Amplification of an adenylosuccinate synthetase gene in alanosine-resistant murine T-lymphoma cells. Molecular cloning of a cDNA encoding the "non-muscle" isozyme.

Adenylosuccinate synthetase (EC 6.3.4.4) catalyzes the initial step in the conversion of IMP to AMP. Two isoforms of this enzyme have been observed in vertebrates. A muscle isozyme is highly abundant in cardiac and skeletal muscle tissue and is thought to play a role in muscle energy metabolism. The non-muscle isozyme, which is present at low levels in most tissues, likely functions in de novo AMP biosynthesis. The analysis of the non-muscle isozyme has been hampered by its low abundance and instability during purification. In this study a genetic selection scheme was used to generate a murine T-lymphoma cell line which was at least 100-fold enriched for the non-muscle isozyme, as a result of amplification of the non-muscle synthetase gene. This cell line made possible the purification of the non-muscle isozyme, and the subsequent isolation of isozyme-specific peptides. Based on peptide sequence information a degenerate oligonucleotide probe was designed and used to screen a mouse kidney cDNA library. A 1.5-kilobase cDNA encoding the non-muscle isozyme was cloned and found to contain an open reading frame of 1368 base pairs encoding 456 amino acids. Gene transfer experiments showed that the cDNA encoded a 50-kDa protein, the size expected for mammalian synthetases, that correlated with the presence of high levels of synthetase activity. The deduced amino acid sequence of the mouse non-muscle synthetase is approximately 75% identical to the previously reported mouse muscle synthetase. Southern blot analysis of mouse genomic DNA with the isozyme-specific cDNA probes revealed that the synthetase isozymes are encoded by separate genes. The non-muscle gene is expressed in most tissues but is virtually undetectable in striated muscle tissues. Three different transcripts (1.7, 2.8, and 3.4 kilobases) are detected for the non-muscle isozyme which show a similar tissue distribution. The availability of a cDNA for the non-muscle isozyme of adenylosuccinate synthetase will facilitate further comparative analyses with the previously cloned muscle isozyme.

Adenylosuccinate synthetase (EC 6.3.4.4) catalyzes the initial step in the conversion of IMP to AMP. Two isoforms of this enzyme have been observed in vertebrates. A muscle isozyme is highly abundant in cardiac and skeletal muscle tissue and is thought to play a role in muscle energy metabolism. The non-muscle isozyme, which is present at low levels in most tissues, likely functions in de novo AMP biosynthesis. The analysis of the non-muscle isozyme has been hampered by its low abundance and instability during purification. In this study a genetic selection scheme was used to generate a murine T-lymphoma cell line which was at least 100-fold enriched for the non-muscle isozyme, as a result of amplification of the non-muscle synthetase gene. This cell line made possible the purification of the non-muscle isozyme, and the subsequent isolation of isozyme-specific peptides. Based on peptide sequence information a degenerate oligonucleotide probe was designed and used to screen a mouse kidney cDNA library. A 1.6-kilobase cDNA encoding the non-muscle isozyme was cloned and found to contain an open reading frame of 1368 base pairs encoding 466 amino acids. Gene transfer experiments showed that the cDNA encoded a SO-kDa protein, the size expected for mammalian synthetases, that correlated with the presence of high levels of synthetase activity. The deduced amino acid sequence of the mouse non-muscle synthetase is -76% identical to the previously reported mouse muscle synthetase. Southern blot analysis of mouse genomic DNA with the isozyme-specific cDNAprobes revealed that the synthetase isozymes are encoded by separate genes. The non-muscle gene is expressed in most tissues but is virtually undetectable in striated muscle tissues. Three different transcripts (1.7, 2.8, and 3.4 kilobases) are detected for the nonmuscle isozyme which show a similar tissue distribution. The availability of a cDNA for the non-muscle isozyme of adenylosuccinate synthetase will facilitate further comparative analyses with the previously cloned muscle isozyme. * This work was supported by National Institutes of Health Grant GM42436, National Cancer Institute Grant CA14030, Robert A. Welch Foundation grants (2-1041 and 6-893, and a grant from the Muscular Dystrophy Association. Peptide and nucleic acid sequencing and synthesis, provided by core facilities at Baylor College of Medicine, were supported by Baylor Mental Retardation Research Center Grant 2 P30 HD24064. 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.

The nucleotide sequence(s) reported in this paper has been submitted to the GenBankTM/EMBL Data Bank with accession nurnbeds) L24554.
8 To whom correspondence should be addressed. Fax: 713-796-9438; e-mail: rkellems@bcm.tmc.edu.
Adenylosuccinate synthetase (AdSS)' catalyzes the first of two steps leading to the synthesis of AMP from IMP. Two isozymes of AdSS have been observed in all mammalian species examined (Matsuda et al., 1977;Stayton et al., 1983). One isozyme (termed the non-muscle isozyme in this study)2 is widely distributed among mammalian tissues and functions at the branch point of purine nucleotide metabolism in de novo synthesis of AMP. The other isozyme (termed the muscle isozyme in this studyI2 is highly abundant in striated muscle tissues and is part of the purine nucleotide cycle (Van Waarde, 1988;Lowenstein, 1990; Van den Berghe et al., 1992). This cycle, which involves AdSS as well as AMP deaminase and adenylosuccinate lyase, is active in cardiac and skeletal muscle where it is believed to play a role in muscle energy metabolism (Van Waarde, 1988;Lowenstein, 1990).
Lower levels of the muscle isozyme of AdSS are found in kidney, brain, and testes, but the role of this isozyme in these tissues is unclear a t this time.
Most previous biochemical studies (see Stayton et al. (1983) for review) have focused primarily on the muscle isozyme because its abundance in muscle allowed purification and characterization of this isozyme. To accurately identify muscle-specific features of this isozyme, however, it is necessary to have adequate quantities of the non-muscle isozyme for structural and functional comparison. Unfortunately, the analysis of the ubiquitous non-muscle isozyme has been hampered by low abundance and instability, making protein purification very difficult. The AdSS isozymes differ in their functional and physical properties (see Stayton et al. (1983) for review). In general, the muscle isozyme has a lower affinity for IMP and is less sensitive to AMP inhibition than its non-muscle counterpart. Also, there appears to be a difference in how divalent cations, especially Ca2+ and M e , affect the isozymes (Cooper et al., 1982;Cooper, 1985). Furthermore, the muscle isozyme is functional as a dimer while the non-muscle enzyme appears to function as a monomer (Stayton et al., 1983). Additionally, the muscle enzyme has been shown to interact with purified myofibrils, a property which strengthens its proposed role in the muscle-specific purine nucleotide cycle (Manfredi et al., 1989). Further comparison of muscle and non-muscle isozymes of AdSS will be aided by a convenient source of the non-muscle isozyme.
To better understand the functional significance of the AdSS isozymes in mammalian tissues, it also will be useful to analyze their pattern of expression. For this purpose it will be necessary to have available cDNAs and/or antibodies specific for each isozyme. Isozyme-specific antibodies would also enable the cellular and subcellular localization of each isozyme to be determined. Furthermore, the availability of isozyme specific cDNAs would enable wild type or mutant forms of each isozyme to be produced in a prokaryotic or eukaryotic expression system for subsequent structural and functional analysis.
We have previously reported the cloning of a mouse cDNA encoding the muscle isozyme of AdSS (Guicherit et al., 1991). Here we present the isolation and initial analysis of a mouse cDNA encoding the non-muscle isozyme. Because of the low abundance of this isozyme in mammalian tissues and available cell lines we developed a genetic selection scheme to isolate a cell line that is enriched for the non-muscle synthetase. Enrichment appears to be the result of gene amplification. The availability of this cell line facilitated the purification of the non-muscle enzyme and the subsequent isolation of isozymespecific peptides. Degenerate oligonucleotide probes based on these peptide sequences were used to screen a mouse cDNA library. Data presented here show that the cloned cDNA encodes a protein with extensive sequence similarity to the previously reported muscle isozyme. The present study shows for the first time that the muscle and non-muscle isozymes ofAdSS are encoded by different genes, which have contrasting patterns of expression. Northern analysis showed multiple transcripts for the non-muscle isozyme which appear to be ubiquitously expressed.
EXPERIMENTAL PROCEDURES Cell Culture and Drug Selection Scheme-The YAC-1 cell line was obtained from the American Type Culture Collection (ATCC TIB 160). Cells were grown in RPMI 1640 media (Life Technologies, Inc.) supplemented with 10% iron-supplemented calf serum (Seru-Max 4, Sigma), 50 pdml gentamicin (Life Technologies Inc.), and 10 m~ HEPES (Sigma). Cells were split into fresh media every 3-4 days. The synthetase overproducer YAC-A16 was derived from YAC-1 by a stepwise selection in increasing concentrations of the drug alanosine (National Cancer Institute), beginning with 0.5 p~ up to 16 p~. At each step the drug was increased by 2-fold, and selection was maintained for at least three culture splits. The alanosine-resistant YAC-A16 cells were maintained in the presence of 16 p~ alanosine.
Purification of the Synthetase from YAC-A16 Cells and Isolation of Synthetase-specific Peptides-Approximately 3 liters of YAC-A16 cells grown in the presence of 16 p~ alanosine were harvested by centrifugation. The cell pellet was resuspended in lysis buffer (10 ml) containing 20 m~ TridHC1, pH 7.5, 0.5 m~ dithiotreitol, 0.5 m~ MgC12, 1 rn EDTA, 10% glycerol, 0.2% detergent (Nonidet P-40, Sigma), and 1 m~ phenylmethylsulfonyl fluoride (Sigma), and cells were lysed by sonication. The crude lysate was subjected to ultracentrifugation at 100,000 x g for 1 h. The synthetase activity was precipitated with (NH,),SO, (between 40 and 70% saturation) from the clarified supernatant (S-100 fraction). The ammonium sulfate pellet obtained after centrifugation was dissolved in the lysis buffer (2.5 ml) and desalted over a PD-10 (Pharmacia LKB Biotechnology Inc.) column. Synthetase activity was applied to an ion-exchange (DEAE-Bio-Gel (Bio-Rad)) column (2.5 x 10 cm) and eluted with a salt gradient of 0-200 m~ KC1 (120 ml) in lysis buffer, at 1.0 mumin. The synthetase-containing column fractions were identified by an activity assay (see below) and by SDS-polyacrylamide gel electrophoresis (Laemmli, 1970). The AdSS-containing fractions were combined and applied to a Procion Red column matrix (Sepharose 4B (Sigma) derivatized with Procion Red H-3B (Sigma) according to protocol (Lowe and Pearson, 1984). The Procion Red column (1 x 7 cm) was eluted with a 0-3 M KC1 gradient (40 ml) in lysis buffer, at 0.4 mumin. The AdSS-containing fractions were again pooled and finally concentrated over a Centricon (30,000 molecular weight cutoff, Amicon) spin-column.
The Promega Probe-DesignTM Peptide Separation System was used to isolate cyanogen bromide-cleaved peptides from the purified synthetase, according to the recommended protocol. Two synthetase-specific peptides were isolated (see "Results") and subjected to sequence analysis.
AdSS Activity Assay-The assay was similar to that described pre-viously (Guicherit et al., 1991). The reaction mixture contained 20 m~ HEPES, pH 7.0, 2.5 m~ aspartate, 5.0 m~ magnesium acetate, 0.2 m~ IMP, and 0.1 m~ GTP. The reaction was started by addition of IMP/GTP mixture and monitored at 280 nm, at 30 "C. The difference extinction coefficient was 11.7 x lo3 M -~ cm".
Isolation and Sequencing of cDNA Clones-A mouse kidney cDNA library was constructed3 in the vector pSPORTl (Life Technologies, Inc.). Approximately 2,000,000 colonies were screened with a [y-32PlATP-end-labeled degenerate oligonucleotide probe (see "Results") deduced from one of the isolated synthetase-specific peptides. Filters were hybridized in a mixture of 5 x SSC (0.15 M sodium chloride, 0.015 M sodium citrate), 20 m~ NaH2P04 (pH 7.5), 2% SDS, 5 x Denhardt's solution, 0.5% Blotto, and 100 pg/ml denatured salmon sperm DNA at 50 "C. Filters were washed twice at 50 "C for 30 min in (3 x SSC, 0.5 x Denhardt's solution, 2% SDS, 25 m~ NaH2P04 (pH 7.5)) and once at 50 "C for 30 min in (1.5 x SSC, 0.25 x Denhardt's solution, 1% SDS, 12.5 m~ NaH2P04 (pH 7.5)). Washed filters were exposed to film for 48 h at -70 "C with an intensifier screen. Positive colonies were further amplified and the cDNA clones were isolated and sequenced according to the dideoxy procedure (Sanger et al., 1977).
Expression of a cDNA Encoding the Non-muscle AdSS IsozymeFor expression of the AdSS cDNA from the kidney library the eukaryotic expression vector pSG5 (Stratagene) which contains a SV40 origin/ promoter, a p-globin intron, and a SV40 poly(A) sequence was used. Standard cloning techniques were applied (Sambrook et al., 1989). An EcoRI-BamHI fragment containing a full-length cDNA was isolated from a pSPORTl (Life Technologies, Inc.) backbone and cloned into pSG5 predigested with EcoRUBglII. The resulting construct (pSGn-mAdSS; see "Results," Fig. 6) was propagated in the Escherichia coli strain HB101, then purified over a CsCl gradient. For expression the construct was transfected into COS (African green monkey kidney) cells by a calcium phosphate-based procedure (Chen and Okayama 1987). Cells were harvested after 48 h, and cell lysates were analyzed for AdSS activity (see above) and by SDS-PAGE. Furthermore, cells were cotransfected with a p-galactosidase expression vector (pCMv-p) (Mac-Gregor and Caskey, 1989) to check for transfection efficiency.
Northern Analysis of Cell Lines and Tissues-Isolation, fractionation, and transfer of total cellular RNA from the cell lines and tissues was as described previously (Guicherit et al., 1991). Probes were randomly labeled with [d2P1dCTP according to the supplier (Boehringer Mannheim). Hybridization was at 60 "C in 4 x SSC, 0.5% SDS, 5 x Denhardt's solution, 0.5% Blotto, and 100 pg/ml denatured salmon sperm DNA, for 1 6 2 4 h. Subsequently, membranes were washed at 60 "C, twice for 30 min in 2 x SSC, 0.1% SDS, once for 30 min in 0.5 x SSC, 0.1% SDS, and twice for 30 min in 0.1 x SSC, 0.1% SDS. After washing, the membranes were exposed to film for several days at -70 "C with intensifier screens.
Southern Analysis of Genomic DNA-High molecular weight genomic DNA was isolated from the murine YAC-1 cell line. Cells were lysed in 25 m~ EDTA, pH 8, 50 m~ TridHC1, pH 8, 0.5% SDS, and 400 pg/ml Proteinase K (Boehringer Mannheim) at 55 "C for at least 3 h (to overnight). Subsequently samples were extracted very mildly with phenol, phenol/chloroform (l:l), and finally chloroform. Genomic DNA was precipitated with isopropanol, washed with 70% ethanol, and resuspended in 10 m~ TridHC1, 1 m~ EDTA, pH 8. After digestion with several restriction enzymes, the DNA was fractionated on a 0.8% agarose (Life Technologies, Inc.) gel and transferred to nylon membranes (Zeta-probe; Bio-Rad) in the presence of 10 x SSC, for 20 h. Hybridizations, washes, and autoradiography were as described for the Northern analysis (see above).

Selective Overproduction ofAdSS in Manosine-resistant Mu-
rine T-lymphoma Cells-The purification of the non-muscle isozyme of AdSS has been hampered by its low abundance in tissues and cell lines. To circumvent this problem and to facilitate the purification of the non-muscle isozyme we isolated a cultured cell line that produces large quantities of this enzyme. This was accomplished by selecting a murine T-lymphoma cell line, YAC-1, for resistance to increasing concentrations of alanosine, an aspartic acid analog (Fig. 1). Alanosine is metabolized to alanosyl-AICAR by SAICAR synthetase (Fig. l), an enzyme of de novo purine nucleotide biosynthesis ("yagi and Cooney, 1984). Alanosyl-AICAR functions as an IMP analog S. Chong and M. Hughes, manuscript in preparation. tase reaction is before the branch point (IMP) for both AMP and GMP biosynthesis. The AJSS reaction is the first of two steps involved in AMP biosynthesis. The aspartate-analog alanosine is a substrate for the SAICAR synthetase which converts it to alanosyl-AICAR, a compound that competes for IMP in the AdSS reaction. and competitively inhibits AdSS (Tyagi and Cooney, 1984;Stayton et al., 1983). We used a stepwise selection procedure to obtain a population of cells, YAC-A16, that grew in 16 J I M alanosine. As shown by the killing curves in Fig. 2 the drugresistant cells were approximately 1,000-fold more resistant to alanosine than the parental cells. Furthermore, the YAC-A16 cells were characterized by the presence of a very high level of AdSS activity (-500 nmol/min/mg) as compared to the original YAC-1 population (not measurable, with lower limit of detection <5 nmol/min/mg). SDS-PAGE analysis of lysates from YAC-1 and YAC-A16 cells showed that the alanosine resistant cells were specifically enriched in a protein of approximately 50 kDa (Fig. 3A, arrow 1, the size expected for AdSS (Stayton et al., 1983). As shown in Fig. 3A the 50-kDa protein appears to be the only protein enriched in the alanosine resistant cell line (YAC-A16). The chromatographic profile (see below) and kinetic properties of the synthetase activity in the drug-resistant cells were characteristic of those expected for the non-muscle isozyme. Furthermore, the 50-kDa protein overproduced in YAC-A16 cells showed no significant cross reactivity with antibody to the muscle synthetase, suggesting that the drug-resistant cells did not produce detectable levels of the muscle i~ozyme.~ Additionally, a cDNA probe of the mouse muscle isozyme did not detect elevated message levels in the AdSS overpr~ducer.~ These results indicate that YAC-A16 cells are considerably enriched for the non-muscle isozyme of AdSS and that these cells could be a convenient source from which to purify this enzyme. Purification and Partial Sequence Analysis of AdSS from YAC-A16 Cells-The synthetase was purified from the alanosine resistant cells by a combination of high speed centrifugation, ammonium sulfate precipitation, and column chromatography as described under "Experimental Procedures." The resulting product, as shown in Fig. 3B (arrow), was >95% pure. The protein was further purified by gel electrophoresis and subjected to chemical cleavage with cyanogen bromide to gen-

: M S I S E S S P A A T S L P N G D C G R P R A R S
158 : ~~A~~~M~c~~c .~~x~G A c G~~w~A

: T C C A T G A m A G C r r m T W A~C A G C M C X A % A G A C A C A 4 5 1 :
S M I Q L F *** ( 4 5 6 )

5 0 8 : C A l T ? C~% A T C T G C M A~A A G M T m A C A T ' X A A A~A G~
( 1 5 5 9 ) ..AAA..A erate peptide fragments. Fragments were isolated by gel electrophoresis and two such fragments of approximately 9-10 kDa were partially sequenced. The sequence of the first 30 amino acids of each peptide was determined (Fig. 4, underlined sequences) and compared with the previously published sequence of the mouse muscle synthetase (Guicherit et aZ., 1991). The deduced sequence of each peptide showed extensive identity to the mouse muscle isozyme of AdSS (Fig. 4 versus Fig. 51, indicating that the isolated peptides were associated with AdSS. Differences between the amino acid sequences presumably reflect isozyme-specific differences between the muscle and nonmuscle isozymes.

Molecular Cloning and Expression of a cDNA Encoding the Non-muscle Isozyme of AdSS-A degenerate oligonucleotide
probe was designed based on amino acid sequence from one of the peptide fragments of the synthetase: 5'-TT(G/A)TC(C/T)T-

GCTC(CPT)GT(G/A)GG(G/A)AA~(C/A)CCAAT(G/A)CC(C/
A)AC-3'. The oligonucleotide mixture was end-labeled and used as a hybridization probe to screen a mouse kidney cDNA library. The oligonucleotide probe hybridized to 10 colonies from a total of 2,000,000 colonies screened. One of these positive cDNA clones was further amplified and completely sequenced in both directions. Sequence analysis revealed an open reading frame of 1368 base pairs (Fig. 4) capable of encoding a poly-peptide of 456 amino acids with a calculated molecular mass of approximately 50 kDa. The most 5' start codon (ATG, nucleotide 83) was preceded by a stop codon (TAA, nucleotide 56) in the 5' non-translated region, a feature often observed for mammalian cDNAs, including the mouse muscle AdSS (Guicherit et al., 1991). At the 3' end of the open reading frame a stop codon is found at nucleotide 1451, followed by a polyadenylation signal at the appropriate position preceding the poly(A) tail.
The deduced amino acid sequence of the kidney cDNA is >70% identical (Fig. 5) to the amino acid sequence of the mouse muscle AdSS (Guicherit et al., 19911, confirming that the kidney cDNA encodes an isozyme of AdSS. As shown in Fig. 5, the sequence similarity extends throughout the two proteins, except for approximately 30 amino acids at the N t e r m i n~s .~ Furthermore, gene transfer studies indicate that the kidney cDNA clone is capable of encoding the synthesis of a BO-kDa protein (Fig. 6, arrow that correlated with the presence of high levels of AdSS activity (Fig. 6). Finally, the amino acid sequences derived from the synthetase-specific peptides isolated from YAC-A16 cells are identical to sequences encoded by the kidney cDNA (Fig. 4, underlined sequences). From the evidence The muscle cDNA sequence has been corrected in the GenBankT EMBL Data Bank. Cloning of Adenylosuccinate Synthetase I I I I I I O I I I I I I I I1 I I I I I I I I I I I  nmAdSS CRCQGGNNAG HTVWDSVEY DFHLLPSGII NPNVTAFIGN   I I I I   I I   I   I  I I   I   I I I I I I I I I G-2, (3-3, and G-4)  presented in Figs. 4-6 we conclude that the kidney cDNA (as in pSGnmAdSS, Fig. 6) contains the protein coding region for a non-muscle isozyme of murine AdSS.  I I I I I I I I I I I I I I I I I I I   G-1   GVVIHLFGLF EEAEKN--EK 118  I I I I I I I I I I I I I I I I  GVVIHLFGLF EEAEKNVQKG

Muscle and Non-muscle Isozymes of AdSS Are Encoded by Separate Genes That Are Differentially Regulated in Mouse
Tissues-Nucleotide sequence comparison of cDNAs encoding the mouse muscle (Guicherit et al., 1991) and non-muscle (Fig.  4) isozymes ofAdSS revealed a significant number of nucleotide mismatches throughout the protein coding region and no significant sequence similarity in the 5'and 3"non-coding regions. These findings suggest that although the two isozymes are clearly related by sequence similarity, they are not likely to be the products of the same gene. To test directly the two gene hypothesis, we probed duplicate genomic Southern blots with either the muscle or non-muscle cDNA probes. As shown in Fig.   7, cDNA probes specific for each isozyme hybridized to very distinct patterns of restriction fragments which clearly differed for each isozyme, providing strong evidence that the murine isozymes of AdSS are encoded by separate genes.
We have previously shown that the mouse muscle isozyme of AdSS is highly abundant in skeletal and cardiac muscle (Guicherit et al., 1991). Northern analysis was used to assess the pattern and level of expression of the non-muscle isozyme in a variety of murine tissues. The results in Fig. 8 indicate that three transcripts (1.7, 2.8, and 3.4 kb) were detected with the non-muscle cDNA. These transcripts were present in all tissues tested, with relatively higher levels in the brain, thymus, kidney and uterus, and lower levels in the liver. The non-muscle synthetase mRNAs were barely detectable in cardiac and skeletal muscle. Thus, the non-muscle synthetase gene shows a distinctly different pattern of expression than that of the muscle-specific AdSS gene (Guicherit et al., 1991).
Amplification of an AdSS Gene in the Alanosine-resistant Cells-To determine the genetic basis for overproduction of AdSS in alanosine-resistant cells, total RNA and DNA was isolated from YAC-1 and YAC-A16 cells and analyzed by blot transfer hybridization using the full-length non-muscle cDNA as a probe. The results of Northern analysis (Fig. 9) indicate that the cDNA probe hybridized to three transcripts of 1.7.2.8, and 3.4 kb that are very abundant in the YAC-A16 cells. The same three transcripts were present in the parental YAC-1 cells (and in the mouse tissues, Fig. 8 The smallest transcript appears to correspond in size to the isolated cDNA clone in pSGnmAdSS (Fig. 6). The relationship between the multiple transcripts is currently under investigation.
Southern blotting analysis of genomic DNA, using the nonmuscle cDNA as a probe, was performed to assess the role of gene amplification in the increased production of AdSS in alanosine-resistant cells. The results of a dilution series compar-ingYAC-A16 DNA with DNAfromYAC-1 cells (Fig. 10) indicate a n approximately 100-fold amplification of a n AdSS gene in the alanosine-resistant cells. From these results we conclude that the overproduction of AdSS in alanosine-resistant cells is the result of a corresponding amplification of the non-muscle AdSS gene.

DISCUSSION
Gene amplification is one of the most successful strategies for achieving high levels of protein production in mammalian cells (Kellems, 1993). In general this is accomplished by selecting cells for the ability to grow in increasing concentrations of a cytotoxic compound, usually an inhibitor or antimetabolite, which is specific for the enzyme of interest. Our initial attempts to select cells with amplified copies of AdSS genes involved the use of hadacidin, a n aspartic acid analog inhibitor of the enzyme (Stayton et d , 1983). Stepwise selection for resistance to Cloning of YAC-A16 -175 nmol/midmg). A more effective selection scheme involved the use of alanosine, also an aspartic acid analog, but not itself an inhibitor ofAdSS. Alanosine is converted by an enzyme of de nouo purine nucleotide biosynthesis to alanosyl-AICAR (Fig. l ) , a cytotoxic metabolite that binds AdSS with high affinity ("yagi and Cooney, 1984;Stayton et al., 1983). Most cells are killed by as little as 0.1 alanosine, making this compound a highly selective agent. A stepwise selection for resistance to increasing concentrations of drug up to only 16 1.1~ alanosine resulted in cells which overproduced AdSS at least 100-fold (specific activity in S-100 cell lysate: -500 nmol/midmg). The increased amounts of synthetase produced by alanosine-resistant cells is fully accounted for by a corresponding amplification of a synthetase gene. The alanosine resistant mouse lymphoma cells we isolated made possible the purification and amino acid sequence analysis of the non-muscle isozyme of murine AdSS.
The existence of mutiple AdSS isoforms has been known for many years (Stayton et al., 1983). However, the genetic basis for the multiple isoforms has not been determined previously. We addressed this issue by isolating cDNAs encoding muscle (Guicherit et al., 1991) and non-muscle (this work) isoforms of the murine enzyme and directly comparing the nucleotide sequence of the two cDNAs. The results revealed nucleotide mismatches throughout the open reading frame, with a n overall identity of -65%.
No significant sequence identity was observed in the 5'-and 3'-untranslated regions. These findings indicate that the mouse AdSS isozymes are not likely to be the products of the same gene, a conclusion that is also supported by Southern blot analysis. In the latter case, duplicate Southern blots of mouse genomic DNA were probed with the fulllength isozyme-specific cDNAs and clearly showed different Adenylosuccinate Synthetase hybridization patterns for each cDNA, indicating that the two mouse proteins are encoded by separate genes. We propose to designate the gene encoding the muscle enzyme, AdSS1, and the gene encoding the non-muscle enzyme, AdSS2.
Northern analysis revealed that the murine AdSS genes have a very distinct and almost mutually exclusive pattern of expression. The muscle-specific AdSSl gene encodes a 1.8-kb transcript found predominantly in skeletal and cardiac muscle (Guicherit et al., 1991). The AdSS2 gene, encoding the nonmuscle isozyme, is active at lower levels in a wide range of tissues. There appear to be three different transcripts (1.7,2.8, and 3.4 kb) specific for the non-muscle synthetase (AdSS2) gene which show the same relative abundance in all tissues and cell lines tested, with the largest of the transcripts being the most abundant. The structural relationship among the three AdSS2 transcripts is unknown at this time. AdSS2 gene expression is virtually undetectable in striated muscle. This may indicate the importance of keeping the two isozymes separated because they could potentially interfere with each others role in metabolism, as a consequence of their regulatory differences. On the other hand it has been reported that muscle has a significant capacity for de nouo AMP biosynthesis (Sheehan et al., 19771, a function normally assigned to the non-muscle isozyme. Additionally, the purine nucleotide cycle which apparently is important for proper muscle physiology has also been shown to operate in non-muscle tissues such as brain, kidney, and liver (Bogusky et al., 1976;Moss and McGivan 1975;Schultz and Lowenstein 1976). In order to better understand the tissue specific role of each isozyme it will be important to investigate the cellular and subcellular localizations of the isozymes in the relevant tissues. With isozyme-specific cDNAs and antibodies i t will be possible to more precisely determine the tissue-specific expression and cellular localization of each isozyme during development and in the adult mouse. This will hopefully lead to a better understanding of the relevance of each isozyme of AdSS in the physiology of the different mammalian tissues. Powell and co-workers (1992) recently identified a human liver AdSS cDNA clone which, like the mouse non-muscle AdSS cDNA, detected multiple transcripts. Even though the human clone was isolated based on cross-hybridization with sequences encoding the murine muscle AdSS, they were unable to assign the human liver synthetase as either a muscle or non-muscle isoform, because liver is believed to contain low levels of each. When we compared the amino acid sequences of the two murine cDNAs with that encoded by the human liver cDNA we found that the deduced amino acid sequence of the human synthetase shares over 90% identity with the murine non-muscle synthetase and less than 75% identity with the murine muscle synthetase. Based on these findings i t is likely that the human liver cDNA clone described by Powell et al. (1992) encodes the non-muscle isozyme of human AdSS. Lai et al. (1991) have made use of somatic cell hybrids between human lymphocytes and AdSS-deficient hamster cells to map a human AdSS gene. They show that a region on human chromosome 1 (lcen-lql2) cosegregates with synthetase activity, suggesting that this region contains an AdSS gene (Lai et al., 1991). Based on our findings and the assumption that the muscle isozyme is not expressed in human lymphocytes it is most likely that the human non-muscle AdSS gene maps to the region of chromosome 1 identified by the studies of Lai et al. (1991).
Multiple genes have also been identified for AMP deaminase, which encode muscle and non-muscle isoforms (Mahnke-Zizelman and Sabina, 1992;Morisaki et al., 1990;Sabina et al., 1990). The muscle isoform ofAMP deaminase, like AdSS, is also a component of the purine nucleotide cycle (Sabina et al., 1989). Several organ-specific and stage-specific transcripts have been identified for AMP deaminase. In adult rat tissue a non-muscle

Amplification and
Cloning of Adenylosuccinate Synthetase the amino acid sequence fiom mammalian AdSS isozymes with consensus sequences of GTP-binding motifs (i.e. G-regions) The consensus motifs are from Dever et al. (1987), and Bourne and Sanders (1991). The distance (in bases) between the motifs is indicated for each isozyme. The number preceding each sequence indicates the position of the first residue. transcript (3.4 kb) appears to be widely expressed but not in muscle tissue, while two alternatively spliced muscle transcripts (each 2.5 kb) are present at relatively high levels in muscle tissue (cardiac and skeletal muscle) and at much lower levels in non-muscle tissues. This is a pattern of expression which is very similar to that of the AdSS transcripts, suggesting that the expression of both enzymes may be regulated by common tissue-specific and possibly developmentally-regulated signals. Additionally, it has been shown that both AdSS and the deaminase can interact with myofibril units (Sabina et al., 1989;Manfredi et al., 1989) which suggests that the enzymes of the purine nucleotide cycle might be colocalized along muscle fibers. Thus, to maintain a precise stoichiometry of the purine nucleotide cycle in muscle one would expect a precisely coordinated regulation of expression of the relevant enzymes. It is interesting to note that the muscle and non-muscle AMP deaminase genes, AMPDl and AMPD2, have been shown to be located on the same chromosome in rat, mouse, and man (Sabina et al., 1990). We have no evidence for linkage of the AdSSl and AdSS2 genes at this time.
The mouse non-muscle isozyme of AdSS, like all previously identified synthetases, contains typical guanine nucleotidebinding motifs (G-regions; 39GDEGKGK, 260DFGTYPF'VT, 292EVYG, 361TKLD), common to all GTP-binding proteins (Bourne and Sanders, 1991;Dever et al., 1987): the N-terminal phosphate-binding loop GXXXXGK (G-1 region), the M?+binding sequence D-(X),-T (G-2 region), the central phosphatebinding sequence (D/EW(G/A) (G-3 region), and the C-terminal guanine-recognition sequence (NPTIQIKXD (G-4 region). The G-1 and G-4 regions are completely conserved in the mouse non-muscle synthetase (Fig. 5, Table I), as in all the other published synthetases, consistent with their proposed functional roles. Furthermore, site-directed mutagenesis studies on the G-1 region of the E. coli synthetase (Liu et al., 1992) have shown that the glycine residues are not only relevant for substrate (GTP) binding but also affect catalysis, confirming the functional relevance of this region. The putative G-2 region is also completely conserved among the mammalian synthetases (Table I) but shows some differences when comparing the mammalian synthetases with the non-mammalian synthetases from D. discoideum (Wiesmuller et al., 1991) and E. coli (Wolfe and Smith, 1988). These differences though do not include the relevant consensus residues, aspartate and threonine. The G-3 region appears to be somewhat variable among the synthetases (Table I), except for possibly the aspartate/glutamate and glycine residues which are usually well conserved among GTPbinding proteins (Bourne and Sanders, 1991). Interestingly the only difference, in consensus residues, between muscle and non-muscle synthetases appears to reside in the G-3 region where the aspartate in the muscle isozyme is replaced by a glutamate in the non-muscle isozyme (Table I). Another common feature among the mammalian synthetases is the 25-30 amino acid N-terminal extension which is absent in the Dictyostelium discoideum (Wiesmuller et al., 1991) and E. coli (Wolfe and Smith, 1988) synthetases. Because of the lack of data concerning the functional relevance of this N-terminal sequence, and because of no obvious sequence conservation, it is not clear what if any functional role this N terminus may fulfill. Furthermore, it remains to be determined which of the above mentioned sequence motifs underlie the differences in kinetic as well as physical properties which have been observed between the muscle and non-muscle isozymes of AdSS. The availability of cDNAs encoding each isozyme of AdSS will allow a more elaborate analysis of the proteins. Toward this goal we plan to overexpress and purify recombinant proteins for each enzyme. Large quantities of purified isozymes will be used for the analysis of funqtional and physical properties for each protein as well as generating isozyme-specific antibodies.
The selection scheme we have described here may have practical use in efforts to achieve high levels of protein production in mammalian cells. In other studies6 we have shown that AdSS minigenes function as dominant selectable and amplifiable genetic markers that confer resistance to alanosine following stable gene transfer into mammalian cells. Thus, AdSS expression vectors may be used to cotransfer and coamplify other genes of interest and in this way obtain mammalian cell lines producing large quantities of desired proteins. The alanosine selection scheme described here represents a significant addition to the list of amplifiable systems (Kellems, 1993) available to the mammalian genetic engineer and will be useful alone or in combination with other amplifiable systems to genetically engineer mammalian cells to produce high levels of proteins of academic interest and/or commercial value.
Acknowledgments-We are grateful to Bill Fanslow who initiated the selection of the cells. We thank Dr. Surjit Datta for his contribution to Fig. 2. We also thank Dr. Richard Cook and his core facility for peptide sequence analysis. Alanosine was kindly supplied to us by the Drug Synthesis and Chemistry Branch, Developmental Therapeutics Program, Division of Cancer Treatment, National Cancer Institute. The kidney cDNA library was a generous giR from Sam Chong in the laboratory of Dr. Mark Hughes. We are also thankful for the technical support of Donna Muzny and Dr. Richard Gibbs at the sequence core facility at Baylor College of Medicine for assisting us in generating the cDNA sequence. We thank Dr. John Winston and Shera Kash for their critical reading of the manuscript and their helpful suggestions.