Previous Article |
Table of Contents
| Next Article 
BIOLOGICAL SCIENCES / BIOCHEMISTRY
Differential P1 arginine and lysine recognition in the prototypical proprotein convertase Kex2
Department of Biochemistry and Molecular Biology, University of Kansas Medical Center, Kansas City, KS 66160
Communicated by Gregory A. Petsko, Brandeis University, Waltham, MA, March 7, 2007 (received for review October 4, 2006)
 |
Abstract |
|---|
 Top Abstract ResultsDiscussionMaterials and MethodsAcknowledgementsReferences |
|---|
chloromethylketone | crystallography | furin
Although the members of the family have been shown or predicted to have slightly different substrate selectivity, the entire family shares a common absolute requirement for an arginine at the P1 position
(5, 12–15). Conservative substitutions of lysine for arginine at P1 have severe catalytic consequences in those enzymes that have been kinetically characterized. In Kex2, depending on the composition of the peptide substrate this substitution results in a decrease in kcat/Km of at least 70-fold (14, 15). Further, presteady-state kinetic studies with Kex2 have revealed that the P1 lysine/arginine selectivity is a kinetic phenomenon that arises during the acylation process (16). Replacement of lysine for arginine at P1 results in a loss of the burst kinetics that corresponds to the rapid acylation that defines Kex2/furin kinetic behavior. This rapid acylation is in contrast to the slow acylation and rapid deacylation that defines the kinetic behavior of the parental subtilases (17, 18). These same presteady-state studies on Kex2 demonstrated that the observed kinetic response upon substitution of lysine for arginine reduces acylation by at least 200-fold. There is an even greater effect with alanine at P4 (>4,600-fold). These same substitutions had little effect on substrate association or the deacylation step (16).
The initial structures of Kex2 and furin provided a great deal of insight into the structural basis for PC selectivity, including the use of a calcium-binding site unique to the PC family that is involved in P1 arginine recognition (19–23). These structures have also formed the basis for modeling studies of the other structurally uncharacterized PC isozymes that have generated new hypotheses on substrate recognition in those enzymes (24). The initial structures unfortunately did not provide any basis for the unique differential recognition of arginine and lysine that defines this protease family and results in the altered kinetic behavior. Therefore, the mechanism of differential selectivity between lysine and arginine is of great interest and is still in question. Inspection of the structures reveals that the S1-binding pocket in Kex2 and furin contains a short loop insertion (residues 256–261 in furin and 275–280 in Kex2) that is absent in the degradative subtilases. This loop insertion acts as a lid capping the S1 pocket and contributes to the coordination geometry of the S1 calcium site. Based on this structural information and the available kinetic data, it was hypothesized that this S1 lid may close down only on a bound P1 arginine and provide a large amount of energy to promote the acylation step that would not be available in the presence of a P1 lysine (23). Therefore, the dynamic nature of this lid element was proposed to be responsible for the observed P1 arginine/lysine selectivity (23).
To clarify the structural basis for the method of kinetic selectivity between lysine and arginine at P1 that defines the PC family, we have crystallized Kex2 in complex with a chloromethylketone inhibitor containing a P1 lysine.
|
Results |
|---|
|
TopAbstractResultsDiscussionMaterials and MethodsAcknowledgementsReferences |
|---|
|
S1 Subsite–Lysine Interactions.
On initial inspection, it appears that the binding of the P1 lysine is identical to the binding of the P1 arginine in the previous structures of Kex2 or furin (20–22). However, on closer inspection, it is revealed that subtle but significant differences exist in the recognition of a P1 lysine as compared with a P1 arginine. Like arginine, the lysine side chain is well ordered in the S1 pocket, with N
of the lysine side chain having a similar B factor (molecule A = 28.95, molecule B = 34.15) to that of the average B factor for the structure (30.30; SI Table 2). The observation that the B factors are slightly higher for the bound inhibitor in molecule B of the crystallographic dimer has been observed for the other structures of Kex2 and is most likely a result of crystal packing. In contrast to S1 arginine binding, which interacts through a salt bridge with D325 and hydrogen bonds between the arginine side chain's NH2 and N
groups with D277, lysine binds in the S1 pocket in a more shallow fashion, interacting primarily with the D277 carboxylate (2.75 Å) and the backbone carbonyl of P275 (2.81 Å; Fig. 2). The primary interaction that the P1 arginine has with D325 is lost on occupancy of S1 by lysine. With lysine occupying the S1 subsite, a water molecule occupies the position occupied by the NH1 amide of the bound arginine. Only an indirect hydrogen bond through this water molecule is mediated by D325 (Fig. 2). This secondary binding site for lysine at S1 results in the displacement of the lysine side-chain terminus by 
1 Å from the position of the bound arginine side chain (SI Fig. 4). Because the position of the C
is constrained to the same location, regardless of the nature of the P1 residue, by covalent linkage of the hemiketal form of the inhibitor to S385 and H213, and S385 and H213 are located in the identical position as found in the P1 arginine inhibitor complex [Protein Data Bank (PDB) ID code 1R64], this displacement decreases in progression from the side-chain terminus (N) toward the backbone (SI Fig. 4).
|
of the P2 residue (2.78 Å; Fig. 3). The interaction with D211 is also retained; however, rather than an indirect interaction although a bound water molecule, a direct interaction between O
1 of D211 and the NH2 amino group of the bound P2 arginine is observed (2.87 Å; Fig. 3). The terminal amino groups of the bound arginine displace two water molecules previously occupying the pocket and, in addition to the previously mentioned interaction with D211, the NH1 and NH2 amino groups interact with the O1 and O2 of D210 (3.49 Å) in a salt bridge, helping to orient the side chain in the pocket (Fig. 3). Therefore, the binding of arginine at P2 appears to satisfy more contacts in the subsite than occupancy by lysine consistent with previous kinetic data (15, 27).
|
|
|
Discussion |
|---|
|
TopAbstractResultsDiscussionMaterials and MethodsAcknowledgementsReferences |
|---|
P1 Recognition.
In the previously reported structures of Kex2 and furin inactivated with peptidyl inhibitors containing P1 arginine residues, the data suggested that the interaction through D325 was responsible for the recognition of a terminal positive charge, which would be present in both arginine and lysine residues (20–24). The additional hydrogen bonds observed between the N and NH2 groups of the bound arginine and the carboxylate of D277 were thought to be important in the discrimination between lysine and arginine, because these hydrogen-bonding interactions would be absent in a lysine-containing substrate. However, based on these interactions, it was not obvious why such a large deficit in catalysis, specifically the observed effect on acylation rates, was manifest in the kinetics of cleavage. Based on the aforementioned structural data, it was hypothesized that perhaps the S1 lid was a mobile element and was responsible for the rapid acylation of P1 arginine substrates (23). The basis for this proposal was the observation that, in contrast to subtilisin, Kex2 and furin contained a conserved loop insertion that acts as a lid domain capping the S1 pocket (20–23). Based on the recognition of arginine through D277, and because this lid domain contains D277, it was reasonable to predict that, in the absence of an S1 arginine, the interactions with D277 would be lost, and this loop would occupy an open, altered, or disordered conformation. Consistent with the kinetic data, this mechanism would predict that occupancy of S1 by arginine would result in lid closure through interactions with D277. Because no interactions between lysine and D277 were predicted to occur, this model would provide energy available for acylation of P1 arginine-containing substrates that would not be available for P1 lysine containing substrates. In contrast to this model, the structural data presented here with the enzyme in complex with a CMK inhibitor containing a P1 lysine residue provide evidence against the S1 lid being a mobile element, because the orientation of the lid and the residues that compose it are unchanged on occupancy of S1 with a lysine residue. Therefore, we believe the observed change in acylation rates between substrates with a P1 lysine or P1 arginine residue is the result of a more simplistic mechanism that is based on the observed altered binding of the P1 lysine reside. This conclusion is based on the observation that, in contrast to what was suggested by the initial structures of Kex2 and furin in complex with inhibitors containing a P1 arginine, D277 does not mediate the selectivity between lysine and arginine at S1. Rather, the interaction between the P1 lysine -amino group and D277 is the only previously observed S1/P1 arginine interaction that is retained. More specifically, because the only significant interactions that occur between the P1 lysine and the S1 subsite occur with the S1 lid D277 carboxylate and the backbone carbonyl of P275, interactions between the S1 lid and the P1 lysine are not only retained but in fact increase on the lysine for arginine substitution. Together with the fact that D277 contributes to the coordination geometry of the S1 calcium ion, these observations suggest a nonmobile nature for the S1 lid in the mechanism of P1 arginine/lysine selectivity.
Interestingly, the terminal positive charge that is common to both lysine and arginine does not result in the conservation of the interaction with D325 that was predicted to occur from the original structures. An indirect interaction between D325 and the P1 side chain remains, mediated through a water molecule that now occupies the position of one of the NH2 groups of the bound arginine in the previous structures (PDB ID codes 1R64 and 1OT5) (Fig. 2). Rather than discriminate against the binding of lysine at P1, it appears that Kex2 contains another site within the S1 pocket that selects for lysine, taking advantage of lysine's increased rotational freedom and lack of a biantennary functional group. Binding of lysine into this site results in the displacement of the side chain by 1 Å from the position occupied by an arginine side chain at S1 (SI Fig. 4). Based on these observations, we hypothesize that the selectivity of arginine vs. lysine that occurs at the step of acylation occurs through the displacement of the scissile bond relative to the serine nucleophile. Although the structure of the acylated enzyme form cannot conclusively demonstrate whether this occurs, because the hemiketal formed constrains the backbone of the inhibitor to the same position as in the P1 arginine-inhibited structures, the displacement of the side chain away from the serine is suggestive of this. We propose that the result of releasing the geometric constraints imposed by the covalent linkage of the inhibitor to S385 and H213 in the actual Michaelis complex would result in the positioning of the scissile carbonyl further from S385 in substrates containing a P1 lysine peptide as opposed to the P1 arginine peptide. In addition to the increased nucleophile attack distance that is suggested by our structural work, it has been noted that the serine attack angle is quite important for efficient acylation (32, 33). Therefore, in addition to affecting the distance between the serine nucleophile and the scissile bond, the misorientation of P1 lysine-containing substrates caused by the differential S1 contacts could distort the attack angle, further reducing acylation rates. Finally, the proposed mechanism predicts that, after acylation occurs, the deacylation process would be unaffected. This is confirmed by the available structures that show the position of the substrate carbonyl is in an identical position regardless of S1 occupancy. Consistent with the presteady-state studies, deacylation would be insensitive to the nature of the P1 residue, because the geometry between the attacking water nucleophile and the acyl intermediate is dictated by the linkage between the substrate and S385 that is identical in the determined structures, regardless of the nature of the P1 residue.
S2-Arginine Recognition.
Kex2 has been demonstrated to exhibit a modest but measurable increase in kcat/Km on substitution of lysine for arginine at P2 (15, 27). The current structure with arginine occupying the S2 subsite provides a structural rationale for this modest difference, because arginine fulfills more enzyme/substrate contacts than is observed with a P2 lysine. Although kinetic studies have shown the importance of the electrostatic interaction at P2, the additional hydrogen-bonding interactions offered to arginine as opposed to lysine could provide a rationale for the modest increase in catalytic efficiency (kcat/Km) observed for substrates containing a P1 arginine compared with P1 lysine (15, 27) through a reduction in the Km for arginine-containing substrates. Alternatively, the extended interactions with a P2 arginine could more precisely locate the substrate scissile carbonyl, leading to more efficient hydrolysis. The decreased importance on basic residues at P2 shown to occur in furin and suggested to occur in the other PCs is also reinforced by the current structure. In furin and presumably the other PCs, the hydrogen bond that occurs between N of the bound arginine and D211 would not occur, because this residue is substituted with a conserved asparagine residue.
Compensation for Poor S1-Lysine Contacts.
The proposed mechanism of selectivity by misorientation provides the opportunity to address the ability of good contacts at P2-P6 in furin compensating for poor (lysine) residues at S1 (13, 34). It has been noted that proteases universally recognize extended 
strands of protein substrates in all protease families (35). One obvious feature lacking in PCs that is present in the parental subtilases is the absence of the bound substrate acting as the center strand of a three-stranded sheet (20). This suggests that perhaps Kex2/PCs place less importance on backbone recognition than the degradative enzymes of the subtilase family. The offset binding of lysine and the proposed misorientation mechanism of selectivity suggest that the lessened importance placed on backbone recognition may have functional implications. Poor positioning by offset binding arising from poor P1/S1 contacts results in a decrease in acylation rates (16). As shown in furin, the defects in the kinetics of hydrolysis are partially overcome by favorable S2-S6 interactions (13). Therefore, based on our working model, the positioning of the bound side chains at sites distant from the scissile bond must be directly transmitted to the positioning of the scissile carbonyl and not only contribute to the stability of the Michaelis complex. This transmission may be facilitated by the reduced constraints on the positioning of the bound substrate by not being included as a central element of protein secondary structure. Thus, the register of the peptide at the active site is not strongly anchored by backbone recognition, and fine tuning is possible by distant subsite interactions that result from the substrate to the geometry imposed by the well ordered predetermined subsites that the structural work suggests exist in Kex2 and the PCs. In this manner, the favorable interactions at sites distant from the scissile bond can be transmitted to the positioning of the scissile bond and improve acylation rates. For example, favorable P2 and P4 interactions would act to correctly position the scissile bond relative to the serine nucleophile and thus offer an explanation toward the partial rescue of activity by favorable interactions at these subsites in furin. The ability of furin to be rescued in a more significant fashion than Kex2 would result from the much more involved set of interactions between the P4 arginine and the S4 site in furin (21, 24). Consistent with this notion that preordered subsites constrain the substrate orientation and effect catalysis is the observation that Kex2 has been shown structurally and kinetically to have a more plastic S4 subsite, allowing for different binding orientations and the recognition of basic and aliphatic residues at this position (14, 21). In contrast, furin places great importance on arginine at the P4 position, similar in magnitude to the importance placed on a P1 arginine in the entire Kex2/PC family (5, 13). Based on our model, this plasticity in Kex2 would result in less coupling between the favorable occupancy of S4 and the positioning of the scissile bond and ultimately the kinetics of hydrolysis. In contrast, favorable S4–arginine interactions should compensate for poor interactions at S1 in furin. Again, this agrees with what is observed kinetically, and it is also reflected in the minimal consensus sites for furin and Kex2 being R-X-X-R and R/K-R, respectively (5, 36, 37). Similar interactions could presumably occur in furin at the S6 and S8 subsites that are predicted to exist in the mammalian enzyme forms (24); however, their location has not been experimentally determined. Kex2 is not predicted to contain well defined subsites beyond S4 (34), because only a modest two-fold preference for arginine at P6 is observed (13), and recent structural work has demonstrated that, at least in the context of one peptide inhibitor, contacts beyond S5 do not exist (T.H., unpublished results). It is therefore predicted that these extended interactions would not be significant in Kex2. Clearly, further work, including structures of the free enzyme and enzyme substrate/product complexes, is necessary to test this proposed mechanism.
In conclusion, we have demonstrated that in contrast to what was suggested by the previous structures of Kex2 and furin, selectivity between P1 arginine and lysine is not mediated by a mobile S1 lid. In contrast, the S1 lid residues D277 and P275 provide the only contacts to a P1 lysine. The work presented here is more consistent with a simple mechanism by which the hydrolysis of substrates containing P1 lysine residues is hampered at the level of acylation through a misorientation of the scissile bond relative to the active site serine because of altered binding of the side chain in the S1 pocket.
|
Materials and Methods |
|---|
|
TopAbstractResultsDiscussionMaterials and MethodsAcknowledgementsReferences |
|---|
Methods. Kex2 expression and purification. Secreted soluble Kex2 (ssKex2) was prepared as described (22). After purification, Kex2 was inactivated by using a 10-fold molar excess of the Ac-R-E-R-K-CMK inhibitor dissolved in DMSO. The inactivation mix was allowed to incubate overnight at 25°C and was subsequently assayed as described to ensure no enzyme activity remained (22). The protein solution was concentrated to a final volume of 1.5–2 ml and loaded onto an S-200 gel filtration column (GE Biosciences, Piscataway, NJ), equilibrated in 40 mM Bis-Tris, pH 7.2/10 mM NaCl/2 mM CaCl2 buffer. Fractions containing protein were run on SDS/PAGE, and only those fractions containing a band corresponding to full-length ssKex2 were retained. The fractions were pooled, concentrated to 25 mg/ml, and stored at 4°C.
Crystallization. ssKex2 protease (25 mg/ml/40 mM Bis-Tris, pH 7.2/10 mM NaCl/2 mM CaCl2) was crystallized and cryoprotected in sodium malonate as described (21, 22).
Data collection. Data were collected on cryocooled crystals maintained at 100 K throughout data collection at the Advanced Photon Source Biocars beam line by using an Area Detector Systems Corporation 14-BM-C Quantum 315 CCD detector. All data were integrated and scaled with HKL-2000 (38). See SI Table 2 for data statistics.
Structure determination and refinement.
The crystals of the inhibited enzyme were isomorphous with the crystals used for the determination of the 2.2-Å resolution crystal structure of the peptidyl boronic acid inhibited enzyme [(21) PDB ID code 1R64]; therefore, this model was used as a starting point for the current structure. All calcium ions, water, sugar, and inhibitor molecules were removed from the model, and this initial model was subjected to a round of rigid body refinement in Refmac5 in the CCP4 suite of programs (39–41), followed by manual model adjustment using COOT (42); metal and water addition and validation were also performed in COOT. Tight noncrystallo-graphic symmetry (NCS) constraints were applied during the initial rounds of refinement. In the final round of refinement, NCS restraints were removed. A final round of translation/libation/screw (TLS) refinement was performed in Refmac5 (43). A total of 20 TLS groups were used per chain, and the groups were determined by submission of the PDB file to the TLSMD server [http://skuld.bmsc.washington.edu/
tlsmd/index.html (44)]. The final model refined to an R factor of 17.9% (Rfree = 20.7% for a test set of 5% of randomly chosen reflections). See SI Table 2 for final model statistics.
ssKex2 protease chloromethylketone inactivation kinetics. The inactivation kinetics of Kex2 by two different CMK peptides was determined by monitoring the rate of enzyme inactivation as a function of inhibitor concentration. To determine the rate constants for inactivation a solution of 43 nM Kex2, 200 mM Bis-Tris, pH 7.2/0.01% Triton X-100/0.5% DMSO/0.1 mM CaCl2 was incubated at 11°C in the presence of varying concentrations of the CMK inhibitor. The stock solutions of the CMK inhibitors were prepared at 10 mM in DMSO and added to the incubation mix to achieve the final concentrations of the decanoyl-R-V-R-K-CMK (0.5, 1, 2.5, 5, 10, 18, 25, 35, 50, and 75 µM) and decanoyl-R-V-K-R-CMK (0.1, 0.25, 0.5, 1, 1.5, and 2 µM) inhibitors. At various time points, a 3-µl aliquot of the incubation mix was removed and added to 550 µl of assay mix (200 mM Bis-Tris, pH 7.2/0.01% Triton X-100/0.5% DMSO/0.1 mM CaCl2/100 µM Boc-Leucine-Arginine-Arginine-AMC). Hydrolysis of the AMC substrate was monitored by following the increase in fluorescence at 465 nm after excitation at 385 nm by using a Photon Technology International (PTI, Birmingham, NJ) fluorometer. Control reactions were run at each inhibitor concentration with the corresponding volume DMSO and found not to result in the loss of enzyme activity. Inactivation rates at each inhibitor concentration were determined in triplicate. The inactivation rate (kobs) at each inhibitor concentration was determined by a plot of the residual activity against time. KI and ki values were determined from the fit of the data to Eq. 1, assuming the inactivation process follows the minimal reaction scheme shown in SI Scheme 1.
 |
|
Acknowledgements |
|---|
|
TopAbstractResultsDiscussionMaterials and MethodsAcknowledgementsReferences |
|---|
|
Footnotes |
|---|
Abbreviations: Kex2, kexin; CMK, chloromethylketone; PC, prohormone (proprotein) convertase; ssKex2, secreted soluble Kex2; PDB, Protein Data Bank.
*To whom correspondence should be addressed. E-mail: tholyoak{at}kumc.edu
Author contributions: T.H. designed research; J.L.W. and T.H. performed research; J.L.W. and T.H. analyzed data; and T.H. wrote the paper.
The authors declare no conflict of interest.
Data deposition: The atomic coordinates reported in this paper have been deposited in the Protein Data Bank, www.pdb.org (PDB ID code 2ID4).
The designation of the substrate residues follows the naming convention of Schechter and Berger (45). Briefly, starting at the scissile bond and counting toward the N terminus, the substrate residues are designated P1, P2, P3.... Conversely, the substrate residues are bound at the corresponding S1, S2, or S3 subsite on the enzyme. 
This article contains supporting information online at www.pnas.org/cgi/content/full/0701983104/DC1.
© 2007 by The National Academy of Sciences of the USA
|
References |
|---|
|
TopAbstractResultsDiscussionMaterials and MethodsAcknowledgementsReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg What's this?
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
for the Ac-R-E-R-K-CMK inhibitor before inclusion of the inhibitor in the model is shown as a white mesh. All figures were prepared by using CCP4MG (
