Human 60-kDa Lysophospholipase Contains an N-terminal l-Asparaginase Domain That Is Allosterically Regulated by l-Asparagine*

Background: The 60-kDa human lysophospholipase comprises an N-terminal domain with predicted, yet uncharacterized l-asparaginase activity and a C-terminal ankyrin repeat-like domain. Results: The N-terminal domain, termed hASNase1, was identified as a functional structural unit possessing catalytic activity. Conclusion: hASNase1 is an allosterically regulated bacterial-type cytoplasmic l-asparaginase. Significance: Domains of multifunctional human proteins harbor homologs of prokaryotic enzymes displaying similar structural and kinetic features. The structural and functional characterization of human enzymes that are of potential medical and therapeutic interest is of prime significance for translational research. One of the most notable examples of a therapeutic enzyme is l-asparaginase, which has been established as an antileukemic protein drug for more than four decades. Up until now, only bacterial enzymes have been used in therapy despite a plethora of undesired side effects mainly attributed to the bacterial origins of these enzymes. Therefore, the replacement of the currently approved bacterial drugs by human homologs aiming at the elimination of adverse effects is of great importance. Recently, we structurally and biochemically characterized the enzyme human l-asparaginase 3 (hASNase3), which possesses l-asparaginase activity and belongs to the N-terminal nucleophile superfamily of enzymes. Inspired by the necessity for the development of a protein drug of human origin, in the present study, we focused on the characterization of another human l-asparaginase, termed hASNase1. This bacterial-type cytoplasmic l-asparaginase resides in the N-terminal subdomain of an overall 573-residue protein previously reported to function as a lysophospholipase. Our kinetic, mutagenesis, structural modeling, and fluorescence labeling data highlight allosteric features of hASNase1 that are similar to those of its Escherichia coli homolog, EcASNase1. Differential scanning fluorometry and urea denaturation experiments demonstrate the impact of particular mutations on the structural and functional integrity of the l-asparaginase domain and provide a direct comparison of sites critical for the conformational stability of the human and E. coli enzymes.

sion of L-asparagine (L-Asn) to L-aspartic acid (L-Asp) and ammonia, although some of them are able to also hydrolyze L-glutamine (L-Gln) to L-glutamic acid (L-Glu) and ammonia. These enzymes are present in bacteria to mammals and play pivotal roles in amino acid metabolism (1,2). Enzymes with L-asparaginase activity can be generally classified into two evolutionary distinct families: the bacterial-type and the plant-type L-asparaginases, which are characterized by different structural and biochemical features (3,4). The bacterial-type enzymes have been further grouped into type I and type II depending on their cellular localization. Type I includes cytosolic enzymes that exhibit low affinity (millimolar K m ) for L-Asn, whereas type II enzymes are localized in the periplasm and show considerably higher affinity (micromolar K m ) for L-Asn (5). These enzymes have been studied extensively over the last 40 years mainly because two of the type II isoforms (L-ASNases from Escherichia coli and Erwinia chrysanthemi encoded by the ansB genes) serve as therapeutics for the treatment of acute lymphoblastic leukemia (6 -8). Conversely, the less studied plant-type L-asparaginases belong to the so-called N-terminal nucleophile (Ntn) hydrolases, which were defined as a new protein structural family in 1995 (9). A major characteristic of this Ntn hydrolase superfamily is a post-translational processing step that generates the active enzyme. The enzymes are expressed as inactive precursors that undergo a slow intramolecular autoproteolytic cleavage reaction at a specific site resulting in two tightly associated subunits, ␣ and ␤, also called protomers. The catalytic residue acting as the nucleophile is exposed at the very N terminus of the newly generated ␤-subunit, which remains complexed with the ␣-subunit during catalysis (10 -12). A well characterized mammalian member of the Ntn nucleophile hydrolase superfamily is the human lysosomal aspartylgluco-saminidase, which catalyzes the hydrolysis of glucosylated L-Asn molecules generated during proteolytic breakdown of glycoproteins (13,14). The human genome encodes another enzyme of this Ntn hydrolase family, variably termed human asparaginase-like protein 1 (15), glial asparaginase (16), CRASH (17), or hASNase3 because its high homology with E. coli L-ASNase3 (encoded by the iaaA gene) (18). Crystal structures of the wild-type form and of a circular permutant version of hASNase3 have been reported recently (19 -22).
The two aforementioned human enzymes and some of their microbial and plant homologs have been characterized as representatives of the Ntn hydrolase superfamily. In contrast, little is known about bacterial-type L-asparaginases expressed in mammalian tissues. An interesting example of such an enzyme is the 60-kDa lysophospholipase first isolated from rat liver (23). This enzyme comprises two domains: an N-terminal domain that is homologous to the E. coli cytoplasmic-type I L-asparaginase (EcASNase1; encoded by the ansA gene) and a C-terminal ankyrin repeatlike domain. This 60-kDa lysophospholipase from rat liver was shown to exhibit three distinct enzymatic activities (lysophospholipase, L-asparaginase, and platelet-activating factor acetylhydrolase) (23); however, no further studies have been reported on this protein, and little is known about its human homolog. In fact, the human 60-kDa lysophospholipase is a poorly characterized protein that shows 79% overall amino acid sequence identity with its rat homolog. In addition, although it has been demonstrated (24) that the protein holds lysophospholipase activity in vitro and plays a key role in cell proliferation and regulation of membrane channels, there are no supporting data to assign L-asparaginase activity to this human two-domain protein.
The present study focuses on the bacterial expression, purification, and biochemical characterization of the N-terminally located L-asparaginase domain (residues 1-369) of the 60-kDa (573-residue) human lysophospholipase. Because of high sequence similarity of this domain with EcASNase1 (47% identity), we designate it as hASNase1. Lysophospholipase activity was detected neither for the hASNase1 enzyme nor for the bacterial homolog. We show that the purified enzyme exhibits characteristics similar to those of bacterial type I L-asparaginase. Our kinetic and mutagenesis studies on hASNase1 unveil strong positive allosteric modulation as a function of L-Asn concentration through the action of a second L-Asn binding site, which we probed by cysteine-specific fluorescence labeling of the allosteric site. These results are in line with a previous crystallographic and mechanistic study on the E. coli cytoplasmic homolog demonstrating L-Asn-dependent allosteric regulation of the bacterial enzyme in vitro (25). Interestingly, we observed a substrate-inducing effect on the oligomeric state of hASNase1; however, the predominant state of the enzyme at assay conditions is monomeric. Our thermodynamic analysis of hASNase1 suggests conformational features and stability distinct from its E. coli counterpart. By mutational analysis, we highlight critical residues located at its putative allosteric and catalytic sites.
Cloning of hASNase1, EcASNase1, and Chaperonin 10-The cDNA region coding for the N-terminal domain (termed hASNase1) of the full-length 60-kDa human lysophospholipase was PCR-amplified using as template a cDNA clone of the ASPG gene (human L-asparaginase homolog, NCBI Reference Sequence NM_001080464; GenBank TM accession number BC035836) isolated from female ovarian tissue (Source Bioscience, UK). Four different C-terminal truncation constructs of hASNase1 were generated and subsequently tested for solubility and enzymatic activity as described in the following sections. NdeI and BamHI sites, respectively, were incorporated in the 5Ј-ends of the amplifying oligonucleotides (primers are listed in the supplemental table). The PCR mixture in a 50-l final volume consisted of 50 ng of template plasmid DNA, oligonucleotide mixture (10 pmol each), KAPA high fidelity buffer, dNTPs (0.2 mM each), and 1 unit of KAPA HiFi DNA polymerase. The PCR fragment was gel-purified and subjected to additional overlap extension PCRs aiming at the elimination by silent mutation of an internal BamHI restriction site (nucleotide sequence position 740) whose presence would limit the unique cleavage at the 3Ј-end. The final PCR product was digested with NdeI and BamHI and ultimately ligated overnight at 16°C into the pET14b-SUMO vector (10) using T4 DNA ligase. The ligation mixture was used to transform DH5␣ E. coli cells. Plasmid DNA isolated from single colonies was digested with NdeI and BamHI to identify positive clones, some of which were sequence-verified. The final constructs include an N-terminal six-histidine tag (His 6 ) followed by the small ubiquitin modifier (SUMO; Smt3p protein from yeast; 101 residues; molecular mass, 11.2 kDa) tag, which has been shown to improve heterologous protein solubility and stability (26). For bacterial expression, the E. coli BL21(DE3) pLysS strain was transformed with each of the four hASNase1 constructs. Expression vectors for EcASNase1 and Chaperonin 10 (GroES; encoded by the Cpn10 gene) were constructed analogously.
Mutagenesis of hASNase1 and EcASNase1-All hASNase1 single site mutants (T19A, T116A, T187S, T187A E266R, E266S, and  C299S) and EcASNase1 mutants (R240E and C273S) were generated by applying the QuikChange methodology (Stratagene) using as templates the cloned wild-type genes and KAPA HiFi DNA polymerase. Mutants T187S and T187A were generated in a single PCR using the same pair of primers by making use of the wobbletype oligonucleotides (G/T) at the first nucleotide of the T187 codon (ACC). Mutants E266R and E266S were constructed in a similar way (the primers are listed in the supplemental table). Sitespecific mutations were confirmed by sequencing the entire coding regions (Seqlab, Goettingen, Germany). All hASNase1 mutants were expressed and purified following the protocol described below for the wild-type enzyme.
Protein Expression and Purification of hASNase1, EcASNase1, and GroES-E. coli BL21(DE3) pLysS cells containing the hASNase1 plasmid were cultured overnight at 37°C in LB medium supplemented with 200 g/ml ampicillin and 35 g/ml chloramphenicol. A fraction of this culture (dilution 1:100) was used to inoculate fresh 2ϫ YT medium (1% yeast extract, 1.6% Tryptone, 0.5% NaCl) supplemented with 200 g/ml ampicillin and 35 g/ml chloramphenicol. When the bacterial culture reached an A 600 of ϳ0.5-0.7, its temperature was lowered to 22°C, and protein expression was induced by adding isopropyl 1-thio-␤-D-galactopyranoside to a final concentration of 0.25 mM. After incubation at 16°C for 18 h, the cells were centrifuged at 4,000 ϫ g for 30 min, resuspended in Ni-NTA-agarose affinity matrix binding buffer A (50 mM Na 2 HPO 4 , 0.3 M NaCl, 10 mM imidazole, 2% glycerol, 1 mM PMSF, pH 8.0), and ultimately lysed by sonication. The cell lysate was centrifuged at 17,200 ϫ g for 45 min, and to the cleared supernatant was added a fraction (1:10 dilution) of 10ϫ ATP buffer B (50 mM Na 2 HPO 4 , 0.3 M NaCl, 100 mM ATP, 200 mM MgCl 2 , 500 mM KCl, pH 8.0). The combined mixture was incubated in a 37°C water bath for 15 min, mixed with preequilibrated Ni-NTA-agarose beads, and incubated at 4°C for 3 h under rotation. Subsequently, the mixture was transferred to a 12-ml polypropylene column and dried by gravity. To remove unspecifically adsorbed material, the nickel resin was first washed with 25 bed volumes of washing buffer C (50 mM Na 2 HPO 4 , 0.5 M NaCl, 20 mM imidazole, 2% glycerol, pH 8.0). The resin was then mixed with a 10-fold diluted buffer B supplemented with GroES at a final concentration of 50 M. The mixture was incubated at 4°C for 2 h under rotation, dried by gravity, and then further washed with 10 bed volumes of buffer B without GroES. Finally, the bound His 6 -SUMO-hASNase1 protein was eluted from the column by applying 300 mM imidazole in buffer A, and fractions were collected dropwise and tested for L-asparaginase activity. All purification steps were performed at 4°C. The fractions that contained active hASNase1 were mixed, buffer was exchanged against buffer D (20 mM Tris-Cl, 5% glycerol, pH 8) using a PD-10 column (GE Healthcare), and then the sample was subjected to DEAE Sephacel anion exchange chromatography. Elution was performed by applying an NaCl gradient (0 -250 mM) in buffer D. Fractions were analyzed by SDS-PAGE, those containing Ͼ90% pure hASNase1 were pooled, and buffer was exchanged against buffer E (50 mM Tris-Cl, 100 mM NaCl, pH 8). The protein was incubated with yeast SUMO protease (molar ratio of protease: protein, ϳ1:100) at 4°C overnight (alternatively 2 h at 30°C) to cleave the N-terminal His 6 -SUMO tag. In the last purification step, the protein was subjected to size exclusion chromatography by passing it through a Superdex 200 column (size, 30 ϫ 1 cm; GE Healthcare) to remove the cleaved tag and remaining impurities. Protein purity was evaluated by SDS-PAGE and was estimated to exceed 95%. The protein sample was aliquoted, mixed with 25% glycerol, and stored at Ϫ20°C at a final concentration of 2 mg/ml until use. The aforementioned purification protocol was followed for the purification of wild-type hASNase1 as well as for all mutants described in this study. The final average yield of the purified human enzymes was ϳ 0.3 mg/liter of 2ϫ YT medium. EcASNase1 and GroES were purified following the protocol described for hASNase1 excluding the washing steps with ATP and GroES as well as the intermediate anion exchange treatment. EcASNase1 and GroES purifications yielded ϳ10 and 2 mg/liter of 2ϫ YT medium, respectively. Protein quantification was performed by using the Bradford method (27) and by the calculated molar absorption coefficient of each enzyme (23,295 M Ϫ1 cm Ϫ1 for hASNase1; 26,360 M Ϫ1 cm Ϫ1 for EcASNase1; 1,490 M Ϫ1 cm Ϫ1 for GroES) at 280 nm based on the amino acid sequence (28). The two quantification methods showed less than 10% difference, and ultimately the values were averaged.
L-Asparaginase and Lysophospholipase Activity Assays-We applied two different methods for monitoring L-asparaginase activity: a single step discontinuous method and a two-step continuous assay for real time measurements of the enzymatic activity. The discontinuous method, the so-called Nesslerization procedure that quantifies ammonia liberated upon deamidation of L-asparagine, has been described previously (29). The continuous assay, which monitors the NADH-dependent conversion of ␣-ketoglutarate plus ammonia to glutamate in a glutamate dehydrogenase-coupled reaction (30), was applied for the kinetic characterization of the enzymes using a Uvikon 943 double beam UV/visible spectrophotometer. All enzymes used for the activity measurements and the kinetic experiments were free of the His 6 -SUMO tag, which was removed during the last size exclusion chromatographic step (see above). For steadystate kinetic analysis, L-Asn concentrations in the range of 0 -10K m were tested in a final volume of 1 ml of 50 mM Tris-Cl, 100 mM NaCl, pH 8 at 37°C. The final enzyme concentration was ϳ0.5 M (20 g in 1 ml). The obtained V/E (velocity/total enzyme concentration) values were plotted against the respective substrate concentrations. Kinetic constants S 0.5 and k cat were calculated from the resulting plots by non-linear regression using the Hill equation (Equation 1) and analyzed by the SoftZymics software (Igor Pro, Wavemetrics).
Here, V is the initial velocity of the enzymatic reaction, V max is the maximal velocity, [S] is the substrate concentration, [S 0.5 ] is the substrate concentration that yields half-maximal velocity, and n is the Hill coefficient. Potential lysophospholipase activity of hASNase1 was tested using a continuous fluorescence assay as described previously (31). Briefly, this assay detects free fatty acid molecules, which are released upon hydrolysis of lysophospholipids by lysophospholipases; free fatty acid then is complexed with the intestinal fatty acid-binding protein conjugated with the fluorescent probe acrylodan, commercially known as a free fatty acid indicator and abbreviated as ADIFAB. Free fatty acid binding to ADIFAB induces a fluorescence signal decrease due to changes of the position of the acrylodan fluorophore, which is located close to the free fatty acid binding pocket of intestinal fatty acid-binding protein. In the present study, we tested palmitoyl-lysoPC and lysoPI as candidate substrates of hASNase1. The assay was performed in a quartz cuvette according to the manufacturer's instructions using final concentrations of 100 M for each substrate, 1 M ADIFAB, and 1.8 M enzyme (15 g) in 200 l of 50 mM HEPES, 140 mM NaCl, 5 mM KCl, 1 mM Na 2 HPO 4 , pH 7.5. Fluorescence intensity was recorded continuously for 10 min at 25°C using a Jasco FP 8300 spectrofluorometer in the high sensitivity mode (excitation at 386 nm and emission at 432 nm with 2.5-nm bandwidths). The assay was initially standardized by testing the fluorescence signal decrease upon binding of palmitic acid (one of the expected products of substrate hydrolysis; lysoPI contains a mixture of fatty acid derivatives; see "Materials") with the ADIFAB sensor protein as well as by measuring the activity of phospholipase A 2 against L-␣-phosphatidylcholine (also provided as a mixture of different fatty acid derivatives). Both hASNase1 and EcASNase1 were tested for lysophospholipase activity.
Homology Modeling of hASNase1-A homology model of the N-terminal L-asparaginase domain of the 60-kDa human lysophospholipase based on the recently determined structure of the homologous EcASNase1 enzyme (25) was obtained from the protein structure prediction service Phyre (32) (Job code 77beadcf4ff84dfc, Fold Library ID c2p2dA, identity of 47%, estimated precision of 100%). The resulting hASNase1 structure was overlaid on the EcASNase1 crystal structure (Protein Data Bank code 2HIM) using PyMOL (33). The putative allosteric and catalytic sites of hASNase1 were analyzed for the presence of critical residues previously identified in EcASNase1.
Probing the Putative Allosteric Site of hASNase1 by Fluorescence Labeling of Cys 299 -Cysteine residue Cys 299 of the catalytically inactive T19A mutant of hASNase1 was labeled using the compound Atto 465-maleimide as a fluorescent dye (Sigma-Aldrich). Maleimides show excellent reactivity with thiol groups, and therefore, they can serve as site-specific labeling probes in cases where reduced cysteines are available. Labeling was done in the following way. Highly purified enzyme (Ͼ95% pure as determined by SDS-PAGE) in 20 mM Tris-Cl, 100 mM NaCl, pH 7.2 was mixed under stirring with a 10-fold molar excess of the dye (dissolved in DMSO). The final concentrations in the reaction mixture (0.5 ml) were 20 M enzyme and 200 M dye. The mixture was incubated at 22°C for 2 h in the dark, and subsequently 1 mM of glutathione (GSH) was added to remove excess thiol-reactive dye. The labeled enzyme mixture was dialyzed against 50 mM Tris-Cl, 100 mM NaCl, pH 8. The labeling of the enzyme was confirmed by SDS-PAGE analysis applying 5 g each of unlabeled enzyme and of the final enzyme-dye conjugate because it has been shown (34) that labeled proteins migrate more slowly than unlabeled species.
For measuring interactions of hASNase1 with L-Asn, L-Asn in the concentration range of 0.1-40 mM was incubated with labeled enzyme (1.5 M final concentration) in a final volume of 100 l. The samples were left for 2 h on ice to equilibrate. Fluorescence intensity was recorded using a Jasco FP 8300 spectrofluorometer in the high sensitivity mode (excitation at 463 nm and emission at 508 nm with 2.5-nm bandwidths). All measurements were done in triplicate. Because hASNase1 contains overall 6 cysteines, the mutant C299S was also labeled and served as a negative control (background fluorescence) for signal changes upon L-Asn binding. The resulting saturation binding curve was fitted using SoftZymics software (Igor Pro, Wavemetrics) according to Equation 2, where F is the fluorescence intensity, F 0 is the fluorescence intensity at zero concentration of ligand, F max is the plateau fluorescence intensity, K d is the dissociation constant, and [S] is the ligand concentration. Thermodynamic Characterization of hASNase1 Conformational Stability-The conformational stability of hASNase1 and EcASNase1 was studied by monitoring changes in intrinsic fluorescence of the stepwise urea-denatured proteins (35,36). The final urea concentrations ranged from 0.5 to 8 M using a stock solution of 10 M urea dissolved in 50 mM Tris-Cl, 100 mM NaCl, pH 8. The final concentration of the enzymes was adjusted to 50 g/ml (ϳ1.25 M for both enzymes) in a final volume of 100 l. The two tryptophan residues of hASNase1 (Trp 93 and Trp 150 ) exhibited intrinsic fluorescence upon excitation at 295 nm (this wavelength was chosen to reduce excitation of tyrosines), showing an emission maximum at 343 nm. The same excitation and emission maxima were obtained for EcASNase1 containing a single tryptophan residue (Trp 68 ). Before measurements, equilibrium of the unfolding reactions was confirmed by monitoring the fluorescence signals at different time points until no further change was observed. Data were analyzed assuming a two-state model for reversible protein unfolding. Isothermal urea-induced unfolding experiments were carried out at 10, 15, 20, 25, 30, 35, and 40°C. For each temperature, data analysis using the linear extrapolation method (37) yielded three parameters: (i) the difference of Gibbs free energy between the native and the unfolded state of the protein (⌬G H2O ); (ii) the dependence of ⌬G H2O on denaturant concentration (parameter m, which is the slope of the equation ⌬G ϭ ⌬G H2O Ϫ m[urea]), which can be thought of as a measure for the sensitivity of the protein toward the unfolding agent; and (iii) the concentration of urea at which the protein is half-unfolded (C 1/2 ). Subsequently, the calculated ⌬G H2O values (free energy change at zero concentration of urea) resulting from each temperature were fitted to the Gibbs-Helmholtz equation (Equation 3) (37) as a function of temperature to obtain the ⌬C p values. T m and ⌬H m values as fitting parameters were calculated from temperature-induced denaturation experiments (range, 15-85°C) following the intrinsic tryptophan fluorescence changes of protein similar to urea experiments.
⌬G(T) is the Gibbs free energy difference between the folded and unfolded state at temperature T, T m is the melting temperature, ⌬H m is the enthalpy difference at T m , and ⌬C p the heat capacity difference of the folded and unfolded states at constant pressure.
Thermal Stability Determination of Wild-type and Mutant hASNase1 Enzymes by Differential Scanning Fluorometry-The effect of point mutations on thermal stability of hASNase1 was studied by applying differential scanning fluorometry techniques (38,39). Enzyme samples were dialyzed (Slide-A-Lyzer, Pierce; 10,000 molecular weight cutoff) against 50 mM Tris-Cl, 0.1 NaCl, pH 8 to remove glycerol and subsequently mixed with SYPRO Orange (Sigma-Aldrich) in a final volume of 20 l. The final concentrations of the enzyme and the dye were 2 M and 10% (v/v), respectively; the DMSO stock solution of the dye (5,000ϫ) was prediluted in H 2 O, giving a 100ϫ solution, from which aliquots were used according to the experimental needs. The samples were mixed in a 96-well plate suitable for real time PCR measurements, centrifuged at 500 rpm for 30 s, and finally sealed with heat-resistant membranes (Microseal B adhesive sealer) to prevent evaporation. The enzyme melting experiments were performed using a CFX96 real time PCR machine (Bio-Rad) with the following settings: 2-min prewarming step at 30°C and subsequent gradient between 31 and 95°C with 1°C/min increments. SYPRO Orange fluorescence was monitored using the filters provided with the machine: FAM (492 nm) for excitation and ROX (610 nm) for emission. Data were exported as an Excel-based worksheet and further analyzed by Igor Pro (Wavemetrics). Melting temperatures (T m ) were obtained by plotting the first derivative d(AFU)/dT (where AFU represents arbitrary fluorescence units) of the raw data as a function of temperature increase (40).
Effects of pH, Divalent Metal Ions, and DTT on Enzyme Activity and Stability-Effects of pH on enzymatic activity and stability were assessed in a broad range of pH values using several buffers: sodium acetate (pH 3-5), sodium phosphate (pH 6 -7), Tris-Cl (pH 7-8.5), and CAPSO (pH 9 -10), all at 50 mM concentration in the presence of 100 mM NaCl. For the investigation of the pH effect on the activity of the enzyme, ϳ50 g of enzyme were assayed in a final volume of 1 ml at saturating substrate concentration (20 mM L-Asn) applying the Nesslerization method. The discontinuous method was preferred to the NADH-dependent continuous assay because the different pH values could influence the activity of the auxiliary glutamate dehydrogenase enzyme, resulting in unforeseeable artifacts. The stability of hASNase1 under various pH conditions was studied by incubating the enzyme (at 0.1 mg/ml final concentration) in several buffers for 12 h at 4°C. Aliquots of these mixtures were assayed for residual L-asparaginase activity by Nesslerization using as standard buffer the one at which the enzyme showed the highest activity based on the previous activ-ity analysis. To study metal ion effects, purified hASNase1 (0.1 mg/ml) was preincubated with a final concentration of 1 mM for each of several divalent metal ions, EDTA, and DTT for 3 h at 4°C. Subsequently, aliquots were tested for residual activity by the Nesslerization method.

RESULTS
Expression and Purification of the hASNase1 Domain-Four different C-terminal truncations (numbered 1, 2, 3, and 4; see Fig. 2A) of full-length human lysophospholipase (gene code ASPG; asparaginase homolog) were designed based on sequence alignments of the N-terminal putative L-asparaginase domain and different procaryotic (E. coli, EcASNase1, and Pyrococcus horikoshii, PhASNase1) and eucaryotic (Rattus norvegicus, RnASNase1, and Saccharomyces cerevisiae, ScASNase1) cytoplasmic L-asparaginases (Fig. 1). All four constructs were tested for protein expression in E. coli and solubility under a spectrum of different conditions (culture medium, E. coli strain, isopropyl 1-thio-␤-D-galactopyranoside concentration, and temperature). Interestingly, truncation 3, which appears to fit best to the bacterial cytoplasmic enzymes according to the amino acid sequence comparison, was not soluble, and trials to purify the enzyme under denaturation/renaturation conditions (unfolding by up to 8 M urea and refolding by dialysis) ultimately resulted in a catalytically inactive protein. Similarly, truncations 1 and 2 failed to produce soluble, active hASNase1. In contrast, truncation 4 (369 amino acid residues), containing 5 additional amino acids (VEERR) downstream of truncation 3, produced the protein in soluble form (Fig. 2B), and most importantly, the protein purified under native conditions exhibited L-asparaginase activity.
However, when the protein purity was analyzed by SDS-PAGE after the last gel filtration step (see "Experimental Procedures"), a prominent band with an apparent molecular mass of ϳ60 kDa and an estimated stoichiometric 1:1 ratio to the hASNase1 band was constantly observed (Fig. 3). Multiple attempts to remove this associated protein species by altering standard purification protocols were unsuccessful. This prompted us to analyze this protein band by mass spectrometry by which it was finally identified as the endogenous E. coli chaperone GroEL. Indeed, contamination by co-purified chaperones can be a hurdle in recombinant protein expression and purification setups (41). Although it has been described previously (42) that incubation of the crude cell extract with 10 mM ATP at 37°C and washing of the affinity column with an ATPcontaining buffer facilitate the removal of GroEL, this approach did not prove beneficial in our case. Moreover, attempts to use an engineered E. coli strain (43) (a kind gift from Prof. Ulrich Hartl's laboratory, Max Planck Institute for Biochemistry, Munich, Germany) carrying the GroEL/ES operon under an arabinose-inducible promoter, which allows tuned regulation of the expression levels of these chaperones, failed to produce the hASNase1 protein (data not shown). Ultimately, we managed to remove the co-purified chaperone by including in the washing buffer the natural binding partner of this protein, GroES, at ϳ10-fold excess over GroEL (44). It appears reasonable to assume that the presence of ATP-Mg 2ϩ -GroES weakened the binding between GroEL and hASNase1, thus facilitat-ing their separation upon exhaustive washing (Fig. 3). It is worth mentioning that the bound GroEL did not affect the activity of hASNase1 because activity measurements on both enzyme preparations (with and without bound GroEL) yielded similar kinetic constants. We emphasize that all hASNase1 variants produced in this work were expressed at levels similar to the wild-type enzyme and showed no aggregation or precipitation tendency at any purification step. GroEL contamination was present during purification of all mutant enzymes as well.
Size exclusion chromatography experiments allowed us not only to obtain a highly pure protein as evidenced by SDS-PAGE analysis (Fig. 3) but also to gain insight into the oligomeric state of hASNase1. Strikingly, when we first analyzed native hASNase1, we observed that the enzyme (ϳ3 M) was running as a monomer (molecular mass, ϳ40 kDa) on a Superdex 200 gel filtration column (Fig. 4C). As the shift between oligomeric states of enzymes can be drastically influenced by their substrates and/or other interacting partners (45,46), we decided to perform gel filtration analysis in the presence of the substrate of hASNase1. Interestingly, with 20 mM L-Asn in the running buffer, we witnessed a slightly shifted shoulder peak of the previously observed chromatographic peak assigned to the 40-kDa hASNase1 monomer, indicative of a higher molecular weight species, although the monomer remained the predominant form (Fig. 4D). This finding suggested that L-Asn could trigger the association of monomeric hASNase1 molecules and consequently induce the formation of dimers and even tetramers, which are the characteristic molecular species reported for EcASNase1 (25) as we confirmed in the present study (Fig.  4B). However, at enzymatic assay conditions, hASNase1 (ϳ0.5 M) is expected to be monomeric because this concentration is 6-fold lower than the concentration of the eluted monomeric protein shown in Fig. 4 (0.5 mg of enzyme loaded on the gel filtration column eluted in a final volume of ϳ4 ml of buffer, i.e. ϳ3 M final concentration of the enzyme).

The hASNase1 Enzyme Shows Non-Michaelis-Menten Kinetics for L-Asn Hydrolysis Similar to Its E. coli Homolog but
Lacks Lysophospholipase Activity-The N-terminal domain of human 60-kDa lysophospholipase, which we have designated as hASNase1, shares 47% sequence identity with E. coli cytoplasmic L-asparaginase (encoded by the ansA gene; EcASNase1). The crystal structure of EcASNase1 has been reported recently (25), shedding light on distinct structural features associated with L-Asn binding to the catalytic site and to the site responsible for allosteric regulation. Based on this structure, we modeled the hASNase1 structure using the Phyre program. Fig. 5 shows the active and the allosteric sites in the predicted hASNase1 structure overlaid by the homologous regions of EcASNase1. Residues critical for catalysis in the E. coli enzyme, such as Thr 14 , Thr 91 , Asp 92 , and Lys 163 , overlay with high accuracy with those of hASNase1, which has identical residues at these positions (Thr 19 , Thr 116 , Asp 117 , and Lys 188 ) (Fig. 5A). Importantly, resi-

FIGURE 1. Alignment of the amino acid sequences of hASNase1 (UniProt accession number Q86U10), R. norvegicus (UniProt accession number O88202), E. coli (UniProt accession number P0A962), P. horikoshii (UniProt accession number O57797), and S. cerevisiae (UniProt accession number P38986) L-asparaginases type I.
The displayed sequence of hASNase1 corresponds to truncation 4, which showed the best expression pattern of the four protein constructs produced in this work (see Fig. 2). Asterisks indicate the two threonine residues that are critical for L-asparaginase activity, and red arrows indicate residues that are located close to the allosteric sites. The alignment was performed using ClustalW (67), and the graph was generated using JalView (68). Blue shading indicates highly conserved amino acid residues.

JOURNAL OF BIOLOGICAL CHEMISTRY 12967
dues Thr 14 and Thr 91 of the bacterial enzyme are considered to be the primary nucleophiles for the attack on the substrate L-asparagine (25). The presence of threonine residues at equivalent positions in the human enzyme is in favor of the view that the primary nucleophile in bacterial L-asparaginases (both cytoplasmic and periplasmic) is a threonine residue (47). This prompted us to investigate this assumption for hASNase1 by mutagenesis studies. Indeed, when point mutants T19A and T116A of hASNase1 were assayed, no L-asparaginase activity was detected (Table 1), thus indicating the critical catalytic role of these threonine residues in hASNase1.
Steady-state kinetic characterization of wild-type hASNase1 using L-Asn as substrate (there was no detectable L-glutaminase activity) revealed that the enzyme did not follow Michaelis-Menten kinetics. Instead, it exhibited a pronounced sigmoidal kinetic behavior, which is a hallmark of allosteric enzymes. The kinetic data were fitted using the Hill equation (Equation 1 under "Experimental Procedures") from which we estimated a Hill coefficient (n H ) of 3.9 and an S 0.5 value of 11.5 mM (Fig. 6). The modeled structure of the allosteric site is shown in Fig. 5B in direct comparison with that of EcASNase1. Significant differences between the two allosteric sites are seen in two key residues that directly interact with the substrate L-Asn: Glu 266 and Ser 327 in hASNase1 as opposed to Arg 240 and Val 302 in the bacterial enzyme. Aiming at further characterization of the allosteric sites of both enzymes, we generated a series of mutants and expressed, purified, and tested them for catalytic activity. Strikingly, hASNase1 tolerated none of the mutations introduced in the predicted allosteric site as evidenced by the total lack of activity of the mutant enzymes (Table 1). Given the high structural homology between hASNase1 and EcASNase1, we reasoned that the mutation E266R could potentially lower the S 0.5 value of the human enzyme by mimicking the Arg 240  site of the bacterial homolog. However, this mutation totally inactivated hASNase1. Other mutations introduced around the putative allosteric site of hASNase1, such as E266S, C299S, T187S, and T187A, also abolished activity (Table 1). Unlike the human enzyme, the E. coli enzyme tolerated mutations of its allosteric site. Mutation R240E resulted in an ϳ5-fold increase of S 0.5 and 2-fold increase of k cat while simultaneously lowering the Hill coefficient n H from 3.55 to 1.9. The C273S mutation increased even further the S 0.5 value (7-fold with respect to wild type), whereas the n H value slightly decreased (Fig. 6). All kinetic data of wild-type hASNase1 and EcASNase1 as well as of mutants generated in this work are summarized in Table 1. Taken together, these results suggest that despite the high similarity of sites predicted to be critical for their activities these two enzymes display different degrees of tolerance toward mutations of residues at these sites.
We additionally tested for potential hydrolysis of palmitoyl-lysoPC and lysoPI by hASNase1 and EcASNase1. We chose these two substrates because the rat 60-kDa lysophospholipase has been reported to hydrolyze palmitoyl-lysoPC (23), whereas lysoPI could serve as a substrate for the human homolog (24). However, we were unable to detect any activity of either hASNase1 or EcASNase1 using these two substrates under conditions where the control enzyme phospholipase A 2 hydrolyzed L-␣-phosphatidylcholine. The rate of the fluorescence decrease in the ADIFAB assay was the same for the blank (ADIFAB plus substrate without enzyme) and the sample (ADIFAB plus sub- Red highlights amino acids that are critical for activity of bacterial L-asparaginases: the catalytic triad formed by residues Thr 91 , Asp 92 , and Lys 163 that ensure proper orientation of the substrate L-Asn (dark gray) and Thr 14 , which is responsible for subsequent nucleophilic attack (23,69). Blue highlights the respective amino acids of hASNase1: Thr 116 , Asp 117 , Lys 188 , and Thr 19 . B, modeled allosteric site of hASNase1 (magenta residues) overlaid by the EcASNase1 allosteric site (blue residues). Key differences in homologous sites of the human enzyme are residues Glu 266 versus Arg 240 and Ser 327 versus Val 302 . Other residues of the allosteric sites of the two enzymes are highly similar. Also shown is an L-Asn molecule (dark gray), which interacts with allosteric site residues. The interactions between L-Asn and the human enzyme were predicted based on the structure of the bacterial enzyme (Protein Data Bank code 2HIM).

TABLE 1 Kinetic data on wild-type and mutants of hASNase1 and EcASNase1
Steady-state kinetic analysis was performed by applying a continuous NADH-dependent assay (30). Enzyme concentrations were ϳ0.5 M in a final volume of 1 ml of 50 mM Tris-Cl, 100 mM NaCl, pH 8 at 37°C; the tested substrate concentrations covered the range of 0 -10K m . Parameters were calculated from non-linear regression of a V/E versus ͓L-Asn͔ plot using the software SoftZymics (Igor Pro, Wavemetrics). ND, non-detectable; the limit of detection for the applied assays in the present study is 0.5 M of ammonia produced in a 1-ml reaction volume. Data are shown as mean values ϮS.D. of triplicate measurements.  strate and enzyme) and accounted for ϳ10 arbitrary fluorescence units/min (data available upon request). This background activity is in line with another study (31), which also reported a constant slight fluorescence signal decrease in the absence of enzyme. In conclusion, our results demonstrate that the N-terminal domain of the 60-kDa full-length human lysophospholipase can exist as a distinct folding unit that resembles the bacterial-type I L-asparaginases, lacking the capacity to hydrolyze substrates other than L-Asn. Fluorescence Labeling of hASNase1-Cysteine labeling allowed us to investigate the presence of an L-Asn binding site distinct from the catalytic site in the monomer. This additional binding site could play the role of an allosteric site of hASNase1, and its existence is supported not only by the sigmoidal kinetic behavior of enzyme described in the previous paragraph but also by our results obtained from substrate titration to a fluorescence-labeled, catalytically inactive version of the protein.

Enzyme
Binding of the substrate L-Asn to the putative allosteric site was monitored by fluorescence signal changes of the Atto dye-labeled enzyme. The dye was covalently attached to cysteine residue Cys 299 , which according to the structural model built on the E. coli enzyme directly interacts with L-Asn bound to the allosteric site (Fig. 5B). The labeling of hASNase1 was confirmed by SDS-PAGE analysis as shown in Fig. 7 based on the observation that labeled proteins migrate more slowly than the unlabeled species (34). Fig. 8 shows a pronounced fluorescence signal decrease upon incubation of the catalytically inactive T19A mutant of hASNase1 with L-Asn in the concentration range of 0.1-40 mM. We did not label the wild-type enzyme because upon titration of L-Asn notably at high concentrations it would hydrolyze the substrate without allowing us to solely evaluate binding phenomena (the T19A mutant is assumed to maintain intact its allosteric site similarly to the wild-type enzyme). From the resulting binding curve, we estimated a K d of ϳ1 mM, which is about 10-fold lower than the S 0.5 (ϳ11 mM). Unlike T19A, the C299S mutant, which served as a negative control for labeling with the Atto dye, yielding a constant background fluorescence (due to labeling of other cysteines), showed no fluorescence signal changes upon incubation with L-Asn. This indicates that signal changes observed for T19A can be attributed to direct interaction of the substrate L-Asn with either Cys 299 or other neighboring residues located within the allosteric cavity. Conversely, we cannot exclude potential interactions between L-Asn and the active site residues leading to conformational alterations that are transmitted to the allosteric site and reflected by fluorescence signal changes. Incubation of labeled enzyme with aspartate induced no fluorescence signal change (data not shown).
Thermodynamic Characterization of hASNase1 and EcASNase1-The free energy difference ⌬G H2O of folded and unfolded states, which is a measure of protein stability, was determined for both enzymes from urea denaturation experiments at different temperatures (37). Non-linear regression analysis was applied to the experimentally obtained ⌬G H2O values using the Gibbs-Helmholtz equation as a function of temperature, which allows the calculation of the ⌬C p parameter (48). Fig. 9 shows the stability curves for both enzymes studied in this work for direct comparison of their characteristic features under identical conditions. At 25°C, the bacterial enzyme has a 2-fold higher ⌬G H2O in comparison with the human enzyme, although this difference appears to decline at higher temperatures (35 and 40°C) as has also been observed in other proteins (49). Table 2 summarizes thermodynamic parameters obtained for hASNase1 and EcASNase1. Interestingly, the predicted ⌬C p values are considerably different for the two enzymes, possibly pointing to variations in the surface area that is exposed upon denaturation (50) and distinct mechanisms of enthalpy and entropy changes occurring upon unfolding (51). The two enzymes exhibited similar melting temperatures T m with hASNase1 showing higher ⌬⌯ m and ⌬S m values, indicating higher enthalpy changes upon unfolding and a higher degree of disorder as compared with the E. coli enzyme (52).
Differential Scanning Fluorometry of hASNase1-To analyze whether allosteric and active site mutations influenced the thermal stability of hASNase1, several of these hASNase1 mutants were studied by differential scanning fluorometry. The melting curves were all monophasic (Fig. 10). The melting tem- FIGURE 9. Temperature dependence of free energy change during unfolding of human and bacterial ASNase1. Shown is a plot of ⌬G H2O as obtained at various temperatures versus the respective temperatures for wild-type hASNase1 (OE) and EcASNase1 (F). Data were fitted to the Gibbs-Helmholtz equation (Equation 3 in text) after determining T m and ⌬H m by thermal denaturation experiments (see "Experimental Procedures"). The standard deviation for ⌬G H2O (free energy change at zero concentration of urea), evaluated from three independent determinations on the wild-type enzymes at 25°C, was found to be 0.3 kcal/mol.

TABLE 2 Thermodynamic parameters for wild-type hASNase1 and EcASNase1
The     Table 3. Our results suggest that three point mutations (T19A, T187A, and T187S) had a significant stabilizing effect on the enzyme . In contrast, mutations T116A and C299S considerably lowered protein stability as compared with wild type, whereas the two mutations (E266S and E266R) introduced at the allosteric site Glu 266 had almost no effect on the stability of hASNase1.
Dependence of hASNase1 Activity on pH and Metal Ions-The optimum activity of hASNase1 was investigated in the pH range from 2.5 to 10. Fig. 11A shows that the enzyme exhibited the highest activity in a relatively narrow alkaline pH range as evidenced by a sharp increase from pH 7 to 8 followed by an immediate decline from pH 8.5 to 10. In addition, the pH dependence for stability of the enzyme showed a maximum peak at pH 8 (Fig. 11B). These results suggest that protonation of certain residues that are involved in catalysis might partially impair the activity of hASNase1. Furthermore, certain divalent metal ions, such as Cd 2ϩ and Cu 2ϩ at 1 mM, totally inhibited hASNase1 activity, whereas DTT, Co 2ϩ , Ca 2ϩ , Ni 2ϩ , Mg 2ϩ , Mn 2ϩ , Fe 2ϩ , and Li ϩ only weakly influenced the hydrolysis of L-Asn. Such metal-selective inhibition by Cd 2ϩ and Cu 2ϩ possibly indicates an interaction between these metal ions and a thiol group pivotal for activity (53). A candidate might be the Cys 299 residue, which is located at the allosteric site of the enzyme and plays an important role in L-Asn binding. In addition, EDTA at concentrations up to 1 mM had no noticeable negative effect on the activity of hASNase1 (Table 4).

DISCUSSION
The present study focuses on the biochemical analysis of the N-terminal domain of a human 60-kDa protein designated as lysophospholipase. The full-length version of this two-do-  . Effect of pH on hASNase1 activity (A) and stability (B). Buffers used were: sodium acetate (pH 3-5), sodium phosphate (pH 6 -7), Tris-Cl (pH 7-8.5), and CAPSO (pH 9 -10), all at 50 mM concentration in 100 mM NaCl. Enzymatic activities were determined as described under "Experimental Procedures" applying the Nesslerization assay for determining ammonia generated by L-asparagine hydrolysis. Error bars represent S.D.

TABLE 3
Melting temperatures of wild-type and mutant hASNase1 as determined by differential scanning fluorometry Samples of 2 M highly purified enzyme solution were mixed with a 10% (v/v) final concentration of SYPRO Orange in a final volume of 20 l in a 96-well plate. Enzyme melting experiments were performed using a CFX96 RT-PCR machine (Bio-Rad) and the following settings: 2-min prewarming step at 30°C and subsequent gradient between 31 and 95°C with 1°C/min increments. SYPRO Orange fluorescence was monitored using FAM excitation (492 nm) and ROX emission (610 nm) filters. Melting temperatures (T m ) were calculated by plotting the first derivative d(AFU)/dT (where AFU represents arbitrary fluorescence units) of the raw data as a function of temperature increase (40).  main protein has been only poorly characterized in one report (24). Our primary motivation to produce and functionally characterize the N-terminal domain of this protein, which structurally and according to its catalytic in vitro properties significantly resembles E. coli cytoplasmic L-asparaginase (ansA; EcASNase1), originated from our previous work on human enzymes that possess L-asparaginase activity (10,19,21). The discovery, molecular engineering, and in vitro evolution of catalytically efficient human L-asparaginases are thought to lay the basis for the replacement of bacterial L-asparaginases presently used as antileukemia therapeutics despite adverse side effects mainly attributed to their bacterial origins (54). An early study on the rat 60-kDa lysophospholipase (23) assigned three distinct activities to this enzyme acting as lysophospholipase, L-asparaginase, and acetylhydrolase on the platelet-activating factor. Several years later, a report on the human 60-kDa lysophospholipase (24) revealed a specific role of this enzyme as an interacting partner of the serum-and glucocorticoid-induced serine/threonine protein kinase Sgk1, an enzyme involved in various cell proliferation pathways. In contrast to its rat homolog, the human enzyme was not reported to bear L-asparaginase activity (24). However, in both these studies, the distinct enzymatic activities were determined only qualitatively using lysates from cells overexpressing the respective genes rather than purified enzyme preparations.

T m°C
Here, we show that the N-terminal L-asparaginase domain (residues 1-369) of the 573-residue human protein, that we termed hASNase1, can be produced in soluble form in E. coli. This protein showed L-asparaginase activity with a maximum catalytic rate of about 7 s Ϫ1 , which is very similar to that of EcASNase1, implying that this N-terminal domain forms a stable and functional folding unit in the absence of the C-terminal putative ankyrin repeat domain. Kinetic and mutational characterization of hASNase1 revealed strong positive allosteric modulation in the velocity versus substrate plot similar to its E. coli homolog. To validate our experimental strategy for characterizing this human counterpart of the cytosolic E. coli L-asparaginase, we also recombinantly produced the bacterial protein to directly compare the two enzymes under identical assay conditions. Our kinetic data obtained on the bacterial enzyme are consistent with a previous study that centered on its structural analysis and allosteric regulation (25).
Given the pronounced sigmoidal kinetic behavior that we observed in steady-state kinetics of L-asparagine hydrolysis, we hypothesized that hASNase1 possesses an allosteric site that could act as "sensor" for the presence of substrate. Our dye labeling approach targeting the Cys 299 residue indicated fluorescence signal changes upon binding of L-Asn, thus strengthening the idea of the existence of such an allosteric site that can be regulated by the substrate L-Asn. In the absence of information on the three-dimensional structure of hASNase1, which could provide more detailed insight into the catalytic mechanism and allosteric regulation of the enzyme, we referred to a structural homology-model based on the EcASNase1 crystal structure, which allowed us to define the active site and an allosteric site of the human enzyme. The S 0.5 value of 11.5 mM determined for hASNase1 is about 10-fold higher than that reported for the E. coli enzyme, raising questions about the physiological role of the full-length protein and its N-terminal domain displaying L-asparaginase activity as shown in this work. The relatively high S 0.5 value determined for the human enzyme falls well within the range of free concentrations of intracellular amino acids (Ͼ10 mM), including L-Asn, reported for mammalian cells (55,56), thus making the enzyme operate efficiently, particularly at elevated substrate concentrations.
The homo-oligomeric EcASNase1 displays sigmoidal kinetics, which can be explained by positive cooperativity induced by the substrate L-Asn. In contrast, hASNase1, which shows non-Michaelis-Menten kinetics similar to its E. coli homolog, is monomeric under the conditions of activity measurements. In fact, a number of monomeric enzymes exhibiting allosteric behavior have been reported of which the most thoroughly characterized is human glucokinase (57,58). However, glucokinase (also called hexokinase IV), unlike L-asparaginase, is a twosubstrate (ATP plus glucose) enzyme and displays a moderate degree of allostery (59). However, monomeric enzymes with single binding sites, like hASNase1, can also show allostery (60). Two basic models have been put forward aiming at the mechanistic explanation of monomeric allostery: the mnemonic model (61) and the ligand-induced slow transition model (62). Both models assume the existence of two different enzyme conformations that are characterized by distinct affinities for the substrate (low and high affinity states). Depending on the substrate concentration, certain conformational changes may occur, thereby perturbing the equilibrium of the two states in favor of the high affinity state and an increased catalytic activity. Based on the observed sigmoidal kinetic behavior and the monomeric state of hASNase1, it appears plausible to assume that this is another example of a monomeric enzyme exhibiting positive allosteric regulation. Importantly, given the fact that L-Asn plays a dual role of being both substrate and regulator of hASNase1, L-Asn can be considered as a homotropic allosteric effector of this enzyme, adding to the steadily growing number of allosterically regulated proteins (63).
The role of the putative ankyrin repeat structure located in the C-terminal part of the full-length protein remains unknown at present. In numerous other proteins, ankyrin repeats were shown to mediate protein-protein interactions, and therefore, they may play crucial roles in cellular signaling events (64 -66). The study of the kinetics of the full-length 60-kDa human lysophospholipase could provide information on the potential influence of the ankyrin repeat domain on the L-asparaginase activity predicted to reside in the N-terminal part. On the other hand, when we tested palmitoyl-lysoPC and lysoPI as potential substrates for hASNase1, we detected no lipase activity. Our biochemical data clearly support the view that the N-terminal domain bears L-asparaginase activity that exhibits positive allosteric regulation by the substrate L-Asn. A particular characteristic of hASNase1 is its monomeric state in conjunction with its pronounced sigmoidal steady-state kinetics. We would like to emphasize that the distinct enzymatic activities assigned to rat and human 60-kDa lysophospholipases warrant further analyses at the cellular level to elucidate the physiological role of these two-domain proteins. Our studies on L-asparaginase activity inherent to the N-terminal domain of this 60-kDa protein expands the basis of our work aiming at the identification