The Serine Protease Plasmin Cleaves the Amino-terminal Domain of the NR2A Subunit to Relieve Zinc Inhibition of the N-Methyl-d-aspartate Receptors*

Zinc is hypothesized to be co-released with glutamate at synapses of the central nervous system. Zinc binds to NR1/NR2A N-methyl-d-aspartate (NMDA) receptors with high affinity and inhibits NMDAR function in a voltage-independent manner. The serine protease plasmin can cleave a number of substrates, including protease-activated receptors, and may play an important role in several disorders of the central nervous system, including ischemia and spinal cord injury. Here, we demonstrate that plasmin can cleave the native NR2A amino-terminal domain (NR2AATD), removing the functional high affinity Zn2+ binding site. Plasmin also cleaves recombinant NR2AATD at lysine 317 (Lys317), thereby producing a ∼40-kDa fragment, consistent with plasmin-induced NR2A cleavage fragments observed in rat brain membrane preparations. A homology model of the NR2AATD predicts that Lys317 is near the surface of the protein and is accessible to plasmin. Recombinant expression of NR2A with an amino-terminal deletion at Lys317 is functional and Zn2+ insensitive. Whole cell voltage-clamp recordings show that Zn2+ inhibition of agonist-evoked NMDA receptor currents of NR1/NR2A-transfected HEK 293 cells and cultured cortical neurons is significantly reduced by plasmin treatment. Mutating the plasmin cleavage site Lys317 on NR2A to alanine blocks the effect of plasmin on Zn2+ inhibition. The relief of Zn2+ inhibition by plasmin occurs in PAR1-/- cortical neurons and thus is independent of interaction with protease-activated receptors. These results suggest that plasmin can directly interact with NMDA receptors, and plasmin may increase NMDA receptor responses through disruption or removal of the amino-terminal domain and relief of Zn2+ inhibition.

N-Methyl-D-aspartate (NMDA) 2 receptors are one of three types of ionotropic glutamate receptors that play critical roles in excitatory neurotransmission, synaptic plasticity, and neuronal death (1)(2)(3). NMDA receptors are comprised of glycinebinding NR1 subunits in combination with at least one type of glutamate-binding NR2 subunit (1,4). Each subunit contains three transmembrane domains, one cytoplasmic re-entrant membrane loop, one bi-lobed domain that forms the ligand binding site, and one bi-lobed amino-terminal domain (ATD), thought to share structural homology to periplasmic amino acid-binding proteins (4 -6). Activation of NMDA receptors requires combined stimulation by glutamate and the co-agonist glycine in addition to membrane depolarization to overcome voltage-dependent Mg 2ϩ block of the ion channel (7). The activity of NMDA receptors is negatively modulated by a variety of extracellular ions, including Mg 2ϩ , polyamines, protons, and Zn 2ϩ ions, which can exert tonic inhibition under physiological conditions (1,4). Several extracellular modulators such as Zn 2ϩ and ifenprodil are thought to act at the ATD of the NMDA receptor (8 -14).
Zinc is a transition metal that plays key roles in both catalytic and structural capacities in all mammalian cells (15). Zinc is required for normal growth and survival of cells. In addition, neuronal death in hypoxia-ischemia and epilepsy has been associated with Zn 2ϩ (16 -18). Abnormal metabolism of zinc may contribute to induction of cytotoxicity in neurodegenerative diseases, such as Alzheimer's disease, Parkinson's disease, and amyotrophic lateral sclerosis (19). Zinc is co-released with glutamate at excitatory presynaptic terminals and inhibits native NMDA receptor activation (20,21). Zn 2ϩ inhibits NMDA receptor function through a dual mechanism, which includes voltage-dependent block and voltage-independent inhibition (22)(23)(24). Voltage-independent Zn 2ϩ inhibition at low nanomolar concentrations (IC 50 , 20 nM) is observed for NR2A-containing NMDA receptors (25)(26)(27)(28). Evidence has accumulated that the amino-terminal domain of the NR2A subunit controls high-affinity Zn 2ϩ inhibition of NMDA receptors, and several histidine residues in this region may constitute part of an NR2A-specific Zn 2ϩ binding site (8,9,11,12). For the NR2A subunit, several lines of evidence suggest that Zn 2ϩ acts by enhancing proton inhibition (8,11,29,30).
Serine proteases present in the circulation, mast cells, and elsewhere signal directly to cells by cleaving protease-activated receptors (PARs), members of a subfamily of G-protein-coupled receptors. Cleavage exposes a tethered ligand domain that binds to and activates the cleaved receptors (31,32). Protease receptor activation has been studied extensively in relation to coagulation and thrombolysis (33). In addition to their circulation in the bloodstream, some serine proteases and PARs are expressed in the central nervous system, and have been suggested to play roles in physiological conditions (e.g. long-term potentiation or memory) and pathophysiological states such as glial scarring, edema, seizure, and neuronal death (31, 34 -36).
Functional interactions between proteases and NMDA receptors have previously been suggested. Earlier studies reported that the blood-derived serine protease thrombin potentiates NMDA receptor response more than 2-fold through activation of PAR1 (37). Plasmin, another serine protease, similarly potentiates NMDA receptor response (38). Tissue-plasminogen activator (tPA), which catalyzes the conversion of the zymogen precursor plasminogen to plasmin and results in PAR1 activation, also interacts with and cleaves the ATD of the NR1 subunit of the NMDA receptor (39,40). This raises the possibility that plasmin may also interact directly with the NMDA receptor subunits to modulate receptor response. We therefore investigated the ability of plasmin to cleave the NR2A NMDA receptor subunit. We found that nanomolar concentrations of plasmin can cleave within the ATD, a region that mediates tonic voltage-independent Zn 2ϩ inhibition of NR2A-containing NMDA receptors. We hypothesized that plasmin cleavage reduces the Zn 2ϩ -mediated inhibition of NMDA receptors by removing the Zn 2ϩ binding domain. In the present study, we have demonstrated that Zn 2ϩ inhibition of agonist-evoked NMDA currents is decreased significantly by plasmin treatment in recombinant NR1/NR2Atransfected HEK 293 cells and cultured cortical neurons. These concentrations of plasmin may be pathophysiologically relevant in situations in which the blood-brain barrier is compromised, which could allow blood-derived plasmin to enter brain parenchyma at concentrations in excess of these that can cleave NR2A. Thus, ability of plasmin to potentiate NMDA function through the relief of the Zn 2ϩ inhibition could exacerbate the harmful actions of NMDA receptor overactivation in pathological situations. In addition, if newly cleaved NR2A ATD enters the bloodstream during ischemic injury, it could serve as a biomarker of central nervous system injury.
NR2A Amino-terminal Domain (NR2A ATD ) Protein Production-The NR2A ATD was produced as a thrombin-cleavage glutathione S-transferase (GST) fusion protein in Escherichia coli. NR2A ATD (Ala 33 -Asp 421 ) was generated by polymerase chain reaction with the following primer pairs: sense, 5Ј-acgattccatggcgctgaacattgcggtgctg-3Ј and antisense, 5Ј-atcgaagctttatcagaggtcctcttcggatatcagcttctgttcgtcctctacgatgacgaagg-3Ј. The product then was cloned into a pGEX-KG vector. A c-myc tag (underlined in antisense primer; EQKLI-SEEDL) was incorporated in the COOH terminus of NR2A ATD for identification purposes. After transformation into BL21-CodonPlus-RIL cells (Stratagene, La Jolla, CA), cells at OD 0.8 were induced by 0.2 mM isopropyl ␤-D-thiogalactopyranoside for 3 h at 37°C. Pelleted cells were lysed using French Press (Thermo Spectronic, Madison, WI), and GST-NR2A ATD protein was solubilized with detergents. The fusion protein was purified through batch chromatography using reduced glutathione-Sepharose 4B matrix (Amersham Biosciences). NR2A ATD protein was cleaved from the GST by 10 units of thrombin (Calbiochem, San Diego, CA) at room temperature for 1 h.
Zn 2ϩ -IDA-Agarose Affinity Binding Assay-Fifty microliters of immobilized iminodiacetic acid (IDA) resin (Pierce) equilibrated in Buffer A (10 mM HEPES, 50 mM NaCl, 0.5% Triton X-100, pH 7.5) were incubated at room temperature for 30 min with a freshly prepared 10 mM ZnCl 2 solution. The fully Zn 2ϩcharged IDA-agarose was thoroughly washed with Buffer A. The pelleted Zn 2ϩ -IDA-agarose was resuspended in 25 l of Buffer A and incubated with 25 l of thrombin-cleaved NR2A ATD protein (room temperature for 30 min). The NR2A ATD -bound to Zn 2ϩ -IDA-agarose was pelleted, and the supernatant was collected (designated as FT). The agarose resin was washed with Buffer A six times, and the first wash solution (W1) was collected for analysis. The bound NR2A ATD was eluted from Zn 2ϩ -IDA-agarose with several volumes of 50 mM EDTA or 1 mM L-histidine in Buffer A (designated as E1, E2, or E3). Negative controls were performed with IDA-agarose that was not charged with ZnCl 2 .
Sequencing of NR2A ATD Plasmin Cleavage Products-The proteins were separated by gel electrophoresis, blotted onto PVDF membrane, and stained with Coomassie Brilliant Blue R-250. The amino-terminal sequence after plasmin cleavage was determined using Edman degradation, which was performed in the Applied Biosystems model cLC-Procise protein sequencer (41) at the Emory Microchemical Facility.
Molecular Modeling of NR2A Subunit-The homology model of NR2A ATD was built using the previously published alignment between NR2A ATD and mGluR1 (42). The closed form of mGluR1 found in Protein Data Bank code 1EWK (43) provided the structural template. The mGluR1 structure and alignment were imported into Prime (version 1.5, 2006, Schrodinger, LLC, New York), and the NR2A ATD homology model was subsequently built using default settings. The model was refined initially using the Prime loop prediction tool on the two central loops (41-45 and 100 -105) of the ATD. The entire structure was then energy minimized, and multiple rounds of energy minimization and side chain optimization were performed on all residues within 5 Å of the two central loops.
Transfection of HEK 293 Cells and Plasmin Treatment-Human embryonic kidney cells 293 (HEK 293 cells; CRL 1573, ATCC, Rockville, MD) were plated at low confluency on glass coverslips (Warner Instruments, Hamden, CT) coated in 200 g/ml poly-D-lysine and incubated overnight at 37°C in humidified 5% CO 2 in media (Dulbecco's modified Eagle's medium; Invitrogen) supplemented with 10% fetal bovine serum, 10 units/ml penicillin, and 10 g/ml streptomycin. Cells then were transiently transfected using the FuGENE transfection reagent (Roche Diagnostics, Basel, Switzerland) with 2.0 g of plasmid cDNA encoding green fluorescent protein, NR1, and wild-type NR2A, green fluorescent protein, NR1, and mutant NR2A-K317A, or green fluorescent protein, NR1, and mutant NR2A-⌬ATD-K317 at a 5:1:1 ratio. After transfection, the cells were incubated overnight at 37°C in humidified 5% CO 2 in media supplemented with 200 M DL-2-amino-5-phosphonovalerate and 200 M 7-chlorokynurenic acid. Following transfection and overnight incubation, media was removed and cells were washed in external solution for patch clamp recording. Some washed coverslips then were removed from their wells and gently placed in a new, untreated 24-well plate that contained the external recording solution (500 l, composition seen below) plus 100 nM plasmin. Clean, untreated plates without cells or bovine serum albumin from the media were used for plasmin treatment to reduce the amount of proteins that could compete with NR2A for cleavage and thus lower the apparent effectiveness of plasmin. The treated coverslips were incubated at 37°C for 10 min and were transferred to new wells containing only external solution for patch clamp recording. We performed the recordings for both the control and the plasmintreated groups on the same day.
Neuron Culture-Primary cultures of cortical neurons were prepared from 1-to 3-day postnatal wild type (C57BL/6) or PAR1 Ϫ/Ϫ mice, which were created by breeding PAR1 ϩ/Ϫ mice, a gift from Dr. Shaun Coughlin (University of California, San Francisco, CA), with C57BL/6 wild-type mice from The Jackson Laboratory. Heterozygous littermates were bred to generate littermate homozygous null mutants and wild-type controls, which were subsequently used to generate homozygous colonies. PAR1 Ϫ/Ϫ mice were Ͼ99% C57Bl/6. All procedures involving animals were approved by the Emory University Institutional Animal Care and Use Committee. The cortex was dissected and chopped into small cubes. The tissues were digested by 24 units/ml papain for 1 h in a 37°C incubator, washed twice with 5-7 ml of inactivation solution (Neurobasal medium with 10% fetal bovine serum and 40 g/ml DNase), and triturated gently with at least three serial fire-polished Pasteur pipettes. Cells were plated onto glass coverslips coated with 100 g/ml poly-D-lysine. Cultures were maintained for 5-15 days in Neurobasal medium supplemented with 5% fetal bovine serum, B-27, 1.0 mM sodium pyruvate, 0.5 mM L-glutamine, 0.6% dextrose, 10 units/ml penicillin, and 10 g/ml streptomycin at 37°C in humidified 5% CO 2 . On the recording day, some washed coverslips were transferred to wells of a new plate that contained the plasmin solution (300 nM), incubated at 37°C for 10 min, and then washed by external recording solution (composition given below).
Two-electrode Voltage-clamp Recordings from Xenopus laevis Oocytes-Preparation and injection of cRNA, as well as twoelectrode voltage-clamp recordings from X. laevis oocytes, were performed as previously described (27). Briefly, oocytes were injected with 5-10 ng of cRNAs synthesized in vitro from linearized template cDNA and stored at 15°C in Barth's solution. The ratio of NR1 to NR2 injected cRNA was 1:2. Two-electrode voltage-clamp recordings were made 2-4 days postinjection at room temperature (23°C). The recording solution contained (in mM) 90 NaCl, 1 KCl, 10 HEPES, 0.5 BaCl 2 , plus 2-10 M EDTA. In the experiments for agonist potency, pH was adjusted to 7.6 with NaOH, and EDTA was added to chelate contaminant divalent ions. In the experiments for the concentration-response curve of zinc inhibition, pH was adjusted to 7.2 and Tricine (10 mM) was used to buffer Zn 2ϩ concentrations as described previously (27). Solution exchange was computer-controlled through an 8-modular valve positioner (Digital MVP Valve, Hamilton, CT). Voltage and current electrodes were filled with 0.3-3.0 M KCl, and current responses were recorded at a holding potential of Ϫ20 to Ϫ40 mV at 23°C. Data acquisition and voltage control were accomplished with a twoelectrode voltage-clamp amplifier (OC-725, Warner Instrument, Hamilton, CT). Only currents greater than 50 nA were included in the analysis. 50 M Glutamate and 30 M glycine were used in all oocyte experiments unless otherwise stated.
Whole Cell Voltage-clamp Recording from Transfected HEK 293 Cells-Voltage-clamp recordings (V HOLD Ϫ60 mV) were performed on transfected HEK 293 cells using an Axopatch 200B amplifier (Molecular Devices, Union City, CA) at room temperature (23°C). Recording electrodes (3.5-7 M⍀) were made from thin wall glass pipettes (TW150F-4, World Preci-sion Instruments, Sarasota, FL) pulled using a vertical puller (Narishige PP-830, Tokyo, Japan), and were filled with an internal solution containing (in mM) 110 D-gluconate, 110 CsOH, 30 CsCl, 5 HEPES, 4 NaCl, 0.5 CaCl 2 , 2 MgCl 2 , 5 BAPTA, 2 NaATP, and 0.3 NaGTP, pH 7.35. Cells were bathed continually at 23°C in an external solution composed of (in mM) 150 mM NaCl, 10 HEPES, 30 D-mannitol, 3 KCl, 2 CaCl 2 , and 0.01 EDTA, pH 7.4. Glutamate and glycine agonist solutions were made in external solution that contained 10 M EDTA or 1 M Zn 2ϩ . Once the cell membrane was ruptured by applying suction, the whole cell was lifted into the path of a 2-barrel fast perfusion system controlled by a piezoelectric translator (Burleigh Instruments, Fishers, NY) with 5-8-ms exchange times for the solutions (29). Cells were exposed first to glycine (50 M), and subsequently glutamate (50 M in 50 M glycine) was applied rapidly for 1 s, after which the cells were again washed in glycine. Steady-state amplitudes and deactivation time constants were measured using ChanneLab (Synaptosoft, Decatur, GA).
Whole Cell Perforated Patch Recording from Cultured Cortical Neurons-Whole cell voltage-clamp current recordings (V HOLD Ϫ60 mV) were performed on 5-10-day cultured cortical neurons at room temperature (23°C). The recording chamber was continually perfused with recording solution composed of (in mM) 150 NaCl, 3 KCl, 2 CaCl 2 , 5.5 glucose, and 10 HEPES (pH 7.2 by NaOH; osmolality adjusted to 315-320 mosmol with sucrose). 10 M EDTA was added to chelate contaminant divalent ions except for Zn 2ϩ -containing solutions. Thin wall glass pipettes (TW150F-4, World Precision Instrument) were used to make recording electrodes (4 -5 M⍀) by a vertical puller (Narishige PP-830, Tokyo, Japan) and filled with (in mM) 150 CsMeSO 4 , 10 NaCl, 0.5 CaCl 2 , 10 HEPES, and 25-50 g/ml gramicidin D (pH adjusted to 7.3 with CsOH and osmolality adjusted to 310 mOsm with sucrose). Fresh stock solution of gramicidin D (1 mg/80 l in dimethyl sulfoxide) was prepared daily. The final internal solution containing gramicidin D was used within 1 h. It took 15-30 min to achieve acceptable perforation with series resistance ranging from 30 to 60 M⍀. Drugs were applied directly by gravity and controlled by solenoid valves (Lee, Westbrook, CT) with tubes placed above the recorded cell. The solution exchange achieved a complete local perfusion of the recorded cell in ϳ400 ms (10 -90% rise time), which was measured by applying a high K ϩ solution onto cultured cortical neurons and measuring the time course of the instantaneous change of the leak current. All drugs were purchased from Sigma except papain and DNase (Worthington, Lakewood, NJ), Neurobasal medium, fetal bovine serum, L-glutamine and sodium pyruvate (Invitrogen), CNQX disodium salt and ifenprodil (Tocris Cookson, Bristol, UK), plasmin (Hematologic Technologies, Inc., Essex Junction, VT), and 2-furoyl-LIGRLO (Emory Microchemical Facility).
Data are expressed as mean Ϯ S.E., and analyzed statistically using unpaired t test or one-way analysis of variance with Tukey's post hoc test. Significance for all tests was set at p Ͻ 0.05. Error bars in all figures are S.E.

RESULTS
Plasmin Cleaves the Native NR2A Subunit-To detect whether plasmin can cleave the NR2A subunit in native tissue, membrane preparations from rat cortex were treated with increasing concentrations of plasmin (200 -500 nM) for 20 min at room temperature. Membrane proteins were separated using gel electrophoresis and transferred to PVDF membranes, which were probed with a primary antibody that recognizes the NR2A COOH-terminal. These experiments (four experiments from four separate membrane preparations) showed that 200 -500 nM plasmin cleaves a ϳ40-kDa fragment from the aminoterminal domain, leaving a truncated 140-kDa NR2A subunit (Fig. 1, A and B). Lower plasmin concentrations (30 nM) showed less cleavage of NR2A, suggesting concentration dependence (data not shown). These data indicate that higher concentrations of plasmin can cleave the NR2A subunit of native tissues at the ATD.
To investigate whether plasmin can cleave the NR1 subunit, Western blots were run on plasmin-treated rat neuronal membranes. Western blots probed with the NR1 monoclonal antibody 54.1 showed several fragments following plasmin treatment (supplemental Fig. S1A). Probing these same Western blots with an anti-COOH-terminal antibody yielded no apparent cleaved fragments but reveals decreasing signal in the original band of the plasmin-treated groups (supplemental Fig.  S1B). These data suggest that plasmin cleaves NR1 efficiently at a site near the COOH terminus. In addition, there was very modest cleavage at a site within the S2 domain of the NR1 subunit, which appears less sensitive to plasmin than NR2A. The COOH-terminal site is inaccessible to plasmin in intact cells, and the weakly sensitive NR1-S2 cleavage site is distant from the Zn 2ϩ binding site, which resides within the aminoterminal domain of the NR2 (not NR1) subunit. This result is consistent with the proposed plasmin cleavage sites on NR1 in the recently published report by Samson et al. (44).
Plasmin Cleaves Recombinant NR2A ATD Protein-The finding that plasmin cleaves the NR2A ATD in neuronal membranes FIGURE 1. Plasmin cleavage of the NR2A subunit in brain tissue. A, Western immunoblot was probed with a COOH-terminal NR2A antibody (Upstate Biotech). Rat brain membranes were incubated with plasmin for 20 min at room temperature prior to loading in a 10% Tris glycine gel. Note that plasmin cleaves a ϳ40-kDa fragment from the NR2A amino-terminal domain. B, arrow indicates approximate plasmin cleavage site on NR2A subunit. Line above the COOH-terminal side indicates antibody epitope. may be significant because this region mediates high affinity voltage-independent Zn 2ϩ inhibition of NR2A-containing receptors. To further investigate plasmin cleavage of the NR2A ATD , recombinant NR2A ATD protein was expressed in bacteria and purified. To accomplish this, the NR2A ATD construct (residues 33-421; Fig. 2A, upper panel, dot boxed) was cloned into pGEX-KG plasmid and COOH-terminal tagged with c-myc. The size of the expressed GST-NR2A ATD fusion protein was ϳ74.3 kDa (Fig. 2A, middle panel). A thrombin cleavage site was inserted between GST and NR2A ATD (Fig. 2A,  lower panel). The NR2A ATD construct was expressed in E. coli, and the recombinant NR2A ATD proteins were purified through both the batch method and column chromatography. The purified recombinant NR2A ATD protein was treated with 10 nM plasmin for 20 min at room temperature, and a PVDF membrane containing at least three bands of NR2A ATD protein frag-ments was stained with Coomassie Blue (Fig. 2B). To determine the plasmin cleavage site(s), the cleaved protein fragments were cut, purified, and subjected to amino-terminal peptide sequencing. Sequencing showed that only one isolated fragment (indicated by thick red arrow in Fig. 2B) had a sequence matching the NR2A subunit at residues downstream of Lys 317 (Fig. 2C, underlined amino acids); other isolated peptide fragments either did not yield interpretable sequences or were from other unidentified proteins. These results suggest that plasmin can cleave recombinant NR2A ATD protein at a site located at Lys 317 , which is similar to plasmincleavage sites on other plasmin substrates (Table 1).
Homology Model of NR2A ATD -To evaluate the accessibility of the proposed cleavage site at the NR2A ATD to plasmin (Fig. 2D), a homology model of the NR2A ATD was constructed using the alignment between NR2A ATD and mGluR1 (43). This model predicts that Lys 317 is near the surface of the protein (Fig. 2E) and thus is potentially accessible by plasmin. This model also indicates that the Zn 2ϩ -binding residues and binding pocket (magenta arrow in Fig. 2E) are upstream of the plasmin cleavage site (red arrow in Fig.  2E). This implies that the fragment containing the Zn 2ϩ -binding pocket within the ATD (gray color in Fig. 2E, right panel) may dissociate from the receptor after plasmin treatment. Therefore, plasmin treatment may relieve Zn 2ϩ inhibition on NR2A-containing NMDA receptors.
Zinc Ions Bind to Recombinant NR2A ATD Protein-There is considerable evidence showing that the region containing the amino-terminal domain of the NR2A subunit controls highaffinity Zn 2ϩ inhibition of NMDA receptors, and several histidine residues in this region may constitute part of the NR2Aspecific Zn 2ϩ binding site (8,11). To validate that the recombinant protein was in the correct conformation, we assessed its ability to bind Zn 2ϩ . Recombinant NR2A ATD protein binds to Zn 2ϩ -charged IDA-agarose but not to uncharged IDA-agarose (Fig. 3A). The bound NR2A ATD protein could be eluted using the metal-chelating agent EDTA (50 mM), but not low affinity Zn 2ϩ binding agents imidazole and histidine. Several histidine residues in the ATD may constitute part of the NR2A-specific Zn 2ϩ binding site (8,11,12). When H44G and H128A mutations were inserted concurrently into the recom-  3 kDa). B, Coomassie-stained PVDF membrane containing NR2A ATD protein fragments. Arrows indicate the potential plasmin (10 nM)-cleaved fragments subjected to amino-terminal sequencing. Only the band indicated by the thick red arrow yielded an interpretable sequence. C, amino acid Lys 317 in red indicates the plasmin cleavage site. Note that the underlined amino acids were determined from amino-terminal peptide sequencing on the cleaved protein fragment (the bottom protein fragment in panel B, indicated by the red arrow). D, diagram of the NR2A subunit shows autonomous glutamate and Zn 2ϩ binding domains. E, homology model of the NR2A ATD . Left and middle panels, plasmin cleaves (indicated by the red arrow) between Lys 317 (CPK colors) and Ala 318 (first residue loop of red portion) in the NR2A ATD protein (initiating methionine is 1). Zn 2ϩ and its coordinating residues are indicated in the left panel (arrow). In the right panel, the red amino acids are downstream of the plasmin cleavage site and presumably remain attached to the receptor. The light gray portion of the protein harboring the Zn 2ϩ site is free to dissociate from receptor after plasmin treatment. binant NR2A ATD protein, the binding of the NR2A ATD (H44G/ H128A) protein to Zn 2ϩ -IDA-agarose was weaker than the NR2A ATD protein. NR2A ATD (H44G/H128A) could be fully displaced by low affinity Zn 2ϩ binding agents imidazole and L-histidine, which we interpreted as evidence that Zn 2ϩ is coordinated by His 44 and His 128 (Fig. 3B). These data suggest that recombinant NR2A ATD might exist in a conformation that resembles the native NR2A subunit and thus is a suitable substrate for analyses of plasmin proteolysis.
Deletion of NR2A ATD (⌬ATD) at Lys 317 Is Functional and Zn 2ϩ -insensitive-These biochemical data suggest that plasmin can cleave the Zn 2ϩ binding domain of NR2A, raising the possibility that the remaining receptor may be functional but resistant to Zn 2ϩ inhibition. To evaluate the functional change by deletion of the NR2A ATD following plasmin cleavage, we mutated NR2A to delete all of the ATD up to Lys 317 (NR2A-  (Fig. 4B). Zn 2ϩ sensitivity was assessed by measuring the percentage of inhibition of NMDA receptor agonist-evoked steady-state currents in the presence of 1 M Zn 2ϩ , a concentration that saturates the high affinity site. Deletion of the amino-terminal domain of NR2A at Lys 317 significantly reduced Zn 2ϩ inhibition (12 Ϯ 1.4%, n ϭ 16) compared with wild-type NR2A (47 Ϯ 3.0%, n ϭ 15; p Ͻ 0.001, unpaired t test) (Fig. 4C). To further investigate the effects of ATD deletion from the plasmin cleavage site on zinc inhibition, the concentration-response curve for zinc inhibition was generated from recombinant NMDA receptors expressed in oocytes. High affinity Zn 2ϩ inhibition was present in wild-type NR1/NR2A receptors, but was absent in the ATD deletion construct NR2A-⌬ATD-K317. The low affinity site for Zn 2ϩ inhibition was present in both wild-type NR2A and NR2A-⌬ATD-K317, and presumably reflects voltage-dependent channel block by Zn 2ϩ (Fig. 4D). These data demonstrate that recombinant NR1/NR2A receptors lacking the residues upstream of the plasmin cleavage site are functional. These data also are consistent with a previous report (9) that NR1/NR2A receptors lacking the ATD are less sensitive to Zn 2ϩ inhibition.

Plasmin Cleavage of NMDA Receptors Relieves Zn 2ϩ Inhibition in Recombinant NR1/NR2A-transfected HEK 293 Cells-
To test the hypothesis that plasmin cleavage relieves Zn 2ϩ inhibition, agonist-evoked currents were recorded from recombinant NR1/NR2A-transfected HEK 293 cells by whole cell voltage-clamp recordings in the absence and presence of 1 M Zn 2ϩ with and without plasmin treatment. Zn 2ϩ inhibition was evaluated as the percentage of inhibition produced by 1 M Zn 2ϩ on steady-state currents evoked by glutamate/glycine (50 M). We found that 1 M Zn 2ϩ inhibited agonist-evoked cur- Size of side chain (Å) 51 43 64 86 39 67 52 rents by 78.0 Ϯ 4.6% (n ϭ 6, Fig. 5A, left panel), but only inhibited current responses in HEK 293 cells by 19 Ϯ 13% following a 10-min treatment with 100 nM plasmin (n ϭ 3, Fig. 5A, right panel). There is a significant difference (p Ͻ 0.001, unpaired t test; Fig. 5B) between these two groups, suggesting that Zn 2ϩ sensitivity is reduced by plasmin treatment.
To further confirm the hypothesis that the relief of Zn 2ϩ inhibition is due to plasmin cleavage at Lys 317 on the NR2A amino-terminal, Lys 317 was mutated to alanine (NR2A-K317A), and Zn 2ϩ inhibition was measured on current responses in recombinant NR1/NR2A-K317A-transfected HEK 293 cells in the control and plasmin treatment groups. Consistent with our previous data, we found that 1 M Zn 2ϩ inhibited agonist-evoked currents on NR1/NR2A-K317Atransfected HEK 293 cells by 72 Ϯ 6.5% after 10 min of plasmin treatment, similar to 77 Ϯ 5.3% (n ϭ 3) inhibition without plasmin treatment (Fig. 5C). There is no significant difference in Zn 2ϩ sensitivity between the control and plasmin-treated HEK 293 cells transfected with NR1/NR2A-K317A (p ϭ 0.64, unpaired t test). These data suggest that plasmin-induced removal of Zn 2ϩ sensitivity requires cleavage of the NR2A amino-terminal at Lys 317 .
Plasmin Cleavage of NMDA Receptors Relieve Zn 2ϩ Inhibition in Cultured Cortical Neurons-To investigate whether the relief of Zn 2ϩ inhibition by plasmin cleavage also occurs in native NMDA receptors, whole cell perforated patch voltageclamp recordings were performed on mouse cortical neurons maintained in culture for 5-10 days (Fig. 6A). Neurons were distinguished by inward sodium currents evoked by depolarizing voltage steps (Fig. 6B) that were reversibly abolished by 0.5 M tetrodotoxin (data not shown). Zn 2ϩ inhibition was evaluated as the percentage of inhibition of 1 M Zn 2ϩ observed at steady-state on NMDA-evoked whole cell currents from control and the plasmin treatment groups. CNQX and ifenprodil were used to block non-NMDA receptors and NR2B-containing NMDA receptors.
Plasmin treatment had no significant effect on the steadystate amplitude of NMDA receptor currents in cultured cortical neurons (data not shown; control group: 124 Ϯ 17 pA, n ϭ 15 versus plasmin-treated group: 143 Ϯ 20 pA, n ϭ 14; p ϭ 0.48; unpaired t test). This result suggests that plasmin itself has no direct effect on NMDA receptor function and is consistent with our previous studies (38,45). We subsequently evaluated the effect of plasmin on Zn 2ϩ sensitivity of NMDA receptor currents from cultured cortical neurons by recording the NMDA current response in the presence and absence of Zn 2ϩ . In these experiments, extracellular Zn 2ϩ (1 M) inhibited NMDAevoked whole cell currents by 53 Ϯ 2.6% (n ϭ 6) before and 29 Ϯ 4.2% (n ϭ 7) after treatment with 300 nM plasmin (Fig. 6, C and  D). These data show that Zn 2ϩ causes significantly less inhibition following treatment with the serine protease plasmin (p Ͻ 0.05, unpaired t test). These data suggest that voltage-independent Zn 2ϩ inhibition is relieved by submicromolar concentrations of plasmin, consistent with plasmin-mediated removal of the Zn 2ϩ binding site within the amino-terminal domain.
It is known that proteases, including plasmin, potentiate the NMDA receptor response through activation of PAR1 (37,38). To evaluate whether relief of Zn 2ϩ inhibition by plasmin involves PAR1 activation, we prepared cultured cortical neurons from PAR1 Ϫ/Ϫ mice. Similar to wild-type neurons, extracellular Zn 2ϩ (1 M) inhibited agonist-evoked NMDA receptor currents by 40 Ϯ 4.4% (n ϭ 10), whereas it had less effect after treatment with 300 nM plasmin (18 Ϯ 4.0%, n ϭ 6; p Ͻ 0.001, unpaired t test; Fig. 7, A and B). These results indicate that the relief of Zn 2ϩ inhibition by plasmin is independent of PAR1 activation. In addition to activating PAR1, plasmin is also capable of activating other PARs, such as PAR2 (46). Therefore, we characterized the effect of PAR2 activation on Zn 2ϩ inhibition. Cultured cortical neurons  FT1 and FT2 under Ϫ zinc). When the NR2A ATDbound Zn 2ϩ -IDA-agarose was eluted with 50 mM EDTA, all bound NR2A ATD protein appeared in the elutes (lanes E1 and E2 under ϩ zinc). NR2A ATD proteins were probed with an anti-c-Myc antibody (1:10,000, Sigma) against the COOH-terminal c-myc inserted into the NR2A ATD construct. All experiments (n ϭ 2) were performed in duplicate (e.g. FT1 and FT2, E1 and E2). The right panel shows a shorter exposure of the same Western blot in the left panel to highlight that NR2A ATD proteins that were eluted as a clear single band. M, protein standards (kDa). W1, first wash after incubation of NR2A ATD and Zn 2ϩ -IDA-agarose. B, a competitive Zn 2ϩ -IDA-agarose binding assay showed NR2A ATD proteins containing H44G and H128A double mutations bound weaker than the wild-type NR2A ATD when challenged with 2 low affinity divalent chelating agents, imidazole and L-histidine. The immunoblot was labeled with the anti-c-Myc antibody (1:10,000, Sigma). L, WT, NR2A ATD . FT, flowthrough. MT, NR2A ATD -(H44G/H128A). E, elute with consecutive agent. Concentrations of each agent are indicated.
were pretreated with the PAR2 agonist peptide 2-furoyl-LI-GRLO for 30 min at room temperature, and Zn 2ϩ inhibition was evaluated. We found that 1 M Zn 2ϩ inhibited agonistevoked NMDA receptor currents by 50 Ϯ 2.2% (n ϭ 3). There is no significant difference (p ϭ 0.54; unpaired t test; Fig. 7C) between the wild-type control and the 2-furoyl-LI-GRLO treatment. These data indicate that activation of PAR2 receptors does not influence Zn 2ϩ inhibition. In addition, the relief of Zn 2ϩ inhibition by plasmin treatment is independent of PAR2 activation.

DISCUSSION
Serine proteases play important roles not only in coagulation and thrombolysis, but also in inflammatory and proliferative responses triggered by tissue injury, as well as in signal pathways in the central nervous system (31,33). Recent studies indicate serine proteases modulate the function of NMDA receptors by indirect or direct mechanisms (37,38,40). The present study demonstrates that nanomolar concentrations of plasmin cleave the NR2A subunit of native and recombinant NMDA receptors, and the cleavage occurs after Lys 317 on the NR2A aminoterminal domain. Plasmin cleavage significantly reduced the Zn 2ϩ inhibition on agonist-evoked currents of NMDA receptors from both recombinant transfected HEK 293 cells and cultured mouse cortical neurons. The relief of Zn 2ϩ inhibition by plasmin treatment was rescued by exchange of Lys 317 to alanine. These data suggest that plasmin cleavage at Lys 317 removes the amino-terminal domain that harbors the high affinity Zn 2ϩ binding site. This finding holds implications for NMDA receptor-mediated toxicity, tPA treatment of stroke, and potential identification of biomarkers of central nervous system injury.
Plasmin Cleavage of NMDA Receptors-Plasmin can cleave numerous proteins, including fibrinogens, PARs, as well as probrain-derived neurotrophic factor (47)(48)(49). In our current study, plasmin can cleave NR2A ATD recombinant proteins and the NR2A amino-terminal in membrane preparations from rat cortex. A ϳ40-kDa fragment is removed by plasmin that encodes the NR2A ATD (Fig. 1A). Following plasmin treatment, sequencing the cleavage products of the NR2A amino-terminal revealed only one fragment with readable sequence that matched NR2A. Lys 317 was identified as the plasmin cleavage site by sequencing the cleavage fragments of NR2A ATD recombinant proteins. This cleavage site is similar to the cleavage site of other plasmin substrates ( Table 1). The ability of a point mutation to block the effects of plasmin suggests that plasmin cleaves the NR2A amino-terminal at only one site, although we cannot exclude the possibility that there are other plasmin cleavage sites on the NR2A. Some residual inhibition by 1 M Zn 2ϩ is still observed after plasmin treatment in both recombinant NR1/NR2Atransfected HEK 293 cells and cultured cortical neurons by 20 -30%. We interpret this remaining inhibition to be a combination of incomplete cleavage of the ATD by plasmin and the presence of other NR2B, -2C, or -2D subunit-containing NMDA receptors in cultured cortical neurons. For example, Zn 2ϩ inhibits NR2B with an IC 50 of 800 nM, suggesting some inhibition might be caused by NR1/NR2B (27,50). In addition, a small amount of inhibition caused by voltage-dependent Zn 2ϩ inhibition may contribute to the residual inhibition.
Interaction between Proteases, PARs, and NMDA Receptors-Recent studies (45) suggest that activation of astrocytic PAR1 triggers G-protein-mediated release of glutamate, which causes the activation of neuronal NMDA receptors and potentiation of synaptic NMDA receptor function by a Mg 2ϩ -dependent mechanism that appears secondary to depolarization. The protease thrombin potentiates NMDA receptor currents in rat hippocampal neurons (37). Plasmin also potentiates the NMDA receptor-dependent component of miniature excitatory postsynaptic currents and increases NMDA-induced whole cell receptor currents recorded from CA1 pyramidal cells (38). These studies showed that the protease modulates NMDA receptor function through an indirect mechanism, such as PAR1 activation. Comparison of the level of NMDA receptor potentiation produced in thrombin-treated hippocampal slices, TFLLR-treated hippocampal slices, and plasmintreated hippocampal slices shows striking differences. The peptide TFLLR robustly activates PAR1 without cleaving NR2A, and potentiates NMDA receptor function 1.4-fold. By contrast, treatment of hippocampal slices by plasmin, a weak activator of PAR1, leads to an enhancement of NMDA responses of 2.4-fold, considerably more than observed with more robust PAR1 activation by TFLLR. Our interpretation of this difference in the levels of potentiation is that plasmin treatment both activates PAR1 and cleaves the Zn 2ϩ binding domain, removing tonic high affinity inhibition by ambient concentrations of Zn 2ϩ present. Thus, the two different mechanisms engaged by plasmin enhance NMDA receptor function considerably more than activation by PAR1 alone.
In addition, functionally relevant cleavage of the extracellular portion of NMDA receptors by proteases has been reported by several laboratories. tPA appears to enhance NMDA receptor function and has been suggested to cleave the ATD of the NR1 subunit at arginine 260 (39,40). Thrombin and plasmin also have been suggested to cleave the NR1 subunit (37,44). Calpain proteolysis of the COOH termini of NR2A, -2B, and -2C subunits results in NMDA receptor degradation and reduced receptor activity (51)(52)(53). Therefore, in addition to indirectly mediating the functional response of NMDA receptors, proteases also can directly modulate NMDA receptor responses through cleavage of the receptors' subunits.
We do not yet fully understand the relative contribution of PAR1 activation and plasmin cleavage of NMDA receptors in normal and pathological situations. For thrombin, it is clear that PAR1 activation occurs at a lower concentration than NR1 cleavage (37). Plasmin appears to activate PAR1 receptors (38) in the same concentration range as it cleaves NR2A. The tPA/ plasminogen/plasmin system is thought to play important roles in normal brain functions, including synaptic plasticity, learning, and memory, as well as in a number of pathophysiological conditions, including ischemic stroke, neurodegenerative disease, and seizures (54 -57). Our study suggests that if tPA can generate a significant amount of plasmin, this could lead to the cleavage of NR2A to reduce Zn 2ϩ inhibition, which can occur either through tonic levels of extracellular Zn 2ϩ or activity-dependent release of Zn 2ϩ (58).
The Significance of Plasmin Cleavage of the NR2A ATD -tPA converts inactive plasminogen into the active protease plasmin, which in turn catalyzes cleavage of fibrin into soluble degraded fragments to facilitate clot dissolution. A successful study by the NINDS, National Institutes of Health, in 1995 using recombinant tPA within 3 hours of acute stroke (59) led to the approval of recombinant tPA (Alteplase) by the FDA as a thrombolytic therapy in acute stroke to reperfuse ischemic tissue. However, the use of tPA is limited by the narrow time window of application, high risk of hemorrhagic transformation, as well as the potential harmful extravascular side effects (60,61). Plasmin, the catalyzed end product of tPA that facilitates blood clot dissolution, might play important roles in excitotoxic neurodegeneration by proteolytic degradation of the extracellular matrix and immune activation (62,63). Injection of plasmin or plasminogen into the brain significantly increased the number of apoptotic neurons and the injury area of the intracortical hemorrhage model in rats (63). Harmful consequences of tPA and plasmin in ischemic animal models have been reported (Refs. 37 and 64 -66, but see Ref. 67). Our results add an additional potential mechanism that could mediate the harmful effects of tPA/plasmin in the brain. Although tPA can cross the intact blood-brain barrier (68), high concentrations of blood-derived proteases such as tPA and plasmin are unlikely to develop in brain tissue while the blood-brain barrier is intact. Blood-derived tPA can enter in brain tissue in situations in which the blood-brain barrier is disrupted (such as hemorrhagic stroke, aneurysm rupture, penetrating head wound, infection, inflammation, status epileptics, or tPA treatment of stroke) (35). Increased levels of plasmin may cleave the NR2A ATD , and the cleaved NR2A fragments may enter into the peripheral circulatory system. Detecting these NR2A ATD fragments may offer the possibility to evaluate some features of ischemia. It has previously been proposed that NMDA receptor NR2 subunits are degraded, and the proteolytic fragments enter into the peripheral blood of patients with transient ischemic attack and stroke, resulting in the formation of autoantibodies to the extracellular ligand binding domains of NR2A and NR2B that can be detected by enzymelinked immunosorbent assay (United State patent 6896872) (69 -71). Because these fragments are detectable, the plasmincleaved NR2A ATD fragments may be capable of entering the bloodstream. Our findings showing that plasmin cleavage of the NR2A ATD provide one potential mechanism whereby NMDA receptor fragments could be generated in ischemic stroke. These data may also provide a rationale for monitoring the serum levels of extracellular NR2A fragments in patients receiving tPA.
In conclusion, these data together suggest that plasmin can cleave the amino-terminal domain of the NR2A subunit in recombinant receptors and native tissues and subsequently relieve Zn 2ϩ inhibition. Our results indicate that the protease plasmin may play a role in the modulation of NMDA receptor function and may provide a possible mechanism underlying the harmful effects of tPA and plasmin in the brain. In situations where plasmin may enter brain, such as during tPA treatment for ischemic stroke or break-down of the blood-brain barrier, removal of Zn 2ϩ inhibition could exacerbate NMDA receptormediated neuronal damage, as well as provide a biomarker (NR2A ATD ) that might enter into blood.