Structures of Amidohydrolases AMINO ACID SEQUENCE OF A GLUTAMINASE-ASPARAGINASE FROM ACZNETOBACTER GLUTAMINASIFZCANS AND PRELIMINARY CRYSTALLOGRAPHIC DATA FOR AN ASPARAGINASE FROM E R WZNZA CHR YSANTHEMP

The complete amino acid sequence of a glutaminase- asparaginase from Acinetobacter glutaminasificam, for which a preliminary tertiary structure is available from crystallographic analysis, has been determined by automated Edman degradation of fragments pro- duced by chemical and proteolytic cleavages. The protein consists of 331 amino acid residues and has a molecular weight of 35,500. The pattern of hydrophilic and hydrophobic regions is typical of a globular protein. A new crystal form of an Erwinia chrysanthemi 1125 asparaginase is reported. The space group is monoclinic C2, with p i t cell parameters of: a = 107,8, b = 91.7, c = 129.2 A and /3 = 91.7’. A V,,, of 2.25 As/ dalton was calculated for one tetramer of 35,100-dal- ton subunits per asymmetric unit. X-ray intensity data have been obtained to 2.2 A resolution. The point group symmetry of the Er. chrysanthemi tetramer is 222 from self-rotation

their utilization in cancer therapy. Several amidohydrolases have been under investigation as potential anti-tumor agents over the last 20 years. The administration of L-asparaginase (L-asparagine amidohydrolase, EC 3.5.1.1) leads to the regression of certain lymphomas and leukemias in experimental animals and in humans (Oettgen et al., 1967;Kidd, 1970;Broome, 1981;Prager et al., 1982). The Escherichia coli and Er. chrysanthemi (previously known as Erwinia carotouora) enzymes are used clinically in the treatment of acute lymphoblastic leukemia (Oettgen et al., 1967;Hruschesky et al., 1976), and other amidohydrolases have been in clinical trials. Asparaginase-sensitive tumor cells generally show a diminished capacity to synthesize L-asparagine, because of relatively low levels of L-asparagine synthetase, and therefore require an exogenous supply of the amino acid for protein biosynthesis. It is presently unclear, however, exactly how asparaginase kills sensitive tumor cells, and the biochemical basis for asparaginase sensitivity has not been firmly established (Keffer et al., 1985).
L-Glutamine, like L-asparagine, is not an essential component in the human diet. The activity of a glutaminase-asparaginase against asparaginase-resistant cells (Schmid and Roberts, 1974) and in asparaginase-resistant patients (Spiers and Wade, 1976) has been demonstrated. Recently, a combination of a glutaminase-asparaginase from Pseudomonas 7A, 6-diazo-5-oxonorleucine (DON),' and acivicin has been shown to be active against human mammary and colon tumors in uitro, and in tumors growing in nude mice (Roberts, 1983). The glutaminase activity (approximately 3-5%) of E. coli asparaginase has a role in the inhibition of cell growth and in cell toxicity (Wu et al., 1978;Hakimi and Boseman, 1979). The simultaneous elimination of both amino acids might be therapeutically advantageous.
Preliminary crystal data and molecular replacement studies based on the tentative model of A. glutaminasificans glutaminase-asparaginase have been published for the enzymes from Pseudomonas 7A (Ammon et al., 1983), Vibrio succinogenes (Ammon et al., 1985), and E. coli (Ammon et al., 1988). Complete primary structures now are available for asparaginases from E. coli A-1-3 (Maita et al., 1974(Maita et al., , 1979Maita and Matsude, 1980) ,and Er. chrysanthemi 1066(Minton et al., 1986. The Er. chrysanthemi 1125 sequence is identical to 1066, apart from 4 residues.2 The NH2-terminal sequences of glutaminase-asparaginases from Acinetobacter glutaminasificans and Pseudomonas 7A have been reported (Holcenberg et al., 1978). The enzymes are homotetramers; in the case of E. coli asparaginase, for example, the subunit contains 321 residues and has a calculated molecular weight of 34,080.
In this paper we report the primary structure of the glutaminase-asparaginase from A. glutaminasificans, for which a preliminary x-ray crystal strycture is available (Ammon et al., 1988), and the growth of new, high quality crystals of the Er. chrysanthemi 1125 asparaginase, for which an amino acid sequence is known. In addition, the results of rotation and translation function studies for Er. chrysanthemi asparaginase and A. glutaminasificans glutaminase-asparaginase versus Er. chrysanthemi asparaginase are reported, and secondary structure predictions for A. glutaminasificans glutaminase-asparaginase, Er. chrysanthemi asparaginase, and E. coli asparaginase are compared.

RESULTS AND DISCUSSION
Amino Acid Sequence of A. glutaminasificans Glutaminase-Asparaginase-The sequencing results for A. glutaminasificans glutaminase-asparaginase are summarized in Fig. 1. The amino acid sequence of A. glutaminusificans glutaminaseasparaginase consists of 331 residues; the calculated molecular weight is 35,500. The sequence was obtained from analyses of the intact enzyme, and from characterization of peptides generated by cleavages at methionyl, lysyl, or arginyl residues (designated CB, K, and CT, respectively). In addition, two CB fragments were subdigested at aspartyl residues (designated D).
Cleavage of A. glutaminasificans glutaminase-asparaginase with cyanogen bromide was the major mode of fragmentation. One fragment, corresponding to residues 151-153, was not recovered. The other fragments were sequenced in part or in entirety. Sequence analysis of peptides K-2 through K-7 and peptides CT-1 through CT-4 served to align the CB fragments and complete their sequences. The sequence of CB-l-D2 joined K-2 and CT-1 and identified residues 66-72. The sequence of CB-6-D2 overlapped CT-3 and CT-4. The fractionation of the peptides, the amino acid compositions, and the sequence data are presented in Figs. 1M-5M and Tables  1M-4M, in the Miniprint. The results presented here differ in two positions from the amino-terminal sequence (residues 1-60) previously reported (Holcenberg et al., 1978). Residues 34 and 50 were identified as asparagine and aspartic acid, respectively, versus aspartic acid and threonine in the present work. We identified residue 34 as aspartic acid in the degradation of CB-1; no asparagine was detected. In addition, cleavage of CB-1 with the aspartylendopeptidase Pseudomonas fragbprotease yielded CB-1-D2 commencing with Asp-34. Residue 50 was identified as threonine by us in both the degradation of CB-1-D2 and of K-2. There is no apparent explanation for these discrepancies.
Er. chrysanthemi Asparaginase Crystals-The crystallization of an Erwinia asparaginase in the orthorhombic space group P212121 was reported nearly 20 years ago (North et al., 1969) is monoclinic, space group C2, with unit cell parameters of a = 107.8 A, b = 91.7 A, c = 129.2 A, / 3 = 91.7". The space group and cell parameters were determined from x-ray data collected with a Nicolet Proportional Counter and were confirmed by precession photography. Four 35,000-dalton subunits (Minton et al., 1986) per crystal asymmetric unit yield a V, (Matthews, 1968) of 2.25 A3/dalton. The value is similar to those observed for other amidohydrolases, which range from V, = 2.26 for E. coli asparaginase (Ammon et al., 1988) to 2.75 for A. glutaminasificans glutaminase-asparaginase (Ammon et al., 1985).
Rotation and Translation Function Calculations-The rotation function (R) is a measure of the overlap of two Patterson functions (Pl, P2), following rotation of one (P2) of them (Rossman and Blow, 1962). We have utilized the self-rotation function to investigate the noncrystallographic (molecular) symmetry of the Er. chrysanthemi asparaginase tetramer, and the cross-rotation function to deduce the relative orientations of the Er. chrysanthemi asparaginase and A. glutaminasificans glutaminase-asparaginase tetramers. Additionally, translation function calculations were used to determine a probable location for the Er. chrysanthemi asparaginase tetramer in the unit cell. A three-dimensional, self-rotation function for Er. chrysanthemi asparaginase was calculated with a Crowther fast rotation program (Crowther, 1972) with all Eh2 -1 coefficients (5,706 terms) in the 10-4.7-A range and a radius of integration of 30 A. Since the average value of Et,' is 1, this corresponds to the use of an origin-removed Patterson function. The Ehz -1 R map was superior to other maps calculated with different sets of Eh2 coefficients. The R map was clear and contained only three maxima, roughly one-third the height of the origin, self-superposition peak ( Table 1). Conversion of the fast rotation function a,&y peak positions to the K,$,+ polar coordinate system showed that the peaks corresponded to rotation dyads; the angles between the dyads were 90 f 0.3". These data are consistent with three mutually perpendicular 2-fold rotation axes and indicate that the Er. chrysanthemi asparaginase tetramer has 222 point group symmetry in the crystal.
The Crowther self-rotation function results were confirmed with the Rossmann program (Rossmann and Blow, 1962). This program performs calculations directly in polar coordinates ( K , $, d), and a search for %fold symmetry in the selfrotation function is accomplished by fixing K at 180", and scanning the $ and 4 coordinates. The calculations used the 1000 largest Eh2 coefficients in the 10-4.7-A resolution range with a radius of integration of 30 A. Each point in the map was weighted by sin($), which is proportional to the area of the R map represented by the ($,4) point. The asymmetric unit of the R map contained the three peaks shown in Table  2 which, following appropriate $,4 symmetry and coordinate system transformations can be seen to correspond closely to the Crowther self-rotation function maxima in Table 1.

55
60 65 70 75 Gln -Ala . Leu ~ Gln -Val . Ala ~ Ser -Glu -Set . Ile . Thr -Asp -Lys -Glu -Leu. Leu -Ser -Leu -Ala -Arg -Gln -Val -Asn -Asp - Leu     Fast rotation function angles. The alignment of the Er. chrysanthemi asparaginase reciprocal axes with the rotation function axes was b* along Z, and a* along Y. The application of the IY,,!?,~ Eulerian angles is termed the "ZYZ" convention to indicate the order in which the new axes are selected for successive rotation.

P h e -A l a -L y s -A l a -G l y -V a l -L y s -A l a -Ile -Ile -H~s -A l a -G l y -T h r -G l y -A s n -G l y -S e r -M e t -A l a -A s n -T y r -L e u -V a l -P r o -G~~-
Spherical polar angles. K is the extent of rotation about an axis whose direction is defined by $ and 4. The definition of $ and 4 with respect to the rotation function axes is that shown in Fig. 4 of Rossmann and Blow (1962).   The orientation and location of the model in the unit cell were refined with the program TRAREF (Huber and Schneider, 1985). The calculations used structure factors for the Er. chrysanthemi asparaginase tetramer in an orthogonal P1 cell The a,,!?,y matrix formalism used here is the transpose of the matrix given in Table l

Comparison of Asparaginase Sequences-The amino acid sequences of A. glutaminasificans glutaminase-asparaginase,
Er. chrysanthemi asparaginase (Minton et d, 1986), and E. coli asparaginase (Maita et al., 1974(Maita et al., , 1979(Maita et al., , 1980 have been aligned as illustrated in Fig. 2. Although the E. coli asparaginase sequence (321 residues) is 10 residues shorter than A. glutaminasifians glutaminase-asparaginase (331 residues), their alignment shows no insertions or deletions larger than 2 residues. The Er. chrysanthemi asparaginase sequence (327 residues) has deletions in three places, the largest being 5 residues long. There are 103 residues that are identical in all three sequences, and 34 residues are similar. A. glutaminasificans glutaminase-asparaginase and E. coli asparaginase are the most similar pair with 149 identical and 42 homologous amino acids. E. coli asparaginase and Er. chrysanthemi asparaginase have 143 identical and 43 similar residues; while A. glutaminasificans glutaminase-asparaginase and Er. chrysanthemi asparaginase are less alike, with 137 identical and 32 similar residues. This is surprising since Er. chrysanthemi asparaginase and E. coli asparaginase are classed as asparaginases with relatively low glutaminase activity, while A. glutaminasificans glutaminase-asparaginase is active as both an asparaginase and a glutaminase. There are several places where the three sequences have a series of identical residues. Near the amino terminus, for example, there are eight consecutive identical amino acids (residues 8-15 in A. glutaminasificans glutaminase-asparaginase) which are also found in the amino-terminal residues of Pseudomonas 7A glutaminase-asparaginase (Holcenberg et d, 1978), and Proteus uulgaris asparaginase? Residues 87-94 are also identical in all + ~/ 2 , 82 = 8, and = yr/2. The central matrix is required to convert from the Crowther program coordinate system used for lponoclinic unit cells (i.e. a*ll Y and b*IIZ) to the system (b*ll Y, c*I(Z) used in TRAREF.
Structures of Amidohydrohes a587 three sequences. Many identical residues are found in the region from 85 to 139 and at the carboxyl terminus from residue 302 to 331. Overall, the amino-terminal halves of the proteins are more similar in sequence than the carboxylterminal halves.
Prediction of Secondary Structure-The method of Chou and Fasman (1978) was used to predict the secondary structures for the three sequences (Fig. 3). The insertions and deletions shown for the sequence alignment mostly occur between or near the ends of elements of predicted secondary structure. The two exceptions are the single-residue deletion in E. coli asparaginase corresponding to A, glutaminasificans glutaminase-asparaginase 112 which is in the middle of a @strand; the single-residue deletion in Er. chrysanthemi asparaginase at A. glutaminasificans glutaminase-asparaginase 265, and the single-residue deletion in E. coli asparaginase corresponding to A. glutaminasificans glutaminase-asparaginase 313, which are predicted to be in a-helices. Five a-helices are predicted in very similar positions in all sequences. Eleven @-strands of at least 3 residues in length are predicted to be in similar positions. We place greater significance on predicted elements of secondary structure that are common to all sequences, since the enzymes fold into similar secondary  structures as shown by the correlation of the diffraction patterns of these enzymes.
Correlation with the Preliminary Crystal Structure of A. glutaminasificans Glutaminase-Asparaginase-A preliminary structure for the Acinetobacter enzyme has been determined from 2.9-A resolution x-ray data (Ammon et al., 1988). The present model for the A. glutaminasificans glutaminase-asparaginase structure has 331 residues in two chains, numbered as residues 1-273 and 300-357. This numbering is arbitrary at present, because the connectivity is not yet fully established. In particular, several of the surface connections between elements of secondary structure are still ambiguous. The A. glutaminasificans glutaminase-asparaginase subunit folds into two domains; residues 1-130 and 255-273 form the amino-terminal domain, while residues 131-254 and 300-357 form the carboxyl-terminal domain. The amino-terminal domain consists of a five-stranded @-sheet surrounded by five a-helices, a common feature in other protein structures (Richardson, 1981). The carboxyl-terminal domain was difficult to interpret, although it clearly contains three a-helices. The ahelical content is 27%, and at least 12% of the residues form &strands in the preliminary model. The predicted secondary structure (Fig. 2) shows 41% a-helix and 35% @-strand. The sizes and shapes of the amino acid side chains in the aminoterminal sequence of A. glutaminasificans glutaminase-asparaginase are consistent with those initially assigned on the basis of the x-ray structure. We expected that the common helices which were predicted for all three sequences would aid in establishing the connectivity for the A. glutaminasificans glutaminase-asparaginase structure. When the amino terminus of the A. glutaminasificans glutaminase-asparaginase sequence was aligned with the 1st residue in the A. glutaminasificans glutaminase-asparaginase structure, the positions of five of the predicted a-helices for A. glutaminasificans glutaminase-asparaginase were close to helices (a2, a4,a6, a7, and a8) in the preliminary tracing of the A. glutaminasificans glutaminase-asparaginase structure. Only the predicted common helices starting at A. glutaminasificans glutaminaseasparaginase 259 and at A. glutaminasificans glutaminaseasparaginase 304 appear in similar positions in the preliminary structure (Ammon et al., 1988), so that the secondary structure predictions were not very helpful. Even if there are a-helices in the five common positions predicted, a total of eight are seen in the preliminary structure. The connectivity has not yet been fully established and will require refitting the correct sequence into the preliminary A. glutaminasificans glutaminase-asparaginase structure.
The Actiue Site of Amidohydrolases-Possible active site residues in amidohydrolases have been identified from the reaction of E. coli asparaginase with an asparagine analog, DONV (Peterson et al., 1977;Handschumacher, 1977), and the reaction of A. glutaminasificans glutaminase-asparaginase and Pseudomonas 7A glutaminase-asparaginase with the glutamine analog DON (Holcenberg et al., 1978). The DONreactive Thr-12 of A. glutaminasificans glutaminase-asparaginase has been tentatively assigned to the a1-helix at the start of the amino-terminal domain; there is evidence that the DON binding site is part of the catalytic site for glutamine and asparagine (Holcenberg et al., 1978;Steckel et al., 1983). This residue is within the conserved 8-residue fragment found at the amino termini of A. glutaminasificans glutaminaseasparaginase, Pseudomonas 7A glutaminase-asparaginase, Er. chrysanthemi asparaginase, E. coli asparaginase, and the P. vulgaris asparaginase sequences, but the corresponding residue in E. coli asparaginase is not labeled by either the DONV or DON reagents (Handschumacher, 1977). A portion of the Structures of Amidohydrobes E. coli asparaginase active site has been tentatively identified as involving a Ser-Thr-Ser fragment at residues 117-119, which our secondary structure predictions indicate should be in a region of coil. The E. coli asparaginase-DONV labeling experiment suggests that the most probable site for DONV binding is Thr-118 or Ser-119. Ser-119 in E. coli asparaginase is aligned with Ala-126 in Er. chrysanthemi asparaginase, and with Ala-123 in A. glutaminusificans glutaminase-asparaginase, and Thr-118 aligns with threonine in both Er. chrysanthemi asparaginase and A. glutaminasificans glutaminase-asparaginase. Since all three sequences have threonine on the amino terminus side of Ser or Ala, this could be inferred as evidence that it is actually Thr-118 and not Ser-119 that undergoes reaction with DONV in the E. coli asparaginase labeling experiment. While these differences could preclude the involvement of the 121-123 region of A. glutaminusificans glutaminase-asparaginase in an asparaginase-like catalytic site, they do not, of course, address the question of why the E. coli enzyme does not show hydrolytic activity from the 8residue conserved region at the NH2 terminus. The answer to these questions and others must await the completion of the x-ray crystal structures of A. glutaminasificans glutaminaseasparaginase and Er. chrysanthemi asparaginase which are in progress in our laboratories.