Structural basis for the Zn²⁺ inhibition of the zymogen-like kallikrein-related peptidase 10

An extended N-terminus is required for activity of KLK10

The first trials to obtain active human KLK10 with the standard residue number 16, a glutamate at the N-terminus of the mature enzyme were not successful. This finding is not surprising, as all other KLKs possess a hydrophobic residue in this position, typically Ile (nine cases), Leu (3), Val (KLK8), Ala (KLK9). Thus, the first hydrophobic residue before Glu16 was chosen, namely Leu13, which resulted in an active KLK10 peptidase, exhibiting typical activity against fluorogenic and chromogenic substrates, which allowed for substrate specificity profiling from the P4 to P1 positions (Debela et al., 2006b). Taking into account this unusual behavior, it is suggested that Leu13 is designated Leu16 and the following three amino acids Asp16A, Pro16B, and Glu16C.

KLK10 was active against Tos-Gly-Pro-Lys-AMC, however, it seemed to prefer P1-Arg, as corroborated by the following examples of chromogenic and fluorogenic substrates with increasing catalytic efficiency k_cat/K_M: Z-Gly-Gly-Arg-AMC (2 m⁻¹·s⁻¹), BOC-Phe-Ser-Arg-AMC (95 m⁻¹·s⁻¹), BOC-Val-Pro-Arg-AMC (701 m⁻¹·s⁻¹), and Z-Gly-Pro-Arg-AMC (2076 m⁻¹·s⁻¹) (Debela, 2007). Z and BOC are the carboxybenzyl and tert-butyloxycarbonyl protecting groups of the peptide N-terminus, which correspond to aromatic and aliphatic hydrophobic P4 side chains and may occupy the S4 subsite. By measuring the KLK10 activity against Z-Gly-Pro-Arg-AMC with increasing Zn²⁺ concentrations, no activity was observed beyond 100 μm. The inhibition constant was determined as IC₅₀=1.6±0.1 μm, which is below the one of KLKs 2–5, 7 and 8 (Goettig et al., 2010).

Overall structure

Despite the presence of the small inhibitor PABA in the crystallization screen, KLK10 crystallized with a ligand-free active site. The corresponding structure will be designated KLK10 and the one that was obtained from crystal soaks with 10 mm ZnCl₂ as KLK10-Zn. Both crystals belong to space group P2₁ and contain two KLK10 molecules per asymmetric unit, with only insignificant deviations between the two independent copies (Figure 2A, B). Already the molecular replacement searches indicated that the target structure of KLK10 differed to some extent from other structures of mature KLKs. The calculation of root mean square deviations using 195 aligned residues resulted for KLK4 (PDB code 2BDG) in a RMSD of 1.63 Å, whereas KLK2 (4NFE) reached 1.43 Å. The zymogen-like conformation of mature KLK3 (3QUM) exhibited an RMSD of 1.47 Å and the only true zymogen structure among the KLKs, namely pro-KLK6 (1GVL), also had a relatively low RMSD with 1.45 Å for 187 aligned residues (Figure 2C). A similarly low RMSD of 1.44 Å was obtained with bovine trypsinogen (2TGT), which eventually was the best basis for a molecular replacement search.

Figure 2:

Overall structure of KLK10.

(A) Both KLK10 (yellow) and KLK-Zn (not shown) exhibit a similar dimer in asymmetric unit, with the active sites facing each other. The 99-loops of molecules A and B, shown in red, intercalate to some extent. A stretch of 10 residues from His95 to Arg96 is little defined in the electron density, comprising the whole insertion from Gln95A to Pro95H. The catalytic triad residues Ser195, His57, and Asp102 are depicted as ball-and-stick models and labeled with their sequence numbers. (B) Monomer of KLK10 in standard orientation, with the catalytic triad as ball-and-stick models. In addition to the 99-loop, the other flexible regions of KLK10 are highlighted. The N-terminus (blue) is largely disordered and does not form the activating salt bridge to Asp194. Both the 75-loop (green), which is one residue shorter than most other KLKs, and the 148-loop (aquamarine) lack well-defined electron density for at least six residues. (C) Stereo representation of superposed KLK10-Zn (deep purple), pro-KLK6 (salmon), and active KLK2 (light gray). The disordered regions of KLK10 or of KLK2 have been omitted, e.g. parts of the 99-loop. Catalytic triad residues, the specificity determining Asp189 and Pro190, as well as Asp194, which potentially forms the stabilizing activating salt bridge to the N-terminus, are depicted as stick models and labeled with sequence numbers. Trp141 is also shown, whose backbone amide forms a hydrogen bond to Asp194, as observed for other trypsin-like zymogens. The most relevant loops are labeled regardless of their disorder.

Consequently, the overall fold of KLK10 resembles the prototypic serine protease trypsin, even more its zymogen trypsinogen, with some ingredients from other KLKs (Figure 2B). Active KLK10 consists of 232 amino acid residues, starts with Leu13 (16) and terminates with Asn246. This single polypeptide chain is, as in most other KLKs cross-connected by six disulfide bridges and exhibits the kallikrein-typical cis-Pro219 in the 220-loop (Bode et al., 1983). Otherwise, the structure of KLK10 differs considerably from all other active KLKs, due to extended disordered stretches, starting with the N-terminus and the three major surface loops surrounding the active site (Figure 2B).

The flexible N-terminus and the active site loops of KLK10

Whereas classical KLKs, such as KLK1 or KLK2 mainly exhibit disorder in the 11-residue 99/kallikrein loop, mature KLK10 displays several disordered regions, including the N-terminus, as well as the 75-, 99-, and 148-loops (Laxmikanthan et al., 2005; Skala et al., 2014). Similar to the pro-form of KLK6, the first defined residue in the electron density is Ser20 in chain A and Pro21 in chain B, right before Cys22, which forms a stabilizing disulfide to Cys157 (Gomis-Ruth et al., 2002). This finding confirms that in the present crystal structures the N-terminal residue Leu16 (Leu13 in the common alignment) did not insert into the activation pocket (Figure 3A). In addition, the density around Asp194 does not indicate any salt bridge formation with the α-ammonium group, which would allow to favor the typical rigid conformation of active, mature trypsin-like serine proteases (Bode et al., 1978; Huber and Bode, 1978). Asp194 adopts a similar position as in pro-KLK6 or both bovine trypsinogen and chymotrypsinogen, with a hydrogen bond to the Trp141 backbone amide (Figure 2C). In contrast to pro-KLK6, whose activation pocket is largely occupied by the 191-loop from Lys188 to Gln192, only Gln192 is located in the corresponding relatively open pocket of KLK10 (Figure 3A). These facts alone could classify the present KLK10 variant as zymogen, albeit a rather active one. Zymogen activity is well known for another human serine protease, single-chain tissue type plasminogen activator (sc-tPA), which can adopt an active conformation by insertion of the Lys156 side chain into the activation pocket, with formation of a salt bridge between the Nζ atom and the carboxylate of Asp194 (Renatus et al., 1997). Perhaps Lys150 or Lys153 of KLK10 might fulfill the role of such an activating side chain, but they seem not ideally positioned with respect to the activation pocket.

Figure 3:

The disordered N-terminus and surface loops of KLK10.

Similar to the same stretches in KLK10-Zn, both molecules A and B lack display disorder without significant electron density in these loops. Gray spheres indicate the likely trace of the disordered main chain segments. (A) The N-terminus of KLK10 is flexible and disordered as in zymogens, e.g. pro-KLK6. Electron density is missing from Leu13 (16) to Gly19, while Ser20 and Pro21 are better defined. (B) The 75-loop exhibits the unusual hydrophobic sequence of four Leu residues, with disorder from Asp71 to Leu77, comprising one deletion with respect to chymotrypsinogen. (C) The 99-loop exhibits the longest disordered stretch of KLK10 from His95 to Arg96, comparable to the same loop of KLK2, which is three residues longer. (D) The 148-loop lacks electron density from Ala146 to Tyr152 in both molecules A, whereas in molecules B the disorder extends to Gly155. Presumably, this loop adopts a wheel-like conformation, which distinguishes it strongly from active conformations and resembles much more the conformation in zymogens.

However, the unusual sequence Leu16-Asp16A-Pro16B-Glu16C does not qualify as typical propeptide either, and may be capable of inserting into the activation pocket in solution, as a series of acetylation experiments of the N-terminus indicates. According to N-terminal sequencing KLK10 possesses a 100% free N-terminus, which upon acetylation is blocked to at least 75%. By contrast, in the presence of the chloromethyl ketone inhibitor PPACK acetylation of KLK10 results in only 25% blocked N-terminus. In addition, the activity against Z-Gly-Pro-Arg-AMC drops to zero turnover for acetylated KLK10, which might represent the trapped zymogen state as present in the crystal structures.

The short 37 and 62-loops of KLK10 do not exhibit significant alterations compared to KLK2 and pro-KLK6, in stark contrast to the larger loops that surround the active site cleft of KLK10. Nevertheless, the massive disorder in the major surface loops of KLK10 is neither typical for a zymogen nor easily explained. For example, bovine chymotrypsinogen (2CGA) is extremely well defined in the electron density, most likely because the N-terminal Cys1 forms a disulfide to Cys122, which is not conserved in the KLKs and appears to suffice in rigidifying the otherwise flexible regions of the activation domain, including the longer loops around the active site and the S1 pocket. The 75-loop of KLK10 with a single residue deletion compared to chymotrypsinogen comprises the unusual sequence DDHLLLLQGE, in which only Asp70 and Glu80 are conserved residues. The electron density in all four copies of KLK10 is missing for Asp71 to Leu77, whereas pro-KLK6 is only disordered from residues 78–80 (Figure 3B). All KLKs, except KLK10, possess mainly charged or polar residues in this loop, which is in general rather well defined and can serve either for Zn²⁺ or Ca²⁺ binding, as in most trypsin-like pancreatic/coagulation proteases (Bode and Schwager, 1975; Goettig et al., 2010). For KLK4 an activity regulating allosteric link between the N-terminus and the 75-loop has been observed similar to coagulation factor VIIa (Debela et al., 2006a).

Remarkably, the 99-loop of KLK10 exhibits an eight residue insertion, which is nearly as long as the 11 residue insertion of the three classical KLKs 1, 2 and 3. Seemingly, such long 99-loops can be relatively rigid upon glycosylation on the conserved Asn95 as in KLK1, while the binding of substrates or protein contacts may rigidify it as in KLK3, whereas it is largely disordered in glycan-free KLK2 (Laxmikanthan et al., 2005; Menez et al., 2008; Skala et al., 2014). Since the 99-loop of KLK10 is disordered from His95 to Arg96, including most of the kallikrein loop insertion from Gln95A to Pro95H, its conformation remains hypothetical for this part (Figure 3C). Most of this stretch seems to block the active site, making a wide turn above the active site with the segment Arg97 to Asp99 facing the catalytic His57. Moreover, the 99-loop appears to contribute to the dimer formation of KLK10 in the crystal structure by mutually intercalating 99-loops near the active sites (Figure 2A).

Also, the 148-loop, which is usually well ordered in active trypsin-like proteases, exhibits extended disorder. This loop is known as autolysis loop of the coagulation factors, whose cleavage can lead to inactivation as recently found for KLK2 (Guo et al., 2016). Both molecules A in the KLK10 and KLK10-Zn crystals lack electron density for Ala146 to Tyr151, whereas in molecules B the density is missing for Ala145 to Gly154 (Figure 3D). Again, the corresponding less defined stretch of pro-KLK6 is shorter, ranging only from residues 148–151. However, the disordered segment of bovine trypsinogen extends even from residues 142–151 and beyond the following defined secondary structural elements, comprising the 191- and 220-loops from residues 184–193 and 217–223, respectively.

By contrast, the C-terminal segments of KLK10 are remarkably rigid and only minor disorder occurs in some side chains. Accordingly, the 176-, 191-, 205-, and 220-loops are well defined in the electron density in all four copies of KLK10, which is also independent of the Zn²⁺, bound in the catalytic center. Essentially, the 176-loop coincides with the conformations of active KLKs, such as KLK2 or KLK4. By contrast, the 191-loop, which shapes large parts of the S1-pocket adopts a zymogen conformation, e.g. with the Asp189 turned towards the molecular surface (Figure 4). Then, the shortened 205-loop displays the expected conformation as in the active KLKs, and also the 220-loop is not too much altered when compared to active KLK2. Due to the disulfide Cys191–Cys220 the 220-loop conformation depends to some extent on the respective conformation of the 191-loop.

Figure 4:

Stereo representation of the 191-loop in KLK10-Zn.

Side chains are represented as sticks and the most relevant residues are labeled, including the catalytic triad residues Ser195 and His57. It is worth noting that the loop is well defined in the 2Fo-Fc electron density (contour 1 σ), similar to the neighbouring 220-loop, to which it is linked by the disulfide Cys191–Cys220. Asp194 is in the typical zymogen-like position turned away from the activation pocket, as in pro-KLK6 or trypsinogen. Nevertheless, the region of the potential salt-bridge formation with the N-terminus near Gln192 is not occluded as in typical zymogens. Ser193 is relocated from the oxyanion hole position, while Asp189 is turned from the bottom of the S1 pocket to the bulk solvent, in contrast to Pro190. This Pro residue might contribute to the mixed tryptic and hydrophobic P1-specificity of KLK10.

Active site and substrate binding pockets

In order to judge the structural characteristics of KLK10 with respect to its active peptidase conformation, the two KLK coordinate data sets with the smallest RMSD are further discussed. As mentioned before, active KLK2 and pro-KLK6 bear the highest similarity to KLK10. Their surface loops share similar conformation or disorder to some extent. Intriguingly, the secondary structure scaffold of all three molecules is essentially the same, even the catalytic triad residues are at their standard position (Figure 2C). However, the catalytic Oγ of Ser195 of molecule A in KLK10-Zn points directly to the Nε2 of His57, while in the other copies it is turned by about 90° towards the S1 pocket as in KLK2, or as in pro-KLK6 it is turned away by 180° with respect to the active protease conformation. Moreover, the backbone amide of Ser193 in KLK10 is relocated from an ideal position as part of the transition state stabilizing oxyanion hole, while its side chain points to the bottom of the S1 pocket without occluding it (Figure 4). Active serine proteases do not require a Gly193 for stabilization of reaction intermediates, as shown by a mutational analysis of coagulation factor XIa elucidating that mainly β-branched residues in place of Gly193 hamper substrate and inhibitor binding (Schmidt et al., 2004, 2008). This finding is corroborated by the example of Phe193 in the kallikrein-like, plasminogen activating snake venom TSV-PA, which does not bind extended polypeptidic substrates and inhibitors (Parry et al., 1998).

Despite the distortion of the KLK10 active site, which depends mainly on the zymogen-like conformations of the N-terminus and the 191-loop, the S1 pocket is relatively wide open and could accommodate a large side chain such as an Arg with little rearrangements (Figure 5). In the current conformation, the primary specificity determining Asp189 is not in the proper position to bind basic side chains. In case the S1 subsite adopts the standard conformation of a serine protease, the Asp189 at its bottom should prefer P1-Arg and P1-Lys residues, in agreement with the profiling results, which show nearly equal acceptance of both basic residues (Debela et al., 2006b). The unusual KLK10 residues, Pro190 and Ala226, which line the distorted S1 pocket, substitute the common Ser190 and Gly226 of KLKs and explain very well that this S1 subsite additionally accepts aliphatic P1 residues, such as Met or probably Cys (Figure 4). The S2 pocket is relatively wide open: only the flexible Arg97 side chain seems to come close to potential P2 residues. Intriguingly, this residue could be the major determinant of the reported preference for Asp and, to a lesser extent, Glu in the S2 subsite (Debela et al., 2006b). The additional preference for Lys and Arg may depend on Asp99 at the back wall of this pocket. Tyr218 is the residue that defines most of the S3 subsite and may be responsible for the overall unspecific character. The slight preference for acidic P3 residues could depend on Arg96 and Arg97, if the little defined, disordered part of the 99-loop extends over the active site cleft. Finally, S4 was very unspecific in profiling experiments and slightly disfavors basic residues, such as P4-His, Lys and especially Arg, while hydrophobic residues rank higher. The S4 subsite is shaped like in other mature KLKs, with a Trp215 as bottom and a Tyr172 as the western wall viewed in standard orientation.

Figure 5:

Surface representations of KLK10 with electrostatic potentials.

(A) The electrostatic potential is contoured from −10 k_BT/e (red) to +10 k_BT/e (blue). Disordered regions were included for the electrostatics calculation and contribute to an overall positive electrostatic potential (KLK10 pI=8.7), which is strongest around the 75-loop. This region corresponds to the anion binding exosite I of thrombin. The specificity pockets S4 to S1 are largely formed as in other active proteases, with S3 partially occupied by the 99-loop and considerable distortions of S1. (B) Disordered regions were omitted for electrostatics calculation and from the molecular surface representation. The flexible N-terminus and the disordered stretches of the 75-, 99-, and 148-loops are depicted as beige tubes.

Zn²⁺ binding site at the catalytic center

Upon zinc soaking, one site per molecule of the KLK10-dimer becomes visible in the electron density, which is most likely occupied by a Zn²⁺ ion (Figure 6A). An anomalous Fourier map identified unambiguously an anomalous scatterer at both sites, which confirms the presence of a transition metal. The metal cation is situated on the surface of both KLK10 molecules between the two catalytic triad residues, His57 and Asp102, and is furthermore liganded by Asp99, together forming a trigonal first coordination sphere (Figure 6B). In order to bind this Zn²⁺, the His57 side chain slightly moves away from Asp102, which also adjusts its side chain towards the metal ion to complete the Zn²⁺ coordination with Asp99 (Figure 6C). The movement of the imidazolyl side chain of His57 is very subtle: it rotates modestly towards metal cation, not even disconnecting itself from Ser195 (Figure 6D). Average metal ion distances in both KLK10 molecules are 2.1 Å to His57 Nδ1, 1.9 Å to Asp99 Oδ2, and 2.0 Å to Asp102 Oδ2, which is in the range of comparable Zn²⁺ ligand distances in known structures (Figure 6C). According to the recent literature, the ideal distance of Zn²⁺ to His is 2.03 Å, and to monodentate Asp 1.99 Å, while bidentate distances range around 2.2 Å (Harding, 2001; Harding et al., 2010).

Figure 6:

Zn²⁺ binding to the catalytic center of KLK10-Zn.

(A) Both Zn²⁺ sites in the KLK10-Zn dimer, with the metal ions displayed as gray sphere. Zn²⁺ is bound directly in the active site and coordinated by His57, Asp99 and Asp102. The trace of the partially disordered 99-loops is shown in red. (B) Close-up view of the Zn²⁺ site as stereo representation. The 2F_o-F_c electron density map can be contoured upto 7.0 σ, while two strong anomalous peaks are present, which coincide with Zn²⁺ sites. An anomalous Fourier map is depicted as orange grid, contoured at 3.5 σ. (C) The trigonal Zn²⁺ coordination sphere. Average ligand distances of both Zn²⁺ ions are 2.1 Å for His57, 1.9 Å for Asp99 and 2.0 Å for Asp102. (D) Superposition of the KLK10 Zn²⁺ structure (deep purple) and Zn²⁺ free KLK10 (yellow). The catalytic triad undergoes only minor relocations upon Zn²⁺ binding. The hydrogen bond network of the Zn²⁺-free structure is depicted as yellow dots. It should be noted that Ser195 and His57 in both strucutures do not exhibit active triad conformations. The Asp102 conformation is stabilized by the highly conserved Ser214, which has been termed the fourth residue of a catalytic tetrad. Zn²⁺ blocks the triad by interrupting the charge relay system in the region of the hydrogen that is shared by the His57 Nδ1 and the Asp102 Oδ2.

The mechanism of inhibition can be explained by a direct disruption of the catalytic triad, similar to the observations for other KLKs. However, the IC₅₀ of 1.6 μm is the lowest for all investigated Zn²⁺-inhibited KLKs, which suggests that direct inhibition of the catalytic center is more efficient than for the other examples, which require additional conformational changes such as rotation of His57 out of its catalytic triad position and connection to ligands in the 99-loop (Goettig et al., 2012). For example, the Zn²⁺-bound structure of KLK10 is different from the crystal structure of the Zn²⁺-inhibited tonin, where the Zn²⁺ is liganded by the imidazoyl side chains of His99, His97, and of the active site His57, which has rotated out of its normal site between Ser195 and Asp102, to complete the Zn²⁺ coordination (Fujinaga and James, 1987). Thus, tonin is more similar to KLK7 and KLK5, which have His99 at the Asp position representing one of the major ligands for Zn²⁺ binding close to the active site (Debela et al., 2007a,b). However, KLK10 represents the first human serine proteases that holds a metal ion in the core of the active site and involves the two catalytic triad residues, His57 and Asp102. Depending on the respective tissue, the KLK10 activity might be tightly regulated by Zn²⁺.

Purification and characterization of the active protease

Expression, refolding and chromatographic purification of KLK10 have been described before in detail (Debela et al., 2006b; Debela, 2007). KLK10 was expressed in E. coli M15(pREP4) cells, using a pQE30 vector (Qiagen). Inclusion bodies of KLK10 were dissolved in 8 m urea and the unfolded protein harboring an N-terminal His₆-tag was purified via a nickel-nitrilotriacetic acid-Sepharose column. Afterwards, the protein was dialyzed against 4 m urea, 100 mm NaCl, 50 mm Tris/HCl (pH 8.0), and 0.005% Tween-20 for 12 h at room temperature. Refolding took place in 350 mm NaCl, 50 mm Tris/HCl (pH 8.0), 20 mm sucrose, 2 mm CaCl₂, 100 mm L-arginine, 5 mm glutathione (reduced), 0.5 mm glutathione (oxidized), 0.005% Tween-20, 0.05% PEG 3350 for 36 h at 4°C. KLK10 samples were further purified or stored in 100 mm NaCl, 50 mm Tris/HCl (pH 8.0) and 0.005% Tween-20, such as during cleavage of the artificial pro-peptide by enterokinase and p-aminobenzamidine-Sepharose affinity chromatography, in which the bound protease was eluted by increasing PABA concentration.

The identity of KLK10 samples was confirmed by N-terminal sequencing. Moreover, the enzymatic activity of KLK10 was initially measured with Z-LR-AMC and Z-FR-AMC, using 20 nm protease in 150 mm NaCl, 50 mm Tris-HCl (pH 8.0), and 0.005% Tween 20 and 1% DMSO at 25°C. The standard substrate concentration was 250 μm and was varied for determining Michaelis-Menten kinetics. For Zn²⁺ inhibition kinetic measurements with Z-GPR-AMC Tris/HCl buffer (pH 7.0) was employed. Acetylation of the N-terminus was performed with 70 μl samples of 10 nmol KLK10, 18% (v/v) acetic anhydride, 54% (v/v) methanol, 50 mm NH₄(HCO₃)₂ (pH 6.5) for 30 min incubation at room temperature. Afterwards the samples were analyzed by N-terminal sequencing or used in activity assays, respectively.

KLK10 crystallization and data collection

Using the sitting-drop vapor diffusion method, native KLK10 crystals were grown at 18°C from 2 μl protein solution, 10 mg/ml KLK10, which was concentrated from a solution of 30 nm KLK10 with 20 μm PABA, and 2 μl of 100 mm (NH₄)₂SO₄, 100 mm MES (pH 5.6), equilibrated against 500 μl reservoir solution. These crystals belong to the monoclinic space group P2₁ and contain two molecules per asymmetric unit. For preparation of KLK10-Zn crystals, these crystals were soaked for 2 min in mother liquor with 10 mm ZnCl₂. Upon soaking, the crystals remained in monoclinic space group P2₁ with two molecules per asymmetric, but changed their cell constants to some extent (Table 1). Data were collected at the beamline BW6 of the DESY synchrotron at a wavelength of 1.05 Å (Hamburg, Germany).

Model building and refinement

The collected diffraction images were integrated with MOSFLM and scaled with AIMLESS of the CCP4 software package (Leslie and Powell, 2007; Winn et al., 2011; Evans and Murshudov, 2013). Molecular replacement search was performed using the KLK4 chain A (PDB code 2BDG) in PHASER and resulted in the following scores: RFZ=9.0, TFZ=7.2, and LLG=+172 and an R-factor of 55.1 (Mccoy et al., 2007). In the following steps, the sequence of KLK10 (O43240, KLK10_HUMAN) from the UniProtKB database was employed (Uniprot, 2015). Using a model comprising 222 residues of bovine trypsinogen (2TGT) improved these values overall to RFZ=8.5, TFZ=7.7, LLG=+320, and an R-factor of 53.1. The correctness of the solution was checked by inspection of proper packing in the asymmetric unit and the unit cell, respectively, and composite omit maps calculated with the CCP4 software. Molecular replacement with the refined solution for the unbound KLK10 coordinates afforded an RFZ=9.6, a final TFZ=27.9, LLG=+3288 and an R-factor of 37.1 for the KLK10-Zn data.

Iterative model building was performed in MAIN and COOT, followed by refinement in PHENIX (Adams et al., 2010; Emsley et al., 2010; Turk, 2013). The polypeptide main chain was overall well defined for all secondary structure elements, while the N-terminus and several surface loops were largely undefined. R_cryst and R_free of the KLK10 data reached 22.3% and 28.5% at 2.70 Å resolution and for KLK10-Zn 22.0% and 27.8%, at 2.65 Å resolution. For the latter data an anomalous Fourier map was calculated in PHENIX with full model phases and the anomalous differences Δano based on Friedel pairs. More than 90% of the residues of both KLK10 and of KLK10-Zn lies in the favored Ramachandran plot regions and the rest in the allowed regions according to MOLPROBITY (Table 1) (Chen et al., 2010). The coordinates and structure factors were deposited in the protein database under the accession codes 5LPE and 5LPF. RMSD values were calculated using the Gesamt option of the SUPERPOSE software (Krissinel, 2012). All Figures were created with PyMOL (Delano, 2002).