Identification of Membrane-bound Variant of Metalloendopeptidase Neurolysin (EC 3.4.24.16) as the Non-angiotensin Type 1 (Non-AT1), Non-AT2 Angiotensin Binding Site*

Background: Angiotensin II, the renin-angiotensin system effector peptide, interacts with a recently discovered binding site that is distinctly different from its classic receptors. Results: A radioiodinated angiotensin II photoprobe bound to a ∼75-kDa membrane protein, enabling its isolation and identification. Conclusion: Membrane-bound metalloendopeptidase neurolysin (EC 3.4.24.16) is the novel angiotensin-binding protein. Significance: This metalloendopeptidase may be a crucial component of the renin-angiotensin system. Recently, we discovered a novel non-angiotensin type 1 (non-AT1), non-AT2 angiotensin binding site in rodent and human brain membranes, which is distinctly different from angiotensin receptors and key proteases processing angiotensins. It is hypothesized to be a new member of the renin-angiotensin system. This study was designed to isolate and identify this novel angiotensin binding site. An angiotensin analog, photoaffinity probe 125I-SBpa-Ang II, was used to specifically label the non-AT1, non-AT2 angiotensin binding site in mouse forebrain membranes, followed by a two-step purification procedure based on the molecular size and isoelectric point of the photoradiolabeled binding protein. Purified samples were subjected to two-dimensional gel electrophoresis followed by mass spectrometry identification of proteins in the two-dimensional gel sections containing radioactivity. LC-MS/MS analysis revealed eight protein candidates, of which the four most abundant were immunoprecipitated after photoradiolabeling. Immunoprecipitation studies indicated that the angiotensin binding site might be the membrane-bound variant of metalloendopeptidase neurolysin (EC 3.4.24.16). To verify these observations, radioligand binding and photoradiolabeling experiments were conducted in membrane preparations of HEK293 cells overexpressing mouse neurolysin or thimet oligopeptidase (EC 3.4.24.15), a closely related metalloendopeptidase of the same family. These experiments also identified neurolysin as the non-AT1, non-AT2 angiotensin binding site. Finally, brain membranes of mice lacking neurolysin were nearly devoid of the non-AT1, non-AT2 angiotensin binding site, further establishing membrane-bound neurolysin as the binding site. Future studies will focus on the functional significance of this highly specific, high affinity interaction between neurolysin and angiotensins.

The renin-angiotensin system (RAS), 2 one of the phylogenetically older hormonal systems, is most known for its essential role in regulation of hydromineral and cardiovascular homeostasis (1). The system was introduced to the scientific community by the discovery of renin, the main enzyme generating effector peptides of RAS, by Tigerstedt and Bergmann in 1898 (2). With the discovery of the existence of the two major angiotensin II (Ang II) receptor subtypes (AT 1 and AT 2 ) in the late 1980s, most of the questions related to the biochemical, pharmacological, and physiological features of the RAS seemed to be resolved (3). Based upon extensive studies of the RAS for over a century, three classes of drugs targeting different levels of the system (angiotensin-converting enzyme inhibitors (-prils), angiotensin receptor blockers (-sartans), and renin inhibitors (-kirens)) are in worldwide clinical use. However, new discov-eries extending fundamental aspects of RAS continue to occur. Among such discoveries are agonistic autoantibodies against angiotensin type 1 receptor (4), the existence of a homolog of angiotensin-converting enzyme (ACE2) (5,6), the novel role of (pro)renin acting at a specific receptor (7), physiologically important actions of angiotensin 1-7 and angiotensin IV through their specific receptors (8 -10), the finding of angiotensin 1-12 as an alternative precursor of effector peptides of the RAS (11), and the presence of a physiologically relevant intracellular RAS (12)(13)(14).
Recently, in the course of radioligand receptor binding studies of brain angiotensin receptors, we discovered a novel, non-AT 1 , non-AT 2 angiotensin binding site in rat brain membranes (15), which was also confirmed in mouse and human brain membranes (16,17). The non-AT 1 , non-AT 2 binding site has nanomolar affinity and high specificity for angiotensins I, II, and III and is pharmacologically, biochemically, and anatomically different from classical angiotensin type 1 (AT 1 ) and type 2 (AT 2 ) receptors, G-protein coupled Mas receptor, and neprilysin (EC 3.4.24.11, also known as neutral endopeptidase), a protease involved in the processing of angiotensins (15)(16)(17)(18)(19). Increased expression of the novel binding site was strongly associated with neuronal cell death following oxidative stress (20), whereas it is reduced in spontaneously hypertensive rat brains (21). However, little is known about the functional significance of these changes and molecular mechanisms of their potential effects. A unique feature of the novel angiotensin binding site is that it is unmasked (i.e. attains high affinity for angiotensins) in the presence of optimal concentrations of organomercurial sulfhydryl reagents p-chloromercuribenzoate (PCMB) or p-chloromercuribenzenesulfonate (15,20). This effect is reversed by glutathione and disulfide-reducing agents, suggesting the involvement of cysteine residues (thiol groups) in unmasking and function of the binding site, with a potential role of extracellular redox state in regulation of the function of this protein (15,20).
The purpose of this study was to identify the non-AT 1 , non-AT 2 angiotensin binding site. Here we report experiments involving photoradiolabeling, stepwise electrophoretic, and mass spectroscopic methods, followed by molecular-genetic and receptor pharmacological approaches used to isolate, characterize, and establish the identity of the binding protein.

EXPERIMENTAL PROCEDURES
Animals-Postnatal day 10 (P10) mouse brains were collected from in-house-born pups of both genders from timed pregnant CD-1 female mice (Charles River Laboratories) and stored at Ϫ80°C until their use. Animals were maintained in 12-h light/dark cycle and fed ad libitum. All procedures were carried out according to a protocol approved by the Texas Tech University Health Sciences Center institutional animal care and use committee (IACUC). The choice of P10 mouse forebrains as a source for isolation and identification of the non-AT 1 , non-AT 2 angiotensin binding site was based on our recent observation indicating ϳ5-fold higher density of the binding site in P10 compared with adult mouse forebrain membranes (20). Neurolysin (EC 3.4.24.16) knock-out mice were generated using the commercially available embryonic stem cell line NPX481 (Baygenomics). A detailed description of the generation and characterization of these and wild-type mice will be published elsewhere.
Cell Culture and Transfections-HEK293 (human embryonic kidney) cells were purchased from American Type Culture Collection and maintained according to their recommended protocol. Transfection-ready plasmids with mouse cDNA clone of neurolysin or thimet oligopeptidase (EC 3.4.24.15) and empty vector (pCMV6-Kan/Neo) were obtained from OriGene Technologies. Cell transfections were carried out with Lipofectamine TM LTX (Invitrogen) as described by the manufacturer. After 24 h, cell medium was replaced by fresh medium containing 1 mg/ml G418 for about 3 weeks to select stably transfected cells. To eliminate individual colony artifacts, colonies of cells resistant to G418 were pooled together for propagation and maintained under a selection pressure of 0.5 mg/ml G418 (22).
Photoaffinity Labeling and Radioligand Binding-A photoprobe analog of Ang II, sarcosine 1 -p-benzoyl-L-phenylalanine 8 -angiontensin II (SBpa-Ang II), which was extensively used for molecular-structural studies of AT 1 and AT 2 receptors (23,24), was prepared at the Department of Pharmacology, Université de Sherbrooke. Radioiodination of SBpa-Ang II and sarcosine 1 -isoleucine 8 -angiotensin II (SI-Ang II) was carried out at American Radiolabeled Chemicals or the University of Florida.
Photoradiolabeling of the non-AT 1 , non-AT 2 angiotensin binding site with 125 I-SBpa-Ang II was conducted using an established procedure (25) except that neurolysin knock-out and wild-type or P10 CD-1 mouse forebrains or HEK293 cells were used for preparation of crude membranes, and 2 M final assay concentration of ZD7155 and 20 M PD123319 (Tocris Bioscience) were used to block AT 1 and AT 2 receptors, respectively. Photoradiolabeling experiments designed for purification of the angiotensin binding site utilized a mixture of 127/125 I-SBpa-Ang II (95 and 5%, respectively), to have a minimal amount of radioactivity in the final samples intended for LC-MS/MS analysis.
Radioligand binding studies of the non-AT 1 , non-AT 2 binding site in HEK293 cells and neurolysin knock-out and wildtype mouse forebrain membranes using 125 I-SI-Ang II were carried out according to our established procedures (15,20). In these experiments, 2 M ZD7155 and 20 M PD123319 were used to block AT 1 and AT 2 receptors, respectively. All photoradiolabeling and radioligand binding experiments of the novel binding site were carried out in the presence of a 150 M final assay concentration of p-chloromercuribenzoate (PCMB) to unmask the non-AT 1 , non-AT 2 angiotensin binding site (15,20).
Photoradiolabeling and radioligand binding studies of AT 1 and AT 2 receptors in neurolysin knock-out and wild-type mouse brains were carried out essentially as described previously (26) in combined membrane preparations of hypothalamus, thalamus, and brainstem using 125 I-SBpa-Ang II or 125 I-SI-Ang II, respectively. In these experiments, an additional mixture of protease inhibitors (o-phenanthroline, puromycin, phenylmethanesulfonyl fluoride, and glutamate phosphonate) was used to protect the ligands and receptors from proteolytic degradation (26,27). To avoid aggregation of AT1 and AT2 receptors, photoradiolabeled membrane pellets were dissolved in modified Laemmli sample buffer (containing 5 M urea and 0.125 M dithiothreitol instead of 2-mercaptoethanol), periodically vortexed for 30 min at room temperature, and resolved on 7.5% Tris-HCl Ready Gels (Bio-Rad) without prior boiling. For all experimental procedures, protein concentration was determined by the BCA assay (Pierce) using bovine serum albumin as a standard.
In Vitro Receptor Autoradiography-In vitro receptor autoradiography studies in neurolysin knock-out and wild-type mouse forebrain coronal sections using 125 I-SI-Ang II were carried out essentially as described (18).
Protein Purification and Two-dimensional Gel Electrophoresis-Crude membrane preparations of P10 mouse forebrains (ϳ4 g total wet weight starting material) were pelleted after photoradiolabeling and multiple washes, solubilized in SDS sample buffer, and separated in 10% Tris-HCl preparative Criterion gels (Bio-Rad). Gel sections corresponding to a ϳ75-kDa region were combined from multiple gels, and the radioactivity was extracted into Tris-glycine SDS-PAGE running buffer at 4°C for 4 days (Ͼ90% recovery of the iodine-125). The extracted sample was concentrated using centrifugal filtering units (Amicon Ultra and Nanosep Omega), and an aliquot was saved for two-dimensional gel electrophoresis. The sample was further separated by isoelectric focusing using one-dimensional pH gradient strips (pH 3-10, 11-cm ReadyStrip TM IPG; Bio-Rad). Strip sections from the pH ϳ5.5-7.0 region were combined, the radioactivity was extracted, and the sample was concentrated as described above.
Representative gels after SDS-PAGE or two-dimensional gel electrophoresis were stained with Bio-Safe Coomassie Blue (Bio-Rad), dried in a vacuum gel drier, and incubated with x-ray film and intensifying screen at Ϫ80°C for 2-5 days for autoradiographic visualization of the photoradiolabeled proteins.
Mass Spectrometry Analysis and Identification of Proteins-After two-dimensional gel electrophoresis of the final purified sample, the region of the Coomassie Blue-stained gel containing radioactive signal was cut and stored at 4°C to decay the radioactivity to background levels. Mass spectrometry analysis was performed on trypsin-treated gel segments. In brief, gel pieces were diced into 1-mm squares, rinsed with water and 50 mM ammonium bicarbonate buffer, and dehydrated. Reduction of disulfide bonds was conducted with dithiothreitol, followed by alkylation with iodoacetamide. Proteins were digested by rehydrating the gel slices in 20 g/ml trypsin (Promega) in ammonium bicarbonate buffer plus 10% acetonitrile for 1 h at 24°C, followed by overnight incubation at 37°C and a second addition of trypsin the next day for 3 h. The digested material was extracted from the gel, combined, and dried, using a vacuum concentrator. 10 -20% of the digest was loaded on a Magic C18 AQ (Michrom) nanospray tip on a Thermo LTQ mass spectrometer and washed with 5% methanol, 0.1% formic acid for 10 min before peptide elution began, using a 5-60% meth-anol gradient. The LTQ ion trap mass spectrometer was equipped with a nanoelectrospray ionization source, running a full MS survey scan every 3 s in the data-dependent mode to collect the MS/MS fragmentation spectrum. The MS and fragmentation spectrum data were used in a Mascot search of the whole mouse proteome. Mascot search parameters included precursor and fragment ion mass tolerance of 1.5 and 0.8 daltons, respectively, one 13 C incorporation, one missed trypsin cleavage site, fixed carbamidomethyl-cysteine modification, and variable methionine oxidation, against the complete mouse proteome (NCBI 20100930). The ion score/Expect cut-off score was set for 15, and peptides with an Expect score less than 0.05 were considered positive identification if more than one peptide was identified for a given protein.
Western blotting of these proteins was carried out essentially as described by Alfonso et al. (29), using ␤-actin as a loading control (Sigma; product number A5441). The only exception was that ␣-adducin and thimet oligopeptidase samples were not boiled prior to SDS-PAGE.
Statistical Analyses-Determination of B max (fmol of radioligand bound/mg of protein) and K d (nM) values were carried out using one-site saturation binding models of Prism software (GraphPad). Values reported were significantly different from zero (95% confidence interval) and are presented as mean Ϯ S.E. Comparisons of radioactivity in immunoprecipitates, B max and K d values obtained from radioligand binding experiments in membrane preparations of HEK293 cells, and specific binding to AT 1 and AT 2 receptors in neurolysin knock-out and wild-type control animal brains were carried out by one-way analysis of variance followed by the Tukey-Kramer test. Values reported are mean Ϯ S.E.

RESULTS
Photoradiolabeling of the non-AT 1 , non-AT 2 angiotensin binding site with 125 I-SBpa-Ang II in mouse forebrain membranes was demonstrated by SDS-PAGE analysis of samples. Specific and reproducible photoradiolabeling of a ϳ75-kDa protein was observed, with minor labeling at ϳ50 kDa (Fig. 1). As expected, nonspecific (NS) samples, which contained saturating concentrations of Ang II to prevent 125 I-SBpa-Ang II binding to the binding site, showed negligible radioactive signal at ϳ75 and ϳ50 kDa (Fig.  1). Therefore, all experiments were carried out with parallel use and comparison of total and nonspecific samples to avoid tracking of nonspecifically photoradiolabeled protein(s).
To obtain sufficient material for purification and identification of the binding protein, multiple preparative SDS-PAGE separations were performed using photoradiolabeled mouse forebrain membranes. Gel slices from the ϳ75-kDa region were pooled together from all preparative runs, and proteins were extracted, concentrated, and separated by two-dimensional gel electrophoresis (Fig. 2a). The photoradiolabeled protein was reproducibly detected by autoradiography after two-dimensional gel electrophoresis (Fig. 2b). To further purify the photoradiolabeled protein, the concentrate of proteins from the 75-kDa regions was subjected to preparative isoelectric focusing on one-dimensional pH 3-10 gradient strips. Next, slices of strips containing the radioactive signal (pH ϳ5.5-7.0) were collected and combined. Proteins were extracted and concentrated, followed by two-dimensional gel electrophoresis of the sample (Fig. 2c). Upon reproducible detection of the photoradiolabeled protein following the purification steps (Fig. 2d), the region of the two-dimensional gel containing the radioactive signal was cut out and used for identification of proteins by LC-MS/MS analysis.
Five independent LC-MS/MS analyses were performed using the two-dimensional gel excisions, where 32-54 proteins were identified in each (supplemental Table 1). Because all two-dimensional gel excision samples contained radioactive signal (i.e. photoradiolabeled non-AT 1 , non-AT 2 binding site), we expected to see the protein of interest detected in all LC-MS/MS analyses. For this reason, only the proteins repeatedly identified in all five LC-MS/MS runs were shortlisted for further analysis (Table 1). These proteins included neurolysin (EC 3.4.24.16, accession number gi͉28077051), collapsin response mediator protein 2 (accession number gi͉40254595), thimet oligopeptidase (EC 3.4.24.15, accession number gi͉239916005), ␣-adducin (accession number gi͉6851286), collapsin response mediator protein 1 (accession number gi͉74186635), G-protein signaling modulator 1 (accession number gi͉32189378), long-chain fatty acid-CoA ligase (accession number gi͉16716465), ezrin (accession number gi͉74179741), and serine protease 1 (accession number gi͉16716569). Of these nine proteins, serine protease 1 had an estimated molecular mass of 26.8 kDa and was considered unlikely to be the angiotensin-binding protein and consequently excluded from further analysis. The remaining eight protein candidates were ranked using their Mascot "Score" and "Peptide query" (or spectral counts) values. These data indicated consistently higher amounts of neurolysin, thimet oligopeptidase, collapsin response mediator protein 2, and ␣-adducin in the analyzed samples. Therefore, these proteins were prioritized for investigation to test the candidacy of the non-AT 1 , non-AT 2 binding site (supplemental Fig. 1).
The results of immunoprecipitation studies using specific antibodies against the top four candidates (neurolysin, thimet oligopeptidase, collapsin response mediator protein 2, and ␣-adducin) in photoradiolabeled mouse forebrain membranes are summarized in Fig. 3. "Total" samples of immunoprecipitates with anti-neurolysin antibody contained a substantial amount of radioactive signal (Fig. 3a), indicating that the non-AT 1 , non-AT 2 angiotensin binding site is probably neurolysin. Total samples of anti-thimet oligopeptidase immunoprecipitates contained a marginal level of radioactivity, which could be due to cross-reactivity of the anti-thimet oligopeptidase antibody with neurolysin or a small degree of radiophotolabeling of thimet oligopeptidase. Total samples of collapsin response mediator protein 2 and ␣-adducin showed negligible amounts of radioactive signal. To ensure that the proteins of interest were indeed immunoprecipitated from photoradiolabeled membrane samples, their presence in the immunoprecipitates was documented by Western blotting (Fig. 3, b-e).
Stable cell lines overexpressing mouse neurolysin or thimet oligopeptidase in HEK293 cells were established to further test the identity of the angiotensin-binding protein. Overexpression of both proteins in HEK293 cells was confirmed by Western blotting (Fig. 4, a and b). Radioligand binding assays in membrane preparations of these cells using 125 I-SI-Ang II showed dramatically increased density of the novel angiotensin binding sites only in HEK293 cells overexpressing neurolysin (B max ϭ 533 Ϯ 69 fmol/mg protein) compared with non-transfected, empty vector-transfected, and thimet oligopeptidaseoverexpressing HEK293 cell membranes (108 Ϯ 25.2, 94.6 Ϯ 20.1, and 90.4 Ϯ 7.9 fmol/mg protein respectively; n ϭ 3) (Fig. 4,  c and d). Notably, affinity of the radioligand for the novel binding site did not differ significantly among the four groups (K d ϭ 3.35 Ϯ 0.41, 2.11 Ϯ 0.57, 1.75 Ϯ 0.42, and 2.53 Ϯ 0.58 nM, respectively). Photoradiolabeling experiments carried out in membrane preparations of the same cells also indicated substantially higher photoradiolabeling of the binding site in cell membranes overexpressing neurolysin compared with thimet oligopeptidase or control groups (Fig. 4, e and f).
Forebrain membranes from neurolysin knock-out (KO) mice were compared with wild-type (WT) mice to determine if the non-AT 1 , non-AT 2 binding site was neurolysin. Absence of neurolysin in the neurolysin KO mouse forebrains was documented by Western blotting (Fig. 5a). Saturable binding of 125 I-SI-Ang II to the non-AT 1 , non-AT 2 angiotensin binding site in neurolysin KO forebrain membranes was not statistically significantly different from zero based on 95% confidence interval values (Fig. 5b). In contrast, there was a substantial amount of high affinity, saturable binding of 125 I-SI-Ang II (K d ϭ 1.13 Ϯ 0.21 nM, B max ϭ 54.4 Ϯ 12.7 fmol/mg protein, n ϭ 3) in wildtype forebrain membranes (Fig. 5c). Photoradiolabeling experiments in the same membrane preparations confirmed the presence of the expected radioactive signal from WT control samples and the absence of radioactive signal in neurolysin KO forebrain membranes (Fig. 5, d and e). In vitro receptor autoradiography studies in coronal sections of neurolysin KO and WT mouse forebrains also verified these observations, indicating high intensity binding of 125 I-SI-Ang II in WT and a paucity of binding in neurolysin KO brains (supplemental Fig. 2).
The presence of classic AT 1 and AT 2 receptors in neurolysin KO and WT animal brains was tested by photoradiolabeling and radioligand binding experiments. Ang II and/or ZD7155 (AT 1 receptor antagonist) and PD123319 (AT 2 receptor antagonist) blocked photoradiolabeling of a diffuse band of protein at ϳ55-60 kDa by 125 I-SBpa-Ang II (supplemental Fig. 3). The presence of AT 1 and AT 2 receptors in neurolysin KO and WT mouse brains was clearly distinguished by radioligand binding experiments using 125 I-SI-Ang II (supplemental Fig. 4).

DISCUSSION
The recently discovered non-AT 1 , non-AT 2 angiotensin binding site in rodent and human brain membranes has dis-  ; collapsin response mediator protein 1 (CRMP1); G-protein signaling modulator 1 (Gpsm1); long-chain fatty acid-CoA ligase (LCFA-CoA-L); ezrin, and serine protease 1. b Rank number reflects top Mascot score position in a sample. c emPAI value reflects protein abundance, using spectral counts and normalizing by protein size and expected peptide coverage. d Also known as spectral counts.

Identification of Non-AT 1 , Non-AT 2 Angiotensin Binding Site
tinctly different biochemical properties and tissue distribution from classic angiotensin receptors and key angiotensin-processing enzymes (15)(16)(17)(18)(19). Many features of this protein (high affinity and pharmacological specificity for angiotensins; conservation in brains of numerous species, including humans; localization on the plasma membrane, where it can interact with extracellular ligands; and association of expression of this protein with neuronal cell death and neurogenic hypertension) strongly favor its functional significance (16,20). However, the unknown identity of the novel angiotensin binding site has limited our understanding of the (patho)physiological function of this hypothesized new member of the RAS.
The photoaffinity probe 125 I-SBpa-Ang II specifically and covalently photoradiolabeled the non-AT 1 , non-AT 2 angiotensin binding site in mouse forebrain membranes (Fig. 1). The radioactive signal was tracked and purified through electrophoretic procedures based on molecular size and isoelectric point of the binding protein (Fig. 2). Two-dimensional gel electrophoresis and LC-MS/MS analysis of the radioactive protein spot revealed eight candidates (Table 1 and supplemental Table  1), of which the four most abundant (neurolysin (EC 3.4.24.16), thimet oligopeptidase (EC 3.4.24.15), collapsin response mediator protein 2, and ␣-adducin) were immunoprecipitated from photoradiolabeled mouse forebrain membranes. Immunoprecipitates of neurolysin contained substantial radioactive signal, in contrast to negligible radioactivity in the other three protein immunoprecipitates (Fig. 3).
Radioligand binding and photoradiolabeling experiments independently showed "gain of phenotype" (i.e. significantly increased density of the angiotensin binding site) in mem-branes of HEK293 cells overexpressing neurolysin but not thimet oligopeptidase or control groups (Fig. 4).
Radioligand binding, photoradiolabeling, and in vitro receptor autoradiography experiments performed in neurolysin KO and WT mouse forebrains revealed a "loss of phenotype" (i.e. loss of high affinity specific binding to the angiotensin binding site) in neurolysin KO animals ( Fig. 5 and supplemental Fig. 2), further establishing membrane-bound neurolysin as the non-AT 1 , non-AT 2 angiotensin binding site. Although the existence of another angiotensin binding protein cannot be excluded, our data indicate that any other binding site has much lower affinity for angiotensins and is expressed in substantially lower levels than neurolysin, precluding its quantification by the methods used in our study.
Separate photoradiolabeling (in the absence of PCMB) and radioligand binding assays indicated that AT 1 and AT 2 receptors were still present in the neurolysin knock-out mouse brains, endorsing earlier observations distinguishing this novel binding site from classic Ang II receptors (16).
Additional evidence supporting neurolysin as the angiotensin binding site is the lack of glycosylation of neurolysin (30,31) and the non-AT 1 , non-AT 2 angiotensin binding site (25), similar pI values for both proteins (31), and its presence in membrane preparations of primary neurons but not astroglial cells (20,32).
A minor photoradiolabeling of a ϳ50-kDa protein was observed in addition to the major ϳ75-kDa protein (Figs. 1b,  4d, and 5d). For a number of reasons, it is likely that the photoradiolabeled protein at ϳ50 kDa is a degradation product or a shorter form of neurolysin. First, similar to the ϳ75-kDa protein, the radioactive signal of the ϳ50-kDa photoradiolabeled protein was substantially increased in membrane preparations of HEK cells overexpressing neurolysin compared with all other groups (Fig. 4d). Second, the radioactive signal of the ϳ50-kDa photoradiolabeled protein, like the ϳ75-kDa protein, was negligible in membrane preparations of neurolysin KO mouse forebrains (Fig. 5d). Third, at least four sequences for mouse neurolysin are published to date (NCBI protein database) with 752-, 704-, 683-, and 533-amino acid-long sequences and 85.3-, 80.4-, 77.9-, and 60.7-kDa predicted masses, indicating that different isoforms of this protein may exist.
Neurolysin (EC 3.4.24.16) is a zinc metalloendopeptidase belonging to the M3 family that is ubiquitously distributed in central nervous system and peripheral tissues of mammals (33). It was first detected and purified from rat brain synaptic membranes (34) and is also known as liver-soluble angiotensin-binding protein, mitochondrial oligopeptidase, microsomal endopeptidase, oligopeptidase M, and thimet oligopeptidase II (35). Neurolysin exists primarily as a cytosolic protein; however, depending on cell type, it can also be secreted, bound to the plasma membrane, or targeted to mitochondria (30 -32, 36). Neurolysin has Ͼ60% amino acid sequence identity with thimet oligopeptidase (EC 3.4.24.15) (33). These oligopeptidases share several substrates, including neurotensin, bradykinin, angiotensins, opioids, substance P, somatostatin, and hemoglobinderived peptides, and hydrolyze many of them at the same peptide bond, at comparable rates (33,(37)(38)(39)(40). Additionally, accumulating evidence indicates the role of these peptidases in the intracellular metabolism of peptides (41,42). Functional studies revealed the role of both peptidases in neurotensin- , ␣-adducin, and collapsin response mediator protein 2 (CRMP2) from photoradiolabeled P10 mouse forebrain membrane preparations. Immunoprecipitation of each protein was carried out from parallel nonspecific (NS) and total (Tot) photoradiolabeled groups as detailed in the legend to Fig. 1 and under "Experimental Procedures." a, equal amounts of NS and total sample of each protein immunoprecipitate were used to count the radioactive signal. Bck, background radioactivity; ***, p Ͻ 0.001 versus all other groups; n ϭ 3-4). b-e, representative Western blot confirmation of successful immunoprecipitation of each candidate protein from NS and total samples. related nociception (43,44), bradykinin-mediated regulation of blood pressure and microvascular permeability (45,46), and antigen presentation through major histocompatibility complex class I (MHC-I) (47)(48)(49). Cell type-dependent membrane association of neurolysin, constituting ϳ10% of total activity of this peptidase, was documented in several studies (31, 32, 34, 50 -53). However, membrane-anchoring mechanism(s) and how the membrane-bound, mitochondrial, and cytoplasmic variants differ are not fully understood.
Some similarities of the non-AT 1 , non-AT 2 angiotensin binding site to the liver-soluble angiotensin-binding protein (i.e. cytoplasmic neurolysin) (54 -56) were noted in our first publication documenting the existence of the novel angiotensin binding site in rat brain membranes (15). However, we could not conclude whether the angiotensin binding site was a membrane-bound variant of neurolysin or thimet oligopeptidase or a different protein because of a number of differences between our observations and observations of the soluble angiotensinbinding protein (pharmacological specificity, insensitivity toward selective neurolysin/thimet oligopeptidase inhibitor JA-2, and time sensitivity toward sulfhydryl reagents) and conflicting reports on the identity of the soluble angiotensin-binding protein as neurolysin or thimet oligopeptidase (57) as well as tissue distribution (31,58,59) and cellular localization (32, 52, 60 -62) of these two proteins (15).
In this study, biochemical, molecular-genetic, and receptor pharmacological approaches identify the non-AT 1 , non-AT 2 angiotensin binding site as a membrane-bound variant of metalloendopeptidase neurolysin (EC 3.4.24.16). A key question regarding the identity of the novel angiotensin binding site is thus resolved. As such, neurolysin may be a key enzyme responsible for sequestration or metabolic processing of angiotensins at certain (patho)physiological conditions, perhaps even more important than other peptidases. This notion is supported by recent observations indicating significantly decreased density of this angiotensin binding site in brain cardiovascular centers of spontaneously hypertensive rats (21). Thus, neurolysin may serve to limit the access of Ang II to AT 1 receptors, and its reduction might contribute to the well documented overactivation of brain RAS in spontaneously hypertensive rats. Future studies focusing on metabolic processing of angiotensins by neurolysin will provide insights about the role of this peptidase related to the function of the renin-angiotensin system.
It is noteworthy that the non-AT 1 , non-AT 2 binding site was found in high densities not only in brain areas known to contain AT 1 and AT 2 receptors (e.g. hypothalamic subnuclei, thalamic nuclei, circumventricular organs, nucleus of solitary tract, ventrolateral medulla, and amygdala) but also in regions with low/ negligible density of classic angiotensin receptors (e.g. cerebral cortex, choroid plexus, hippocampus, nucleus accumbens, substantia nigra, and substantia innominata) (18). This correlates well with Ang II immunoreactivity in the brain (63), indicating possible AT 1 and AT 2 receptor-independent and non-cardiovascular functions of RAS (e.g. neurogenesis, plasticity and memory, and cognition). The density of the binding site is substantially higher than AT 1 and AT 2 receptors (e.g. in paraventricular nucleus of hypothalamus, which is rich in AT 1 receptors, density of the binding site is ϳ15 times higher compared with the AT 1 receptor) (18). As mentioned above, neurolysin processes other neuropeptides; hence, it is possible that some functions of angiotensins involve inhibition of the ability of neurolysin to process those neuropeptides. Such a phenomenon is documented for angiotensin IV and its receptor, insulin-regulated aminopeptidase, which metabolizes several bioactive peptides, including vasopressin, CCK-8, oxytocin, and somatostatin (64). Another possibility could be activation of intracellular signaling pathways by membrane-bound neurolysin in response to angiotensins similar to several other peptidases (65).
Two important questions remaining unanswered that warrant detailed investigations include the following. What molecular mechanism(s) unmasks the high affinity angiotensin binding characteristics of neurolysin in (patho)physiological conditions? What is the functional significance of this interaction?  In neurolysin WT, B max ϭ 75.5 Ϯ 3.2 fmol/mg protein, and K d ϭ 1.56 Ϯ 0.17 nM; in neurolysin KO, B max and K d values were not significantly different from 0: B max ϭ 17.7 Ϯ 11.1 fmol/mg protein, 95% confidence interval ϭ Ϫ13.2-48.6; K d ϭ 3.87 Ϯ 4.7 nM, 95% confidence interval ϭ Ϫ9.14 -16.9. c, representative SDS-PAGE analysis of neurolysin KO and WT forebrain membranes after photoradiolabeling of the non-AT 1 , non-AT 2 angiotensin binding site with 125 I-SBpa-Ang II in parallel nonspecific (NS) and total (Tot) groups (Coomassie Blue-stained gel). d, autoradiogram corresponding to c.