Cloning, characterization, and modeling of mouse and human guanylate kinases.

Guanylate kinase catalyzes the phosphorylation of either GMP to GDP or dGMP to dGDP and is an essential enzyme in nucleotide metabolism pathways. Despite its involvement in antiviral drug activation in humans and in mouse model systems and as a target for chemotherapy, the human and mouse primary structures have never been elucidated. Full-length cDNA clones encoding enzymatically active guanylate kinase were isolated from mouse B-cell lymphoma and human peripheral blood lymphocyte cDNA libraries. Multiple tissue Northern blots demonstrated an mRNA species of approximately 1 kilobase for both mice and humans in all tissue types examined. The mouse cDNA is predicted to encode a 198-amino acid protein with a molecular mass of 21,904 daltons. The human cDNA is predicted to encode a 197-amino acid protein with a molecular mass of 21,696 daltons. These proteins share 88% sequence identity with each other and 52-54% identity with the yeast guanylate kinase. Molecular modeling using the yeast diffraction coordinates indicates a high degree of conservation within the active site and maintenance of the overall structural integrity, despite a lack of similarity along the periphery of the enzyme.

Guanylate kinase catalyzes the phosphorylation of either GMP to GDP or dGMP to dGDP and is an essential enzyme in nucleotide metabolism pathways. Despite its involvement in antiviral drug activation in humans and in mouse model systems and as a target for chemotherapy, the human and mouse primary structures have never been elucidated. Full-length cDNA clones encoding enzymatically active guanylate kinase were isolated from mouse B-cell lymphoma and human peripheral blood lymphocyte cDNA libraries. Multiple tissue Northern blots demonstrated an mRNA species of approximately 1 kilobase for both mice and humans in all tissue types examined. The mouse cDNA is predicted to encode a 198-amino acid protein with a molecular mass of 21,904 daltons. The human cDNA is predicted to encode a 197amino acid protein with a molecular mass of 21,696 daltons. These proteins share 88% sequence identity with each other and 52-54% identity with the yeast guanylate kinase. Molecular modeling using the yeast diffraction coordinates indicates a high degree of conservation within the active site and maintenance of the overall structural integrity, despite a lack of similarity along the periphery of the enzyme.
Guanylate kinase (GMK, 1 ATP:GMP phosphotransferase; EC 2.7.4.8) catalyzes the reaction (d)GMP ϩ ATP 3 (d)GDP ϩ ADP where (d)GMP indicates GMP or dGMP. In addition to being a critical enzyme in the biosynthesis of GTP and dGTP, guanylate kinase functions in the recovery of cGMP (cGMP 3 GMP 3 GDP 3 GTP 3 cGMP) and is, therefore, thought to regulate the supply of guanine nucleotides to signal transduction pathway components (1,2).
As with other enzymes involved in nucleotide metabolism, guanylate kinase is a target for cancer chemotherapy and is inhibited by the potent antitumor drug, 6-thioguanine (3)(4)(5). Guanylate kinase activity is also required for the potentiation of antiviral drug activity in virus-infected cells. Activation of the anti-herpes guanosine nucleoside analogs, acyclovir and ganciclovir, after an initial phosphorylation step by the viral thymidine kinase, is carried out by guanylate kinase (6,7).
Although guanylate kinase is a key enzyme for cancer chem-otherapy and antiviral drug activation in humans, the gene encoding this important enzyme has not been described. In this study, we describe the isolation, expression, and functional analysis of both human and mouse guanylate kinases in an effort to aid improvements in antiviral and tumor chemotherapeutic approaches through molecular modeling and rational drug design.

EXPERIMENTAL PROCEDURES
Materials-Restriction enzymes, deoxynucleoside triphosphates, and random-primed DNA labeling kits were purchased from Boehringer Mannheim. Reagents for enzyme assays were purchased from Sigma.
[␣-32 P]dCTP (3000 Ci/mmol) and 14 C-labeled protein molecular weight markers were purchased from Amersham Corp. 35 S Express protein labeling mix was purchased from DuPont NEN. Biotinylated thrombin, pET23d vector, and strepavidin beads were obtained from Novagen (Madison, WI). Multiple tissue Northern blots were purchased from Clontech (Palo Alto, CA, catalog nos. 7760-1 and 7762-1). Bradford assay reagents, chromatography columns, electroporation cuvettes, and prestained protein molecular weight markers were purchased from Bio-Rad. Ni-NTA resin, Qiaprep, spin kits and plasmid purification reagents were purchased from Qiagen (Chatsworth, CA). All other reagents were purchased from Sigma unless otherwise indicated.
PCR Amplification and Cloning of Human Guanylate Kinase cDNA-The PCR primers 5Ј-ACTACTGGATCCATGGCGGGCCCCAG-GCCTGTG-3Ј (DMO437) and 5Ј-TACTACGGATCCTCAGGCGCCGG-TCCTTTGAGC-3Ј (DMO438) were synthesized by Genset (La Jolla, CA). Both primers contain BamHI restriction sites near their respective 5Ј ends. DMO437 also contains an NcoI site. reverse transcription-PCR was performed as described by Kawasaki (8). The resulting 600-bp amplification product was cloned into the BamHI site of pUC118, and the resulting plasmid was designated pUC118:hgmk.
PCR Amplification and Cloning of Mouse Guanylate Kinase cDNA-Approximately 500,000 phage from a random-primed cDNA library of the mouse cell line 70 Z/3 were plated and screened by the method of Sambrook et al. (9). Hybridization was carried out with a 32 P-labeled random-primed human guanylate kinase 600-bp BamHI fragment. The EcoRI fragments from nine positive clones were cloned into the EcoRI site of pUC118.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) U53514.
For the human multiple tissue Northern blot, the random-primed purified hgmk PCR product (600 bp) described above was used as a probe. For the mouse multiple tissue blot, the mouse gmk probe was generated by asymmetric PCR using 10 ng of EcoRI fragment from pUC118:mgmk as a template in the presence of 150 Ci of [␣-32 P]dCTP and 100 pmol of primer 5Ј-ACTACTACTGGATTCAGCAGGCCTTCAG-GCGTG-3Ј (DMO856) in a standard polymerase chain reaction for 40 cycles. Hybridization conditions were as described above. The filters were subsequently stripped and probed with ␤-actin.
An NcoI site was introduced at the mouse gmk initiator codon by PCR amplification using the following two primers: 5Ј-CGCGGACGC-CATGGCAGGACCTAG-3Ј (DMO855) and 5Ј-ACTACTACTGGAT-TCAGCAGGCCTTCAGGCGTG-3Ј (DMO856). PCR conditions were identical to the reverse transcription-PCR protocol listed above, except that the template was 2 ng of pUC118:mgmk DNA. The resulting fragment was restricted with NcoI and BamHI and directionally cloned into the NcoI and BamHI sites of pET23d. The entire insertion and fusion sites were sequenced, and the resulting plasmid was designated pET23d:mgmk.
Construction of pETHT Bacterial Expression Vector-pET23d was modified by PCR using overlapping oligonucleotides to create an aminoterminal tag encoding a stretch of six histidine residues, followed by a 2-amino acid spacer and a consensus sequence for a thrombin proteolytic cleavage site, followed by another 2-amino acid spacer. The primer 5Ј-ACTACTACTAGATCTCGATCCCGCGAA-3Ј (DMO604) was used to prime the sense strand starting at a unique BglII site 82 bases upstream of the initiating methionine, encoded by the NcoI site, at the beginning of the multiple cloning site in pET23d. Three overlapping "antisense" strand-specific primers, 5Ј-ATGATGATGATGATGGCTG-CTAGCCATAGTATATCTCCTTC-3Ј (DMO605), 5Ј-CGGCACCAGGC-CGCTGCTATGATGATGATGATGATGGCT-3Ј (DMO606), and 5Ј-AGTAGTATCCATGGAGCTGCCGCG CGGCACCAGGCCGCTGCT-3Ј (DMO607), were used in consecutive PCR amplifications. The first PCR was performed on 2 ng of pET23d DNA in 0.1 ml with 0.2 mM of each deoxynucleotide triphosphate, 0.5 M each of DMO604 and DMO605, and 0.5 units Taq DNA polymerase. The 110-bp PCR product was isolated and used as the template in a second PCR under the same conditions, except 0.01 M DMO606 and 1 M DMO607 were used instead of DMO605. The resulting 185-p product was digested with NcoI and BglII, gel isolated, and ligated to BglII/NcoI-digested pET23d. The resulting plasmid was sequenced and designated pETHT.
Both human and mouse guanylate kinase cDNAs were cloned as NcoI/BamHI fragments isolated from pET23d:hgmk or pET23d:mgmk into the NcoI/BamHI sites of pETHT. Sequence analysis confirmed 5Ј and 3Ј ligation sites.
In Vitro Transcription and Translation-Plasmid DNAs (pET23d: hgmk, pETHT:hgmk, pET23d:mgmk, and pETHT:mgmk) were subjected to in vitro transcription as described by Black et al. (13). The resulting transcripts and a control (no RNA) were then used to program a rabbit reticulocyte cell-free translation, as described by Black et al. (13). A small fraction of each translation carried out in the presence of [ 35 S]methionine/cysteine was heat denatured and subjected to electrophoresis on polyacrylamide containing SDS gels. After drying, the gel was exposed to film.
Overexpression of Guanylate Kinase in E. coli-pETHT:hgmk and pETHT:mgmk were electroporated into competent E. coli BL21(DE3) tk Ϫ cells. Induction of this strain with 0.4 mM isopropyl-1-thio-␤-Dgalactopyranoside for 3 h was performed by the method of Studier et al. (14) as described in the pET System Manual (Novagen).
Purification of Guanylate Kinases-Purification of MGMK was carried out according to Qiagen, except the cell pellets were resuspended in lysis buffer (20 mM Tris, pH 8, 0.6 M NaCl, 5 mM imidazole, 10 mM 2-mercaptoethanol, 10% glycerol, 0.2% Tween 20, and 1 mM phenylmethylsulfonyl fluoride) containing 100 g/ml lysozyme. Ten mM MgCl 2 , DNase I, and RNase A (10 g/ml each) were added, and the mixture incubated at room temperature for 20 min. The histidinetagged protein was mixed with Ni-NTA resin in batch, rocked for 1 h at 4°C, and applied to a small chromatography column. After passing 5 bed volumes of lysis buffer minus phenylmethylsulfonyl fluoride over the column twice, the column was subjected to two additional washes in 20 mM Tris, pH 8, 0.15 M NaCl, 5 mM imidazole, and 10% glycerol. Proteins were eluted with 20 mM Tris, pH 8, 0.15 M NaCl, 0.125 M imidazole, and 10% glycerol. Fractions containing visible amounts of protein on Coomassie-stained gels were pooled and dialyzed against 20 mM Tris, pH 8.0, and 150 mM NaCl at 4°C.
Cleavage of the amino-terminal tag was done using biotinylated thrombin (Novagen) at 1 unit/100 g of purified protein in elution buffer containing 2.5 mM CaCl 2 at room temperature for 20 h. Streptavidin beads (100 l) and 200 l of Ni-NTA resin were added to the digestion, and the mixture was rocked gently for 1 h at room temperature. The mixture was then spun for 5 min at 2000 rpm in a microcentrifuge at room temperature. The supernatant was aliquoted and stored at Ϫ70°C.
Protein concentrations were quantitated by the Bradford method using reagents supplied by Bio-Rad. Known concentrations of bovine serum albumin were used to generate a standard curve.
Antiserum Production-Purified mouse guanylate kinase (2 mg) was supplied to R and R Rabbitry (Stanwood, WA) for the generation of rabbit polyclonal antiserum according to their standard protocol (15).
Cloning hgmk and mgmk cDNAs into a Mammalian Expression Vector-The mammalian expression vector, pREP8, was purchased from Invitrogen (San Diego, CA). This vector was modified to remove the EBNA and oriP sequences by restriction with BstEII and XbaI and ligation of the blunt-ended vector. The resulting 7-kilobase plasmid was designated pREP8⌬7.
Both guanylate kinase cDNAs were cloned into the HindIII (blunt)/ BamHI sites of pREP8⌬7 as NcoI(blunt)/BamHI fragments. The vector constructs (pREP:hgmk and pREP:mgmk) were sequenced to confirm orientation and 5Ј fusion sites.
Approximately 1 ϫ 10 7 cells of each line were pelleted and resuspended in 50 mM Tris, pH 7.5. The cells were frozen and thawed, and 0.01% Nonidet P-40 (v/v) was added. A small amount (8000 cell eq) was removed and subjected to polyacrylamide gel electrophoresis, followed by immunoblot analysis using a 1:5000 dilution of either preimmune serum or anti-MGMK serum. These lysates were also used for determination of guanylate kinase activity.
Spectrophotometric Assay for Guanylate Kinase Activity-Enzyme activity was measured using a lactate dehydrogenase-pyruvate kinase coupled assay described by Agarwal et al. (16). The change in A 340 was measured over time with cell lysates as the source of GMK enzyme on a Hewlett-Packard model 8452A diode array spectrophotometer (Hewlett-Packard, Fullerton, CA).

Isolation of the Human Guanylate
Kinase cDNA-To ascertain whether any new guanylate kinase cDNAs had been isolated since the publication of the bovine sequence in 1993, the nucleotide and protein data bases were searched using bovine sequences. A BLAST search for homology to a sequence from a conserved region in the cDNA of bovine retina (nucleotides 303-333) revealed a nearly perfect match to a sequence with the accession number a11042. This is described as a human hematopoietic cell growth-potentiating factor (nucleotide sequence 7 from patent EPO274560). Indeed, an alignment of the bovine and putative human guanylate kinase sequences displayed 90.4% identity at the amino acid level. Primers (DMO437 and DMO438) were designed from this sequence to amplify the coding region of the putative human guanylate kinase (hgmk) by reverse transcriptase-coupled PCR of total cellular RNA isolated from human peripheral blood lymphocytes. A single band of 600 bp was amplified and cloned into the BamHI site of pUC118. Complete double-stranded DNA sequencing was done on two subclones, and both were found to be identical to the a11042 sequence. A protein of 197 amino acids with a molecular mass of 21,696 daltons is predicted from a translation of the open reading frame. This is in close agreement with the bovine gmk that encodes a 198-amino acid protein with a calculated molecular mass of 22,051 daltons (1).
Northern Blot of Mouse RNA Probed by hgmk cDNA-To extend the number of guanylate kinase sequences for alignment and molecular modeling studies, we sought to isolate the mouse guanylate kinase cDNA. To determine whether the hgmk cDNA could be used to isolate the mouse cDNA, the 600-bp hgmk cDNA probe was used to hybridize to a Northern blot containing total RNA from the SP2/0 mouse myeloma line. This probe hybridized strongly, showing a single band at approximately 1 kilobase (Fig. 1). This is similar in size to the yeast gene (920 bp) and the full-length bovine retina cDNA clone (1042 bp).
Cloning of Mouse gmk cDNA-A random-primed cDNA library (500,000 plaques) from a mouse B-cell lymphoma line (70Z/3) was screened with the 600-bp hgmk cDNA probe. Nine independent clones were identified and plaque purified. All clones were then digested with EcoRI, subcloned into pUC118, and completely sequenced. Of the nine independent clones, one was found to contain a partial open reading frame that had been reverse transcribed from the middle of the putative mgmk coding region. All remaining clones contained the entire mgmk open reading frame as well as 50 -100 bp of both 5Ј-and 3Ј-untranslated regions. One of the clones contained the entire 3Ј-untranslated region as well as a portion of the poly(A) tail. DNA sequence analysis of both strands demonstrated that the nucleotide sequence was identical in the overlapping regions of all nine clones. The nucleotide and deduced amino acid sequences are shown in Fig. 2. The mgmk encodes a 198-amino acid polypeptide with a calculated molecular mass of 21,904 daltons.
Sequence Analysis and Comparison-Translations of the cDNA open reading frames were aligned with each other as well as the peptide sequences of bovine, porcine, yeast, and E. coli guanylate kinases (Fig. 3). The percentage of identity and similarity were calculated using the GCG program Bestfit with a gap weight of 1.0 and are shown in Table I. Human, mouse, bovine, and porcine guanylate kinases share nearly 90% identity, with the highest similarity value of 99.5% between bovine and porcine enzymes. Comparisons between mammalian and yeast GMKs reveal much less conservation, with amino acid identities on the order of 52 to 54%. The yeast and E. coli guanylate kinases share 48.6% identity. The canonical ATP binding motif, GXGXXGK(S/T) (where X is any amino acid), can be found near the amino terminus (at positions 11-18 in the mouse and human sequences). Furthermore, a guanylate kinase signature sequence was identified when the human GMK was used to search the data base using the GCG MOTIFS program. The consensus pattern, TTRXXRXXEXXGXXYX(F/Y)(L/I/V/M), aligns perfectly with positions 39 -56 in the mouse and human sequences. The presence and location of these two motifs in the putative guanylate kinases as well as their high degree of homology with known guanylate kinases is strong evidence that these cDNAs encode functional guanylate kinases.
Molecular Modeling-A high resolution three-dimensional model based on the x-ray diffraction data of the yeast enzyme complexed with GMP has been proposed (17). All guanylate kinases have strictly conserved the residues shown by x-ray diffraction studies to interact with GMP (Ser 34 , Arg 38 , Arg 41 , Glu 44 , Tyr 50 , Glu 69 , Tyr 70 , and Asp 100 : yeast amino acid designations). In addition, the Ser 80 in yeast interacts with GMP but is replaced by threonine in all other guanylate kinases. These contact residues are denoted by asterisks in Fig. 3. We have used the yeast model to highlight differences and similarities between the yeast, human, and mouse amino acid sequences using the Look program from the Molecular Applications Group (Palo Alto, CA). A comparison of these three GMK species is shown in Fig. 4 and is based on the structural model of the yeast GMK (1GKY; Brookhaven protein data base; Ref. 17). All identical amino acids are displayed as purple, partial conservation or similarity is shown as light purple, and all nonconserved residues are highlighted in bright yellow. The substrate, GMP, is also shown and is buried in what Stehle and Schulz (17) describe as a "giant anion hole." Eight of the key residues described above are located in the interface in the domain to the left of GMP. Residue Asp 100 is positioned on the opposite side of GMP at the end of a loop region and is suggested to interact by hydrogen-bonding with N 3 and N 11 of guanine (17). The neighboring Asp 98 holds the magnesium ion and can form a salt bridge with Lys 14 . Lys 14 is a component of the ATP binding site (residues 8 -15 in yeast) that spans the turn designated by Gly 11 .
Overall, the major differences between these enzymes occur in regions that are on the periphery of the active site and form a "golden halo" in Fig. 4. Interestingly, these differences do not appear to alter the overall structural integrity of the enzyme. With only 52-54% identity between mammalian and the yeast guanylate kinases, this is particularly striking.
Multiple Tissue Northern Blots-The human or mouse gmk probes hybridized strongly to a single ϳ1-kilobase transcript in the respective Northern blot (human or mouse RNA). Guanylate kinase transcripts are expressed in all tissue types examined and is indicative of ubiquitous mRNA expression (Fig. 5, A  and B). This type of expression pattern is not unexpected for an essential "housekeeping" gene. Fig. 5, C and D, show the same Northern blots probed with ␤-actin and demonstrate the integrity of the RNA and the relative amounts loaded per lane. Heart and skeletal muscle contain two forms of ␤-actin, 2 and 1.8 kilobases (Fig. 5, C and D, lanes 1 and 6).
In Vitro Expression of Human and Mouse gmks-The coding regions of both hgmk and mgmk cDNAs were directionally subcloned into the bacterial expression vectors pET23d and pETHT as described under "Experimental Procedures." In vitro transcription and translation of these constructs were performed as described in Black et al. (13). Radiolabeled translation products were subjected to acrylamide gel electrophoresis in the presence of SDS, and the dried gel was exposed to x-ray film. The resulting autoradiograph shows an ϳ23-kDa band for both the mouse and human pET23d constructs and shows a slightly larger ϳ25-kDa band for the histidine-tagged mouse and human GMKs (Fig. 6), indicative of full length in vitro translation products. Enzyme assays of the mgmk cell-free translations gave very marginal activity over a high background, and assays of hgmk cell-free translations did not display any detectable activity above background levels.
Protein Overexpression and Purification-To facilitate protein purification, the pETHT vector that encodes an initiating methionine followed by six histidine residues, a thrombin cleavage site, and an NcoI restriction site for cloning the cDNAs in frame was used. Thus, the human and mouse guanylate kinase cDNAs are expressed with an amino-terminal histidine-thrombin (HT) tag. HT-tagged proteins were then purified on nickel resin columns, and the amino-terminal tag was removed by cleavage with biotinylated thrombin. The thrombin was removed with strepavidin beads, and the resulting supernatant was removed from the nickel resin, resulting in only cleaved, purified guanylate kinase. The thrombin cleavage leaves three amino acids at the amino terminus before the initiator methionine (Gly-Ser-Ser).
Relative to mouse gmk expression, the human cDNA was expressed at a very high level in several inductions. Unfortunately, the human GMK was mostly partitioned to an insoluble fraction. Mouse GMK did not share this property and was found predominately in the soluble fraction. Binding of MGMK to the Ni-NTA beads in batch required only 4 -5 mM imidazole for effective blocking of nonspecific binding. An imidazole concentration of 25-30 mM was necessary to block the HGMK lysate, with virtually no proteins eluting up to 75 mM imidazole. Both GMK proteins eluted in 125 mM imidazole. Prior to thrombin cleavage, the imidazole was removed by dialysis in 20 mM Tris, pH 8.0, and 150 mM NaCl. Initial purifications were estimated from Coomassie-stained polyacrylamide gels to be 50% pure for HGMK and 80% pure for MGMK. A gel of fractions from various steps during the mouse GMK purification is shown in Fig. 7A. Despite modifications to the lysis, binding, wash, and elution conditions, the presence of minor contaminating proteins persisted. However, after thrombin cleavage and Ni-NTA chromatography, the HT tag and contaminating proteins were removed from the resulting purified GMK. The thrombin-cleaved MGMK migrated to the expected position in polyacrylamide:SDS gels relative to known molecular weight markers (Fig. 7B). The final preparation of MGMK protein was greater than 99% pure. Attempts to purify HGMK to greater that 50% under native conditions were unsuccessful. However, FIG. 3. Alignment of mouse, human, bovine, porcine, yeast, and E. coli guanylate kinases. The GCG programs PILEUP and PRETTY were used to generate the alignment. The consensus sequence displays amino acids that are conserved in all six sequences. *, the position of the nine residues identified that interact with GMP in the yeast x-ray diffraction structure (17). The gray boxes denote the ATP binding site and GMP binding site consensus sequences. Standard single amino acid nomenclature is used.
Ͼ99% purity was obtained using the denaturing lysis and purification protocols described by Qiagen (data not shown).
Determination of Activities-Because the human GMK did not display activity in E. coli lysates or cell-free translations, we sought to express active gmk cDNAs in mammalian cells. As described under "Experimental Procedures," stable clones of BHK tk Ϫ (ts13) cells transfected with pREP8⌬7, pREP:hgmk, or pREP:mgmk were isolated. Guanylate kinase protein ex-pression was confirmed by immunoblot analysis of transfected cell lysates using antiserum directed against MGMK (Fig. 8). Both human and mouse guanylate kinases cross-react with the polyclonal antiserum, as might be anticipated from the similarity in deduced amino acid sequences. Human GMK migrates slightly slower than the mouse form. The polyclonal antiserum also reacted with a sample of purified HT-tagged MGMK that was cleaved by thrombin. Preimmune serum did not cross-react with any proteins from the lysates under the same blotting and developing conditions (Fig. 8).
Transfected cell lysates were assayed for guanylate kinase activity by the spectrophotometric assay described by Agarwal et al. (16). This method uses a coupled lactate dehydrogenasepyruvate kinase reaction that measures the ADP and GDP produced from ATP and GMP by GMP kinase. Lysates from cells transfected with gmk cDNAs displayed high activity relative to those from cells transfected with vector alone (Fig. 9).  4. Computer modeling of mouse and human guanylate kinases with bound GMP based on the yeast structure. Using the Look program (Molecular Applications Group), the identical residues (purple), similar residues (light purple), and nonconserved residues (yellow) between the mouse, human, and yeast guanylate kinase sequences are displayed on the molecular model for yeast GMK that was elucidated by Stehle and Schulz (17). Certain amino acids are shown to aid in identification of particular regions discussed in the text. The amino and carboxyl termini are denoted by N and C, respectively.

DISCUSSION
Guanylate kinase plays an essential role in generating the nucleotide precursors, GDP and dGDP, for RNA and DNA metabolism, respectively. Despite the importance of guanylate kinase in nucleotide metabolism, the only identified eukaryotic guanylate kinase nucleotide sequences are those from yeast (18) and bovine retina (1). The porcine amino acid sequence was determined by Zschocke et al. (5) by protein sequence analysis of guanylate kinase purified from pig brain. Although the E. coli gmk gene (spoR) has also been cloned and was found to share significant similarity with the amino acid sequence of yeast guanylate kinase, the E. coli enzyme differs both structurally and enzymatically from all other known guanylate kinases (19). Guanylate kinase activity has been observed in a variety of tissues and has been purified from many eukaryotic sources including bovine retina (1), hog brain (16), bakers' yeast (18,20), rat liver (16), and human erythrocytes (16). Although the human gmk (GUK1) was mapped to chromosome 1q32-43 over 15 years ago (21), no primary sequence has been identified.
In this report, we describe the isolation of the cDNA for both human and mouse guanylate kinases. Guanylate kinase mRNA is ubiquitously expressed, as would be anticipated from an essential housekeeping enzyme. Alignment studies indicate a strong homology with related guanylate kinases from mammalian sources and less conservation in amino acid identity with the yeast and E. coli enzymes. Modeling studies in which the yeast structure was used as a template for displaying the conserved, similar, and nonconserved residues between the human, mouse, and yeast GMKs revealed conservation within the "giant anion hole" active site and a loss of conservation in residues comprising the perimeter of the enzyme. Interestingly, although the identity between the human or mouse GMK and the yeast GMK was 52-54%, the overall structure is highly conserved.
Detectable enzyme activity was demonstrated with the mouse GMK when it was expressed in cell-free translations, crude bacterial lysates purified from E. coli, and in lysates from mammalian transfected cells. Preliminary kinetic determinations of mouse guanylate kinase purified from E. coli indicates a K m of ϳ25 M when GMP is used as the substrate (data not shown). This is similar to that reported for porcine GMK (K m ϭ 32 M; Ref. 22) and human erythrocyte GMK (K m ϭ 15-24 M; Ref. 7). Activity from the human GMK was only detected in mammalian cell lysates. Attempts to purify human GMK from E. coli resulted in only ϳ50% purification, and the enzyme was inactive. It is unclear what the bases of these differences between human and mouse GMK are, especially since they share 88% amino acid sequence identity.
In the last several years, sequence data from a number of novel genes have revealed that they encode a motif approxi-FIG. 6. Autoradiograph of rabbit reticulocyte lysate cell-free translation products of gmk transcripts subjected to gel electrophoresis. Transcripts were synthesized in vitro from pET23d and pETHT constructs and used to program a rabbit reticulocyte cell-free translation system. Translation products were subjected to denaturing polyacrylamide gel electrophoresis, followed by autoradiography.  9. Guanylate kinase assays of transfected cell lysates. Ten l of each BHK tk Ϫ (ts13)-transfected cell lysate (pREP8⌬7, pREP: hgmk, and pREP:mgmk) was assayed in the presence of 100 mM GMP for enzyme activity over time. The experiment was repeated with similar results. E, pREP8␦7; f, pREP:hgmk; q, pREP:mgmk. mately ϳ200 amino acids in length with a high level of similarity to the yeast guanylate kinase (77% similarity between dlg-A and yeast guanylate kinase; Ref. 2). These guanylate kinase domains have been identified in a number of related membrane-associated proteins, including the Drosophila disclarge tumor suppressor protein (dlg-A) (2); Z0 -1 and Z0 -2, two closely related proteins which are associated with tight junctions (23); p55, a major palmitoylated membrane protein from human erythrocytes (24); and two homologs, SAP90 or PSD-95 and SAP97, proteins restricted to rat synaptic vesicles (25,26,27). Although it has not been directly demonstrated that these guanylate kinase domains are functional, Mueller et al. (27) have indicated that SAP90 retains the ability to bind GMP, GDP, ATP, and ADP in vitro, despite the lack of a canonical ATP binding site. This suggests that this class of proteins does encode functional guanylate kinase enzymes, although why a guanylate kinase activity is associated with these proteins has not been established. One possibility might be that they serve to supply GDP at the cell membrane to regulate GTPase signal transduction proteins, such as Ras.
A more direct role for guanylate kinase activity in signal transduction comes from its function in guanine nucleotide metabolism. Guanylate kinase participates in the recovery of cGMP and is, therefore, thought to regulate the supply of guanine nucleotides to signal transduction pathways. Little is known about how extensive this role is, but such studies may be facilitated by the preliminary study reported here.
At least two other key areas of study will be aided by the elucidation of human and mouse guanylate kinase sequences. Guanylate kinase is required for certain antiviral drug activation pathways. Of particular note are the well studied antiherpetic drugs, acyclovir and ganciclovir, which must be initially phosphorylated by the herpes-encoded thymidine kinase. Once phosphorylated, these drugs require further phosphorylation to their triphosphate state to become activated and able to act as chain terminators during DNA synthesis. Both subsequent phosphorylation steps are carried out by cellular enzymes: guanylate kinase (monophosphate to diphosphate) and a number of nonspecific nucleoside diphosphokinases (diphosphate to triphosphate). More recently, this activation pathway has been exploited for ablative gene therapy of cancer, where the HSV-1 TK gene is introduced into tumor cells that are specifically killed after administration of ganciclovir. In a sim-ilar fashion, guanylate kinase is involved in activation of chemotherapeutic drugs such as 6-thioguanine (4, 6, 7). Development of novel antiviral drug or chemotherapeutic agents will be greatly aided by rational drug design based on molecular modeling techniques.