Purification, Characterization, and Cloning of a Cytosolic Aspartyl Aminopeptidase*

An aminopeptidase with a preference for N-terminal aspartyl and glutamyl residues but distinct from glutamyl aminopeptidase (EC 3.4.11.7) was purified to near homogeneity from rabbit brain cytosol. Its properties were similar to an enzyme described previously (Kelly, J. A., Neidle, E. L., and Neidle, A. (1983) J. Neurochem. 40, 1727–1734). Aspartyl aminopeptidase had barely detectable activity toward simple aminoacyl-naphthylamide substrates. Its activity was determined with the substrate Asp-Ala-Pro-naphthylamide in the presence of excess dipeptidyl-peptidase IV (EC 3.4.14.5). The native enzyme has a molecular mass of 440 kDa and migrates as a single band of 55 kDa after SDS-polyacrylamide gel electrophoresis. The sequences of three tryptic peptides were used to screen the GenBankTM data base of expressed sequence tags. Human and mouse clones described as “similar to a yeast vacuolar aminopeptidase” and containing full-length cDNAs were identified and sequenced. The human cDNA was expressed in Escherichia coli. The amino acid sequence has significant homology to yeast aminopeptidase I, placing it as the first identified mammalian member of the M18 family of metalloproteinases. Homologous sequences in Caenorhabditis elegans and in prokaryotes revealed three conserved histidines, three conserved glutamates and five conserved aspartates. Aspartyl aminopeptidase is found at relatively high levels in all mammalian tissues examined and is likely to play an important role in intracellular protein and peptide metabolism.

Aminopeptidases catalyze the sequential removal of amino acids from the unblocked N termini of peptides and proteins. These enzymes are widely distributed in eukaryotes and prokaryotes (1, 2) as either integral membrane or cytosolic proteins. Aminopeptidases are generally classified in terms of their substrate specificities, i.e. preference for a neutral, acidic, or basic amino acid in the P1 position. In the case of X-Pro aminopeptidase (aminopeptidase P; E.C. 3.4.11.9), it is the amino acid in the P1Ј position that governs specificity. Most aminopeptidases are metalloenzymes, although cysteine and serine aminopeptidases have been described (2). In addition to their role in general protein and peptide metabolism, aminopeptidases have more specific functions. These include activation (3) and inactivation (4) of biologically active peptides, removal of the N-terminal methionine of newly synthesized proteins (5) and possibly in the trimming of antigens for presentation by the major histocompatibility complex-1 system (6).
The removal of N-terminal aspartyl and glutamyl residues from proteins and peptides in eukaryotes is catalyzed by glutamyl aminopeptidase (aminopeptidase A; EC 3.4.11.7), an enzyme first described by Glenner and Folk in 1961 (7) (for a review, see Ref. 8). This membrane-bound ectoenzyme is a member of the metalloproteinase family M1 and contains the HEXXH ϩ E zinc binding ligands (9). Glutamyl aminopeptidase, first cloned as the murine BP-1/6C3 antigen (10), is a protein of 945 amino acids with a molecular mass of 107.8 kDa. The purified porcine, human, rat, and mouse enzymes are isolated as homodimers (8). A distinguishing feature of glutamyl aminopeptidase is its stimulation by Ca 2ϩ (11). The ability of glutamyl aminopeptidase to degrade angiotensins I and II is of considerable interest. The action of glutamyl aminopeptidase on angiotensin II (Asp-Arg-Val-Tyr-Ile-His-Pro-Phe) yields des-Asp angiotensin II, also known as angiotensin III. There is evidence that angiotensin III may mediate some of the effects of the renin-angiotensin system in the brain (12). A recent study utilizing 3-amino-4-thio-butyl sulfonate, a specific inhibitor of this enzyme, provided evidence for a predominant role of angiotensin III in the control of vasopressin release (13). The same inhibitor was used to establish a major role of glutamyl aminopeptidase in the metabolism of cholecystokinin-8 (14).
The literature also contains two reports of a cytosolic acidic amino acid preferring aminopeptidase, which is distinct from glutamyl aminopeptidase. Cheung and Cushman (15) described the partial purification of such an enzyme from the soluble fraction of a dog kidney extract. It is activated by preincubation with Mn 2ϩ and has a preference for Asp-2-naphthylamide (Asp-NA) 1 over Glu-NA. Kelly et al. (16) have more extensively characterized a high molecular weight (450,000) acidic amino acid specific aminopeptidase in mouse brain cytosol. One of the features of the mouse brain enzyme is its inability to cleave simple NA substrates such as Asp-NA and Glu-NA. Due to its instability, this enzyme was not previously purified to homogeneity. In view of our interest in the role of angiotensin III in the brain renin-angiotensin system (17), we sought to further study the latter enzyme. A new assay was developed to measure its activity. We report on the purification, specificity, properties, subunit structure, molecular cloning, and expression of the enzyme we call aspartyl aminopeptidase.
Triethylamine and Boc-Ala-N-hydroxysuccinimide ester (0.5 mmol each) were added to a stirred solution of 0.5 mmol of compound 1 in 10 ml of DMF. The mixture was allowed to stir overnight at room temperature. The solvent was then removed by evaporation, and the residue was dissolved in chloroform. The chloroform was sequentially washed with saturated NaHCO 3 , H 2 O, 10% citrate, and H 2 O. After drying over Na 2 SO 4 , the solvent was evaporated to yield 220 mg of Boc-Ala-Pro-NA. Treatment with trifluoroacetic acid for 30 min followed by evaporation and ether precipitation, yielded 180 mg of Ala-Ala-Pro-NA⅐trifluoroacetate (yield, 36%; M ϩ 1 ϭ 384).

Asp-Ala-Pro-NA
N-Benzyloxycarbonyl-Asp(O-t-Bu)-OH and triethylamine (0.5 mmol each) were added to a stirred solution of 0.5 mmol of compound 1 in 10 ml of DMF, followed by 0.5 mmol each of 1-hydroxybenzotriazole and dicyclohexylcarbodiimide after the solution had cooled to 4°C. The reaction was allowed to proceed at 4°C for 18 h, after which time the mixture was filtered. The filtrate was evaporated to dryness, the residue was dissolved in chloroform, and the chloroform extract was washed sequentially with saturated NaHCO 3 , H 2 O, 10% citrate, and H 2 O. After drying over Na 2 SO 4 , the chloroform was removed by evaporation, and the residue was purified by silica gel chromatography. The product N-benzyloxycarbonyl-Asp(O-t-Bu)-Ala-Pro-NA was eluted with a solution of 95% chloroform, 5% ethanol. The purified product was hydrogenated in the presence of 10% Pd/C to yield 139 mg of Asp(O-t-Bu)-Ala-Pro-NA. Treatment of this product with trifluoroacetic acid for 30 min followed by evaporation and ether precipitation yielded Asp-Ala-Pro-NA (M ϩ1 ϭ 427).

Glu-Ala-Pro-NA
Equimolar amounts of Boc-Glu(O-t-Bu)-ONS (Bachem) and 1-hydroxybenzotriazole (0.5 mmol each) were added to 0.5 mmol of compound 1 dissolved in 20 ml of DMF. The mixture was stirred for 1 h at room temperature and then poured into a flask containing 200 ml of ethyl acetate and 30 ml of H 2 O. The organic layer was washed with 2 ϫ 20 ml of saturated NaHCO 3 , 2 ϫ 20 ml of H 2 O, and 20 ml of saturated NaCl. After drying over Na 2 SO 4 , the solvent was removed in vacuo. The residue was washed with 50 ml of ethyl ether and then treated with trifluoroacetic acid for 1 h. After removal of the trifluoroacetic acid under high vacuum, the product Glu-Ala-Pro-NA was obtained in 81% yield. M ϩ 1 ϭ 441.

Fmoc-Asp-Ala-Pro-NA
Equimolar amounts of triethylamine and Fmoc-Asp-(O-t-Bu)-OH (0.5 mmol) were added to a stirred solution of 0.5 mmol of compound 1 in 10 ml of DMF. The solution was cooled to 4°C, and then 0.5 mmol each of dicyclohexylcarbodiimide and 1-hydoxybenzotriazole were added. The solution was stirred at this temperature for 18 h and then filtered. The solvent was evaporated, and the residue was dissolved in chloroform and then washed with 10% citrate and H 2 O. After drying over Na 2 SO 4 and evaporation of solvent, a hygroscopic solid was obtained. Treatment with trifluoroacetic acid for 30 min followed by ether precipitation yielded 186 mg of Fmoc-Asp-Ala-Pro-NA as white crystalline material (yield 29%; M ϩ 1 ϭ 650).

Asp-Ala-Pro-SM
Equimolar amounts of triethylamine and Boc-Ala were added to 1.6 mmol of Pro-SM⅐trifluoroacetic acid (prepared by coupling Boc-Pro to SM and deprotecting with trifluoroacetic acid) and dissolved in 20 ml of THF. The solution was cooled to 4°C, equimolar amounts of dicyclohexylcarbodiimide and 1-hydroxybenzotriazole were added, and the solution was stirred at this temperature for 18 h. The reaction mixture was filtered, the filtrate was evaporated to dryness, and the residue was dissolved in chloroform and sequentially washed with saturated NaHCO 3 , H 2 O, 10% citrate, and H 2 O. After drying over Na 2 SO 4 , evaporation yielded 220 mg (0.42 mmol) of Boc-Ala-Pro-SM. Treatment with 4 N HCl in dioxane, evaporation, and crystallization by ether addition yielded 151 mg of Ala-Pro-SM⅐HCl (0.33 mmol).
Equimolar amounts of triethylamine and Fmoc-Asp(O-t-Bu)-OH were added to 0.33 mmol of Ala-Pro-SM dissolved in 10 ml of THF plus 10 ml of DMF. Coupling with dicyclohexylcarbodiimide and 1-hydroxybenzotriazole as described above yielded 201 mg of Fmoc-Asp(O-t-Bu)-Ala-Pro-SM (0.24 mmol). The Fmoc group was removed by treatment for 2 h with 5 ml each of DMF and diethylamine to yield 102 mg (0.16 mmol) of Asp(O-t-Bu)-Ala-Pro-SM. The final product was obtained in quantitative yield by treatment with trifluoroacetic acid. M ϩ 1 ϭ 537.

Peptide Synthesis
A series of tetrapeptides were synthesized on an Applied Biosystems model 430A peptide synthesizer by Fmoc chemistry. Peptide purity was evaluated by HPLC (see below).

Measurement of Enzymatic Activity
The activity of aspartyl aminopeptidase was measured by a coupled enzymatic assay similar to that described for the measurement of pyroglutamyl peptidase II (19). In this assay, the product of the action of aspartyl aminopeptidase is a substrate of dipeptidyl-peptidase IV (EC 3.4.14.5).
The reaction sequence is as follows, where C represents a chromogen. Initially, we used 2-naphthylamine as chromogen, and more recently we have used the less toxic and more soluble sulfamethoxazole. The released chromogen is measured by a colorimetric procedure following its diazotization (20). The reaction mixture contained 10 l of a 10 mM solution of substrate in Me 2 SO, 0.5 units of dipeptidyl-peptidase IV, and 50 mM Tris-HCl, pH 7.5, in a total volume of 250 l. Partially purified enzyme preparations were assayed in the presence of 0.4 mM puromycin to eliminate the possible contribution of puromycin-sensitive aminopeptidase (EC 3.4.11.14) to substrate hydrolysis.

Measurement of Enzymatic Activity in Rat Tissues
The substrate used to measure aspartyl aminopeptidase can also be cleaved in crude tissue samples by glutamyl aminopeptidase and possibly by the puromycin-sensitive aminopeptidase. Therefore, to measure aspartyl aminopeptidase in the supernatant fraction of homogenates of rat tissues, incubation buffers also contained Zn 2ϩ and puromycin, each at a final concentration of 0.4 mM. We have determined that in crude tissue extracts, 0.4 mM Zn 2ϩ totally inhibits glutamyl aminopeptidase but does not inhibit aspartyl aminopeptidase. Rat tissues were homogenized in 5 volumes of 25 mM Bicine buffer, pH 7.0, 5% glycerol, in a Potter-Elvejhem homogenizer fitted with a Teflon pestle. Homogenates were centrifuged for 30 min at 15,000 ϫ g, and the supernatant was taken for analysis. Incubation mixtures (final volume of 125 l) contained 5 l of tissue supernatant, 0.5 units of dipeptidylpeptidase IV, 5 l of 10 mM Zn 2ϩ , 5 l of 10 mM puromycin, 5 l of 10 mM Asp-Ala-Pro-SM in Me 2 SO, and Tris-HCl buffer (0.05 M, pH 7.5). Tubes were incubated at 37°C for 5 min, and the reaction was stopped by the addition of 125 l of 10% trichloroacetic acid. After centrifugation, 125 l of supernatant was removed for analysis of chromogen.

Measurement of Protein
Protein was quantitated by the method of Lowry (21). Protein elution from columns was monitored by absorbance at 280 nm.

HPLC Analyses
HPLC was performed on a Waters 600E liquid chromatograph equipped with a Vydac C4 protein column protected by a Vydac C4 5 guard column. For analysis of peptide purity, the column was equilibrated with 15% acetonitrile, 0.05% trifluoroacetic acid at a flow rate of 1 ml/min. Elution was carried out by linearly increasing the acetonitrile concentration to 70% over a period of 34 min. The products were monitored by measurement of absorbance at 210 nm.
For analysis of angiotensin degradation, HPLC was run under isocratic conditions at a flow rate of 1 ml/min in a buffer consisting of 170 ml of acetonitrile and 5 ml of phosphoric acid diluted to 1 liter with H 2 O and adjusted to pH 3.0 with triethylamine. Peptides were detected by absorbance at 210 nm.

Gel Electrophoresis
SDS-PAGE was run on 10% gels according to the procedure of Laemmli (22).

Mass Spectrometry
Electrospray mass spectrometry was run at The Mount Sinai School of Medicine.

Determination of Molecular Mass of the Native Enzyme
The molecular mass of the native enzyme was determined by FPLC on a calibrated Superose 6 column. The column was equilibrated with 50 mM Bicine, pH 7.0, 0.1 M NaCl, 10% glycerol and calibrated with a high molecular weight calibration kit consisting of thyroglobulin, catalase, ferritin, and aldolase (Amersham Pharmacia Biotech).

Determination of Subunit Molecular Mass
The subunit molecular mass was determined by SDS-PAGE. A calibration kit supplied by Amersham Pharmacia Biotech contained marker proteins ranging in molecular mass from 94.4 to 14.4 kDa.

Kinetic Analyses
K m and V max were obtained by a least squares analysis of the Lineweaver-Burk plot. Determination of K i was by the method of Dixon (23). Each determination was conducted at two substrate concentrations and at six concentrations of competing peptide.

Enzyme Purification
Step 1: Preparation of Rabbit Brain Supernatant Frozen rabbit brains (50 g) were partially defrosted and homogenized in a Waring blender with 4 volumes of 25 mM Bicine, pH 7.0, 5% glycerol (buffer A). The homogenate was centrifuged for 1 h at 100,000 ϫ g, and the supernatant was retained. The pellet was washed with an equal volume of buffer A and centrifuged for 1 h at 100,000 ϫ g, and the supernatant fractions were combined.
Step 2: Affi-Gel Blue Chromatography The supernatant was passed over a 2.5 ϫ 20-cm column of Affi-Gel blue, 100 -200 mesh, equilibrated with buffer A. The enzymatically active fractions of the effluent were combined.

Step 3: Q-Sepharose Chromatography
The enzymatically active fractions were applied to a 5 ϫ 15-cm Q-Sepharose column, equilibrated with buffer A, and the column was then washed with 1500 ml of buffer A. The enzyme was eluted with a 2-liter 0 -0.4 M NaCl gradient in buffer A. The active fractions were combined and concentrated in an Amicon ultrafiltration cell fitted with a PM-10 membrane.

Step 4: AcA-22 Gel Filtration
The concentrated enzyme was applied to a 2.5 ϫ 90-cm AcA-22 column equilibrated with 50 mM Bicine, pH 7.0, 10% glycerol, 0.1 M NaCl. The active fractions were combined, and NaCl was added to a final concentration of 0.3 M.

Step 6: Hydroxylapatite Chromatography
The dialyzed enzyme was applied to a 2.5 ϫ 1-cm column of hydroxylapatite equilibrated with buffer C, and the column was washed with this buffer. The effluent was collected, and fractions containing active enzyme were concentrated and then stored at Ϫ80°C.

Sequence Determination of Tryptic Peptides
Microsequencing was conducted at the W. M. Keck Foundation Biotechnology Resource Laboratory of Yale University. Gel slices containing protein were digested with trypsin, and the eluate was subjected to microbore HPLC. Matrix-assisted laser desorption/ionization mass spectrometry was used to evaluate peak purity and identify candidate peaks suitable for sequencing. Two sequences were obtained, and a third sequence of a tryptic peptide was obtained from the Baker Medical Research Institute (Prahran, Australia).

Nucleotide Sequencing
Nucleotide sequencing of ATCC clone 355377 was performed by the Biotechnology Center, Utah State University (Logan, UT) using sequence-specific oligonucleotide primers synthesized by Integrated DNA Technologies, Inc. (Coralville, IA).

Construction of the pSE-6HDAP Expression Plasmid
The cDNA coding for aspartyl aminopeptidase was excised from ATCC clone 355377 (lafmid BA plasmid) by first digesting with NotI and then partially digesting with BstXI to release a 1.55-kilobase pair piece containing the entire coding region minus 174 bp from the 5Ј-end. An adaptor was synthesized containing a 5Ј-overhang complementary to an NcoI digestion site on the 5Ј-end and a 3Ј-overhang complementary to the BstXI-generated overhang on the 3Ј-end. This adaptor contained the ATG start codon followed by six histidine codons, a glycine codon, and then the codon for asparagine normally present after the putative start codon in the aspartyl aminopeptidase sequence. The sequences of the oligonucleotides annealed to make the adaptor were 5Ј-CATGCACCATCACCACCATCACGGGCAG-3Ј and 5Ј-GGCACTAC-CACCACTACCAC-3Ј. The Escherichia coli expression plasmid pSE-420 (a gift from Dr. J. Brosius) was digested with NcoI and NotI. The digested pSE-420, the BstXI-NotI insert, and the adaptor were ligated to yield the plasmid pSE-6HDAP. The plasmid was transfected into XL-2 blue cells (Stratagene). The presence of the adaptor and insert was confirmed by restriction digestion. The cells containing the plasmid were grown in the presence of 2 mM isopropyl-1-thio-␤-D-galactopyranoside, and aspartyl aminopeptidase activity was determined.

Northern Blot
Total RNA was prepared essentially by the method of Chomczynski and Sacchi (24) from freshly frozen rat tissues. Formaldehyde gel electrophoresis and transfer of 20 g of RNA/lane were performed using the methods described in Sambrook et al. (25). The RNA was transferred to a Hybond-N nylon membrane (Amersham Pharmacia Biotech) and hybridized with 40 Ci of a mouse aspartyl aminopeptidase cDNA probe from ATCC clone 355377, which had been labeled to a specific activity of approximately 0.1 Ci/ng with [␣-32 P]dCTP using the NEBlot kit from New England Biolabs. Prehybridization and hybridization at 42°C were carried out in 20 ml of 6ϫ SSPE, 5ϫ Denhardt's solution, 50% formamide, and 100 g/ml sheared salmon sperm DNA. Washes were in 1ϫ SSC, 0.1% SDS at 42 and 55°C followed by a high stringency wash in 0.1ϫ SSC, 0.01% SDS at 55°C An RNA ladder from Life Technologies, Inc. was also electrophoresed for calculation of molecular size.

RESULTS
The cytosol of a rabbit brain homogenate contained an enzyme that cleaved the substrate Asp-Ala-Pro-NA and that, in the presence of excess dipeptidyl-peptidase IV, liberated free naphthylamine. Enzymatic activity was optimal in the neutral pH range (Fig. 1). This enzyme having the properties of an aspartyl aminopeptidase (see below) was purified 722-fold by conventional chromatographic techniques from a rabbit brain supernatant, with an overall yield of 3.8% (Table I). Starting with 50 g of rabbit brain, 112 g of enzyme protein was obtained. Examination of the enzyme by SDS-PAGE revealed a highly purified preparation containing a major protein band and trace amounts of higher molecular weight bands (Fig. 2).
The molecular mass of the native enzyme was determined by gel filtration chromatography on a calibrated FPLC Superose 6 column. The enzyme eluted identically with ferritin, establishing a molecular mass of 440 kDa. This value is virtually identical to the estimate of 450 kDa reported by Kelly et al. (16). SDS-PAGE gave a monomer molecular mass of 55 kDa.
The specificity of the enzyme was explored with a series of peptides. Asp-Ala-Pro-SM and Asp-Ala-Pro-NA used for assay and purification were cleaved at the Asp-Ala bond, since release of the chromogen was dependent upon the presence of dipeptidyl-peptidase IV. The k cat /K m ratio for Asp-Ala-Pro-NA exceeded that of the corresponding glutamyl peptide. The difference was primarily reflected in K m values with the glutamyl peptide binding very poorly to the enzyme. The enzyme required the presence of a free ␣-amino group. When the N terminus of the substrate Asp-Ala-Pro-NA was blocked by an Fmoc group, there was no cleavage. The enzyme also required the presence of an acidic amino acid at the N terminus for optimal activity. There was no measurable cleavage (Ͻ1% of Asp-Ala-Pro-NA) for Ala-Ala-Pro-NA, Lys-Ala-NA, Pro-Leu-Gly-NH 2 , leucine enkephalin (Tyr-Gly-Gly-Phe-Leu), Ile-His-Pro-Phe or Ser-Ala-Ala-Leu (also see below). An asparaginyl peptide (Asn-Ala-Ala-Leu) was cleaved at 5% of the rate of Asp-Ala-Pro-NA. Simple naphthylamides or p-nitroanilides such as Asp-NA, Glu-NA, Phe-NA, and Leu-pNA were also negligibly cleaved.
The subsite specificity of aspartyl aminopeptidase was explored with a series of aspartyl tetrapeptides and compared with the subsite specificity of glutamyl aminopeptidase (Table  II). HPLC was used for analysis of the products of reaction and quantitation of the rate of substrate disappearance. The K i of the peptides competing with the chromogenic substrate Asp-Ala-Pro-NA (for aspartyl aminopeptidase or Glu-NA (for glutamyl aminopeptidase) was determined by the method of Dixon (23). These K i values were used as a measure of K m . The subsite specificities of both enzymes are similar. Neutral or hydrophobic amino acids in the P1Ј position are preferred. For this series of tetrapeptides, a positively charged amino acid in the P1Ј position has a particularly adverse effect on the specificity constant (however, see below). Moreover, substitution in the P2Ј position can also strongly affect the k cat /K m ratio. For   example, with aspartyl aminopeptidase, the k cat /K m ratio for Asp-Ala-Phe-Leu is 13-fold greater than the k cat /K m ratio for Asp-Ala-Lys-Leu. In general, the specificity constants for substrates acted upon by glutamyl aminopeptidase were greater than the specificity constants of the same substrates cleaved by aspartyl aminopeptidase.
The effect of glutamyl and aspartyl aminopeptidases on the degradation of angiotensin II was also compared. Incubation of angiotensin II with either enzyme led to the production of angiotensin III (des-Asp-angiotensin II) as the only product. Prolonged incubation with aspartyl aminopeptidase, resulting in a total degradation of angiotensin II, did not lead to the formation of products other than angiotensin III (Fig. 3). These studies therefore directly demonstrate the specificity of aspartyl aminopeptidase for N-terminal acidic amino acids. Both the k cat and K m of angiotensin II measured with aspartyl aminopeptidase exceeded the corresponding values measured with glutamyl aminopeptidase; however, the k cat /K m ratios were similar (Table II). It should be noted that angiotensin II contains a positively charged amino acid in the P1Ј position yet has a 20-fold greater specificity constant compared with the tetrapeptide Asp-Lys-Ala-Leu (Table II). This implies that the enzyme may prefer octapeptides to tetrapeptides. However, to substantiate this conclusion, the effect of peptide length on the specificity constant would have to be systematically explored, since the substituents in the P1Ј-P3Ј positions of angiotensin II differ from the corresponding substituents in the tetrapeptide.
The effect of a series of proteinase inhibitors or potential activators on enzyme activity was studied (Table III). In these experiments, each compound was tested without preincubation and after 15 min preincubation. The enzyme was sensitive to longer preincubation periods. A 30-min preincubation led to loss of about one-third of the activity. Unlike glutamyl aminopeptidase, aspartyl aminopeptidase was not stimulated by Ca 2ϩ , and, unlike the activity described by Cheung and Cushman (15), aspartyl aminopeptidase was not stimulated by Mn 2ϩ . Zn 2ϩ at a concentration of 0.4 mM inhibited activity by   approximately 50% without preincubation but totally inhibited after 15 min of preincubation. An interesting feature of aspartyl aminopeptidase is its sensitivity to dithiothreitol (DTT). A concentration of 1 mM DTT inhibited enzymatic activity by 88%. We considered the possibility that disulfide bonds were necessary for the quaternary structure of the enzyme and that DTT may inhibit by dissociating the oligomer. Accordingly, we examined the electrophoretic profile of aspartyl aminopeptidase in buffers containing DTT. No evidence of dissociation was observed (data not shown). The inhibition by DTT may be due to metal ion chelation. Preincubation with o-phenanthroline led to virtually a total loss of activity; however, aspartyl aminopeptidase was relatively insensitive to EDTA. Prolonged dialysis against EDTA buffers was necessary to observe inhibition. After 120 h of dialysis against 10 mM EDTA, an approximate 67% inhibition was observed. Activity could be fully restored by the addition of Zn 2ϩ . In these experiments, the concentration of EDTA in the incubation mixtures was 0.8 mM. Activity was fully restored by Zn 2ϩ at final concentrations of 1.2, 1.6, and 2.4 mM. The aminopeptidase inhibitors amastatin and bestatin were without effect. At the concentration of amastatin used, membrane alanyl aminopeptidase was totally inhibited, and glutamyl aminopeptidase was 50% inhibited. Bacitracin unexpectedly stimulated enzymatic activity (Table  III; Fig. 4). A Lineweaver-Burk analysis of the experiment shown in Fig. 4 demonstrated that 200 M bacitracin lowered the K m from 2.2 to 1.6 mM while moderately increasing V max by 15%. Activation was moderately enhanced by preincubation. The enzyme was sensitive to p-chloromercuribenzoate, a feature commonly observed with cytosolic proteinases. We also measured the effect of chloride ions, since chloride was reported to stimulate yeast aminopeptidase I (see below) (26). It should be noted that all metal ions reported in Table III were tested as chloride salts. No activation by 40 mM calcium chloride was observed, nor was there activation by 10 mM NaCl.
The activity of the enzyme was determined in the cytosolic fraction of rat tissues. These assays were conducted in the presence of 0.4 mM Zn 2ϩ to inhibit any contaminating glutamyl aminopeptidase. Although this concentration of zinc inhibits purified aspartyl aminopeptidase by about 50% (Table III), it does not inhibit the enzyme in crude tissue extracts. Aspartyl aminopeptidase was present in all tissues studied, and its distribution was fairly uniform. The highest activity was found in testes. There was an approximately 5-fold differential in specific activity of highest to lowest values (Table IV).
Cloning-The electrophoretically homogeneous protein was subjected to trypsinization and the fragments separated by microbore HPLC. Two sequences were obtained from the Yale University facility. These were GFFELFPSLSR and LLQAGF-HELK. One sequence, LVQVEFPILR, was obtained by the Baker Medical Research Institute. No matches were found in the Protein Data Bank for any of these sequences. Some homology was noted for the peptide LVQVEFPILR and a Mycobacterium leprae sequence described as "vacuolar aminopeptidase I precursor" (LVRIDDPILR). The same data base when screened with the sequence LLQAGFHELK revealed a sequence in the Caenorhabditis elegans genome (accession number U13070) described as similar to yeast vacuolar aminopeptidase (accession number M25548). The C. elegans sequence was used to search the expressed sequence tag data base dbest of GenBank TM , and this search revealed mouse and human expressed sequence tags described as being similar to yeast vacuolar aminopeptidase. It was possible to locate sequences homologous to all three tryptic peptides of the rabbit brain  Clones containing the 5Ј-end of the sequence were identified. The human clone (ATCC number 355377) contained an insert size of 1.791 kilobase pairs, sufficiently large to encode the entire reading frame of aspartyl aminopeptidase. This clone was sequenced, and its sequence is shown in Fig. 5. The molecular weight of the protein was calculated as 52,065, in good agreement with the value of 55,000 of purified aspartyl aminopeptidase determined by SDS-PAGE. When the full-length human sequence was entered into the blastp nr data bank, sequences were found with high homology from yeast, M. leprae (accession number U15182), and from a Lyme disease spirochete described as an aminopeptidase I-like protein (accession number S44086). One of the yeast sequences was that of aminopeptidase I (EC 3.4.11.22), and the other sequence was described as "hypothetical 54.2-kDa protein in CDC12-ORC6 intergenic region" (accession number P38821). An alignment of the yeast, human, mouse, C. elegans, M. leprae, and Lyme disease spirochete sequences is shown in Fig. 6.
An examination of these sequences revealed three conserved histidines corresponding to amino acids 94, 170, and 440 of the human sequence. There are also three conserved glutamates corresponding to amino acids 52, 301, and 302 of the human sequence and five conserved aspartates corresponding to amino acids 96, 264, 299, 356, and 431 of the human sequence (Fig. 6).
The cDNA for the cloned human enzyme was expressed as a His 6 fusion protein in E. coli (Table V). Hydrolysis of the substrate Asp-Ala-Pro-SM by the nontransfected bacteria was very low, and activity was diminished when dipeptidyl-peptidase IV was omitted from the incubation mixture. The specific activity of Asp-Ala-Pro-SM in cells transfected with the cDNA of the cloned human enzyme varied from 5.5 to 47.6 units/mg in different groups. The mean activity was 90-fold greater than nontransfected cells and exceeded the activity found in crude extracts of testes (3.48 units/mg; Table IV). The K m for the expressed enzyme measured with the substrate Asp-Ala-Pro-SM (2.3 mM) was similar to that of the purified rabbit brain enzyme (2.0 mM). Hydrolysis was dependent on dipeptidylpeptidase IV, demonstrating that the expressed enzyme cleaved the Asp-Ala bond of the substrate. The activity was moderately inhibited by 0.4 mM Zn 2ϩ and unaffected by 0.4 mM puromycin. When the ␤-carboxyl-blocked compound Asp(O-t-Bu)-Ala-Pro-NA was substituted as substrate, activity was reduced by 98%. Thus, the expressed enzyme has all the characteristics of aspartyl aminopeptidase.
Northern blot analysis indicated a single mRNA species of 1.9 kilobase pairs in all tissues tested (Fig. 7). The largest amount of RNA was in testes; intermediate amounts were in kidney and lung; and lesser but significant amounts were in spleen, liver, and brain. These results were roughly comparable with enzyme activity, with the exception of brain, which showed similar enzyme activity to lung but considerably less RNA. Additional experiments (not shown) also demonstrated that the same message was present in intestine and heart.

DISCUSSION
The presence of an acidic amino acid-preferring aminopeptidase in mouse brain cytosol was documented more than a decade ago by Kelly et al. (16). This enzyme was described as the major aminopeptidase in brain homogenates that degrades N-terminal acidic amino acid-containing peptides. These investigators noted the instability of this enzyme. We have found that the activity could be stabilized by 5-10% glycerol-containing buffers. It was therefore possible to purify the enzyme to near homogeneity by conventional chromatographic procedures. The final preparation is stable when stored at Ϫ80°C in the presence of 10% glycerol. The inability of this enzyme to cleave simple substrates such as Asp-NA or Glu-NA led us to develop a coupled enzyme assay in which the product of cleavage of the substrate Asp-Ala-Pro-NA or Asp-Ala-Pro-SM is further degraded by excess dipeptidyl-peptidase IV to yield free chromogen. Its inability to cleave Asp-NA and Glu-NA probably explains why this enzyme has been largely overlooked. Aminopeptidases are routinely assayed by the use of simple aminoacyl-arylamide substrates. For example, Iribar et al. (27) report on the presence of a cytosolic aspartyl aminopeptidase in rat frontal cortex, which was measured with Asp-NA. The activity reported in pmol/min/g of tissue is 3 orders of magni- tude lower than activity measured with Asp-Ala-Pro-NA (Table  IV). Since aspartyl aminopeptidase constitutes more than 0.1% of the soluble protein in a rabbit brain homogenate, it must be viewed as a major intracellular proteinase.
Substrate specificity studies confirm that the enzyme is an acidic amino acid-preferring aminopeptidase favoring aspartyl residues over glutamyl residues. Its inability to cleave simple substrates such as Asp-NA and Glu-NA most likely reflects an extended substrate binding site with a requirement of subsites in addition to S1 and S1Ј for effective substrate binding. In this respect, it is of interest that aminoacyl-p-nitroanilides are rather poor substrates of the related enzyme yeast aminopeptidase I (28).
The specificity constants for substrates acted upon by aspartyl aminopeptidase are generally about 1 order of magnitude lower than the specificity constants of the same substrates acted upon by glutamyl aminopeptidase (aminopeptidase A). This difference can primarily be attributed to the relatively high K m values of most substrates of aspartyl aminopeptidase (millimolar range) (Table II). One notable exception is the substrate Asp-Ala-Asp-Leu, which has a K m of 50 M but is only slowly hydrolyzed. This tetrapeptide, which is a moderate inhibitor, should serve as a lead compound for the development of more potent aspartyl aminopeptidase inhibitors. Exploration of the physiological significance of aspartyl aminopeptidase would be facilitated by the availability of a specific and potent inhibitor. In view of the relatively high K m values of the substrates tested, one must consider the possibility that this enzyme is regulated in the cell by endogenous activators. This may be reflected in the moderate stimulation produced by the cyclic peptide antibiotic bacitracin, a compound generally employed as a protease inhibitor.
Aspartyl aminopeptidase was expressed in E. coli as a His 6 fusion protein. The properties of the expressed enzyme are similar to the enzyme isolated from brain (Table V) (8) and that the His 6 linker does not affect the catalytic properties of the enzyme. The level of expression is very high, and the presence of the His 6 group makes it possible to rapidly purify the enzyme on a nickel-nitrilotriacetic acid column (Qiagen). We have found that an approximately 90% pure preparation can be obtained by adsorption of the His 6 protein in a crude cell lysate on the nickel-nitrilotriacetic acid column followed by a 20 mM imidazole buffer wash and elution with a 200 mM imidazole buffer (not shown). Therefore, it is now possible to obtain large amounts of recombinant enzyme for further studies including crystallization.
The Northern blot analysis indicates a wide tissue distribution for aspartyl aminopeptidase, with the same 1.9-kilobase pair RNA species expressed in all tissues tested but with some tissue variation in the amount in relation to enzyme activity (i.e. brain versus lung). This may be due to the presence of differing amounts of endogenous inhibitors or activators or to variations between experimental animals. In any case, the ubiquitous nature of aspartyl aminopeptidase in various tissues and in evolution indicates that it must serve an important metabolic function or functions.
The sequences of the human and mouse aspartyl aminopeptidase cDNAs have significant homology to yeast aminopeptidase I (Fig. 6). Yeast aminopeptidase I is a zinc metalloproteinase found in the vacuole (28). Almost all of the aminopeptidase activity in the yeast can be attributed to this enzyme (28). It has been cloned (29), and its structure places it as the sole member of the M18 family of metalloproteinases (9). In con-trast to the substrate specificity of aspartyl aminopeptidase, yeast aminopeptidase I resembles leucine aminopeptidase (EC 3.4.11.1) in that it preferentially cleaves leucyl and other hydrophobic aminoacyl-peptide bonds (28). Therefore, aspartyl aminopeptidase is not the mammalian homolog of yeast aminopeptidase I but rather the first mammalian member of the M18 family (Fig. 8). Moreover, yeast aminopeptidase I is strongly inhibited by bestatin ((2S,3R)-3-amino-2-hydroxy-4phenylbutanoyl-L-leucine) (30), whereas aspartyl aminopeptidase is not affected by this aminopeptidase inhibitor (Table  III). This is probably due to a requirement for an acidic side chain in the P1 position for effective binding to aspartyl aminopeptidase. However, the yeast genome does contain a sequence encoding a hypothetical 54.2-kDa protein (accession number P38821). This protein is more closely related to human aspartyl aminopeptidase (42.7% identity, Gene Stream align) than it is to yeast aminopeptidase I (30.2% identity) (Fig. 8). This hypothetical protein likely encodes the yet to be characterized yeast aspartyl aminopeptidase.
We have found that expressed sequence tags encoding aspartyl aminopeptidase have been mapped to chromosome 2q by both the Stanford Human Genome Center (markers SHGC 114 and SHGC 32022) and The Whitehead Institute Center for Genome Research (marker SGC-32022). The protein has been designated as "highly similar to hypothetical 54.2 kDa protein in CDC12-ORC6 intergenic region (Saccharomyces cerevisiae)" (55% similarity). The Stanford mapping places the gene at 209 -220 cM, whereas the Whitehead mapping places the gene at 226 -228 cM. The reason for the discrepancy is not clear.
On the basis of its homology to yeast aminopeptidase I, aspartyl aminopeptidase can be tentatively designated as a zinc metalloproteinase. The putative catalytic zinc ion of aspartyl aminopeptidase is very tightly bound. This is shown by the resistance of the enzyme to inactivation by EDTA. Moreover, the enzyme is purified from Bicine-buffered solutions, and Bicine itself is a strong metal ion chelator (28). Prolonged dialysis against EDTA is required to demonstrate significant inhibition, and enzymatic activity can be fully restored by the addition of Zn 2ϩ . Preincubation with o-phenanthroline, however, does result in strong inhibition.
Although the subunit sizes of yeast aminopeptidase I and aspartyl aminopeptidase are similar, the native molecular mass of the yeast enzyme (640 kDa) is considerably higher than that of aspartyl aminopeptidase (440 kDa). Yeast aminopeptidase I has been described as a dodecameric protein (28). Aspartyl aminopeptidase may be composed of eight identical subunits, similar to ovine brain glutamine synthetase (31); however, further studies will be required to determine its quaternary structure. High molecular masses and oligomeric structures are common features of cytosolic aminopeptidases. Aminopeptidase H with a molecular mass of 400 kDa and a subunit molecular mass of 52 kDa appears to have a similar structure to aspartyl aminopeptidase (32). Bleomycin hydrolase (a cysteine proteinase (33)) and leucine aminopeptidase (a metalloproteinase (34)) have homohexameric structures. The active sites of bleomycin hydrolase have been localized to the central channel of the hexamer (35). It is not known if a similar localization will be found for the active sites of aspartyl aminopeptidase.
Mammalian aspartyl aminopeptidase has high homology to a sequence in the C. elegans genome and also to sequences found in M. leprae and the Lyme disease spirochete. The latter three sequences are only putative aminopeptidases, since their enzymatic activities have not been determined. Although these proteins are likely to be members of the M18 metalloproteinase family, it is not known if their substrate specificities resemble aspartyl aminopeptidase or yeast aminopeptidase I (a leucyl aminopeptidase). The nature of the active site of yeast aminopeptidase I is unknown, since this zinc metalloproteinase lacks the classical zinc signature HEXXH (29). Amino acids serving as ligands for zinc are histidine, cysteine, glutamate, and aspartate (36). An alignment of the seven related sequences reveals three conserved histidyl residues, three conserved glutamyl residues, and five conserved aspartyl residues (Fig. 6). Site-directed mutagenesis should reveal whether any of these conserved residues are essential for enzymatic activity. Such studies should provide valuable information on the mechanism of action of a new class of metalloproteinase and lead to the rational design of specific inhibitors.
An examination of the sequence by PROSITE reveals putative phosphorylation sites for cAMP-dependent protein kinase, casein kinase II and protein kinase C. It is not known at this time if aspartyl aminopeptidase is phosphorylated at any of these sites and if so whether phosphorylation serves a regulatory function. The sequence contains no transmembrane regions and appears to represent a cytosolic protein.
The question of the physiological significance of aspartyl aminopeptidase awaits further study; however, several possibilities can be considered. The relatively high concentration of this enzyme in the cytosol of various tissues (Ͼ0.1% of soluble protein) and its fairly uniform tissue distribution suggest a role in general intracellular protein and peptide metabolism (37). An interesting exogenous substrate is aspartame (aspartylphenylalanine methyl ester) (16). A more specialized role in specific tissues may be its metabolism of biologically active peptides such as angiotensin II, cholecystokinin-8, and neuropeptide K. The activity of cytosolic aspartyl aminopeptidase was reported as decreased in the frontal cortex of aged rats (27). In this respect, it is of interest to note that ␤-amyloid peptide contains an N-terminal aspartyl residue, and a reduction in the removal of this residue has been proposed to facilitate ␤-amyloid deposition in Alzheimer's disease (38). It has recently been proposed that aminopeptidases mediate the N-terminal trimming of peptides presented by the major histocompatibility complex-1 system (39). Thus presentation of N-terminal extensions of the immunodominant chicken ovalbumin peptide SIINFEKL is not blocked by proteasome inhibitors. The amino acid immediately preceding SIINFEFL is glutamate, and ESIINFEKL should be an excellent substrate for aspartyl aminopeptidase but a very poor substrate for cytosolic leucine aminopeptidase. Identification of endogenous substrates of aspartyl aminopeptidase will be of obvious importance.