyme. As a result, E255I-Kex2 maintained a 30-fold preference for Nle versus Ala at P 4 (WT Kex2 exhibits a 77-fold preference). The ratios of catalytic efficiencies for processing of AcVYKK2MCA versus AcAYKK2MCA and AcYKK2MCA versus AcAYKK2MCA were not significantly altered by the E255I substitution (Fig. 9). Thus, E255I Kex2 exhibited reduced recognition of a basic residue at P 4 without a concomitant loss of recognition of aliphatic residues. As a result, whereas WT Kex2 exhibits similar k cat /K m values for P 4 Arg and aliphatic substrates having Lys at P 1 , the mutant enzyme exhibited 5-fold preference for Nle versus P 4 Arg and a 13-fold preference for cyclohexylalanine versus P 4 Arg.
The preferences shown by E255I-Kex2 for P 4 aliphatic versus basic residues largely disappeared with Arg at P 1 . Relative to WT Kex2, E255I-Kex2 exhibited slightly diminished activity and slightly relaxed P 4 specificity in cleavage of the hexapeptide and tetrapeptide substrates having a P 1 Arg (Table II). Pairwise comparison of AcAARYKR2MCA and AcAAAYKR2MCA reveals a 3.4-fold preference of WT Kex2 for P 4 Arg versus Ala. In contrast, E255I-Kex2 did not discriminate between the two substrates. The ratio of k cat /K m values for cleavage of AcRYKR2MCA and AcAYKR2MCA was reduced from 10-fold for WT Kex2 to 3-fold for E255I-Kex2. Similarly, the ratio of k cat /K m values for cleavage of Ac␤YKR2MCA and AcAYKR2MCA was reduced from 10-fold for WT Kex2 to 4-fold for E255I-Kex2. P 6 recognition was not significantly affected, however. Comparison of the ratios of k cat /K m values for cleavage of AcRAAYKR2MCA and AcAAAYKR2MCA revealed that WT Kex2 exhibited a 4.3-fold and E255I-Kex2 exhibited a 3.3-fold preference for P 6 Arg versus Ala. DISCUSSION In this work, we performed site-directed mutagenesis of residues initially predicted to be key elements of the S 4 and S 6 subsites of Kex2 and examined their effects on P 4 and P 6 recognition. The specificity of the Kex2 mutants was tested by measuring k cat /K m for cleavage of a series of substrates with systematic substitutions at P 4 and P 6 , and none of the substitutions significantly affected the stability or activity of the mutant enzymes. Although, in general, recognition of the tetrapeptide substrates by the mutant enzymes only differed slightly from that by wild type Kex2, experiments with  Table I. hexapeptide substrates revealed specific recognition of P 6 Arg by T252D substituted Kex2 mutants. To further investigate the nature of this specific effect on P 6 Arg substrates, P 4 recognition in the context of different P 6 residues was examined. The presence of a favorable, basic P 6 significantly affected the enzyme-substrate interaction at P 4 by T252D substituted Kex2 mutants. In a complementary set of experiments, an attempt to limit the recognition of P 4 basic substrates, the wild type Glu at 255 was substituted with Ile. Indeed, E255I diminished recognition of P 4 Arg in substrates having Lys at P 1 .
After these experiments were completed, crystal structures of Kex2 and furin became available (34 -36). The recent crystallographic data consist of the Kex2 subtilisin and P-domains complexed with a tripeptidyl and tetrapeptidyl boronic acid inhibitors, acetyl-Ala-Lys-Arg-Boro and acetyl-Arg-Glu-Lys-Arg-Boro, and furin inhibited by decanoyl-Arg-Val-Lys-Argchloromethylketone. All of the residues mutated in this study, Thr 252 , Glu 255 , and Gln 283 , reside on the surface of a shallow, solvent-exposed groove just beyond the P 3 residue, consistent with these residues contributing to extended substrate selec-tivity. The ␥ carboxylate of Glu 255 interacts directly with the substrate P 4 side chain in both the Kex2 and furin structures, but the orientation of the P 4 Arg residue and the nature of the P 4 -Glu 255 contact are different between the two structures. A FIG. 5. T252D-Kex2 can distinguish between Arg and Ala at P 6 . The bars represent the k cat /K m ratio of AcRAAYKR2MCA/ AcAAAYKR2MCA. The k cat /K m ratios with Arg or Lys at P 4 (AcRARYKR2MCA/AcAARYKR2MCA or AcRAKYKR2MCA/ AcAAKYKR2MCA) both demonstrate a 10-fold preference for P 6 Arg (14). All of the k cat /K m data are listed in Table I. FIG. 6. T252D/Q283E-Kex2 does not exhibit substrate inhibition with AcRAKYKR2MCA. As demonstrated previously (14), furin exhibits substrate inhibition with hexapeptide substrates. Our most furin-like enzyme, T252D/Q283E-Kex2, was tested for substrate inhibition, and none was observed. (r ϭ 0.96.)  a These data were previously published (14). b These data were previously published (10).

TABLE II Comparison of steady state kinetics for wild type and E255I Kex2
The standard deviations were Ϯ Յ 15% of each k cat /K m value.
Substrate a These data were previously published (14). b These data were previously published (10).
direct interaction between P 4 and either Thr 252 or Gln 283 is not observed in the Kex2 structure. Although these structures provide information about the S 4 -P 4 interaction, there is no direct structural data in regards to the S 6 -P 6 interaction. However, some general aspects of P 6 recognition have been postulated (34,35). Because Thr 252 and Gln 283 are located on the protein surface and are solventaccessible, it is possible that either substitution at these positions or binding of an extended substrate could facilitate reorientation of the side chains within the groove to maxi-mize contact with the substrate. On the other hand, although the putative binding site for P 6 is not clear from this structure, Thr 252 is oriented in a groove distal to the active site and could potentially interact with a P 6 residue. Further analysis will require a structure of Kex2 in complex with a hexapeptidyl adduct.
The furin crystal structure revealed structural evidence for the critical requirement for P 1 and P 4 Arg substrate residues (34). In this structure, residues equivalent to Glu 255 and Gln 283 in Kex2 interact specifically with the P 4 Arg (Glu 236 and Asp 264 in furin). The analogous residue to Thr 252 in furin is Asp 233 and appears to orient the Glu 236 toward the P 4 Arg. Because this structure consists of furin complexed with a tetrapeptide chloromethylketone inhibitor, no direct information is provided as to the P 6 -S 6 interaction. Furthermore, the location of a S 6 binding pocket is not obvious, even though biochemical evidence suggests a separate binding site for P 6 Arg residues in the context of a P 4 Arg substrate (14). Henrich et al. (34) predict that Glu 230 and the furin residue that is equivalent to T252D in Kex2, Asp 233 , may interact with the P 6 Arg directly.
This study revealed that wild type Kex2 does exhibit a modest preference for Arg at P 6 , a fact that was previously under appreciated (14). However, T252D-Kex2 and T252D/Q283E-Kex2 both displayed much stronger recognition of P 6 Arg. Moreover, the P 6 specificities of T252D-Kex2 and the double mutant were similar to furin in their preference for Arg versus Ala at P 6 (14). Thus, the T252D substitution was able to discern a favorable P 6 substrate to the same degree as furin. Although T252D was initially predicted to interact primarily with P 4 and not P 6 in the previous models of Kex2, the crystal structure of furin suggested that the analogous residue in furin, Asp 223 may form the S 6 subsite (34). Our biochemical data indicate that T252D imparts discrimination at P 6 and support that this residue may be involved in identification of P 6 residues. In an effort to further compare the mode of recognition for hexapeptide substrates by furin and the Kex2 mutants, saturation kinetics were performed with T252D/Q283E-Kex2 and a hexapeptide substrate. In previous experiments, furin displayed substrate inhibition at high concentrations of substrate. FIG. 7. The substrate residue at P 4 can affect the P 6 specificity of T252D-Kex2 substituted Kex2, and Q283E-Kex2 exhibits P 4 specificity that was relatively independent of the context of P 6 . The filled bars represent the k cat /K m ratio of AcAARYKR2MCA/ AcAAAYKR2MCA, and the open bars are the ratio of AcRARYKR2MCA/AcRAAYKR2MCA. All of the k cat /K m data are listed in Table I. FIG. 8. Comparison of relative processing of Ac-RYKK2MCA versus Ac-AYKK2MCA by wild type and E255I Kex2. WT Kex2 exhibits 100-fold preference for Arg versus Ala at P 4 in the context of a P 1 Lys tetrapeptide substrates. In contrast, E255I-Kex2 preference for Arg versus Ala in the same context is only 6-fold. E255I-Kex2 has a greatly diminished recognition for Arg, which is consistent with the model that Glu 255 is involved in P 4 Arg substrate recognition. All of the k cat /K m data are listed in Table II. FIG. 9. The pattern of recognition of aliphatic P 2 residues by E255I-Kex2 is relatively unchanged as compared with wild type. The catalytic efficiency of processing either Ac␤YKR2MCA (filled bars), AcVYKK2MCA (open bars), or AcYKK2MCA (hatched bars) relative to AcAYKK2MCA is shown. The ratios were derived from the data in Table II. This inhibition was potentially due either to an inhibitory binding mode of the substrate within the active site or to substrate binding at a second site that was outside of the active site and inhibited processing via a putative allosteric mechanism (14). Although no substrate inhibition was observed with T252D/Q283E-Kex2 under saturating conditions, the hexapeptide substrate may not bind to the active site of T252D/Q283E-Kex2 in an inhibitory fashion, or Kex2 may not possess the second, inhibitory binding site. However, this finding did not necessarily indicate that T252D/Q283E-Kex2 recognized P 6 Arg in a different manner than furin because the substrate identification by furin and its inhibition by substrate may be two unrelated, independent events.
Because T252D-substituted Kex2 enzymes specifically recognized P 6 Arg, the effect of P 6 identity on P 4 recognition was investigated. When substrates were extended nonspecifically with Ala at P 5 and P 6 , all of the mutants, but not wild type Kex2, maintained their P 4 specificity for Arg versus Ala with both the tetrapeptide and hexapeptide substrates. This suggested that, unlike wild type, interaction of the mutant Kex2 enzymes with neither the Ala side chain nor the peptide backbone contributed significantly to substrate discrimination. Although in the case of the tetrapeptide substrates, T252D-Kex2 and Q283E-Kex2 displayed slightly less specificity for P 4 Arg than did wild type enzyme, the mutant enzymes exhibited a greater preference for P 4 Arg in the context of hexapeptide substrates than did wild type. The specificity for P 4 by T252D/ Q283E-Kex2 was not only unaltered by extending the substrate nonspecifically, but the degree of specificity was similar to that of wild type Kex2 with the tetrapeptide substrates. These data indicated that the recognition of both P 4 and P 6 by the mutant Kex2 enzymes differed from wild type and that the structures of their S 4 and S 6 subsites were distinct from those of wild type Kex2.
The presence of an Arg residue at P 6 did not influence the specificity at P 4 for Arg versus Ala in Q283E-Kex2. The relatively small increase in k cat /K m for the P 6 Arg substrates as compared with those having Ala at P 6 suggested that Glu 283 may not interact directly with P 6 as was initially predicted by our model. Instead, Glu 283 may function in the S 4 -P 4 interaction, making direct contact with P 4 similarly to furin (34). Alternatively, the interface of the S 4 -S 6 subsite may be rearranged in Q283E-Kex2 such that S 4 is shielded from the consequences of the S 6 interaction with substrate. This explanation is supported by the fact that Q283E-Kex2 did not exhibit a strong preference for either Ala or Arg at P 4 . Moreover, the crystal structure of Kex2 indicated that Gln 283 may play an indirect role in substrate recognition by altering the structure of the pocket and orienting the substrate toward Glu 255 and Glu 249 of the S 4 subsite (35).
In contrast to Q283E-Kex2 and furin, the P 4 specificities of wild type Kex2, T252D-Kex2, and T252D/Q283E-Kex2 were affected by the presence of Arg at P 6 . The energetically favorable contact provided by P 6 Arg seemed to diminish P 4 recognition by the T252D-containing enzymes. In effect, cross-talk between S 4 and S 6 , which was seen at a low level with wild type Kex2, was amplified by Asp at 252. The addition of Q283E further increased the degree of cross-talk between the subsites. This interdependence between the S 4 and S 6 subsites was characterized by the ability of the P 6 -S 6 interaction to influence the S 4 subsite interaction with its corresponding P 4 residue. These data suggest a fundamental difference between Kex2 and furin, which, from biochemical data, clearly displays independent recognition for P 4 and P 6 (14).
The interplay between the S 4 and S 6 subsites in T252D-Kex2 and T252D/Q283E-Kex2 and, to a lesser degree, in wild type Kex2 suggests that a degree of plasticity exists between the S 4 and S 6 subsites. When a P 6 Arg is present, S 4 and S 6 may reorganize to form the largest possible number of electrostatic and hydrophobic contacts to drive catalysis (Fig. 10). Moreover, the T252D substitution in Kex2 enhanced the ability for the S 4 and S 6 subsites to adapt to substrates with various P 4 and P 6 residues, possibly by increasing the number of possible local conformations of the enzyme. Because residue 252 was modeled to be at the interface of S 4 and the putative S 6 , it may contact directly or enhance other interactions with substrate side chains in either the S 4 or S 6 pocket. Also, the loop formed by residues 249 -252 in Kex2 is in an alternate conformation than is observed for furin, and this repositioning of the loop has been hypothesized to contribute to the differences in P 4 substrate recognition between Kex2 and furin (34,35).
Multiple modes of substrate recognition may result from these malleable, extended subsites. Similar flexible binding sites were also observed when mutations were made in the ␣-lytic protease S 1 subsite. Crystallographic data of M192A and M213A ␣-lytic protease mutants, with and without substrate analogues bound, demonstrated that there were local changes in conformation to accommodate the various P 1 residues; consequently, the mutant appeared to have relaxed specificity (43). Furthermore, crystal structures of Savinase with a Gly substitution residue, Ile 107 , which is analogous to Glu 255 in Kex2, demonstrated that the Savinase S 4 subsite was structurally flexible. Other residues in the binding pocket, as well as the protein backbone, rearranged to adapt to the mutation of I107G (24). These examples demonstrate that the crystal structures of the native enzyme can only serve as a tool when analyzing mutagenesis data because substitutions may alter the structure of an enzyme.
The other Kex2 mutant with an altered P 4 specificity profile was E225I. Previously, a dual mode of binding aliphatic and basic P 4 residues was observed with wild type Kex2 (10), and the E255I substitution in Kex2 affected each binding mode differentially. Recognition of substrates with Arg at P 4 was greatly decreased by E255I, which suggests that this residue is involved in recognition of P 4 basic residues. Indeed, Glu 255 in the crystal structure of Kex2 also exhibits a direct contact with the P 4 Arg substrate (35). However, instead of a dramatic increase in specificity for substrates with aliphatic amino acids at P 4 , E255I displayed a relaxed preference for Ala, Nle, Val, and cyclohexylalanine at P 4 . These results support the hypothesis that recognition of P 4 basic and aliphatic residues may involve independent modes of binding. The ability of E255I- FIG. 10. A schematic model to illustrate the cross-talk between P 4 and P 6 . When the enzyme does not have a substrate bound, then the conformation and intramolecular electrostatic interactions are in a "resting" state. However, upon the addition of a substrate with a P 4 and P 6 residue, then the enzyme may rearrange its electrostatic and hydrophobic contacts to best interact with the favorable P 4 and/or P 6 . The local conformations of the S 4 and S 6 subsites will not necessarily be the same if a favorable P 4 residue and/or P 6 residue is present.
Kex2 to accommodate different P 4 residues also indicates that the P 4 pocket may be flexible.
Although members of the Kex2/furin family of proteases are highly homologous and many of the mammalian members have overlapping tissue expression profiles, genetic studies with mice have indicated that the function of the mammalian proteases are quite distinct from one another (44 -46). Although all of these proteases are believed to cleave preferentially Cterminal to a P 1 Arg, not much is known about their extended specificity except in the case of Kex2 and furin. However, preliminary data indicate that the substrate recognition at extended residues diverges among the mammalian members of the Kex2/furin family of proteases. For example, both PC7 and furin are expressed in a wide variety of tissue types and require a P 4 Arg to process peptide substrates. However, unlike furin, PC7 cleaved pyr-ERTKR2MCA more efficiently than the other tetrapeptide substrates tested, suggesting that PC7 may have recognition for residues beyond P 4 (47). Modulation of extended substrate recognition might enable the evolution of an expanded repertoire of protease specificities without impacting either the catalytic mechanism or the proximal elements of substrate recognition. Indeed the present work demonstrates that the specificities of extended subsites are quite malleable and that amino acid substitutions can be made in the substrate binding site distal to the active site residues without sacrificing the overall activity of Kex2 protease.