Glycosaparaginase from Human Leukocytes INACTIVATION AND COVALENT MODIFICATION WITH DIAZO-OXONORVALINE*

The apparent active site of human leukocyte (N4-(~-acetylglucosaminy~)-~-aspa- raginase has been studied by labeling with an asparagine analogue, 5-diazo-4-oxo-~-norva- line. Glycoasparaginase was purified 4,600-fold from human leukocytes with an overall recovery of 12%. The purified enzyme has a K,,, of 110 p ~ , a Vma, of 34 pmol X 1” X min”, and a specific activity of 2.2 units/ mg protein with N4-(~-N-acetylglucosaminyl)-~-as-paragine as substrate. The carbohydrate content of the enzyme is 15%, and it exhibits a broad pH maximum between 7 and 9. The 88-kDa native enzyme is com- posed of 19-kDa light (L) chains and 25-kDa heavy (H) chains and it has a heterotetrameric structure of L2Hz- type. The glycoasparaginase activity decreases rapidly and irreversibly in the presence of 5-diazo-4-oxo-~-

pH 7.5. The enzyme activity is competitively protected against this inactivation by its natural substrate, aspartylglucosamine, indicating that this inhibitor binds to the active site or very close to it. The covalent incorporation of [5-'4C]dia~~-4-~~~-~-norvaline paralleled the loss of the enzymatic activity and one inhibitor binding site was localized to each L-subunit of the heterotetrameric enzyme. Four peptides with the radioactive label were generated, purified by high performance liquid chromatography, and sequenced by Edman degradation. The sequences were overlapping and all contained the amino-terminal tripeptide of the L-chain. By mass spectrometry, the reacting group of 6-diazo-4-oxo-L-norvaline was characterized as 4oxo-L-norvaline that was bound through an wketone ether linkage to the hydroxyl group of the aminoterminal amino acid threonine.
The lysosomes of mammalian cells contain a glycoasparaginase (N4-(@-acetylglucosaminyl)-~-asparaginase, aspartylglycosylaminase, aspartylglucosaminidase, glycosylasparaginase, EC 3. 5.1.26) that is involved in the degradation of the * This work was supported by the Finnish Academy of Sciences and by National Science Foundation, Science and Technology Center Cooperative Grant 8809710 (to J. R. Y. and to L. E. HJ. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement'' in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. ( 1 To whom correspondence should be addressed. N-glycosidic linkage between asparagine and N-acetylglucosamine in various asparaginylglycopeptides. The substrate N4-(P-N-acetylglucosaminy1)-L-asparagine (aspartylglucosamine) is hydrolyzed by the enzymes to aspartate and l-amino-N-acetylglucosamine. The latter product is further cleaved nonenzymatically to ammonia and N-acetylglucosamine (1).
Human liver has been reported to contain a monomeric aspartylglucosaminidase of 80 kDa (2) and a trimeric enzyme of 60 kDa consisting of three nonidentical polypeptides (3), both with a pH maximum at 6.1, in addition to an undetailed aspartylglucosaminidase with a pH maximum at 7.5 (4). The rat liver aspartylglucosaminidase is a heterodimeric protein of 43 kDa that contains two noncovalently bound subunits and has a broad maximum between pH 6.5 and 10 (5). The amino-terminal sequences of its subunits have been determined ( 5 ) . Lack of the glycoasparaginase activity in humans results in a lysosomal storage disease called aspartylglycosaminuria (McKusick 20840) characterized with severe psychomotor retardation and accumulation of glycoasparagines in tissues and body fluids (1). Bacterial (6) and yeast (7) asparaginases can hydrolyze Dor L-asparagine with free cy-amino and a-carboxyl groups to yield aspartate and ammonium ion. In addition to asparagine, bacterial glutaminase-asparaginases are able to hydrolyze glutamine to yield glutamic acid and ammonium ion (8). Inhibitors of these related amidohydrolases have proved useful in understanding the characteristics of the enzymes. An b a sparagine analogue 5-diazo-4-oxo-L-no~aline (DONV)' has been used to label the active site of Escherichia coli L-asparaginase (9). Glutaminase-asparaginases are similarly inhibited and labeled by the next larger homologue of DONV, 6diazo-4-oxo-~-norleucine (DON), which binds to threonine hydroxyls in an 8-residue segment identical in both Acinetobacter and Pseudomonas 7A glutaminase-asparaginase enzymes (10). DONV has been shown to irreversibly inhibit hen oviduct (11) and rat liver glycoasparaginase (12), but no further characterization of the adduct was done. The effect of the amidohydrolase inhibitors on human glycoasparaginases has not been studied. In this paper, we describe purification of heterotetrameric glycoasparaginase from human leukocytes and report our findings that 5-diazo-4-oxo-~-no~aline is an irreversible inhibitor of the enzyme. We also show that the compound is an active site-directed inhibitor or affinity label The abbreviations used are: DONV, 5-diazo-4-oxo-~-norvaline; HPLC, high performance liquid chromatography; ONV, 4-0XO-Lnorvaline; DON, 6-diazo-4-oxo-~-norleucine; PTC, phenylthiocarbamyl; TEA, triethylamine; PITC, phenylisothiocyanate; DTT, dithiothreitol; PMSF, phenylmethylsulfonyl fluoride; PAGE, polyacrylamide gel electrophoresis; SDS, sodium dodecyl sulfate. and that a threonine hydroxyl is alkylated to form an ether linkage with the reacting group of the inhibitor.

5-Diazo-4-oxo-~-norvaline at each concentration caused
progressive, irreversible inhibition of glycoasparaginase activity in a pseudo-first-order reaction under conditions that maintained the activity of the enzyme in the absence of the reagent (Fig. 1). The activity of the enzyme could not be restored by extensive dilution or dialysis of the inhibited enzyme. The hyperbolic relationship between the pseudofirst-order constant and concentration of DONV (data not shown) indicates that the inhibitor forms a reversible, prealkylation complex with the enzyme (16, 17). Thus, the inhibition corresponds to the mechanism k, K3 k, where E -X is the irreversibly inactivated enzyme and follows the form The second-order rate constant corresponding to [ I ] << Kr is given by k3/K1.
A plot of l/kob, against l/(DONV) showed a linear relationship typical of hyperbolic saturation (Fig. 1, inset). The steady-state constant (&) and the rate constant ( k 3 ) in 50 mM phosphate buffer, pH 7.5, calculated from the slope were K, = 80 p~, k3 = 0.10 min" and the second-order rate constant = 1.25 X lo3 "' min". In order to determine if DONV reacts with some group at the active site of the enzyme, we studied the effect of the substrate, aspartylglucosamine, on the inactivation rate. The presence of substrate in the same medium with the inactivator competitively inhibited the inactivation of glycoasparaginase with an inhibition constant KL = 98 p~ (Fig. 2).
To determine the region of glycoasparaginase modified by Portions of this paper (including "Materials and Methods," part of "Results," Figs. 6-11, and Tables IV-VI) are presented in miniprint at the end of this paper. Miniprint is easily read with the aid of a standard magnifying glass. Full size photocopies are included in the microfilm edition of the Journal that is available from Waverly Press. Effect of substrate and inhibitor concentrations on the pseudo-first-order rate constant kob at 7.5, 25 "C. Glycoasparaginase (5 pg, 10 milliunits) was incubated without substrate ( . ) and in the presence of 50 pM (O), 100 pM (A), and 200 pM (A) aspartylglucosamine and 5-300 p~ inhibitor ( I ) DONV in 50 mM phosphate buffer. Inset, a graph of the slopes against substrate concentration. An inhibition constant (KL) for substrate to protect against inactivation by DONV was calculated from the intercept on the S axis (-KL) (derived from Refs. 16,17). DONV, the enzyme was radioactively labeled in the presence of [5-'4C]DONV for 24 h. The completely inhibited enzyme was further incubated for another 2 h in the presence of 1,000fold excess of unlabeled DONV and injected onto the reversephase HPLC column. Monitoring of the eluent by the absorbance at 225 nm and radioactivity showed that only the light chain of the enzyme eluting at 34% of solvent B was effectively radiolabeled (Fig. 3). Background levels of radioactivity were found with the heavy subunit eluting at 43% of solvent B and any of the other components in the chromatograms. Taking into consideration that the glycoasparaginase-DONV adduct lost about 20% of its label during repeated chromatographic runs in 0.1% trifluoroacetic acid, 1.3 mol of [5-'4C]DONV was bound/mol of enzyme or 0.7 mol/mol of L-subunit.
The lower panel presents the radioactivity from 5% of each fraction. Arrows indicate the radioactive peptides at the elution times of 62 min (CH1) and 80 min. B, rechromatography of the radioactive peptides ( A ) obtained at 80 min of elution time. The chromatography was carried out as described in the legend of A with the exception that gradient was 1530% B (120 rnin). The lower panel presents the radioactivity from 20% of each fraction. The arrow indicates the radioactive peptide CH2. C, HPLC separation of CNBr-peptides from 1 nmol of [5-I4C]DONV-labeled L-chain. Column and eluents were as described in A. Gradient: 2% B (10 rnin); 2-30% B (10-70 rnin); 30-6075 B (70-125 min); 60-80% B (125-155 min). The lower panel presents the radioactivity from 20% of each fraction. The arrows indicate the radioactive peptides CB1 and CB2 at the elution times of 13 and 27 min. phase HPLC. Most of the radioactivity was localized in two peptide peaks eluting at 18% (CH1) and at 23% of solvent B, respectively (arrows in Fig. 4A). Since the shape of the latter peak was indicative of a peptide mixture, these fractions were combined and applied to the same column in 15% solvent B. Rechromatography with a ' shallower gradient resulted in a single peak (CH2) eluting at 21% of solvent B (Fig. 4B). In a parallel experiment the modified protein was subjected to chemical cleavage in the presence of cyanogen bromide. The DONV-labeled L-chain was stable in 70% formic acid and after cyanogen bromide cleavage, two major peaks of radioactivity coeluted with two small peptide peaks CB1 and CB2 at 3 and 10% of solvent B, respectively (Fig. 4C).
The purified, radioactive peptides were lyophilized and subjected to sequence analysis by Edman degradation. The absence of significant secondary amino acid or unknown peaks indicated high purity of the peptides. The chymotryptic peptide CHI was a tripeptide of Thr-Ile-Gly. The peptide CH2 consisted of seven amino acids with the same Thr-Ile-Gly sequence at its amino-terminal end ( Table I). The radioactive cyanogen bromide fragments CB1 and CB2 showed identical tripeptide sequences of Thr-Ile-Gly in Edman degradation ( Table I). All the sequences obtained from the peptides were identical to the amino-terminal end of the intact L-subunit ( Table I). Further localization of the inhibitorbinding site by this method was unsuccessful, since the radioactivity could not be localized to any particular cleavage cycle presumedly because it was either flushed away by 100% trifluoroacetic acid before the actual sequencing cycles or found to be retained on the sample filters after the sequencing procedure. The combined evidence demonstrates that the reacting site of 5-diazo-oxo-norvaline to human leukocyte glycoasparaginase is within a few amino-terminal amino acids of its Lsubunit.
To identify the residues involved in the covalent modification of the glycoasparaginase by 5-diazo-oxo-norvaline, the enzyme was labeled with nonradioactive DONV. The L-subunit was purified under conditions identical to those used for [5-14C]DONV labeling, digested with trypsin, and the peptides were isolated by HPLC as described earlier. The mass spectrum of the HPLC fraction collected at 30% of solvent B (data not shown) consisted of two strong (M+H)' signals at m/z 997 and 1,125. The ion at m/z 997 matches the value expected for the N-terminal tryptic peptide from the 0-subunit. Formation of the DONV-peptide adduct by nucleophilic substitution would increase the mass of this peptide by 128 daltons to m/z 1,125. Analysis of the ion at m/z 1,125 by collision-activated dissociation (22)(23)(24) produced the mass spectrum in Fig. 5. The predicted fragment ions, type b and y" (22), are displayed with the structure in the figure, and those ions observed in the mass spectrum are underlined. To provide additional support for the localization of the label to His (65) Val (412)  9 Lys (568) Asn (92) 10 Thr (240) Thr (31)  11 Gly (160) Try (47)  12 His (44) Pro (153)  13 Ile (170) Phe (152)  14 Ala (230) Lys (131)  15 Ala (260) c -c -c -c -o -  ' Methyl ester collision-activated dissociation mass spectrum.

Fragment ions and ion intensities observed in the collision actiuated dissociation mass spectra of 4-oxo-~-norvaline-Thr-Ile-Gly-Met-Val-Val-His-Ile-Lys and 4-oxo-~-norvaline-Thr-Ile-Gly-Met-Val-Val-His-Ile-Lys methyl ester
the threonine residue, this peptide was converted to its corresponding methyl ester derivative. Formation of the methyl ester would shift the (M+H)+ of the labeled peptide by 28 daltons to m/z 1,153, since the addition of the DONV label would place a free acid at the N terminus of the peptide.
Additionally, the m/z values for the y" ions series in the collision-activated dissociation mass spectrum would increase by 14 daltons (reflecting the addition of a methyl ester at the C terminus) except for the y" ions associated with fragmentation of the amide bond on the N-terminal side of the modified amino acid. These y" ions will increase by an additional 14 daltons since the a-carboxylic acid of the DONV label also formed the methyl ester. This shift in mass is observed between y"8 (cleavage of the amide bond between Thr and Ile) and the (M+H)+ ion (Table 11), clearly showing the label is located on the threonine residue.

DISCUSSION
In this work, glycoasparaginase was purified 4,600-fold from human leukocytes. Throughout the purification of the en- zyme, we monitored its activity with its natural substrate using a new, specific assay for glycoasparaginase activity (15). The purified enzyme was obtained with a 12% overall yield. Gel electrophoresis under reducing conditions revealed that the enzyme protein runs as two separate bands nominated as light (L) and heavy (H) subunits with apparent M, of 19,000 and 25,000, respectively. Since the amounts of Land Hsubunits according to SDS-PAGE and reverse-phase HPLC are very similar, and on native PAGE the enzyme runs as a single band at 83,000, a quaternary structure of LzH2-type with an apparent M, of 88,000 for the native glycoasparaginase is suggested. The pH optimum of glycoasparaginase of human leukocytes and chorionic villous cells is 7-9 (15), which is extraordinarily high for lysosomal enzymes that usually have an acidic optimum. At pH 4-5, the leukocyte glycoasparaginase had only 0-10% of its activity at 7.5, which may reflect unknown catabolic characteristics of the enzyme different from most lysosomal exoglycosidases. The enzymatic activity of the dissociation intermediates of human leukocyte glycoasparaginase suggests that its action requires association of at least one of each chain, which by themselves are enzymatically inactive. A recent report indicates the presence of a heterodimeric glycoasparaginase of M, 43,000 with a broad pH maximum between 6.5 and 10 in rat liver (5). The aminoterminal sequences of the rat enzyme have high similarity to those of the human leukocyte enzyme.
Since the free a-amino and a-carboxyl groups of asparagine were essential for the hydrolysis of the substrate, we synthesized an asparagine analogue 5-diazo-oxo-~-norvaline and studied its action on human leukocyte glycoasparaginase. Incubation of the compound with glycoasparaginase resulted in irreversible loss of its enzymatic activity. The specificity of the interaction between the asparagine analogue and the enzyme is demonstrated in several ways: the glycoasparaginase activity is inhibited by DONV under conditions that maintain the activity of the enzyme without inhibitor, the inhibition follows pseudo-first-order kinetics at any one concentration of the inhibitor, the enzyme activity is partially IV 20 I V protected by its natural substrate, and a single amino acid residue is covalently labeled by the inhibitor. The stoichiometry of [5-'4C]DONV binding to glycoasparaginase was 0.7 mol of DONV/mol of L-subunit indicating the existence of two inhibitor-binding sites/one enzyme molecule or one each L-chain. The labeling efficiency is in accordance with the sequencing finding that some of the inhibitor-binding sequence was found to be without the label. The inhibitor-binding site to glycoasparaginase could be localized within the three most amino-terminal amino acid residues of the L-chain by automated Edman degradation sequencing of [5-"C]DONV-labeled chymotryptic and cyanogen bromide peptides. The cyanogen bromide cleavage produced two peptides, CB1 and CB2, that had different retention times in reverse-phase HPLC, but generated identical sequences upon Edman degradation. The small amounts of the peptides limited mass spectrometric analysis, and no structure-related mass spectra could be recorded for them. The structural difference between CB1 and CB2 is most likely ascribed to a mixture of DONV-containing tetrapeptides with homoserine and homoserine lactone at their carboxyl-terminal end. Cyanogen bromide cleavage is known to produce both compounds, and it has been demonstrated that even octapeptides with identical sequences are separated from one another by reverse-phase HPLC due to the presence of either homoserine or homoserine lactone at the carboxyl terminus (25). The final identification of the amino acid modified by the inhibitor was accomplished by mass spectrometry of tryptic peptides. The reacting group of DONV was shown to be 4-0xo-Lnorvaline bound through an a-ketone ether linkage to the hydroxyl group of the amino-terminal amino acid, threonine. More generally, our data show the applicability of mass spectrometry in characterization and localization of modifications in proteins. Active-site-directed irreversible inhibitors of enzymes produce labile adducts often difficult to characterize by chemical means (9, 10) and especially in such cases, mass spectrometry should be considered as the method of choice.
The binding of the substrate analogues to the hydroxyl group of threonine or serine residues is a common feature between glycoasparaginase and other amidohydrolases. The diazo compounds, like the actual substrates of the enzymes (26), are obviously attached to the active center of amidohydrolases through their a-amino-and a-carboxyl groups. This is followed by a nucleophilic attack in which the diazo nitrogens are cleaved off and the positively charged carbonium ion reacts with the negatively charged oxygen and a covalent oxygen-carbon bond is formed (10). Increased acidity of the hydroxyl group is suggested by this reaction in absence of catalyst. The acidity of amines is not great enough to react with the diazo group without a catalyst (27). Since both glycoasparaginase and asparaginases recognize the asparagine moiety with free w-amino and a-carboxyl groups and are inhibited with the same substrate analogue, it is reasonable to assume that this general reaction scheme applies to glycoasparaginase as well.
The amino-terminal sequence of the human leukocyte glycoasparaginase L-subunit shows considerable homology to that of bacterial amidohydrolases, and it is highly similar to the light subunit of the rat liver glycosylasparaginase ( Table  111). The primary structure of the L-subunit of the human glycoasparaginase has two significant differences compared to the other sequences: it is the only enzyme to contain an amino-terminal threonine as well as to lack threonine at the position 13 that corresponds to the DON-binding threonine at position 12 in Acinetobacter glutaminase-asparaginase (10, 28) located within the conserved 8-residue fragment found also in Pseudomonas 7 A glutaminase-asparaginase (lo), Erwinia chrysanthemi asparaginase (29), and E. coli asparaginase (30). The presence of the amino-terminal threonine in human leukocyte glycoasparaginase is apparently of functional importance, since this particular residue forms the covalent adduct with [5-14C]DONV and is replaced in the rat liver enzyme by glutamine (5). DONV is known to inhibit rat liver glycoasparaginase, but whether it is bound to the threonine at position 13 or elsewhere remains to be shown.   (2:l:l by valj solvent system and the plates were stained with ninhydrin and orcinol. The kinetlc studies were carried out by a spectrophotometric assay ( I I j using IL Mullistat 111 centrifugal analyzer (Instrumentation Laboramry. Lexington, MA,

Details of the incubation candstions and
reagent Concentraf~onS are found in the legend, of Fig. 1 . .

. .
The purified protein was hydrolyzed in 6 M HCI at 110 OC for 16 hours and the liberated amino acids were analyzed as their a-phtalaldehyde derivatives with an automatic LKB 4151 Alpha Plus amino acid analyzer (LKB-Products AB, Bromma, Sweden) using norleucine as an internal standard. Proline was measured by HPLC using precolumn PTC derivatization. e c dwith the PhastSystemTM according to the manufacturer's instructions using prc-cast 8.
The native PAGE electrophoresis was performed 25 46 gradient gels (PhartSystemTM Separation Technique file No. 120). Protein bands and molecular weight standards were localized by silver staining. SDS-PAGE war carried out with 12 % gels and protein markers with known molecular weight. The samples were denatured for 10 mm at 90 OC in 2.5 96 SDS and 5 46 mercaptoethanol, and loaded on SDS polyacrylamide gel according IO Laemmli (19). After migratmn at 30 mA. the gels were stained with silver. Isoelectric focusing was performed with the proteins were localized with silver staining.
PhastSystemTM according to the manufacturer's instructions (pH gradient 3-9) and the

QttLcr analvtlcal WLhQd5
Protein assays were performed with the Bio-Rad protcin assay kit according to the manufacturer's instructions. The purified glycoasparaginase preparation was tested for the activity of U-N-acetylglucoraminidare, a-mannosidase, 0corresponding 4-methylumbelliferone derivatives as substrates (20). The pH optimum fucoaidase, a-iduronidase. &galactosidase and a-N-acetylneuraminidase using Of glycoasparaginase was determined over a pH range of 3.0 to 9.5 in intervals of 0.5 pH Units. The following buffers were used: Brittan-Robinson's universal buffer (50 mM. In each step. the fractions Containing glyeoasparaginare activity wcre pooled and utilized for the next step. The buffy c o~t was allowed to sediment with an equal volume of dextran (6 %) and saline for 2-3 hours. The upper layer containing Icukocytes was centrifuged (IO minutes at 500 x g). The contaminating erythrocytes were removed by hemolyzing in water and the lcvkocyte mixture were stored at -20 OC. 1 liter of packed leukocytes was frozen and thawed three times.
I liter of 50 mM phosphate buffer. pH 6.3, was added and the suspension was centrifuged for 30 minutes at 20 000 x g and thc pellet was discarded. The proteins in the solution were precipitated with 0.5 % caprylic acid ( v k ) and the supernatant was dialyzed for 24 hours against 20 mM imidazole-HCI buffer, pH 6.8. The protein mixture was applied to a 70-ml DEAE-Scphadex A-50 column in a glass sintered funnel and the unbound protein was eluted from the gel by washing with 20 mM imidazole-HCI, pH 6.8. The glycoasparaginase. which was bound lo the gel, was rcleascd by 0.5 M NaCl and dialyzed with the starting buffer using an Amicon YM-30 filter and sample concentrator. The enzyme concentrate was then applied to a column of DEAE-Sepharose CI-6B (2.6 x 27 cm) and eluted with a linear NaCl gradient (0-0.3 M NaCI). The fractions containing the enzyme activity were pooled and dialyzed with EDTA-free starting buffer containing I mM Mg++, MnCC, Ca++ and 0.5 M NaCl (buffer 1. [5-14C1-DONV labeled and HPLC purified L-chain was incubated with cyanogen bromide was added and thc sample was injected onto the HPLC column. The radioactive peptides generated by the cleavages were purified by HPLC on a Vydac 218TP52 nanow bore column. Aliquats from each HPLC fraction were subjected to the B-counting and those containing the radioactivity were evaporated to dryness with a Szvant SpeedVacT and used for peptide sequencing. Details of the chromatography arc given in the legends of Fig. 3 and 4. Mass spectra were recorded on a Finnigan MAT TSQ70 (Finnigan cesium ion gun and a 20 keV conversion dynode as previously described (23). Mass analysis was pcrformcd by adding 1 VI of a 5% acetic acid solution containing the peptide mixture of peptides at 10-50 pmol level to I p1 of monothiglycerol on a gold matrix intn the gar phasc by bombardment with 8-10 keV Cs+ ions generated with the plated. copper probe tip,

I mm in diameter. Peptides
were sputtered out of the liquid cmium ion gun. Typically the mass range of m/z 400-3000 was scanned 10 times at 390 daltonris and the resulting maw spectra were summed together. Scqucncc rclccted through a 3-6 dalton mass window in quadrupole 1. Argon prcsrurc in the analysis of peptides was performed as previously described (22,23). Parent ions were collision cell was 3x10-5 torr and collision energies wried between 30 and 40 cV.

Results
The caprylic acid precipitation and hatch technique an DEAE-Scphader A-SO cnabled us -ion of Glv- The purification scheme is summarized in Table IV. to remove a significant amount of inactive material without loss of enzyme activity. The (Fig. 6A). The ability of the enzymc to interact with Concanavalin A-lectin was utilized glycoasparaginase activity eluted at 80 mM NaCl from DEAE-Sepharosc CI-6B column a-methyl-glucoside and 1.0 M NaCl (buffer 2. Fig. 68). The Sephacryl S-200 HR gel next in the purification. The enzyme was bound to the column and eluted with 1.0 M protein solution completely free from pigments. In hydrophobic interaction filtration (Fig, 6C) eliminated the low and high m~l e c~l a r weight proteins and yielded a chromatography on Alkyl Superose. the enzyme activity eluted at 0.5 M ammonium sulphate in thc major protein fraction (Fig. 6D).
In the final s a p of anion exchange chromatography on a Mono Q column (Fig. 6E)

(AGA=plycoasparrginns~).
The yield from I L of packed leukocytes was 340 L g of pure and active enzyme with a rpeific activity of 2.2 Ulmg. The yield of purification was I2 % and the speclfie activity increased 4600-fold as comDared with that of the crude extract (Table IV) hand wtth an rpparent M r of 8 3 . W (Fig. 7A). By gel fIltralion on I On native PAGE. glycoasparaginase ran 1s a Scphacryl S-200 FIR column calibrated with marker proteins. the Mr was estimated as 76.000 (data not shown) When the enzymc prepantian was incubated in 2.5 ' % SDS and > % mereaptoethmol far IU minutes at 90 OC. the protein completely dissociated to give two protein bands w t h the rclntiw molecular weight of 25.000 and IY.000 in I? % SDSPAGI.: (Fig. 7B). In a pH gradient. the native enzyme dirtrlhutcd betwecn pH 4.6. 5.2 on iroclcctrx focusing (Fic. 7C) and conmined at least five separate bands indicating charge heterogeneity.

. .
A. Polyacrylamide gel cleclrophoreris of the purified enzyme (lane I) and the molecular weight markers (lane 2). Electrophoresis was carried out on a 8.2s Lk gradient gel. which was stained with silver.
B. SDS-polyacrylamide gel electrophoresis of the purified enzyme (lane I ) and the carried out on I 12 % gel. which was stained with silver. molecular weight markers (lane 2) under reducing conditions. Electrophoresis was C. lswlectric focusing or the purified enzyme (lane I) and the marker proteins (lane 2). IEF was performed on I pH 4.5.9.0 gradient gcl. which was stained with silver.
Glycoarparaginare activity was lost rimultrncovsly with the disappearance of the native enzyme durmg the incubation before SDS-PAGE electrophorcsir (Fig. 8). The intensity of the ncw bands at 19.000 and 25.000 increased rapidly with the progression of dissociation.
At thc same time. protcin bands with glycoasparaginase activity were consistently observed at 67.000 and 55.000 regions untll the native enzyme band had completcly disappeared. The proteins 11 19.000 and 25.000 were enzymatically inactive (Fig.  8).
On thcsc basts we conclude that glycovrparaginare of human leukocytes is a polymeric and most obviously a heterotetramcric protein of L2H2-typc with an apparent M r of about 88,000 and i t is composed of light and heavy subunits with M r o f 19.000 and 25.000.
The linkages between the subunits were ruptured in 0.1 % TFA and the liberrted subunits were isolated by reverse-phase HPLC (Fig.  9). The isolated. enzymntically inactive fractions. wcce subjcctcd to SDS-PAGE. which indicated that the flrst mqar peak eluting wnth retention time of 9.3 minutes represcnted the L-subunit (MI l9,OlNl) and the second major perk with retention time of 13.8 minulcI reprcrented the If.