Crystal Structure of Neuropsin, a Hippocampal Protease Involved in Kindling Epileptogenesis

Abstract

Neuropsin is a novel serine protease, the expression of which is highly localized in the limbic areas of the mouse brain and which is suggested to be involved in kindling epileptogenesis and hippocampal plasticity. The 2.1-Å resolution crystal structure of neuropsin provides the first three-dimensional view of one of the serine proteases highly expressed in the nervous system, and reveals a serine protease fold that exhibits chimeric features between trypsin and nerve growth factor-γ (NGFγ), a member of the kallikrein family. Neuropsin possesses anN-glycosylated “kallikrein loop” but forms six disulfide bonds corresponding to those of trypsin. The ordered kallikrein loop projects proline toward the active site to restrict smaller residues or proline at the P2 position of substrates. Loop F, which participates in forming the S3/S4 sites, is similar to trypsin rather than NGFγ. The unique conformations of loops G and H form an S1 pocket specific for both arginine and lysine. These characteristic loop structures forming the substrate-binding site suggest the novel substrate specificity of neuropsin and give a clue to the design of its specific inhibitors. Neuropsin is a novel serine protease, the expression of which is highly localized in the limbic areas of the mouse brain and which is suggested to be involved in kindling epileptogenesis and hippocampal plasticity. The 2.1-Å resolution crystal structure of neuropsin provides the first three-dimensional view of one of the serine proteases highly expressed in the nervous system, and reveals a serine protease fold that exhibits chimeric features between trypsin and nerve growth factor-γ (NGFγ), a member of the kallikrein family. Neuropsin possesses anN-glycosylated “kallikrein loop” but forms six disulfide bonds corresponding to those of trypsin. The ordered kallikrein loop projects proline toward the active site to restrict smaller residues or proline at the P2 position of substrates. Loop F, which participates in forming the S3/S4 sites, is similar to trypsin rather than NGFγ. The unique conformations of loops G and H form an S1 pocket specific for both arginine and lysine. These characteristic loop structures forming the substrate-binding site suggest the novel substrate specificity of neuropsin and give a clue to the design of its specific inhibitors. Proteases have been shown to play essential roles in the nervous system, including those of neurite outgrowth (1Monard D. Trends Neurosci. 1988; 11: 541-544Abstract Full Text PDF PubMed Scopus (293) Google Scholar), neural degeneration (2Tsirka S.E. Gualandris A. Amaral D.G. Strickland S. Nature. 1995; 377: 340-344Crossref PubMed Scopus (590) Google Scholar), and synaptic plasticity (3Liu Y. Fields R.D. Festoff B.W. Nelson P.G. Proc. Natl. Acad. Sci. U. S. A. 1994; 91: 10300-10304Crossref PubMed Scopus (138) Google Scholar). These actions are thought to be mediated by the proteolytic cleavage of zymogen precursors, the activation of specific cell surface receptors, or the degradation of extracellular matrix proteins (4McGuire P.G. Seeds N.W. Neuron. 1990; 4: 633-642Abstract Full Text PDF PubMed Scopus (88) Google Scholar). Neuropsin was cloned from a mouse hippocampal cDNA library using sequences for key regions of the serine protease domain of nerve growth factor (NGF)-γ 1The abbreviations used are: NGF, nerve growth factor; tPA, tissue plasminogen activator; GlcNAc, N-acetylglucosamine; STI, soybean trypsin inhibitor; MCA, 4-methylcoumaryl-7-amide; MSP, myelencephalon-specific protease; r.m.s., root mean square. (5Chen Z.-L. Yoshida S. Kato K. Momota Y. Suzuki J. Tanaka T. Ito J. Nishino H. Aimoto S. Kiyama H. Shiosaka S. J. Neurosci. 1995; 15: 5088-5097Crossref PubMed Google Scholar). In the brain, no NGFγ has been identified so far, though NGFβ is present. Neuropsin is one of the serine proteases highly expressed in the nervous system (6Gschwend T.P. Krueger S.R. Kozlov S.V. Wolfer D.P. Sonderegger P. Mol. Cell. Neurosci. 1997; 9: 207-219Crossref PubMed Scopus (105) Google Scholar, 7Scarisbrick I.A. Towner M.D. Isackson P.J. J. Neurosci. 1997; 17: 8156-8168Crossref PubMed Google Scholar, 8Yamashiro K. Tsuruoka N. Kodama S. Tsujimoto M. Yamamura Y. Tanaka T. Nakazato H. Yamagichi N. Biochim. Biophys. Acta. 1997; 1350: 11-14Crossref PubMed Scopus (165) Google Scholar). The expression of neuropsin is localized at highest concentration in the hippocampus and the amygdala, which are important for acquisition of memory and emotional memory, respectively. This localization is in contrast to that of tissue plasminogen activator (tPA), which is well documented to play a crucial role in the nervous system by mediating plasticity but is distributed more uniformly across the other brain regions and throughout other organs (9Qian Z. Gilbert M.E. Colicos M.A. Kandel E.R. Kuhl D. Nature. 1993; 361: 453-457Crossref PubMed Scopus (644) Google Scholar). Activity-dependent changes in expression of neuropsin have been observed upon direct hippocampal stimulation and induction of kindling, which is a model for epilepsy and neuronal plasticity characterized by the progressive development of electrographic and behavioral seizures (5Chen Z.-L. Yoshida S. Kato K. Momota Y. Suzuki J. Tanaka T. Ito J. Nishino H. Aimoto S. Kiyama H. Shiosaka S. J. Neurosci. 1995; 15: 5088-5097Crossref PubMed Google Scholar, 10Okabe A. Momota Y. Yoshida S. Hirata A. Ito J. Nishino H. Shiosaka S. Brain Res. 1996; 728: 116-120Crossref PubMed Scopus (57) Google Scholar). A single intraventricular injection of monoclonal antibodies specific to neuropsin reduces or eliminates the epileptic pattern (11Momota Y. Yoshida S. Ito J. Shibata M. Kato K. Sakurai K. Matsumoto K. Shiosaka S. Eur. J. Neurosci. 1998; 10: 760-764Crossref PubMed Scopus (67) Google Scholar). Moreover, oxidative stress is shown to effect the expression of neuropsin in the limbic areas, which might be related to the disturbance in shock-avoidance learning of mice (12Akita H. Matsuyama T. Iso H. Sugita M. Yoshida S. Brain Res. 1997; 769: 86-96Crossref PubMed Scopus (21) Google Scholar). These activity-dependent changes and the specific localization of neuropsin indicate the involvement of this protease in hippocampal plasticity and its pathogenesis. Knowledge of the three-dimensional structure of neuropsin provides clues to the biological activity of this protease and also is important to the design of inhibitors that might be useful in treatment of pathological conditions such as epilepsy. Neuropsin was over-expressed in baculovirus-infected High Five insect cells, purified, and crystallized as described previously (13Kishi T. Kato M. Shimizu T. Kato K. Matsumoto K. Yoshida Y. Shiosaka S. Hakoshima T. J. Struct. Biol. 1997; 118: 248-251Crossref PubMed Scopus (11) Google Scholar). The resulting sample and the crystallized protein were verified with N-terminal analysis using an Applied Biosystem automatic analyzer 476A. Neuropsin has a putative glycosylation site at Asn95 of the kallikrein loop. Time-of-flight mass spectroscopy with PerSeptive JMS-ELITE matrix-associated laser desorption/ionization time-of-flight indicated its heterogeneous glycosylation. Two-dimensional high performance liquid chromatography mapping (14Tomiya N. Awaya J. Kurono M. Endo S. Arata Y. Takahashi N. Anal. Biochem. 1988; 171: 73-90Crossref PubMed Scopus (390) Google Scholar) revealed that theN-glycans contained 89% paucimannosidic structures with and without attached fucose residue(s) at the innermost GlcNAc residues but the glycosylation pattern exhibited high heterogeneity as found on many glycoproteins (15Friedrich A. Trends Glycosci. Glycotech. 1996; 40: 101-114Google Scholar). The detailed procedures and obtained structures will be described elsewhere. The crystals belong to space groupP1 (a = 38.15 Å, b = 54.95 Å, c = 64.29 Å, α = 95.72°, β = 90.03°, and γ = 110.29°). X-ray diffraction data were collected with Rigaku imaging plate area-detectors, R-Axis IV and R-Axis IIc, using Cu-Kα radiation and also with a Weissenberg camera at the BL-18B beamline station of the Photon Factory, Tsukuba using 1-Å radiation. Intensities were evaluated with the program DENZO/SCALEPAK (16Otwinowski Z. Minor W. Methods Enzymol. 1996; 276: 307-326Crossref Scopus (38617) Google Scholar), which yielded 25,778 independent combined reflections corresponding to 90.7% completeness at 2.1-Å resolution (74.5% in the highest resolution bin), an R merge of 6.0% (19.8%), a mean ratio of intensity, and ς of 8.5 (3.1). Initial phases were calculated by molecular-replacement with the program AMoRe (17Navaza J. Acta Crystallogr. Sec. A. 1994; 50: 157-163Crossref Scopus (5030) Google Scholar) using a search model based on the structure of bovine pancreatic β-trypsin (PDB code 4PTP) (18Finer-Moore J.S. Kossiakoff A.A. Hurley J.H. Earnest T. Stroud R.M. Proteins Struct. Finct. 1992; 12: 203-222Crossref PubMed Scopus (110) Google Scholar). Rigid body refinements of the searched model were performed with the program X-PLOR (19Brunger A.T. Krukowski A. Erickson J.W. Acta Crystallogr. Sec. A. 1990; 46: 585-593Crossref PubMed Scopus (600) Google Scholar), followed by density averaging/histogram matching with the program DM (20Cowtan K.D. Main P. Acta Crystallogr. Sec. D. 1996; 52: 43-48Crossref PubMed Scopus (289) Google Scholar). Six regions of insertions and deletions were inspected on the resulting 2Fo − Fc map, which was generated with the program O (21Jones T.A. Zou J.-Y. Cowan S.W. Kjeldgaard M. Acta Crystallogr. Sec. A. 1991; 47: 110-119Crossref PubMed Scopus (13014) Google Scholar). The structure was built and refined through alternating cycles using the programs O and X-PLOR, respectively. The kallikrein loop had large but mostly poor density connects to the side chain of Asn95. After several cycles of refinements incorporating solvent water molecules located at the regions other than the kallikrein loop, we defined one residue ofN-acetylglucosamine (GlcNAc) residue bonded to Asn95, as well as weaker density for additional sugar residues that we have been unable to identify definitively. The residual weak density is extending toward a large solvent channel in the crystal. Two regions were poorly defined in the map. The first is at the loop residues, Arg74 and Asp75, and the second is at the three C-terminal residues. These have uninterpretable densities implying complex disorder. The current structure contains 194 water molecules. The R-factor is 18.6% (an R free of 22.7%) for all reflections to 2.1-Å resolution. The root-mean-square (r.m.s.) deviations from target values are 0.008 Å for bond lengths, 1.535° for bond angles, and 1.139° for the peptide torsion angles. The averaged B-factor is 30.8 Å2. There is no residue in disallowed regions as defined in PROCHECK (22Laskowski R.A. MacArthur M.W. Moss D.S. Thornton J.M. J. Appl. Crystallogr. 1993; 26: 283-291Crossref Google Scholar), but 89.4% residues in the most favorable regions and 10.6% residues in the additional allowed regions. Neuropsin consists of fourteen β-strands (designated as β1–β14) that are extensively twisted, two α-helices (designated as α1 and α2), and one short 310-helix (Fig.1 a). Each seven-β-strand forms an antiparallel β-sheet folded in a β-sandwich with a cleft where the catalytic triad (Asp57, His102, and Ser195) is located (Fig. 1, a and b). This overall structure is homologous to those of the chymotrypsin-type serine proteases, which share an identical catalytic mechanism, but among which the substrate specificity varies. Many known structures of these proteases delineate a clear framework, demonstrating that this variety is a function of evolved diversity in the structures of surface loops that surround the substrate-binding site. Because the loops of neuropsin, which contains eight prominent loops (A–H in Fig. 1 a), are conserved in their relative positions with respect to the active site, general themes for their individual functions can be derived. One of the characteristic features of neuropsin is theN-glycosylated loop D that corresponds to the so-called “kallikrein loop.” This loop, having an Asn-X-Ser sequence, is typical for members of the kallikrein family that contains NGFγ, which exhibits relatively high (46%) sequence identity to neuropsin (Fig. 2). Neuropsin, however, forms six disulfide bonds corresponding to those of trypsin with an additional disulfide bond (SS3 between Cys128 and Cys232 in Fig. 1 a) that is missing in members of the kallikrein family. Large differences exist in the loop regions surrounding the substrate-binding site, whereas the core region contains only minor variations. Excluding the insertion and deletion residues, the main-chain atoms of neuropsin superimpose on the corresponding atoms of bovine pancreatic trypsin (18Finer-Moore J.S. Kossiakoff A.A. Hurley J.H. Earnest T. Stroud R.M. Proteins Struct. Finct. 1992; 12: 203-222Crossref PubMed Scopus (110) Google Scholar), mouse submaxillary gland NGFγ in 7 S NGF (23Bax B. Blundell T.L. Murray-Rust J. McDonald N.Q. Structure (Lond.). 1997; 5: 1275-1285Abstract Full Text Full Text PDF PubMed Scopus (65) Google Scholar), an α2β2γ2 complex of NGF, and pancreatic porcine kallikrein (24Bode W. Chen Z. Bartels K. Kutzbach C. Schmidt-Kastner G. Bartunik H. J. Mol. Biol. 1983; 164: 237-282Crossref PubMed Scopus (223) Google Scholar) with r.m.s. deviations of 1.26, 1.43, and 1.84 Å, respectively. The geometry of the catalytic triad is highly similar to those of the serine proteases with r.m.s. deviations in a range of 0.2–0.24 Å. Neuropsin has no prominent structural similarity to tPA, showing a high r.m.s. deviation of 2.74 Å for 79 identical residues (37% sequence identity). Enzyme assay using several 4-methylcoumaryl-7-amide (MCA) derivatives of oligopeptides (25Shimizu C. Yoshida S. Shibata M. Kato K. Momota Y. Matsumoto K. Shiosaka T. Midorikawa R. Kamachi T. Kawabe A. Shiosaka S. J. Biol. Chem. 1998; 273: 11189-11196Abstract Full Text Full Text PDF PubMed Scopus (110) Google Scholar) has shown that neuropsin cleaves peptide bonds C-terminal to Arg or Lys. This specificity is well by the S1 a pocket that is by two G and and at its by the side chain of neuropsin has large changes of loop G with of Å with those of NGFγ, and trypsin The changes in neuropsin to be by the deletion in loop G. In loop H of neuropsin also large from these proteases loop H is with loop G in all the is that loop H of trypsin has a deletion of the that is conserved in neuropsin, NGFγ, and This deletion a of neuropsin loop H from trypsin than from NGFγ or kallikrein the specificity of neuropsin for arginine is with that for lysine. This is in contrast to NGFγ, and in which a for arginine key residues of the S1 of neuropsin has relatively large from NGFγ and kallikrein of neuropsin has a relatively large from trypsin with NGFγ, the changes of the neuropsin loop structures in of and which have been to form bonds to the arginine of NGFβ in 7 S NGF (23Bax B. Blundell T.L. Murray-Rust J. McDonald N.Q. Structure (Lond.). 1997; 5: 1275-1285Abstract Full Text Full Text PDF PubMed Scopus (65) Google Scholar). These differences be for the unique specificity of other serine proteases, which are by cleavage of the bond between neuropsin the of the pocket to form bonds with the of loop and to form an with of loop G. The kallikrein loop of neuropsin from those of NGFγ and In these members of the kallikrein the loop was the highly In the loop of neuropsin is an ordered and relatively without neuropsin has no arginine or residue in the loop. crystal structure Sci. 1997; PubMed Scopus Google Scholar) of mouse an ordered kallikrein loop, but no similarity with the loop of that the to Asn95 participates in the substrate the GlcNAc residue and the residual density are from the active site as in members of the kallikrein family. The kallikrein loop of neuropsin toward the active site cleft with a prominent which its role in substrate of neuropsin on porcine pancreatic trypsin with soybean trypsin S.W. J. Mol. Biol. 1998; PubMed Scopus Google Scholar) revealed between the kallikrein loop of neuropsin and the two loops toward This was by in which high such as or were found to have effect on the neuropsin whereas such as the The kallikrein loop forms a pocket in which is at the and restrict the of the side chain in the P2 position of substrate This is with the of a assay (25Shimizu C. Yoshida S. Shibata M. Kato K. Momota Y. Matsumoto K. Shiosaka T. Midorikawa R. Kamachi T. Kawabe A. Shiosaka S. J. Biol. Chem. 1998; 273: 11189-11196Abstract Full Text Full Text PDF PubMed Scopus (110) Google Scholar), in which high of neuropsin were observed for peptide having smaller residues or proline in the P2 has been well by the crystal structure of that loop of the site with the residues, and to the P2 specificity of the for proline W. U. R. S.R. J. J. PubMed Scopus Google Scholar). of neuropsin on of neuropsin located and of but no between and the proline residue of to which that the P2 for proline be mediated by the kallikrein loop of neuropsin, of loop of but rather weaker than that of at the P2 position reduces the neuropsin which is one of the differences from kallikrein and NGFγ. is the structure of loop that is similar to that of trypsin rather than NGFγ and where changes of loop from neuropsin with large and Å, by of α1 (Fig. Loop is one of the forming the S3/S4 site. is that this loop has which forms bonds with the of loop H and which is conserved in trypsin but is by in NGFγ and This residue has been to be one of the residues for the substrate specificity 1994; PubMed Scopus Google Scholar). Moreover, which is conserved in neuropsin and is one of the key residues for the loop structure this residue is an essential of the at be that the disulfide bond which is conserved in trypsin as to the of loop with that of trypsin through with that with the C-terminal of loop F, is that the differences in the loop also be with the differences in the kallikrein In NGFγ and the highly kallikrein loops with loop the kallikrein loop of neuropsin has no direct with loop F, as with loop D of trypsin. One of the of these in loops D and is that the S3/S4 site of neuropsin is similar to that of trypsin rather than NGFγ. The of and a but for the S3/S4 site as in trypsin and the high of neuropsin observed for having residues in the Moreover, in neuropsin, of loop is toward the S3/S4 site (Fig. 1 whereas of NGFγ is from the S3/S4 site. is that play a role in the with the substrates. one of the cleavage of (25Shimizu C. Yoshida S. Shibata M. Kato K. Momota Y. Matsumoto K. Shiosaka T. Midorikawa R. Kamachi T. Kawabe A. Shiosaka S. J. Biol. Chem. 1998; 273: 11189-11196Abstract Full Text Full Text PDF PubMed Scopus (110) Google Scholar), which is an extracellular matrix protein proteolytic for neuropsin, had the N-terminal sequence of residue at the position with A substrate for has been found to the highest for neuropsin to poor structural between neuropsin and is from the large r.m.s. deviation of Å for identical residues. Moreover, cleaves than which is one of the of whereas the of neuropsin for these are In also exhibits a for arginine at the but no such was observed for neuropsin, as described These with structural differences of several the unique substrate specificity of neuropsin, the P2 for proline is to with the on the N-terminal side of the is known on the C-terminal side at present. a by and of to a site (Fig. 1 b). The of the substrate-binding surface and the surface of neuropsin several differences in from other serine One of the is the of the S1 where residues, and the side to the solvent of the kallikrein loop also is from the These be related to the specificity that are from those of other Neuropsin exhibits weak proteolytic and but cleaves as the extracellular neuropsin its limbic is an neuropsin NGFβ and form 7 S NGF of NGFγ. In 7 S NGF, the active site of NGFγ was by the C-terminal of the as a with of the large kallikrein loop with the C-terminal regions and of suggest that neuropsin most of these the smaller residues, of NGFβ to the and of In the of NGFγ for of NGFβ is missing in Moreover, which is located at the of the active site of NGFγ and forms bonds to chain of and of is by in In 7 S NGF, a was located at the between NGFγ and to the This is also NGFγ is by neuropsin and of NGFγ are by serine and in these suggest that neuropsin is of forming 7 S NGF, neuropsin the NGFβ this will In to neuropsin, three other serine proteases have been to be more highly expressed in the nervous system than in most These protease I.A. Towner M.D. Isackson P.J. J. Neurosci. 1997; 17: 8156-8168Crossref PubMed Google Scholar), K. Tsuruoka N. Kodama S. Tsujimoto M. Yamamura Y. Tanaka T. Nakazato H. Yamagichi N. Biochim. Biophys. Acta. 1997; 1350: 11-14Crossref PubMed Scopus (165) Google Scholar), and (6Gschwend T.P. Krueger S.R. Kozlov S.V. Wolfer D.P. Sonderegger P. Mol. Cell. Neurosci. 1997; 9: 207-219Crossref PubMed Scopus (105) Google Scholar). and sequence to neuropsin of and respectively. which is a serine protease expression is most prominent in the and amygdala, has a protease domain sequence identity to with neuropsin indicate that these proteases have structures of surface loops surrounding the substrate-binding site. loop D of of these proteases has no site and no residues. This of a kallikrein loop in their P2 from that of Moreover, loop G of and has no which structural changes of loops G and H forming the S1 with neuropsin, has a insertion in loop G and a deletion in loop H. These differences these other proteases with substrate from that of In many of neuropsin structure and function that this hippocampal serine protease chimeric structural features of trypsin and NGFγ with novel substrate These give a clue to the design useful in treatment of pathological conditions such as epilepsy and also useful in the of synaptic plasticity. and S. for the of 7 S NGF and soybean trypsin M. Suzuki for data at and S. and J. for with the mass spectroscopy and N-terminal