Structure of Human Cytosolic X-prolyl Aminopeptidase

X-prolyl aminopeptidases catalyze the removal of a penultimate prolyl residue from the N termini of peptides. Mammalian X-prolyl aminopeptidases are shown to be responsible for the degradation of bradykinin, a blood pressure regulator peptide, and have been linked to myocardial infarction. The x-ray crystal structure of human cytosolic X-prolyl aminopeptidase (XPN-PEP1) was solved at a resolution of 1.6Å. The structure reveals a dimer with a unique three-domain organization in each subunit, rather than the two domains common to all other known structures of X-prolyl aminopeptidase and prolidases. The C-terminal catalytic domain of XPNPEP1 coordinates two metal ions and shares a similar fold with other prolyl aminopeptidases. Metal content analysis and activity assays confirm that the enzyme is double Mn(II) dependent for its activity, which contrasts with the previous notion that each XPNPEP1 subunit contains only one Mn(II) ion. Activity assays on an E41A mutant demonstrate that the acidic residue, which was considered as a stabilizing factor in the protonation of catalytic residue His498, plays only a marginal role in catalysis. Further mutagenesis reveals the significance of the N-terminal domain and dimerization for the activity of XPNPEP1, and we provide putative structural explanations for their functional roles. Structural comparisons further suggest mechanisms for substrate selectivity in different X-prolyl peptidases.

leukocytes (4), platelets (5), and rat and guinea pig brains (6,7). The membrane-bound form, first purified from porcine kidney (8) and later purified from bovine and rat lungs (1,9), is attached to the lipid bilayer through a glycosylphosphatidylinositol anchor (10). A previous study has shown that injection of apstatin into mice can reduce myocardial infarction severity (11), suggesting a pathological role of X-prolyl aminopeptidases in mammals.
Human XPNPEP1 consists of 623 amino acid residues with a calculated molecular mass of 69,886 Da. The enzyme purified from human leukocytes exists as a dimer of 140 kDa (4). The catalytic activity of the enzyme is enhanced in the presence of Mn 2ϩ (3). Each 70-kDa subunit of the enzyme was thought to contain only one Mn 2ϩ ion in a previous study (3). A Blast search against the NCBI sequence data base revealed similarity in the C-terminal catalytic domains among X-prolyl peptidases of known structures. In contrast, the overall sequence homology between human XPNPEP1 and other X-prolyl peptidases is low. In particular, XPNPEP1 is significantly larger in size than other members of the X-prolyl peptidases (about 40 -50 kDa).
Here we report the crystal structure of human XPNPEP1 at 1.6-Å resolution. Whereas other X-prolyl peptidases with known structures all contain two domains, the XPNPEP1 structure possesses a novel three-domain organization with a conserved C-terminal catalytic domain. In contrast to previous reports, we identified the presence of a double Mn 2ϩ binding site in the catalytic domain, both in the crystal structure and in solution.

EXPERIMENTAL PROCEDURES
Cloning and Expression-The cDNA of wild type (WT) XPN-PEP1 and that of a domain I-truncated mutant (residues 162-623) were cloned into the pET28a vector (Novagen) between SalI and HindIII sites. E41A (i.e. Glu 41 to Ala substitution) and W477E point mutants were produced from the constructed pET28a-wild type XPNPEP1 plasmid with a one-step overlap extension PCR method by using the Easy Mutagenesis System kit (Transgen).
All XPNPEP1 variants were expressed as an N-terminal His 6tagged protein in Escherichia coli BL21(DE3) in LB medium supplemented with 1 mM MnCl 2 (manganese-rich LB). Sel-enomethionine-substituted WT 3 protein was expressed in a metE Ϫ E. coli host strain B834 (Novagen) in the M9 minimal medium supplemented with 50 mg of selenomethionine per liter and 1 mM MnCl 2 .
For expression, E. coli cells cultured overnight were diluted 100-fold in fresh medium and cultured at 37°C to an optical density of about 0.8 at 600 nm. The cell culture was then cooled down to 16°C and induced with 0.5 mM isopropyl ␤-D-1-thiogalactopyranoside. It was grown for another 20 h at 16°C with shaking at 220 rpm, and then the cells were harvested by centrifugation.
Protein Purification-The harvested cells were resuspended in buffer A (20 mM Tris-HCl (pH 7.9), 500 mM NaCl, and 10% (v/v) glycerol) and lysed by sonication. The released His 6 -tagged protein was purified following standard protocols of nickel-nitrilotriacetic acid resin (Qiagen). It was eluted from the resin with buffer B (20 mM Tris-HCl (pH 7.9), 500 mM NaCl, 10% (v/v) glycerol, and 300 mM imidazole) and dialyzed against a salt-free buffer (20 mM Tris-HCl (pH 8.0)). Further purification was performed with a Hitrap Q HP affinity column (Amersham Biosciences) and the final protein sample was dialyzed against a buffer of 20 mM Tris-HCl (pH 8.0) and 20 mM NaCl.
Crystallization-Crystals of native or selenomethionyl-labeled protein were grown by the hanging-drop vapor-diffusion method. The reservoir contained 20% (v/v) polyethylene glycol (PEG) 400, 0.15 M CaCl 2 , and 100 mM HEPES (pH 7.5). A typical hanging drop consisted of 2 l of protein solution (20 mg/ml) mixed with 2 l of the reservoir solution. Large (over 0.5 mm) colorless block-shaped crystals suitable for diffraction were grown within a week at 16°C.
Data Collection, Phasing, and Model Refinement-Crystallographic data from the crystals of native and selenomethionyllabeled protein were collected on beamlines BL5A and BL17A of the Photon Factory synchrotron facility (KEK, Tsukuba, Japan). The diffraction images were integrated and scaled using HKL2000 (12). A 3.5-Å resolution structure of selenomethionyl-labeled XPNPEP1 was solved by the multiwavelength anomalous diffraction phasing method at the selenium absorption edge using SOLVE/RESOLVE (13). The 3.5-Å resolution phases were extended to 1.6 Å using the native data set and the program ARP/WARP (14). The program automatically built about 80% of residues, and the remainder were built manually with COOT (15). The structure was refined with REFMAC5 (16). Figures were drawn with the program PYMOL (DeLano Scientific, San Carlos, CA).
Enzyme Activity Assays-The enzyme activity was assayed with bradykinin (Phoenix Pharmaceuticals, Inc.) or the Arg-Pro-Pro tripeptide (synthesized by SBS Genetech, Beijing, China) as the substrate in 100 mM Tris-HCl (pH 8.0) and 100 mM NaCl at 37°C for 5 min (100 l final volume). The free amino acid released by the enzyme was detected with the o-phthalaldehyde, 5-mercaptoethanol reagent as previously described (9). Fluorescence of the o-phthalaldehyde derivate was measured with a microplate reader (Type 374, Thermo Electron Corporation). For kinetic analysis, assays were prepared with a range of concentrations of bradykinin (0.01-0.10 mM) or Arg-Pro-Pro (0.02-0.20 mM) and 1 g of the purified WT enzyme. To examine the effects of different factors on the enzyme activity, 0.09 mM Arg-Pro-Pro and 1 g of purified enzyme variant were used in a 100-l assay. To verify the effect of EDTA, purified WT enzyme was incubated in 100 mM Tris-HCl (pH 8.0) and 100 mM NaCl with 50 mM EDTA for 10 min, and dialyzed against 100 mM Tris-HCl (pH 8.0) and 100 mM NaCl prior to measuring its relative activity.
Other Assays-Analysis of the total metal content was carried out using inductively coupled plasma mass spectrometry (ICP-MS, Thermo) at the Tsinghua University Analysis Center (Beijing, China). Purified protein samples without crystallization trial were extensively dialyzed against 100 mM Tris-HCl (pH 8.0) and 100 mM NaCl before ICP-MS analysis.
Analytical ultracentrifugations (AUC) were performed with the sedimentation velocity method at 58,000 rpm at the Institute of Biophysics, Chinese Academy of Sciences (Beijing, China). The protein samples for AUC were prepared at a concentration of about 0.5 mg/ml in a buffer of 100 mM Tris-HCl (pH 8.0) and 100 mM NaCl. The AUC data were processed as a c(M) distribution model (17).

RESULTS
Structure Determination and Refinement-The structure of human XPNPEP1 was solved at 1.6-Å resolution using the multiwavelength anomalous diffraction method. Refinement of the XPNPEP1 structure resulted in a final model with a crystallographic R-factor (R cryst ) of 0.154 and a free R-factor (R free ) of 0.195. One asymmetric unit of this C222 1 crystal form contained a single protein molecule composed of 623 amino acid residues. The polypeptide chain was complete with the exception of the N-terminal His 6 tag, the N-terminal residues 1-2, and the C-terminal residues 620 -623, which were not included in the final model because of poor electron density. For the same reason, Asn 553 , Arg 554 , and the side chain of Phe 509 were assigned zero occupancy. All other residues had excellent electron density, and the final average temperature factor (B) was 27.2 Å 2 . An asymmetric unit also contained 1,055 ordered water molecules, one partial PEG molecule containing six ethylene glycol residues, and four metal ions. Of the four metal ions, two major ones localized at the active site were refined as full occupancy Mn 2ϩ with temperature factors of 14.8 Å 2 (Mn1 in Fig.  1A) and 17.7 Å 2 (Mn2), respectively. The two minor ions were refined as full occupancy Ca 2ϩ and Na ϩ , respectively. Although crystals were grown in 0.15 M CaCl 2 , native XPN-PEP1 treated with EDTA (followed by dialysis before crystallization) yielded no crystal but thick precipitation under the same condition, suggesting that the two active-site metal ions observed in the native crystal protein were unlikely to be calcium ions replacing Mn 2ϩ during crystallization. Besides the mobile Asn 553 , only Glu 434 , affected by the inter-molecule Ca 2ϩ , localizes in the Ramachandran unfavorable region. Experimental structure factors and the coordinates of the refined model have been deposited in the Protein Data Bank (PDB) with access code 3CTZ. Crystallographic statistics are summarized in Table 1.
Domains I and II are primarily held together by hydrophobic interactions. Domains II and III are linked by the residues between helix ␣13 and helix ␣14 (residues 321-323).
Homodimer-Our solution studies, including gel filtration and AUC, revealed that XPNPEP1 proteins primarily exist as 140-kDa dimers (Fig. 3), which is consistent with previous reports (3,4). In the crystal structure, two symmetry-related XPNPEP1 molecules (named as subunits A and B) are related by a dyad to form a homodimer (Fig. 1B). The two subunits are mainly held together by hydrophobic interactions (Fig. 1C). The side chains of Tyr 439 , Leu 481 , Leu 484 , and Tyr 526 in subunit A and Tyr 549 , Phe 551 in the ␤27-␤28 hairpin of subunit B form one hydrophobic core. The symmetry equivalent hydrophobic residues form the second hydrophobic core of the dimer interface. Residue Pro 460 and the side chains of Phe 459 , Leu 468 , Phe 471 , and Trp 477 in subunit A, together with their symmetry equivalents, form a third hydrophobic core. In addition to these hydrophobic interactions, a salt bridge between Glu 442A (i.e. Glu 442 of subunit A) and Lys 548B and their symmetric counterparts, together with two pairs of hydrogen bonds between Glu 442A/B and Tyr 549B/A , and between Leu 467A/B and Ser 470B/A , also help to stabilize the interaction between the two subunits. Approximately 1,600 Å 2 (6%) of the solvent accessible surface area from each subunit is buried upon dimer formation. Among the above discussed residues, Trp 477 plays a vital role for dimerization. Mutating this Trp to Glu abolished the capability of the enzyme to form a native dimer in our AUC studies (Fig. 3). Nevertheless, a small peak appeared at the position of 120 kDa in the AUC studies on the W477E mutant. We speculate that it represents a fraction of monomer or dimer with abnormal molecule shapes.
is the intensity of an individual measurement of the reflection, and ͗I i (h)͘ is the mean intensity of the reflection. b R cryst ϭ ⌺(ʈF obs ͉ Ϫ ͉F calc ʈ)/⌺͉F obs ͉, where F obs and F calc are the observed and calculated structure-factor amplitudes, respectively. c R free was calculated as R cryst using the reflections in a test set not used for structure refinement, which is a randomly selected subset containing 5% of unique reflections. d Calculated using MolProbity. Numbers reflect the percentage of residues in the preferred, allowed, and disallowed regions, respectively.
Active Site-The putative active site is located in the inner (concave) surface of the curved ␤-sheets of domain III (Fig. 1A) on the basis of comparison with homologous structures, such as the structure of E. coli AP-P (20 -22). Two well coordinated metal ions were observed in this active site (Fig. 4). ICP-MS data consistently indicated that the molar ratio between Mn 2ϩ and the 70-kDa XPNPEP1 subunit was 1.79:1, whereas the content of other common metals was negligible ( Table 2). Although our crystallized XPNPEP1 protein sample was expressed with manganese-rich LB media and appeared clear, we carried out similar expression of the protein in plain LB media (i.e. without manganese supplementation) and interestingly obtained some "red protein." ICP-MS analysis on this red protein sample indicated that the protein contains mainly iron (0.73:1), manganese (0.70:1), and magnesium (0.41:1) ions (Table 2). Therefore, the molar ratio between the total metal ion content and the protein remained close to 2:1.
In our crystal structures, both Mn 2ϩ ions are well coordinated. One of the Mn 2ϩ ions (termed Mn1, see Fig. 4) is coordinated by the O␦-1 atoms of Asp 415 (2.15 Å) and Asp 426 (2.14 Å), the O⑀-1 atom of Glu 537 (2.19 Å), and two water molecules (termed W1 and W2) with Mn 2ϩ -ligand distances of 2.23 and 2.27 Å, respectively. These Mn 2ϩ -ligands form an approximate trigonal-bipyramidal coordination geometry, with the O␦-1 atoms of the two aspartate residues and W1 in the equatorial plane, and the O⑀-1 atom of the glutamate residue and W2 on the axis. The coordination sphere of the second Mn 2ϩ ion (termed Mn2) is comprised of the O␦-2 atom of Asp 426 (2.37 Å), the O⑀-2 atoms of Glu 523 (2.27 Å) and Glu 537 (2.26 Å), N⑀-2 atom of residue His 489 (2.27 Å), and two water molecules (W1, 2.31 Å and W3, 2.17 Å), which complete a distorted octahedral coordination. Furthermore, W1 and the carboxylate groups of Asp 426 and Glu 537 act as bridges between the two Mn 2ϩ ions. The side chains of His 395 , His 485 , His 498 , and Glu 41 surrounding the two Mn 2ϩ ions are likely to play roles in recognition and catalysis during the substrate hydrolysis, according to studies on the equivalent residues in the active site of E. coli AP-P (22), which shares an almost identical active site with XPNPEP1.
Activity Assay-Activity assays on the same enzyme protein sample used for the crystallization were performed with the tripeptide Arg-Pro-Pro and bradykinin as substrates. The assays were performed in Mn 2ϩ -free buffer and gave a K m value of 308 (Ϯ8) M and a k cat of 7.7 s Ϫ1 on Arg-Pro-Pro, whereas K m was measured as 78 (Ϯ9) M and k cat as 3.8 s Ϫ1 for bradykinin. Our K m and k cat values on bradykinin were comparable with previous reports (3,5). Meanwhile, the red enzyme (enzyme expressed from plain LB) showed 44% activity of the Mn 2ϩbound enzyme, and the activity of the enzyme dropped to 10% after treatment with 50 mM EDTA.
Although not directly involved in metal binding, Glu 41 is the only residue located outside of the catalytic domain but close to the active site in the three-dimensional structure (Fig. 4). To test its function in catalysis, we made a mutant substituting Glu 41 with Ala. This mutant maintained 91% of the WT activity, suggesting that it only marginally affects the activity of the enzyme. In contrast, another mutant, W477E, designed to block dimer formation and a domain I truncation mutant showed only 6 and 2% of the WT activity, respectively (Fig. 5).

XPNPEP1 Contains Two Mn 2ϩ
Ions in the Active Site-Prior to our structural study, Cottrell and colleagues (3) had reported some significant work on the characterization of recombinant XPNPEP1. In particular, they assayed the effects of different metal ions and chelating agents on the activity of XPNPEP1,  and identified the enzyme as Mn 2ϩ dependent for its activity. It was based on this observation that we chose manganese-rich LB in our experiments to avoid Mn 2ϩ depletion during protein expression. Nevertheless, their conclusion that each 70-kDa XPNPEP1 subunit contains only one metal ion was markedly different from the observation in our XPNPEP1 structure. The XPNPEP1 crystal structure displays two well coordinated metal ions with very strong electron density in each active site. The electron density map for the two metal ions can be observed even when contoured at the level of ϩ20 standard deviations. To further investigate the metal content of the enzyme, an ICP-MS analysis on the XPNPEP1 recombinant protein expressed from manganese-rich LB was performed and revealed that the enzyme contains 1.79 Mn 2ϩ ions per 70-kDa subunit and much lower levels of other metals. This result indicates the molar ratio between Mn 2ϩ and the enzyme is nearly 2:1, which is entirely consistent with our structural observation.
To study the effects of Mn 2ϩ from the cell culture medium on the metal content of recombinant XPNPEP1, we expressed the enzyme in plain LB medium following the method of Cottrell and colleagues (3). ICP-MS analysis of this sample revealed three major metal contents, i.e. magnesium, manganese, and iron ions in the enzyme with the molar ratio 0.41, 0.70, and 0.73 per 70-kDa subunit. Although the molar ratio between Mn 2ϩ and protein dropped to a level comparable with the previous report (0.99:1), the molar ratio between the total metal content and protein remains nearly 2:1, which is consistent with our structural observation. The detailed differences in metal content between our study and that of Cottrell et al. (3) may be due to different expression conditions, including temperature and induction time. For example, we used a "slow" expression method at 16°C for 20 h. In contrast, their method was a quick one at 40°C for 3 h. We believe that slow cell growth allows the recombinant protein to fold properly and to obtain the optimum metal ion content. Moreover, in our case, the difference in metal composition between XPNPEP1 samples expressed from plain and manganese-enriched LB medium was probably caused by the depletion of Mn 2ϩ in the former, because the protein yields from both media were as high as 30 mg/liter of E. coli culture.
The Mn 2ϩ content of XPNPEP1 expressed from plain LB was 39% (0.70:1.79) of that from manganese-rich LB and is fairly close to their activity ratio (44%). This suggests that only the portion of XPNPEP1 in which two Mn 2ϩ ions are coordinated possesses catalytic activity, whereas the remaining portion, which coordinates either two magnesium or iron ions, presents little or no catalytic activity. Here, we made an assumption that XPNPEP1 preferentially binds the same type of ions in the two metal-binding sites, based on available structural evidence. Alternatively, Mn 2ϩ is essential only for one of the two metalbinding sites; the other one is more tolerant to miscellaneous binding. Such a possibility remains to be further verified. Meanwhile, our result is comparable with previous data on the effect of divalent cations (3). Treatment of XPNPEP1 with EDTA greatly reduced its catalytic activity, both in our results and those of Cottrell and colleagues (3). We conclude that XPN-PEP1 is a member of the double metal ion-dependent X-prolyl aminopeptidase, and coordinating two Mn 2ϩ ions in its active site is most favorable for its activity.
X-prolyl Peptidases Have a Conserved Active Site and Catalytic Mechanism-Results from a BLAST search against the PDB sequences in the NCBI data base revealed weak sequence conservation between domain III of XPNPEP1 and the catalytic domains of other X-prolyl peptidases, including E. coli AP-P (PDB code 1wl9), X-prolyl dipeptidases (prolidase) from Pyrococcus furiosus (1pv9), prolidase from Pyrococcus horikoshii OT3 (1wy2), and human prolidase (2iw2). In contrast, no significant sequence homology is found for the two XPNPEP1 N-terminal domains (domain I and II). Sequence alignment of the conserved catalytic domain regions of these X-prolyl peptidases, together with the corresponding region of human XPN-PEP2 revealed 11% identity and 38% similarity for the enzymes (Fig. 6). Although the sequence homology among these X-prolyl peptidases is marginal, residues for chelating metal ions are absolutely conserved in all six listed enzymes.
Superposition of structures of the five available active sites shows that there is a close agreement between the coordination geometries (Fig. 7). The ligand residues (i.e. Asp 415 , Asp 426 , His 489 , Glu 523 , and Glu 537 ) and most of the second shell of surrounding residues (e.g. His 395 , His 485 , and His 498 ) are conserved in both primary and three-dimensional structures.  Among these X-prolyl peptidases, both human XPNPEP1 and E. coli AP-P are X-prolyl aminopeptidases (EC 3.4.11.9) with the same substrate specificity, and their activities require Mn 2ϩ ions. The two Mn 2ϩ ions in both XPNPEP1 and E. coli AP-P are five-coordinated (Mn1) and six-coordinated (Mn2), respectively, with the same amino acid ligands in the two structures. Superposition of the active site of XPNPEP1 with that of E. coli AP-P in complex with a product Pro-Leu dipeptide (PDB code 1a16, Fig. 7A) reveals that the positions of the Mn 2ϩ ions, all ligand residues, and water molecules are identical. Only the side chain of His 395 in XPNPEP1 exhibits a minor change from the equivalent His 243 in E. coli AP-P. The structure and catalytic mechanism of E. coli AP-P have been extensively studied (20 -22). Based on the close similarity of the active sites between XPNPEP1 and E. coli AP-P, they likely share the same catalytic mechanism. In E. coli AP-P His 243 is proposed to form a hydrogen bond with the carbonyl of the proline residue in the substrate to stabilize its binding at the active site. The side chain of the corresponding His 395 in XPNPEP1, which shows a higher temperature factor than the other active site residues and a non-functional rotamer, may regain its active conformation upon substrate binding.
In XPNPEP1, the O⑀-1 atom on the side chain of Glu 41 forms a hydrogen bond with the N␦-1 atom of His 498 . Thus, the side chain of Glu 41 points toward the active site. Equivalent residues can be found in E. coli AP-P (Asp 38 and His 361 , Fig. 7A). In a previous study of E. coli AP-P, mutation of His 361 to Ala caused Asp 38 to change its 1 rotamer and move away from the active site (22). These data provide evidence that an interaction exists between the two residues, and it was consequently hypothesized that the imidazole ring of His 361 can be doubly protonated by the carboxylate group of Asp 38 . However, in our study of XPNPEP1, substitution of Glu 41 with Ala did not significantly reduce its activity (Fig. 5), indicating that Glu 41 is dispensable for the activity. Thus, its role in double protonation remains to be verified. Interestingly, both of the two prolidases from archaeobacteria species lack a residue equivalent to Glu 41 in XPNPEP1 (Fig. 7B), arguing from a different perspective that this acidic residue is dispensable for catalytic activity.
XPNPEP1 Has a Novel Three-domain Structure-Although sequence alignment and structure superposition reveal that XPNPEP1 shares a conserved catalytic C-terminal domain and active site with other X-prolyl peptidases, it features a novel overall structure that distinguishes it from other known members of the X-prolyl peptidase family. Although all other X-prolyl peptidases were found to have only two domains in each subunit, namely one N-terminal domain and one C-terminal catalytic domain, XPNPEP1 uniquely contains three domains in each subunit: an N-terminal domain, a middle domain, and a C-terminal catalytic domain ( Fig. 1A and supplemental materials Fig. S1, top panel). Comparing the XPNPEP1 subunits with those of the canonical X-prolyl peptidases, we discovered that domains II and III of XPNPEP1, respectively, occupy positions similar to the two domains in the canonical FIGURE 6. Multiple sequence alignment of catalytic domains from X-prolyl peptidases. From top to bottom, the sequences are from human XPNPEP1, human XPNPEP2 (GenBank TM number NP_003390), human prolidase (PDB code 2iw2), aminopeptidase P from E. coli (PDB code 1a16), prolidase from P. furiosus (PDB code 1pv9), and prolidase from P. horikoshii OT3 (PDB code 1wy2). Residues for chelating metal ion are highlighted with a blue background; other identical residues with a red background; and conserved residues with a yellow background. Hydrophobic residues on the dimer interface of XPNPEP1 and their equivalent residues on XPNPEP2 are colored green. The conserved region was defined by NCBI BLAST. Sequences were aligned using the program ClustalX, and the alignment was presented using the online ESPript server.
X-prolyl peptidase dimer. However, domain I in XPNPEP1 appears to be additional. As described above, Glu 41 is the only residue in domain I that interacts with the active site residues, yet it plays only a marginal role in catalysis, such that domain I appears not to participate in catalysis. However, our activity assays for a domain I truncation mutant of XPNPEP1 showed almost no catalytic activity (Fig. 5), indicating a crucial role of this domain. To investigate a possible role of this domain, we made a comparison between the XPNPEP1 subunit and canonical X-prolyl peptidase dimers (3, 20, 23, 24) (e.g. the E. coli AP-P dimer, which has 2100-Å 2 solvent accessible surface buried in the dimer interface). The position of domain I of XPNPEP1 is comparable with the positions of neighboring N-terminal domains in canonical X-prolyl peptidase dimers, and is especially similar to the two prolidases from archaeobacteria species (supplemental materials Fig. S1). Further superposition of XPNPEP1 domains I and II reveals that their core regions are similar ( Fig. 2A). Superposition of domains I or II with the N-terminal domain of E. coli AP-P ( Fig. 2B; 2.74 Å root mean square deviation for 124 residues between domain I and the N-terminal domain of E. coli AP-P, and 2.41 Å root mean square deviations for 107 residues between domain II and the N-terminal domain of E. coli AP-P) and other X-prolyl peptidases all indicate structure similarities in their core regions, albeit with no obvious sequence homology. Interestingly domains I and II in XPNPEP1 are related by a pure rotation of 150°, which is similar to but distinct from the dyad symmetry relationship between the two N-terminal domains in canonical X-prolyl peptidase dimers. These observations provide hints that a role of domain I in XPNPEP1 is, at least partially, to mimic the neighboring N-terminal domains in canonical X-prolyl peptidases dimers, so that the enzyme can maintain its active site pocket (Fig. 8).
Although all X-prolyl peptidases, including XPNPEP1, can form symmetrical homodimers, the dimerization of XPNPEP1 is different from those previously reported. In particular, the two XPNPEP1 subunits are arranged parallel to each other, whereas they are nearly perpendicular to each other in the canonical X-prolyl peptidase dimer (Figs. 1B and 8, and supplemental Fig. S1, lower panel). Although structural comparison between an XPNPEP1 subunit and other X-prolyl peptidase dimers suggest an isolated XPNPEP1 subunit might have full activity, our dimer blocking mutant of XPNPEP1, W477E, exhibited little catalytic activity (Fig. 5). We speculate that dimerization of XPNPEP1 plays a significant role in maintaining the correct fold of the enzyme. Interestingly, in all the other X-prolyl peptidases, each active site pocket is directly opened to solvent. However, in the case of the XPNPEP1 dimer, the active site pocket in each protomer is opened toward the partner protomer (Fig. 1B, face view). The result is that substrates have to access the active site from the gap between the two subunits. This feature may endow XPNPEP1 with selectivity for its favored substrate. In our kinetic assays for the nine-residue sub- FIGURE 7. Active site comparison with other X-prolyl peptidases. A, stereo view showing the superposition of active sites from XPNPEP1 (yellow) and E. coli AP-P (green, PDB code 1a16) in complex with a product Pro-Leu dipeptide (orange) and human prolidase (blue, 2iw2). B, stereo view showing the superposition of active sites from XPNPEP1 (yellow) and prolidases from P. furiosus (green, 1pv9) and P. horikoshii OT3 (blue, 1wy2). Only the residues and atoms from XPNPEP1 are labeled. strate bradykinin (sequence: Arg-Pro-Pro-Gly-Phe-Ser-Pro-Phe-Arg) and its N-terminal tripeptide substrate Arg-Pro-Pro, XPNPEP1 showed a K m for bradykinin almost four times smaller than the K m for Arg-Pro-Pro. The stronger binding for the longer peptide substrate suggests the presence of additional binding sites that help its N terminus to access the active site.
Different Substrate Selectivity between X-prolyl Aminopeptidases and Prolidases May Be Due to Distinct Features of Their Active Site Pocket Entrance-All of the X-prolyl peptidases with published structures share nearly identical active sites, suggesting that they all share the same catalytic mechanism. However, X-prolyl aminopeptidases use long peptides with a penultimate proline residue as their substrate, whereas prolidases allow only X-Pro dipeptides. Comparing their active site pockets, we find that the pockets of X-prolyl aminopeptidases are shallow and wide. In contrast, the pockets of prolidases are much deeper and more twisted, such that long peptide substrates would be unable to reach their active sites.
The structure of E. coli AP-P in complex with a product Pro-Leu dipeptide molecule provides a good model for studying their different substrate selectivity on tripeptides among X-prolyl aminopeptidases (supplemental material Fig. 2S). From superposition of their active sites, no residue is found in XPNPEP1 that would have a steric conflict with the Leu residue of the Pro-Leu product. However, the side chain of Arg 399 in human prolidase appears to largely occupy the position of the Leu residue in the dipeptide product, and thus the tripeptide substrate is predicted to be unfavorable as a substrate for the prolidase. Similarly, the side chains of Ser 281 and Arg 295 of prolidase from P. furiosus (Ser 284 and Arg 298 of prolidase from P. horikoshii OT3) may cause steric hindrance with the Leu residue of the product.
XPNPEP2 May Be Another Three-domain X-prolyl Aminopeptidase-Mammals are known to contain two X-prolyl aminopeptidases: one is cytosolic (XPNPEP1) and the other is membrane bound (XPNPEP2). The XPNPEP2 consists of an N-terminal signal peptide to direct its translocation into endoplasmic reticulum and a C-terminal glycosylphosphatidylinositol anchor sequence (10,25). The human XPNPEP2 consists of 674 residues and shares 42% sequence identity and 61% positives with XPNPEP1. Aside from the active site residues, nearly all of the hydrophobic residues that contribute to dimerization in XPNPEP1 have their equivalent residues in XPN-PEP2 (Fig. 6). Based on its size and sequence similarity with XPNPEP1, we predict that XPNPEP2 shares the same threedomain X-prolyl aminopeptidase fold and dimerization as XPNPEP1.
Both XPNPEP1 and XPNPEP2 can hydrolyze bradykinin containing a penultimate proline residue and are inhibited by the same specific inhibitor, apstatin (1-3). Bradykinin is a small vasoactive peptide involved in a variety of biological processes (26,27). Experiments performed in mice in vivo showed that administration of apstatin can reduce myocardial infarction severity (11). Based on the structure of XPNPEP1, designing or screening for new chemical inhibitors for XPNPEP1 and XPN-PEP2 may result in novel therapeutic approaches for the prevention of myocardial infarction.