View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Elsevier - Publisher Connector

Structure, Vol. 12, 1313–1323, July, 2004, 2004 Elsevier Ltd. All rights reserved. DOI 10.1016/j.str.2004.04.013 1.2 A˚ Crystal Structure of the Carboxyl Proteinase Pro-Kumamolisin: Structure of an Intact Pro-

Mireia Comellas-Bigler,1 Klaus Maskos,1 the suffix “e”). It is presumed that the signal peptide Robert Huber,1 Hiroshi Oyama,2 Kohei Oda,2 is cleaved off upon secretion, and that the activation and Wolfram Bode1,* cleavage of the prodomain occurs immediately after re- 1Department of Structure Research lease into the (normally acidic) environment. Max-Planck-Institute for Biochemistry Kumamolisin (previously named kumamolysin or Am Klopferspitz 18a KSCP) is a member of a rapidly growing family of endo- D-82152 Martinsried peptidases, comprising not only a number of bacterial Germany (Oda et al., 1994, 1996; Shibata et al., 1998) but also 2 Department of Applied Biology eukaryotic such as the human lysosomal tri- Faculty of Textile Science peptidyl-peptidase I/CLN2 (Ezaki et al., 2000; Lin et al., Kyoto Institute of Technology 2001; Rawlings and Barrett, 1999a). Members of this Sakyo-ku, Kyoto 606-8585 proteinase family had originally been designated as pep- Japan statin-insensitive carboxyl proteinases (Murao et al., 1993; Oda et al., 1994, 1996). The crystal structure of the homologous Pseudomonas sp. serine-carboxyl pro- Summary teinase (Wlodawer et al., 2001), also called PSCP/sedo- lisin, showed that these peptidases exhibit a - Kumamolisin, an extracellular proteinase derived from like fold, equipped with a Ser-Glu-Asp , an acido/thermophilic Bacillus, belongs to the sedo- however. The structural evidence led to the reclassifi- lisin family of characterized by a sub- cation as serine-carboxyl proteinases (SCP), which tilisin-like fold and a Ser-Glu-Asp catalytic triad. In comprise the S53 “” family of kumamolisin, the Asp82 carboxylate hydrogen bonds (Wlodawer et al., 2003) clan SB (MEROPS database) to Glu32-Trp129, which might act as a proton sink (Rawlings and Barrett, 1999b). stabilizing the catalytic residues. The 1.2/1.3 A˚ crystal Previously, our X-ray structure analysis of kumamoli- structures of the Glu32→Ala and Trp129→Ala mutants sin (Comellas-Bigler et al., 2002) not only confirmed the show that both mutations affect the active-site confor- characteristic sedolisin features such as a Ser278e- mation, causing a 95% activity decrease. In addition, Glu78e-Asp82e catalytic triad with particularly short in- the 1.2 A˚ crystal structure of the Ser278→Ala mutant terconnecting hydrogen bonds, but showed an addi- of pro-kumamolisin was determined. The prodomain tional Glu32e-Trp129e pair tightly hydrogen bond linked exhibits a half-␤ sandwich core docking to the cata- to the catalytic Asp82e carboxylic acid. We suggested lytic domain similarly as the equivalent subtilisin pro- that these residues might facilitate the proton delocal- domains in their catalytic-domain complexes. This ization during nucleophilic attack, in particular at high pro-kumamolisin structure displays, for the first time, temperature. In order to check this hypothesis, we pre- the uncleaved linker segment running across the ac- pared kumamolisin variants with either Glu32e or Trp129e tive site and connecting the prodomain with the prop- mutated to an Ala residue, characterized their proteo- erly folded catalytic domain. The structure strongly lytic activities, and determined their three-dimensional points to an initial intramolecular activation cleavage structures at atomic or near-atomic resolution. The in , as presumed for pro-subtilisin and pro- structures show that the decrease in proteolytic activity . does not result from the replacement of these two side chains alone, but is also a consequence of local struc- tural changes affecting the other catalytic residues. Introduction The prodomain of recombinant pro-kumamolisin seems to be essential for the expression of active kumamolisin Kumamolisin is a thermostable calcium-dependent en- (Oyama et al., 2002) as observed in its homolog sedolisin dopeptidase isolated from Bacillus novosp. MN-32, a (Oda et al., 1994). Thus, the resemble other bacterium found in the acidic hot spring at the Mt. Aso, subtilases (Siezen and Leunissen, 1997) not only with Kumamoto prefecture, Japan (Murao et al., 1988, 1993). respect to the overall fold of their catalytic domains but Kumamolisin is an extracellular proteinase, which is well also in terms of the possible chaperonin function of their adapted to its natural acidic and hot habitat showing prodomains. For the BPNЈ (Gallagher et al., Њ optimal proteolytic activity at 70 C and around pH 3 1995) and E (Jain et al., 1998) but in particular for pro- (Oyama et al., 2002). The is synthesized as a furin (Anderson et al., 2002; Henrich et al., 2003) and 552 amino acid residue precursor, comprising a 188 other proprotein/prohormone convertases (PCs) it has N-prepro-region (with its amino acid residues in the fol- been shown that their prodomains act as chaperonins, lowing designated by sequential numbers with a suffix which seem to stabilize folding intermediates on the “p” added) and a 364 amino acid mature enzyme (resi- path to the fully folded/active catalytic domains (Eder due numbering of the mature kumamolisin as used pre- and Fersht, 1995; Li et al., 1995; Tangrea et al., 2002). viously [Comellas-Bigler et al., 2002] designated with To address the fold and the arrangement of the sedo- lisin prodomains, as well as the assembly of their linker *Correspondence: [email protected] peptides with respect to the active center, we prepared Structure 1314

tical with those of the wild-type enzyme, with 342 ␣-carbon atoms showing rms deviations of 0.18 and 0.26 A˚ using a threshold of 2 A˚ . Like mature kumamolisin, E32A and W129A catalytic domains essentially consist of a central eight-stranded parallel ␤ sheet, flanked by eight helices and a few short strand pairs arranged on both sides (as shown for the catalytic domain of pro- kumamolisin in Figure 2B [Comellas-Bigler et al., 2002]). Helices h3 and h6, carrying the residues Glu78e/Asp82e and Ser278e (see Figure 2A), respec- tively, are mostly buried inside of the hydrophobic core, while the other helices, in particular helices h4 and h5 contacted by the prodomain (see Figures 2B and 2C), Figure 1. SDS-PAGE of the Purified Kumamolisin Mutants are mostly exposed at the molecular surface. The single bound calcium ion is sandwiched between the multiple The purified mutant samples and the wild-type (W.T.) were analyzed by 12.5% SDS-polyacrylamide gel electrophoresis and visualized turn segments of a long open hairpin loop connecting by Coomassie staining. The molecular weight marker (M.W.) indi- strands s10 and s11. cates the position of proteins with apparent molecular weights from The globular catalytic domain exhibits an indent (to- 66.2 to 21.5 kDa. ward the front in Figures 2B and 2C), which accom- modates the substrate binding region and the active site residues. In the wild-type enzyme, a network of and crystallized a Ser278Ala pro-kumamolisin mutant unusually short hydrogen bonds connects the catalytic and determined its X-ray crystal structure at high resolu- Ser278e O␥ atom via the carboxylic acid groups of tion. The structure reveals that the kumamolisin propep- Glu78e, Asp82e, and Glu32e with the indole moiety of tide is folded in a compact structure resembling the Trp129e, making it more nucleophilic. In the E32A mu- prodomains of the pro-subtilisins and of the PCs, not tant, the main chain is virtually unchanged compared only with respect to the overall fold but also to the with native kumamolisin, even in the catalytic center interaction with the catalytic domain. The linker peptide (Figure 3A). The absence of the Glu32e side chain gives between the prodomain and the catalytic domain runs rise to an intramolecular cavity occupied by a solvent through the active site cleft in a substrate-like manner, molecule (Wat17), which hydrogen bonds to both the with the His171p-Phe172p peptide bond docked close Asp82e carboxylic acid (O␦1) and the Trp129e indole to the normally reactive but here mutated Ser278e→Ala nitrogen (N⑀1). The latter indole moiety adopts two residue. This structure is the first atomic model of an slightly different conformations, allowing only one of uncleaved full-length pro-subtilase, which in addition them to form a favorable hydrogen bond to the fixed displays the exact interaction geometry of a peptide water molecule. The side chain of the catalytic Ser278e substrate with the active site cleft. The structure directly oscillates between two different positions, a first one shows that the first activation step is not only autocata- where the O␥ atom still can hydrogen bond to the un- lytic but occurs intramolecularly, and represents thus a changed Glu78e carboxylic acid O⑀1 atom, and a second prototype for the whole subtilase superfamily, including one directed toward the interior of the catalytic domain the proPCs. i.e., away from Glu78e O⑀1. Therefore, the reduction in activity indeed seems to be mainly caused by the lack of the Glu32e carboxylate group preventing formation Results of an optimal proton shuttle. The bound water molecule that replaces this carboxylate group seems to partially Structures of the E32A and the W129A mimic it, possibly explaining the remaining catalytic ac- Mutants of Kumamolisin tivity of this E78A mutant (Table 1). As shown in Figure 1, the kumamolisin mutants E32A The active site region of the W129A mutant, which and W129A, with single residue replacements of Glu32e retains 3.8% of the original hydrolytic activity (Table and Trp129e (see Figure 2A) by Ala residues, represent 1), shows more drastic changes compared to native single-chain molecules with the same apparent molecu- kumamolisin (Figure 3B). The main chain segment lar weight (Figure 1) and similar CD spectra as the mature Ile127e-Gly131e, lacking the bulky Trp side chain, devi- enzyme (data not shown). However, their catalytic activ- ates directing the Ala129e side chain away from the ity/specificity constant toward the peptide-like sub- catalytic triad, while the hydrogen bond pair between strate Lys-Pro-Ile-Ala-Phe*Nph-Arg-Leu is reduced to both Asp179e carboxylate oxygens and the Gly130e and 5.7% and 3.8% of the native enzyme, respectively (Table Gly131e amide nitrogens is essentially maintained. 1) (H.O. and K.O., unpublished data). Both mutants crys- These structural changes and the loss of a hydrogen tallize isomorphously with the previously reported bond donor (Trp129e) do not affect the positioning of monomeric form A of wild-type kumamolisin (Comellas- the side chains of Glu32e and Asp82e. However, these Bigler et al., 2002) (Table 2). Consistent with the low structural rearrangements cause the Glu78e side chain (35% v/v) solvent content, the proteinase molecules are to adopt multiple conformations (see the omit density in likewise tightly packed. The polypeptide chains defined Figure 3B), resulting in the Ser278e side chain switching from the N-terminal Ala1e up to Pro357e are, except between the native and an alternate conformation. Thus, residues 127e–132e in the case of W129A, virtually iden- the lack of Trp129e and the consequent structural Structures of E32A, W129A, and Pro-Kumamolisin 1315

Figure 2. Sequence and 3D Structure of Full-Length Pro-Kumamolisin (A) Amino acid sequence of pro-kumamolisin (Oyama et al., 2002) as stored under the Swiss-Prot accession number Q8RR56. Sequence numbers with the suffix “p” refer to the kumamolisin prodomain, while numbers with the suffix “e” correspond to the kumamolisin catalytic domain. ␤ sheets and ␣ helices observed in pro-kumamolisin are represented by arrows and cylinders, respectively (green for the prodomain and blue for the catalytic domain). The catalytic residues are indicated with red arrows; Ca2ϩ ligands are marked in black. The primary and the final procession sites are indicated by green arrowheads labeled with 1 and 2. The figure was prepared with the program ALSCRIPT (Barton, 1993). (B) Stereo ribbon plot of full-length pro-kumamolisin shown in standard orientation. The peptide linker between the catalytic domain (in blue) and the prodomain (in green) is colored in red for well-defined residues and in black for undefined residues. The catalytic residues and the Ca2ϩ ion are shown as orange stick models and a red sphere, respectively, and helices and ␤ strands are labeled with Greek (propeptide) and Latin letters (catalytic domain). The figure was prepared with Molscript (Kraulis, 1991) and RASTER3D (Merritt and Bacon, 1997). (C) Stereo solid surface representation of pro-kumamolisin shown in standard orientation as in Figure 2B. The colors indicate the electrostatic surface potential at pH 7 contoured from ϩ20 kT/e (dark blue) to Ϫ20 kT/e (dark red). The peptide linker is represented as a yellow stick model visible from P3-Arg169p to P3Ј-Leu174p. The image was generated with GRASP (Nicholls et al., 1993) and RASTER3D (Meritt and Bacon, 1997). Structure 1316

Table 1. Kinetics Parameters of the Kumamolisin Mutants

Ϫ1 Ϫ1 Ϫ1 kcat (s )Km (␮M) kcat/Km (␮M s ) Relative kcat/Km (%) Wild-type 6.02 29 0.21000 100 S278A 0.08 84 0.00099 0.5 E32A 1.24 102 0.01200 5.7 W129A 0.67 89 0.00800 3.8

Substrate: Lys-Pro-Ile-Ala-Phe*Nph-Arg-Leu (*, cleavage site). Assay conditions: 60ЊC, 8.85, 17.9, 29.5, 43.1, and 61.4 ␮M substrate, 0.1 M sodium formate buffer (pH 3.5).

changes primarily seem to cause the reduction of activ- exception of the first four N-terminal residues, a virtually ity in the W129A mutant. identical structure, in agreement with an rms deviation of 0.47 A˚ for 350 ␣-carbon atoms (using a threshold Overall Structure of Pro-Kumamolisin of 2 A˚ ). Accordingly, the catalytic domain is virtually To prevent autoactivation, pro-kumamolisin had been prefolded within the proenzyme. Small but significant prepared and crystallized as an inactive mutant, with differences are detectable in the loop Glu70e-Pro75e the catalytic Ser278e replaced by an Ala residue. This encompassing the S1Ј cavity, and in two loops (Pro101e- single-chain proenzyme, virtually lacking any hydrolytic Asp104e and 134Aspe-Ala137e) in the interface toward activity (Table 1), crystallized in a different space group the prodomain, where the peptide linker enters the ac- (R32) than the catalytic domain (see Table 2). As shown tive site cleft. The prodomain essentially packs along the in Figures 2B and 2C, the prodomain contacts the cata- two surface-located helices h4 and h5 of the catalytic lytic domain just at the nonprimed-side entrance to the domain (Figure 2B), grasping with its long ␣1-␣2 loop active site cleft. The connecting linker runs through the far around the catalytic domain, almost reaching the entire active site cleft ending up in a relatively flexible C-terminal end of the mature enzyme. The surface con- loop, before the chain turns into the globular domain of tact regions of these two compact domains are roughly the mature enzyme. This domain arrangement gives rise complementary forming a combined interface of 1650 A˚ 2 to a bowl-shaped proenzyme (Figure 4) with two laterally (calculated with the Protein-Protein Interaction Server), attached knobs of different size (Figure 2C), the larger leaving only a few cavities where immobilized solvent represented by the prodomain, and the smaller by the molecules are observed. At the extreme tip, the partially exposed partially flexible C-terminal part of the linker. flexible side chain of Arg69p extends into such an in- A superposition of pro-kumamolisin with the active terfacial cavity, and the side chains of Arg48p and proteinase (Comellas-Bigler et al., 2002) reveals that the Asp150e form an interdomain salt bridge. Deeper in the catalytic domain of the proenzyme exhibits, with the interface, both domains provide a number of aliphatic

Table 2. Statistics for Data Collection and Refinement Crystal A B C Content Pro-kumamolisin mutant S278A Kumamolisin mutant E32A Kumamolisin mutant W129A

Space group R32 P21 P21 Cell Constants a, b, c (A˚ ) 158.54, 158.54, 105.72 42.72, 78.57, 49.99 42.89, 78.83, 49.03 ␣, ␤, ␥ 90.0, 90.0, 120.0 90.0, 105.48, 90.0 90.0, 105.41, 90.0 Molecules/asymmetric unit 1 1 1 Resolution (A˚ ) 1.18 (1.20–1.18) 1.18 (1.20–1.18) 1.28 (1.30–1.28) Reflections measured 523,625 207,860 153,826 I/␴ (I) 24.0 (6.9) 13.1 (5.6) 9.5 (4.4) Unique reflections 164,763 98,535 74,340 a Rmerge (%) 5.3 7.7 5.4 Completeness (%) 99.3 (98.2) 97.0 (85.8) 92.7 (72.8)

Refinement Reflections used for refinement 162,816 98,434 74,287 Resolution range (A˚ ) 99.00–1.18 99.00–1.18 99.00–1.28 b c Rfactor (%)/Rfree (%) 21.3/24.4 20.2/22.2 18.3/20.7 Rmsd bonds (A˚ )/angles (Њ) 0.010/1.73 0.011/1.42 0.009/1.49 B factor rms deviation between bonded 0.98 0.85 0.87 atoms (A˚ 2) Average B (A˚ 2) protein ϩ Ca2ϩ/water 19.17/32.31 14.24/28.00 13.10/28.42 Ramachandran core conformations (%) 91.1 88.8 89.5 Additionally allowed conformations (%) 8.9 11.2 10.5

Data in brackets hold for the outermost shell. a Rmerge ϭ⌺hkl|͗I͘ Ϫ I|/⌺hkl|I|. b Rfactor ϭ⌺hkl||Fobs| Ϫ |Fcalc||/⌺hkl|Fobs|. c Rfree is the R value calculated with 500 reflections that were not used for the refinement. Structures of E32A, W129A, and Pro-Kumamolisin 1317

Figure 3. Stereo Plots of a Section around the Active Sites of Kumamolisin Mutants Displayed in Standard Orientation as in Figures 2B and 2C Only some loop segments (ropes) are shown together with selected side chains of particular importance (stick models). Water molecules and possible hydrogen bonds (only in the mutants) are represented by cyan balls and dashed lines, respectively. The images were produced using Molscript (Kraulis, 1991), Bobscript (Esnouf, 1997, 1999), and RASTER3D (Meritt and Bacon, 1997). (A) Wild-type kumamolisin (olive green) superimposed on the mutant E32A (pink). (B) Wild-type kumamolisin (olive green) superimposed on mutant W129A (red and orange). The final 2Fo Ϫ Fc electron density of mutant E32A (beige) contoured at 1.5 ␴ is only shown around important side chains. The Glu78 residue is overlaid by the difference density omit map (blue) contoured at 3.0 ␴. side chains forming a solvent-free intermolecular hy- 2001, 2002), where 40, 39, and 39 ␣-carbon atoms show drophobic core. rms deviations of 1.48, 1.32, and 1.46 A˚ , using 2 A˚ thresh- olds. Compared with these propeptides, the orientation Prodomain of Pro-Kumamolisin of the linker relative to the prodomain differs for about The shell-shaped prodomain of pro-kumamolisin com- 25Њ, reflecting slightly different arrangements of the pro- prises three subdomains: a compact core, which is domains with respect to their catalytic domains (see flanked by a three-stranded multiple turn entity at one Figure 4). Noteworthy, similar overall folds are also found end and by an elongated loop containing helices ␣1 and in the activation domains of other functionally quite un- ␣2 at the opposite end (Figure 4). The globular core related proteins (Gallagher et al., 1995), among them the (39p–47p, 85p–126p, and 147p–161p) exhibits an open- prodomains of zinc pro-carboxypeptidases (Coll et al., sandwich antiparallel-␣/antiparallel-␤ fold, made up by 1991; Estebanez-Perpina et al., 2001; Gomis-Ruth et al., the two ␣ helices ␣3 and ␣4 and the twisted ␤ sheet 1999; Guasch et al., 1992), which block the catalytic consisting of the antiparallel strands ␤5, ␤6, ␤4, and ␤10 centers of their catalytic domains in a quite different (Figures 2A and 4). Topologically, this pro-kumamolisin manner, however. core resembles the prodomains of subtilisin BPNЈ (Gal- The prodomain of pro-kumamolisin is considerably lagher et al., 1995), subtilisin E (Jain et al., 1998), and larger than these homologous prodomains, due to addi- the mouse prohormone convertase PC1 (Tangrea et al., tional core appendices that extend (in the standard ori- Structure 1318

Figure 4. Structural Comparison of the Prodomains of Kumamolisin, Subtilisins, and the Prohormone Convertases Stereo ribbon plot of the kumamolisin prodomain (green) overlaid with the optimally fitted prodomains of subtilisin BPNЈ (pink) as docked to the mature subtilisin BPNЈ (Gallagher et al., 1995), and mouse prohormone convertase 1 (orange) as determined for the isolated propeptide (Tangrea et al., 2002). The dashed red end of the pro-kumamolisin propeptide indicates linker positions P3 and P2. The orientation is obtained starting from the standard orientation (Figure 2B) rotating 90Њ along the x axis and 90Њ along the new z axis. The image was produced with Molscript (Kraulis, 1991) and RASTER3D (Meritt and Bacon, 1997). entation, Figures 2B and 2C) toward the front and toward teinases (Bode and Huber, 1992) (Figure 5; Table 3). the back of the molecule, respectively. The front-sided Thus, the conformation of the P3 and the P2Ј residue is three-stranded multiple-turn appendix comprising the that typical for parallel/antiparallel ␤-pleated sheets, the antiparallel strands ␤2, ␤8, and ␤7 packs against the P2 and the P1Ј residues have the conformation of a linker up to Pro166p, bordering the active site cleft (Fig- polyproline II helix, and the P1 residue attains the confor- ure 2C). The back-sided core appendix, on the other mation of a 310 helix. hand, consists of the two- and the three-turn helices ␣2 The P5-side chain of Val167p slots into the shallow and ␣3, which are connected by an exposed multiple- S5 subsite formed by the side chains of Trp136e and turn loop. These helices are arranged in a V-shaped Phe107e together with the Thr103e-Asp104e main chain. manner along the catalytic domain surface framing helix The P4-Ala168p residue, hydrogen bonded through its h5. The sequence homology of the propeptides is quite amido nitrogen and carbonyl functions with the side low among the sedolisins known so far. It is worth noting chain of Asn102e, sits deeply in the hydrophobic S4 that a short segment of relatively strong similarity exists pocket formed by the side chains of Phe107e, Leu33e, between kumamolisin, sedolisin, and CLN2, mapping to and Trp129e, with its ␤-methyl group extending below the last three regular secondary prelinker elements ␤9, an eyelid-like protrusion formed by the side chain of ␣5, and ␤10, which are, however, not in direct contact Asn102e (Figure 5A). The side chain of P3-Arg169p ex- with the catalytic domain. tends away from the surface. Its guanidyl group, how- ever, forms two N-O side chain hydrogen bonds with the Asp164e carboxylate group, which normally contrib- Activation Peptide/Linker utes to the . Identical to the Pro residue ␤ Leaving the prodomain core after 10, the polypeptide of the Ac-IPF-CHO inhibitor (Comellas-Bigler et al., chain of pro-kumamolisin continues into the linker, 2002), the P2-Pro170p pyrrolidin ring extends into the which forms a short 6 residue kink (Leu160p-Pro166p) hydrophobic S2-depression, demarcated on one side before entering the active site cleft with P5-Val167p. by the aliphatic part of the side chain of the catalytic Ј From the P5-residue Val167p down to the P4 -residue residue Glu78e, and on the other side by the indole Arg175p, the linker runs through the active site cleft in moiety of Trp129e. The amido nitrogen of P1-His171p a largely extended manner, forming seven inter-main hydrogen bonds to Ser128e carbonyl oxygen, while its chain hydrogen bonds and interacting through an inter- side chain slots into the S1 groove, which is bordered by face of 1067 A˚ 2 (Figure 5). The arrangement of the activa- main chain segments 130e-131e and 161e-164e, paved tion peptide in the active site cleft indicating the subsites with two ordered water molecules at the bottom, and from S5 to S4Ј coincides with the peptide substrate pre- delimited by Asp179e, which hydrogen bonds to the viously modeled to the kumamolisin (Comellas-Bigler et His171p imidazole group. The P1 carbonyl group ex- al., 2002), and is in agreement with the two pseudo- tends into the oxyanion hole forming a hydrogen bond iodotyrostatin molecules interacting with sedolisin (Wlo- to the Ser278e amido nitrogen, while the second normal dawer et al., 2004). Each residue exhibits a distinct main hydrogen bond partner in the oxyanion hole, the chain conformation, which is characteristic for the pep- Asp164e carboxylate group, is not in the expected posi- tide positions (on either side of the scissile P1-His171p— tion, due to the above-mentioned salt bridge interaction P1Ј-Phe172p peptide bond) of “canonically binding” with the P3-Arg169p side chain. proteinase protein inhibitors and their complexes The P1-P1Ј scissile peptide bond spans the catalytic formed with -like and subtilisin-like serine pro- center in a manner similar to that known for inhibitor- Structures of E32A, W129A, and Pro-Kumamolisin 1319

Figure 5. Stereo Plots of the Active Site Cleft and the Bound Linker Peptide of Pro-Kumamolisin (A) The pro-kumamolisin S278A mutant represented as a cyan stick model is superimposed with a half-transparent pink-colored Connolly surface. The linker peptide is displayed as a yellow stick model, with nitrogen and oxygen atoms given in blue and red. The dashed cyan lines indicate hydrogen bonds between the active site residues and the linker peptide. The figure was made with Molscript (Kraulis, 1991). (B) Central part of the linker peptide (orange) around the scissile peptide bond (P4 to P2Ј) superimposed with its difference electron density omit map (blue) contoured at 3 ␴. The image was performed with the programs Bobscript (Esnouf, 1997, 1999) and RASTER3D (Meritt and Bacon, 1997). proteinase complexes (Figure 5B). Due to lack of the O␥ parts of Glu78e and Leu81e and main chain segment hydroxyl in this S278A mutant, there is no direct contact 73e-78e (Figure 5A). The P2Ј-Arg173p main chain is between the “catalytic” residue Ala278e and this P1-P1Ј clamped to the Gly275e main chain via two hydrogen peptide bond. However, the O␥ atom of a superimposed bonds, while its side chain extends along a surface Ser278e residue would be only 2.3 A˚ distant from the groove bounded by Pro259e, Gly214e, and Gly215e. The His171p carbonyl carbon, i.e., ready to attack the scis- distal guanidyl group of Arg173p forms one hydrogen sile bond. The P1Ј side chain of Phe172p nestles into bond to the Asp213e carbonyl oxygen. The P3Ј-Leu174p the large S1Ј groove, flanked by the aliphatic side chain side chain inserts into a large hydrophobic surface pocket delimited by residues Pro72e and Asn73e and by the side chains of Ile267, Thr272, and Ile274. P4Ј- Table 3. Main Chain Conformational Angles of the Linker, Arg175p, finally, extends its side chain along Val273e. Compared with the Reactive Site Loop of the Proteinase The following 17 propeptide residues, from Arg176p to Protein Inhibitor Eglin C in Its Complex with Subtilisin Thr4e (see Figure 2B), form a wide loop, which expands Pro-Kumamolisin Reactive Site Loop of into a large intermolecular crystal cavity. Ala178p is the Linker Inhibitor Eglin C last linker residue rigidly placed, while the segment φ ␺ φ ␺ Glu179p-Ala184p is only defined for its main chain and the eight residues from Arg185p to Thr4e are fully disor- Ϫ Ϫ P4 142.59 157.29 71.07 140.08 dered. Only from the Ala5e position onward is the pro- P3 Ϫ114.48 139.01 Ϫ138.53 167.98 P2 Ϫ81.56 123.39 Ϫ62.17 143.06 kumamolisin peptide chain well defined by the electron P1 Ϫ80.94 42.62 Ϫ115.41 44.67 density, exactly following the course observed for the P1ЈϪ116.28 159.39 Ϫ96.72 168.83 mature enzyme. P2ЈϪ140.08 141.08 Ϫ117.41 109.86 P3ЈϪ86.06 139.06 Ϫ121.08 111.94 P4ЈϪ63.27 139.06 Ϫ75.63 Ϫ1.90 Discussion

Angle values of protein inhibitor eglin c extracted from Bode et al. Kumamolisin exhibits, like the other sedolisins, an active (1986, 1987). site triad consisting of a Ser278e, a Glu78e, and an Structure 1320

Asp82e residue. It seems to be unique, however, in that that also in pro-kumamolisin the prodomain stabilizes the catalytic Asp82e further hydrogen bonds to a (by interacting with helices h4 and h5) a folding interme- Glu32e-Trp129e pair. In this study, we have tried to ana- diate of the catalytic domain. In the subtilisins, such lyze the role of both residues. The replacement of either an intermediate is required in the late folding pathway of these residues by Ala resulted in a significant reduc- toward the correctly folded catalytic domain, to over- tion in the catalytic activity/specificity constant (Table come a high-energy barrier considered to be the price 1), suggesting that both residues are not essential, but to be paid for the extreme stability of the mature enzyme nevertheless have some importance for proton shuttling (Eder and Fersht, 1995). In the pro-subtilisins, this refold- during catalysis. However, structure analysis reveals ing resistance of the mature peptidase has been shown that these mutations induced, in addition to the side to be associated with the first calcium site of subtilisin chain replacement, perturbations in the local environ- (Strausberg et al., 1995) (see Figure 2B). Therefore, an ment. The lack of the Glu32e side chain caused the equivalent association of the refolding kinetics with for- catalytic Ser278e to adopt (besides the wild-type con- mation of the single calcium site in the sedolisins will formation) a second presumably inactive nonnative con- have to be investigated in the future. formation. The built-in Wat17 molecule replacing the In addition, the current pro-kumamolisin structure di- Glu32 carboxylate group might, on the contrary, allow rectly proves that the catalytic domain has adopted a a partial reformation of the wild-type constellation (see mature-like conformation already in the proform, and Figure 3A). The active site disturbance caused by the that the linker extends through the active site cleft such removal of the Trp129 side chain in the W129A mutant, that it first will be autocatalytically cleaved at His171p- in contrast, has much more dramatic structural effects Phe172p, i.e., via an intramolecular attack. Both the in the active site as well as at the anchoring segment linker chain and the active site cleft are quite comple- Ser128e-Gly131e. The residual activity of this mutant, mentary, so that this structure mimics an efficient kuma- also manifested itself in the self-activation of the molisin–substrate complex, with a typical substrate proW129 mutant and the further degradation of the pro- cleavage conformation, presumably not disturbed by → peptide, again shows that this Glu32e-Trp129e pair is the small Ser278 Ala replacement (see Carter and Wells, 1988). It is noteworthy that the cleavage sequence not essential for the activity of kumamolisin. In conclu- around the scissile peptide bond agrees well with the sion, we believe that in wild-type kumamolisin the tightly known cleavage preference of kumamolisin (Oda et al., linked Ser-Glu-Asp triad, connected through short low- 2000). The extreme stability of this S278A pro-kumamoli- barrier hydrogen bonds (for references and discussion sin mutant might be due (besides the lack of the Ser278 see Comellas-Bigler et al. [2002]), together with an adja- O␥ atom and the neutral pH value of the crystallization cent proton acceptor/donor (such as the Thr32e-water buffer) to the presence of the P3-Arg169p linker residue, pair in sedolisin, or Wat17 in the E32A mutant) is suffi- whose guanidyl group attracts the Asp164e carboxylate cient to make the Ser278 O␥ nucleophilic at low pH and side chain, preventing the formation of a functional oxy- to shuttle a proton from and to the reactive Ser278e and anion hole by substrate-assisted inactivation. In an the leaving group of the scissile peptide bond of the acidic environment, protonation of this latter carboxyl- bound substrate, respectively. Under the extreme envi- ate group would cause the disruption of this salt bridge, ronmental conditions of the thermophilic Bacillus no- rendering the catalytic apparatus in pro-kumamolisin vosp. MN-32, the Glu32e-Trp129 pair might provide ad- more active. Thus, this P3-Arg residue might be a built- ditional advantage. Certainly, the amino acid packing in in switch, delaying self-activation of pro-kumamolisin the kumamolisin active site region is adapted to both until secretion into the acidic medium. Noteworthy, this side chains, so that a replacement causes unpredictable P3-Arg residue is not conserved among the sedolisins. disturbances in the wider environment. Presumably, the first cleavage product upon pro- Less ambiguous is the interpretation of the pro-kuma- kumamolisin processing will be the inactive complex molisin structure: in spite of a very low (10%) sequence between the N-terminally elongated catalytic kumamoli- identity of topologically equivalent residues, the core of sin domain and its propeptide down to His171p. The its prodomain resembles (besides the prodomains of common migration in the cleaved pro-W129A mutant the zinc pro-carboxypeptidases) the prodomains of the during gel filtration (data not shown) proves that the pro-subtilisins and of the pro-protein convertases (see propeptide and the catalytic domain of the initially Figure 4), and in a similar manner docks to an equivalent cleaved pro-kumamolisin, analogous to the subtilisins site of the catalytic domain. This structural similarity (Ikemura and Inouye, 1988), and to furin (see Anderson between these latter Ser-His-Asp subtilases and the et al., 2002), form a complex, with the propeptide pre- Ser-Glu-Asp-equipped sedolisins is a further indication sumably acting as a competitive inhibitor. As suggested of divergent evolution from common ancestors. Ac- by the slow dissociation of our prodomain-W129A com- cording to newly acquired full-length sequences, other plex (in the acidic crystallization buffer used, see Experi- sedolisins seem to exhibit this pro-enzyme structure as mental Procedures), the heterodimeric complex (maybe well. However, the relationships between both protein- facilitated by acidification or upon additional cleavage[s] ase superfamilies are even closer. Work on sedolisin within the compact prodomain as observed in the other (Oda et al., 1994) has shown that its propeptide (and subtilases) will dissociate releasing the active, N-ter- probably that of kumamolisin) as that of the subtilisins minally elongated enzyme. Because mature kumamoli- (Gallagher et al., 1995) and the PCs (Bhattacharjya et sin is found to start with Ala1e, the exposed, flexible al., 2001) is required for efficient refolding, presumably linker part between Ser188p and Ala1e must be further acting like a chaperonin as well. In the absence of more truncated, presumably not catalyzed by active kuma- detailed kinetic measurements, we can only presume molisin, due to the nonmatching specificity. Structures of E32A, W129A, and Pro-Kumamolisin 1321

In conclusion, this pro-kumamolisin structure is the solution (11.6 mg/ml kumamolisin in 25 mM sodium chloride, 12.5 first crystal structure of an intact pro-subtilase, and MOPS, pH 7.0) and 1 ␮l of the reservoir buffer (1.4 M tri-sodium Њ therefore presumably a structural prototype for the citrate dihydrate, 0.1 M HEPES-Na, pH 7.5). After 1 week at 20 C, the crystals reached a final size of 0.4 ϫ 0.2 ϫ 0.2 mm3 in the case whole subtilase superfamily. Thus, this structure not of full-length S278A and of 0.2 ϫ 0.05 ϫ 0.05 mm3 in case of E32A only has wider implications for the microbial and mam- and W129A. malian sedolisins, but for the subtilisins proper and for furin and the other pro-protein convertases as well. Data Collection and Refinement All crystals were soaked for 10 s in the cryobuffers containing 25% Experimental Procedures (v/v) glycerol before freezing them in a nitrogen stream at 100 K (Oxford Cryosystems Cryostream). High-resolution data to beyond Plasmids, Expression, and Purification 1.2/1.3 A˚ resolution were collected for all kumamolisin mutants at the The E32A, W129A, and pS278A plasmids were constructed by PCR DESY BW6 beamline (Deutsches Elektronen Synchroton, Hamburg, mutagenesis of the pro-kumamolisin pS3-A1 plasmid similar as de- Germany) at a wavelength of 1.050 A˚ . scribed previously (Oyama et al., 2002). After transformation, the Indexing and integration of the diffraction data was carried out E. coli JM109 cells were cultured in ampicillin-containing super broth with DENZO (Otwinowski and Minor, 1997). The data were merged, at 30ЊC, before protein expression was induced with IPTG. The scaled, and truncated using programs supported by the Collabora- washed cells were disrupted and the resultant suspension was cen- tive Computational Project No. 4 (CCP4, 1994). All structures were trifuged. For the E32A and the W129A mutants, the supernatant was solved by molecular replacement using the program suite AMoRe incubated at pH 4.8, 25ЊC, and clarified by centrifugation. (Navaza, 1994) and the native kumamolisin search model (Comellas- E32A was first eluted from a DEAE-Cellulose column in buffer A Bigler et al., 2002). The best solution for the Ser278Ala pro-kumamol-

(20 mM acetate buffer, pH 5.5, containing 5 mM CaCl2) by a linear isin mutant had a correlation coefficient and R factor of 36.7% and NaCl gradient (0–0.6 M NaCl), concentrated, fractionated on Sepha- 46.9% for 15.0 to 3.5 A˚ data. The first electron density map calcu- dex G-75 in buffer B (20 mM acetate buffer, pH 5.5, containing 0.1 M lated with the kumamolisin molecular replacement model nicely

NaCl and 1 mM CaCl2), and further purified on a MonoQ 5/5 column, displayed density also for most of the prodomain. The program from where it was eluted by buffer C (50 mM acetate buffer, pH 5.5, MAIN was used for model building performed on an SGI graphics containing 1 mM CaCl2) using a 0–0.4 M NaCl linear gradient, and workstation (Turk, 1992). Positional and individual temperature fac- stored with an apparent molecular weight of 43 kDa (see Figure 1) tor refinement was carried out with the program CNS (Bru¨ nger, 1998) at Ϫ30ЊC. The E32A mutant sample obtained exhibited the same until convergence, yielding final R factors of 18.3%–21.3% (Table 2). sequence as the recombinant wild-type kumamolisin (see Comellas- In the E32A and W129A structures, the polypeptide chains are Bigler et al., 2002), mostly starting with N-terminal Ala1e, besides defined by electron density from the N-terminal Ala1e up to Pro357e, small amounts with Ser188p and Ser186p (see Figure 2A). except for main chain segment 241e-245e and a few side chains. W129A (Ala1e-Pro364e) in complex with the prodomain as a sepa- The pro-kumamolisin S278A molecule is fully defined by electron rate entity was eluted from a DEAE-Cellulose column in buffer A density from residue Glu12p up to Ala178p. The residues from with a linear gradient from 0 to 0.5 M NaCl, concentrated, and Glu179p to Ala184p are only partially defined, while the eight resi- fractionated on Sephadex G-75 in buffer B. The protein eluted from dues from Arg185p to Thr4e corresponding to the loop connecting a RESOURCE-Q column equilibrated with buffer C with a 0–0.25 M the prodomain with the catalytic domain are fully disordered. The NaCl linear gradient, and was stored at Ϫ30ЊC as a two-chain com- rest of the catalytic domain shows adequate electron density up to plex consisting of the 21 kDa propeptide and the 43 kDa “mature” Pro357e. In all structures, a single Ca2ϩ ion coordinated by six li- W129A mutant (see Figure 1). gands is observed, refining with full occupancy. An additional elec- The S278A pro-kumamolisin mutant, comprising residues Ser2p tron density near Gly200 is interpreted as a sulfate ion in both cata- to Pro364e, was eluted from a DEAE-Cellulose column in buffer E lytic domain structures. During the final steps of refinement, two

(50 mM acetate buffer, pH 5.5, containing 5 mM CaCl2) by a linear alternative conformations were identified and built for a few side gradient from 0 to 0.5 M NaCl, was concentrated, fractionated on chains in all three structures. The main chain angles, calculated with Sephadex G-75 in buffer F (50 mM acetate, pH 5.5, containing 0.1 M PROCHECK (Laskowski et al., 1993), show that all residues in all

NaCl and 5 mM CaCl2), further eluted from a RESOURCE-Q column structures fall into the most favored or additionally favored regions

(in 50 mM MOPS-NaOH buffer, pH 7.0, containing 1 mM CaCl2) of the Ramachandran plot. The crystal data and the refinement using a 0–0.2 M NaCl linear gradient, and stored with an apparent parameters are given in Table 2. molecular weight of 64 kDa (see Figure 1) at Ϫ30ЊC. Kinetic parameters were determined at 60ЊC in 0.1 M sodium Acknowledgments formate buffer, pH 3.5, using KPIAF*NphRL as a substrate. The cleavage of the substrate between Phe and Nph was monitored The financial support by the SFB469 of the LM Universita¨ tMu¨ nchen, spectrophotometrically by following the decrease in absorbance at by the Fonds der Chemischen Industrie, and by the “Training and 300 nm. Initial rates were measured at five substrate concentrations, Mobility” project ERBFMRX-CT98-0193 of the European Union and the KM and the Vmax values were calculated graphically from (W.B.), as well as by a Grant-in-Aid for Scientific Research (B) No. Lineweaver-Burk plots (see Table 1). 15380072 from the Japan Society for the Promotion of Science (K.O.), is kindly acknowledged. We thank G. Bourenkov and H.D. Protein Crystallization Bartunik, MPG-ASMB Hamburg, for help with the data collection. The purified recombinant mutants E32A (catalytic domain only), W129A (in complex with the prodomain as a separate entity), and Received: March 24, 2004 S278A pro-kumamolisin were crystallized using sitting-drop vapor- Revised: April 23, 2004 diffusion procedures. E32A, of space group P2 with one molecule 1 Accepted: April 23, 2004 in the asymmetric unit, was crystallized isomorphously with native Published: July 13, 2004 kumamolisin (Comellas-Bigler et al., 2002). Isomorphous crystals consisting of the W129A catalytic domain alone were obtained from the complex of the W129A catalytic domain and the propeptide References (Figure 1). The S278A pro-kumamolisin crystals belong to space group R32 and contain one molecule per asymmetric unit (Table 2). Anderson, E.D., Molloy, S.S., Jean, F., Fei, H., Shimamura, S., and Thomas, G. (2002). The ordered and compartment-specific autopro- The P21 crystals were grown by mixing 1 ␮l of a 9 mg/ml mutant/ complex solution in 25 mM sodium chloride, 12.5 mM sodium ace- teolytic removal of the furin intramolecular chaperone is required tate, pH 5.5, with 1 ␮l reservoir solution, consisting of 0.4 M ammo- for enzyme activation. J. Biol. Chem. 277, 12879–12890. nium sulfate, 0.1 M sodium acetate (pH 4), and evaporation toward Barton, G.J. (1993). Alscript—a tool to format multiple sequence the reservoir. The R32 crystals grew in a mixture of 1 ␮l protein alignments. Protein Eng. 6, 37–40. Structure 1322

Bhattacharjya, S., Xu, P., Xiang, H., Chretien, M., Seidah, N.G., and E-propeptide complex at 2.0 angstrom resolution. J. Mol. Biol. 284, Ni, F. (2001). pH-induced conformational transitions of a molten- 137–144. globule-like state of the inhibitory prodomain of furin: implications Kraulis, P.J. (1991). Molscript: a program to produce both detailed for zymogen activation. Protein Sci. 10, 934–942. and schematic plots of protein structures. J. Appl. Crystallogr. 24, Bode, W., and Huber, R. (1992). Natural protein proteinase-inhibitors 946–950. and their interaction with proteinases. Eur. J. Biochem. 204, Laskowski, R.A., Macarthur, M.W., Moss, D.S., and Thornton, J.M. 433–451. (1993). Procheck: a program to check the stereochemical quality of Bode, W., Papamokos, E., Musil, D., Seemueller, U., and Fritz, H. protein structures. J. Appl. Crystallogr. 26, 283–291. (1986). Refined 1.2-a crystal-structure of the complex formed be- Li, Y.Y., Hu, Z.X., Jordan, F., and Inouye, M. (1995). Functional analy- tween subtilisin Carlsberg and the inhibitor eglin-c. Molecular-struc- sis of the propeptide of subtilisin-e as an intramolecular chaperone ture of eglin and its detailed interaction with subtilisin. EMBO J. 5, for protein folding: refolding and inhibitory abilities of propeptide 813. mutants. J. Biol. Chem. 270, 25127–25132. Bode, W., Papamokos, E., and Musil, D. (1987). The high-resolution Lin, L., Sohar, I., Lackland, H., and Lobel, P. (2001). The human X-ray crystal-structure of the complex formed between subtilisin CLN2 protein/tripeptidyl-peptidase I is a serine that auto- Carlsberg and eglin-c, an inhibitor from the leech Hirudo- activates at acidic pH. J. Biol. Chem. 276, 2249–2255. medicinalis. Structural-analysis, subtilisin structure and interface Merritt, E.A., and Bacon, D.J. (1997). Raster3D: photorealistic molec- geometry.2. Eur. J. Biochem. 166, 673. ular graphics. Methods Enzymol. 277, 505–524. Bru¨ nger, A.T., Adams, P.D., Clore, G.M., DeLano, W.L., Gros, P., Murao, S., Ohkuni, K., Nagao, M., Oda, K., and Shin, T. (1988). A Grosse-Kunstleve, R.W., Jiang, J.S., Kuszewski, J., Nilges, M., novel thermostable, S-Pi (pepstatin Ac) insensitive acid proteinase Pannu, N.S., et al. (1998). Crystallography & NMR system: a new from thermophilic Bacillus novosp strain Mn-32. Agric. Biol. Chem. software suite for macromolecular structure determination. Acta (Tokyo) 52, 1629–1631. Crystallogr. D Biol. Crystallogr. 54, 938–940. Murao, S., Ohkuni, K., Nagao, M., Hirayama, K., Fukuhara, K., Oda, Carter, P., and Wells, J.A. (1988). Dissecting the catalytic triad of a K., Oyama, H., and Shin, T. (1993). Purification and characterization . Nature 332, 564–568. of kumamolysin, a novel thermostable pepstatin-insensitive car- Coll, M., Guasch, A., Aviles, F.X., and Huber, R. (1991). 3-dimensional boxyl proteinase from Bacillus novosp Mn-32. J. Biol. Chem. 268, structure of porcine procarboxypeptidase-b. A structural basis of 349–355. its inactivity. EMBO J. 10, 1–9. Navaza, J. (1994). Amore: an automated package for molecular re- Comellas-Bigler, M., Fuentes-Prior, P., Maskos, K., Huber, R., placement. Acta Crystallogr. A 50, 157–163. Oyama, H., Uchida, K., Dunn, B.M., Oda, K., and Bode, W. (2002). Nicholls, A., Bharadwaj, R., and Honig, B. (1993). Grasp: graphical ˚ The 1.4 A crystal structure of kumamolysin: a thermostable serine- representation and analysis of surface properties. Biophys. J. 64, carboxyl-type proteinase. Structure 10, 865–876. 166–170. Eder, J., and Fersht, A.R. (1995). Pro-sequence-assisted protein- Oda, K., Takahashi, T., Tokuda, Y., Shibano, Y., and Takahashi, S. folding. Mol. Microbiol. 16, 609–614. (1994). Cloning, nucleotide-sequence, and expression of an isoval- Esnouf, R.M. (1997). An extensively modified version of MolScript eryl pepstatin-insensitive carboxyl proteinase gene from Pseudo- that includes greatly enhanced coloring capabilities. J. Mol. Graph. monas Sp-101. J. Biol. Chem. 269, 26518–26524. Model. 15, 132–134, 112–113. Oda, K., Ito, M., Uchida, K., Shibano, Y., Fukuhara, K., and Taka- Esnouf, R.M. (1999). Further additions to MolScript version 1.4, in- hashi, S. (1996). Cloning and expression of an isovaleryl pepstatin- cluding reading and contouring of electron-density maps. Acta Crys- insensitive carboxyl proteinase gene from Xanthomonas sp T-22. tallogr. D Biol. Crystallogr. 55, 938–940. J. Biochem. (Tokyo) 120, 564–572. Estebanez-Perpina, E., Bayes, A., Vendrell, J., Jongsma, M.A., Oda, K., Ogasawara, S., Oyama, H., and Dunn, B.M. (2000). Subsite Bown, D.P., Gatehouse, J.A., Huber, R., Bode, W., Aviles, F.X., and preferences of pepstatin-intensitive carboxyl proteinases from pro- Reverter, D. (2001). Crystal structure of a novel mid-gut procarboxy- karyotes: Kumamolysin, a thermostable pepstatin-insensitive car- peptidase from the cotton pest Helicoverpa armigera. J. Mol. Biol. boxyl proteinase. J. Biochem. (Tokyo) 128, 499–507. 313, 629–638. Otwinowski, Z., and Minor, W. (1997). Processing of X-ray diffraction Ezaki, J., Takeda-Ezaki, M., Oda, K., and Kominami, E. (2000). Char- data collected in oscillation mode. Methods Enzymol. 276, 307–326. acterization of endopeptidase activity of -I/ Oyama, H., Hamada, T., Ogasawara, S., Uchida, K., Murao, S., Beyer, CLN2 protein which is deficient in classical late infantile neuronal B.B., Dunn, B.M., and Oda, K. (2002). A CLN2-related and thermosta- ceroid lipofuscinosis. Biochem. Biophys. Res. Commun. 268, ble serine-carboxyl proteinase, kumamolysin: cloning, expression, 904–908. and identification of catalytic serine residue. J. Biochem. (Tokyo) Gallagher, T., Gilliland, G., Wang, L., and Bryan, P. (1995). The 131, 757–765. prosegment-subtilisin BPNЈ complex: crystal-structure of a specific Rawlings, N.D., and Barrett, A.J. (1999a). Tripeptidyl-peptidase I is foldase. Structure 3, 907–914. apparently the CLN2 protein absent in classical late-infantile neu- Gomis-Ruth, F.X., Companys, V., Qian, Y., Fricker, L.D., Vendrell, ronal ceroid lipofuscinosis. Biochim. Biophys. Acta 1429, 496–500. J., Aviles, F.X., and Coll, M. (1999). Crystal structure of avian car- Rawlings, N.D., and Barrett, A.J. (1999b) MEROPS: the peptidase boxypeptidase D domain II: a prototype for the regulatory metallo- database. Nucleic Acids Res. 27, 325-331. carboxypeptidase subfamily. EMBO J. 18, 5817–5826. Shibata, M., Dunn, B.M., and Oda, K. (1998). Substrate specificity Guasch, A., Coll, M., Aviles, F.X., and Huber, R. (1992). Three-dimen- of pepstatin-insensitive carboxyl proteinase from Bacillus coagu- sional structure of porcine pancreatic procarboxypeptidase A: a lans J-4. J. Biochem. (Tokyo) 124, 642–647. comparison of the A and B zymogens and their determinants for Siezen, R.J., and Leunissen, J.A.M. (1997). Subtilases: the superfam- inhibition and activation. J. Mol. Biol. 224, 141–157. ily of subtilisin-like serine . Protein Sci. 6, 501–523. Henrich, S., Cameron, A., Bourenkov, G.P., Kiefersauer, R., Huber, Strausberg, S.L., Alexander, P.A., Gallagher, D.T., Gilliland, G.L., R., Lindberg, I., Bode, W., and Than, M.E. (2003). The crystal struc- Barnett, B.L., and Bryan, P.N. (1995). Directed evolution of a subti- ture of the proprotein processing proteinase furin explains its strin- lisin with calcium-independent stability. Biotechnology (NY) 13, gent specificity. Nat. Struct. Biol. 10, 520–526. 669–673. Ikemura, H., and Inouye, M. (1988). In vitro processing of pro-subti- Tangrea, M.A., Alexander, P., Bryan, P.N., Eisenstein, E., Toedt, lisin produced in Escherichia coli. J. Biol. Chem. 263, 12959–12963. J., and Orban, J. (2001). Stability and global fold of the mouse Jain, S.C., Shinde, U., Li, Y.Y., Inouye, M., and Berman, H.M. (1998). prohormone convertase 1 pro-domain. Biochemistry 40, 5488–5495. The crystal structure of an autoprocessed Ser221Cys-subtilisin Tangrea, M.A., Bryan, P.N., Sari, N., and Orban, J. (2002). Solution Structures of E32A, W129A, and Pro-Kumamolisin 1323

structure of the pro-hormone convertase 1 pro-domain from Mus musculus. J. Mol. Biol. 320, 801–812. Turk, D. (1992). Weitereentwicklung eines Programms fu¨ r Moleku¨ lar- graphik und Elektronendichte-Manipulation und seine Anwendung auf verschiedene Protein-Strukturaufkla¨ rungen (Munich: Tech- nische Universita¨ tMu¨ nchen). Wlodawer, A., Li, M., Dauter, Z., Gustchina, A., Uchida, K., Oyama, H., Dunn, B.M., and Oda, K. (2001). Carboxyl proteinase from Pseu- domonas defines a novel family of subtilisin-like enzymes. Nat. Struct. Biol. 8, 442–446. Wlodawer, A., Li, M., Gustchina, A., Oyama, H., Dunn, B.M., and Kohei, O. (2003). Structural and enzymatic properties of the sedolisin family of serine-carboxyl peptidases. Acta Biochim. Pol. 50, 81–102. Wlodawer, A., Li, M., Gustchina, A., Oyama, H., Oda, K., Beyer, B.B., Clemente, J., and Dunn, B.M. (2004). Two inhibitor molecules bound in the active site of Pseudomonas sedolisin: a model for the bi- product complex following cleavage of a peptide substrate. Bio- chem. Biophys. Res. Comm. 314, 638–645.

Accession Numbers

The coordinates have been deposited at the EBI under the accession codes 1T1G (E32A mutant), 1T1I (W129A mu- tant), and 1T1E (pro-kumamolisin). The nucleotide sequence has been deposited at the DDBJ/EMBL/GenBank databases under the accession number AB070740.