Intracellular Localization of Human Cytidine Deaminase

The cytidine deaminases belong to the family of multisubunit enzymes that catalyze the hydrolytic deamination of their substrate to a corresponding uracil product. They play a major role in pyrimidine nucleoside and nucleotide salvage. The intracellular distribution of cytidine deaminase and related enzymes has previously been considered to be cytosolic. Here we show that human cytidine deaminase (HCDA) is present in the nucleus. A highly specific, affinity purified polyclonal antibody against HCDA was used to analyze the intracellular localization of native HCDA in a variety of mammalian cells by in situ immunochemistry. Native HCDA was found to be present in the nucleus as well as the cytoplasm in several cell types. Indirect immunofluorescence microscopy indicated a predominantly nuclear localization of FLAG-tagged HCDA overexpressed in these cells. We have identified an amino-terminal bipartite nuclear localization signal that is both necessary and sufficient to direct HCDA and a non-nuclear reporter protein to the nucleus. We also show HCDA binding to the nuclear import receptor, importin α. Similar putative bipartite nuclear localization sequences are found in other cytidine/deoxycytidylate deaminases. The results presented here suggest that the pyrimidine nucleotide salvage pathway may operate in the nucleus. This localization may have implications in the regulation of nucleoside and nucleotide metabolism and nucleic acid biosynthesis.

The cytidine deaminases belong to the family of multisubunit enzymes that catalyze the hydrolytic deamination of their substrate to a corresponding uracil product. They play a major role in pyrimidine nucleoside and nucleotide salvage. The intracellular distribution of cytidine deaminase and related enzymes has previously been considered to be cytosolic. Here we show that human cytidine deaminase (HCDA) is present in the nucleus. A highly specific, affinity purified polyclonal antibody against HCDA was used to analyze the intracellular localization of native HCDA in a variety of mammalian cells by in situ immunochemistry. Native HCDA was found to be present in the nucleus as well as the cytoplasm in several cell types. Indirect immunofluorescence microscopy indicated a predominantly nuclear localization of FLAG-tagged HCDA overexpressed in these cells. We have identified an amino-terminal bipartite nuclear localization signal that is both necessary and sufficient to direct HCDA and a non-nuclear reporter protein to the nucleus. We also show HCDA binding to the nuclear import receptor, importin ␣. Similar putative bipartite nuclear localization sequences are found in other cytidine/deoxycytidylate deaminases. The results presented here suggest that the pyrimidine nucleotide salvage pathway may operate in the nucleus. This localization may have implications in the regulation of nucleoside and nucleotide metabolism and nucleic acid biosynthesis.
Pyrimidine nucleotides are synthesized either de novo or by salvage pathways from preformed pyrimidine compounds. The pyrimidine salvage pathways serve two different functions. One is to scavenge pyrimidine compounds for nucleotide synthesis. The other is degradative and leads to the generation of compounds that serve as carbon and nitrogen sources. Cytosine, cytidine, and deoxycytidylate deaminases belong to the family of enzymes that deaminate mononucleotide substrates and are involved in maintenance of the pyrimidine nucleotide pool in the cell (1). They catalyze the irreversible hydrolytic deamination of cytosine, cytidine and deoxycytidine and several of their therapeutically useful analogues to the corre-sponding uridine derivative. Based on protein sequence alignment of this family of enzymes and the crystal structure of Escherichia coli cytidine deaminase (2) a conserved active site signature was identified ((C/H)XE----PCXXC) (2,3). The cysteine/histidine and two cysteine residues are involved in zinc binding at the active site, a requirement for catalytic activity, and the glutamine residue is involved in proton shuttling during catalysis (2). The cytidine deaminase family of enzymes tend to be multisubunit complexes. The E. coli enzyme is a homodimer with two active sites formed by contributions from each monomer (2). Human cytidine deaminase (HCDA) 1 forms a homotetramer of 16.2-kDa subunits, each with one active site (4,5). This is similar to the 58-kDa Bacillus subtilis enzyme, reported to be a homotetramer consisting of four identical 14-kDa subunits (6).
The cytidine deaminases are widely distributed in eukaryotic tissues. They are under allosteric and transcriptional control. Induction of transcription has been demonstrated in response to differentiation, vitamin D 3 , and ␥ interferon treatment (7)(8)(9). A role for deoxycytidylate deaminase (dC-MPD) in DNA replication has been suggested by a variety of observations in that activity is high in rapidly dividing tissues, particularly in the S phase of mitosis (10 -13), and dCMP deaminase is a major growth contributor of dUMP for thymidylate synthase (14,15). HCDA is a potent growth inhibitor of granulocyte macrophage colony-forming cells. A likely explanation for this is that the enzyme depletes both cytidine and deoxycytidine pools required for DNA synthesis (16,17). It has also been suggested that growth suppression might be by a receptor-mediated process triggered by cytidine deaminase (17). Cytidine deaminase plays an important role in the metabolism of a number of antitumoral and antiviral cytosine nucleoside analogues, leading to their pharmacological inactivation (18,19). The deamination of cytosine arabinoside, one of the most useful chemotherapeutic agents used in the treatment of cancer, results in a significant loss of antineoplastic activity (18,20).
Although the enzymes are generally considered to be cytoplasmic in intracellular localization, in the sequence of HCDA, we identified two clusters of basic amino acids near the amino terminus with similarity to the classical bipartite nuclear localization signal. Nuclear localization signals (NLSs) are short regions within some proteins that direct import into the nu-* 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 (21). In general, they contain a high content of the basic amino acids arginine and lysine and may contain residues, such as proline, that disrupt helical domains (22). NLSs, unlike other signal sequences, can be located anywhere in the primary sequence of the protein. The best studied NLS is that found within the SV40 large T-antigen, PKKKRKV (23), and defines the first class of NLS known as monopartite NLSs. A second class of NLS, found to be the most common, is the bipartite NLS. This signal consists of two runs of basic amino acids separated by a spacer region (21). The classic example of this type of NLS is found within the Xenopus protein nucleoplasmin (24). Nuclear proteins that carry a basic NLS (mono-or bipartite) are recognized by the heterodimeric importin receptor, composed of importin ␣ and importin ␤ subunits (25). The crystal structure of yeast importin ␣ was recently determined, and two sites that can accommodate the NLS peptides within a helical surface groove were identified (26).
The identification of a putative bipartite NLS in HCDA led us to consider whether this enzyme might be nuclear in its localization. We have gone on to show this, and further, that HCDA does contain a functional nuclear localization signal. Other members of this family of enzymes contain similar sequences. These studies suggest that the nuclear localization of HCDA may be important in providing substrate for RNA biosynthesis, particularly in times of high metabolic activity. An analogous role and nuclear localization can be envisaged in dCMP deaminase and other members of this enzyme family that have an NLS.

EXPERIMENTAL PROCEDURES
Isolation of HCDA cDNA and Genomic Clones-A human heart cDNA library (1 ϫ 10 6 recombinants) and a human genomic library (1.5 ϫ 10 6 recombinants) were screened using a 32 P-labeled oligo probe derived from the HCDA cDNA sequence (27) following standard procedure. Positive clones were plaque purified, and phage DNA was isolated as described previously (28,29). Automated DNA sequencing was carried out using the Perkin-Elmer d-rhodamine cycle sequencing termination kit in accordance with the manufacturer's instructions.
Plasmid Construction-Wild-type and amino-terminal deletion 3ЈFLAG-tagged HCDA constructs (wild-type HCDA and N-10 HCDA) for intracellular localization were amplified by polymerase chain reaction (PCR) with Pfu polymerase (Stratagene) from HCDA cDNA using primers 5HindHCD or 5HindHCD-10 and 3BamFLAGHCD and were cloned into pCMV-blue (Pharmingen) at the HindIII/BamHI sites.
Plasmids glutathione S-transferase (GST) and SV40 GST, encoding GST and the SV40 NLS (PKKKRKV) fused to the amino terminus of GST were created by cloning PCR products amplified from the plasmid pGEX4T3 (Amersham Pharmacia Biotech) using primers 5HindGST or 5SV40GST and 3PstFLAG-GST into pCMV-blue (Pharmingen) at the HindIII/PstI sites.
pCMV-blue encoding amino acids 1-30 of HCDA encompassing the putative NLS of HCDA fused to the amino terminus of GST-FLAG was constructed by inserting the 99-base pair PCR product, amplified from HCDA cDNA with primers 5EcoRVHCDA and 3NotIHCDA, in-frame with GST into EcoRV/NotI sites immediately upstream of GST in the pCMV-blue GST construct thereby creating HCDA (1-30)-GST.
Sequences of all oligonucleotides used in this study are available on request.
Expression of HCDA in E. coli-The pET 14b vector (Novagen) was used for expression of proteins in E. coli. The FLAG-tagged HCDA insert was produced by PCR from the cDNA sequence using oligonucleotides 5NcoHCD and 3BamFLAGHCD using Pfu polymerase (Stratagene). The PCR product was digested with NcoI and BamHI, ligated into pET14b/NcoI/BamHI, and transformed into E. coli JM109 cells, and the DNA sequence of the insert was verified (ABI 377 automated sequencing). HCDA point mutations were created by PCR using mutagenic oligonucleotides (C8A and C14A) and using the splicing by overlap extension technique. Complementing mutagenic oligonucleotides (e.g. E67A and E67A(R)) were used to generate two overlapping PCR products, using the upstream primer 5NcoHCD and the downstream primer 3BamFLAGHCD. The overlapping products were isolated by agarose gel electrophoresis and spliced in a further round of PCR. The product was digested and directionally cloned as described above. For protein expression, plasmids were retransformed into E. coli BL21(DE3)pLysS cells. Cells were grown to an absorbance of 0.6 and induced with 0.4 mM isopropyl-1-thio-␤-D-galactopyranoside for 3 h at 30°C. FLAG-tagged HCDA was purified on a M2-FLAG affinity column (Eastman Kodak Co.) following the instructions of the manufacturer. Protein content was assayed using the Bradford reagent (Bio-Rad).
Preparation of HeLa Cell Extract-About 2 ϫ 10 8 HeLa cells were resuspended in 20 mM HEPES buffer, pH 8.0 (5ϫ packed cell volume), containing 10 mM KCl, 1.5 mM MgCl 2 , 1 mM dithiothreitol, and 1% Triton. The lysate was sonicated on ice for 2 min and centrifuged at 33, 000 rpm for 20 min. Crystalline ammonium sulfate was added to the supernatant to 70% while being stirred at 4°C. The precipitate was allowed to stand at 4°C for 30 min. The precipitate was centrifuged at 33,000 rpm for 30 min, redissolved in 2 ml of 20 mM HEPES buffer, pH 8.0, containing 1 mM EDTA and 10% glycerol, and dialyzed at 4°C against the same buffer. A 20-l sample was resolved by SDS-polyacrylamide gel electrophoresis and analyzed by Western blotting as described below.
Affinity Purification of HCDA Antiserum and Immunoblotting-10 mg of FLAG-tagged HCDA was purified on an M2-FLAG affinity column as described above and coupled to 1 ml of cyanogen bromideagarose (Amersham Pharmacia Biotech) according to the manufacturer's instructions. The HCDA-coupled cyanogen bromide-agarose column was mixed with 1 ml of HCDA rabbit antiserum at 4°C overnight with gentle agitation. The column was then washed with 15 ml of 20 mM HEPES buffer, pH 8.0, followed by 15 ml of 20 mM HEPES buffer, pH 8.0, containing 100 mM KCl. HCDA antibody was then eluted with three volumes of 3 ml of 0.1 M glycine, pH 3.5, dialyzed against phosphatebuffered saline (PBS), and concentrated with PEG 6000. The affinity purified antiserum was tested on Western blots containing HeLa cell extract and purified recombinant FLAG-tagged HCDA. For immunoblot analysis, proteins were separated on 14% SDS-polyacrylamide gels,  HCDA and Importin ␣ Interaction in the Yeast Two-hybrid System-Yeast two-hybrid vectors pJG4 -5, pSB202 (derived from pEG202 as described (30)), and pSH18.3-4 were a gift from Roger Brent and are described (31). Plasmids bearing the importin ␣, importin ␤, and trans- portin cDNAs were kindly provided by I. Mattaj. EcoRI and XhoI restriction sites were created by PCR using primers 5EcoRI-imp ␣ and 3XhoI-imp ␣, 5EcoRI-imp ␤ and 3XhoI-imp ␤, and 5EcoRI-trans and 3XhoI-trans plus importin ␣, importin ␤, and transportin, cloned into pJG4 -5. HCDA was PCR-amplified from HCDA cDNA with primers 5BamHCD and 3SalHCD and cloned in frame into pSB202 BamHI/ SalI. Liquid ␤-galactosidase assays using -nitrophenyl-␤-D-galactopyranoside as a substrate were carried out as described for the CLON-TECH Matchmaker two-hybrid system (PT265-1) by the manufacturer.
Cell Culture and Transfection-COS-7, CCL13, HepG2, and HeLa cells were obtained from the American Type Culture Collection and maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum at 37°C with a 5% CO 2 atmosphere.
Cells were seeded onto 13-mm round coverslips (for immunofluorescence) at about 9 ϫ 10 4 cells per coverslip in 24-well dishes, or onto 100-mm dishes (for cell labeling and immunoprecipitation) at about 4 ϫ 10 5 cells per dish, 18 h before transfection. Cells were transfected with 10 g of purified plasmid DNA per 100-mm dish using the calcium phosphate precipitation method. The cells were incubated with the precipitate overnight, followed by a 2-min incubation with PBS containing 10% (v/v) dimethylsulphoxide. 48 h after transfection, cells were fixed for immunofluorescence or radioactively labeled before harvesting and immunoprecipitation.
Immunofluorescence-Cells on coverslips were fixed for 15 min in 100% methanol or 3.7% paraformaldehyde in PBS. Immunostaining was performed by incubating fixed cells for 1 h at room temperature with 40 g anti-flag M2Ab (Anachem; 1:250 in PBS containing 5% bovine serum albumin) or anti-HCDA affinity purified antibody (1:50 in PBS containing 5% bovine serum albumin), followed by incubation for 1 h at room temperature with fluorescein isothiocyanate (FITC)-conjugated rabbit anti-mouse IgG or FITC-conjugated swine anti-rabbit IgG secondary antibody (DAKO; 1:50 in PBS containing 5% bovine serum albumin). Coverslips were incubated for about 5 s in 1 g/ml 4Ј,6diamidino-2-phenylindole (Sigma), rinsed in PBS and mounted on glass slides with Aquamount improved (BDH). Slides were visualized under a Leica DMRB fluorescence microscope.

Site-directed Mutagenesis of Active Site Residues of HCDA-
HCDA is a homotetramer, previously identified as containing one active site zinc atom per monomeric subunit (32,33). By amino acid sequence comparison of CDA from different species and the crystal structure of E. coli cytidine deaminase a conserved active site signature was identified ((C/H)XE----PCXXC) (2,3,34). The cysteine/histidine and two cysteine residues are involved in zinc coordination at the active site, which is a requirement for catalytic activity. The glutamine residue is involved in proton shuttling during catalysis, and the proline residue plays a role in the orientation of the amino group to be removed by hydration (2,35). The conserved amino acid residues of the active site signature in HCDA, Cys-65, Glu-67, Pro-98, Cys-99, and Cys-102 were each mutated to alanine. Cys-14 and Cys-59, which are not thought to be involved in CDA catalysis, were also each mutated to alanine. FLAG-tagged wild-type HCDA and point mutants, generated by PCR, were expressed in E. coli, and the expressed protein was purified on a FLAG affinity column. Proteins were assayed for cytidine deaminase activity by a spectrophotometric assay as described (27). Mutagenesis of the active site residues completely abolished CDA activity, whereas C14A and C59A had no effect on enzyme activity (Table I). Mutational analysis of the key conserved amino acid residues of the presumed active site has therefore supported their importance in the function of the HCDA enzyme, as has previously been demonstrated for E. coli cytidine deaminase and the editing cytidine deaminase, APOBEC-1 (36,37). Mutagenic studies are often prone to questions concerning protein stability and folding. Mutagenesis of the active site amino acid residues may have resulted in misfolding of the enzyme and hence loss of activity.
Intracellular Localization of Native HCDA-The HCDA antiserum was affinity purified and used on Western blots as well as immunofluorescence analysis. The antibody readily reacted with recombinant HCDA, with a molecular mass of 17.1 kDa, and with one single species of the expected HCDA molecular mass (ϳ16 kDa) from a crude extract of HeLa cells (Fig. 1). This demonstrated that the antibody was highly specific for HCDA.
To look at the intracellular localization of endogenous HCDA, we performed indirect immunofluorescence experiments using the affinity purified anti-HCDA antibody on a variety of cell lines: HeLa ( Fig. 2A), COS7 (Fig. 2B), CCL13, and HepG2 (data not shown). A similar staining pattern was observed in all of the cell lines analyzed. A strong punctate nuclear stain accompanied by a diffuse cytoplasmic stain (Fig.  2, A and B, row c). When the cells were incubated with neutral rabbit serum, there was either no signal or a very low background cytoplasmic stain (Fig. 2, A and B, row d).
Bipartite NLS of HCDA-Nuclear proteins accumulate in the cell nucleus because they contain nuclear targeting signals that allow selective entry through the nuclear pore complex (38). On analysis of the sequence of HCDA, we identified a putative bipartite NLS sharing homology with the classical bipartite NLS. Two clusters of basic amino acid residues are separated by an intervening spacer region (Fig. 3).
Intracellular Localization of Overexpressed HCDA-The role of the putative NLS in the intracellular distribution of HCDA was evaluated by engineering a FLAG tag to its carboxyl ter- minus to provide an epitope tag for detection of the expressed protein in transfected cells. All constructs in this study (wildtype HCDA, HCDA(N-10), GST, SV40NLS-GST, and HCDA (1-30)-GST) were confirmed by DNA sequencing (data not shown). Expression of HCDA wild-type and deletion constructs in COS7 cells was verified by [ 35 S]methionine labeling of cells and immunoprecipitation of expressed FLAG-tagged protein (Fig. 4).
Immunofluorescence microscopy showed that FLAG-tagged wild-type HCDA localized almost exclusively in the nucleus of transiently transfected HeLa cells (Fig. 2A, rows a and b), COS7 cells (Fig. 2B, rows a and b), CCL13 and HepG2 cells (data not shown). In some cells a diffuse cytoplasmic signal was also observed (data not shown). The molecular mass of FLAGtagged HCDA is approximately 18 kDa. This is below the reported molecular diffusion limit of the nuclear pore complex, 40 -60 kDa (39). However, HCDA does act as a homotetramer with a combined mass of 72 kDa, which is above the size exclusion limit for passive transit through the nuclear pore, suggesting that the distribution of HCDA is not due to passive diffusion.
The Amino-terminal Region of HCDA Contains the NLS-In defining a classical NLS, two questions must be answered. First, is the proposed NLS necessary for efficient nuclear targeting of the parent protein? Second, is this sequence sufficient to target a heterologous non-nuclear protein to the nucleus? Deletion of the amino-terminal first 10 amino acids of HCDA, which includes the first cluster of basic amino acid residues of the putative bipartite NLS, resulted in cytoplasmic localization of the mutant protein, with complete exclusion from the nucleus (Fig. 5), confirming that the distribution of HCDA is not due to passive diffusion. The results suggest that the first cluster of basic amino acids in the amino-terminal of HCDA is essential for the effective nuclear targeting of HCDA and that the amino-terminal region of HCDA contains the determinant for nuclear localization.
The Amino-terminal Domain of HCDA Confers Nuclear Localization on GST-We next investigated whether the proposed NLS is sufficient for the nuclear localization of HCDA. GST (26 kDa) is a reporter protein that has previously been used for this purpose (40). GST was localized in the cytoplasm when expressed in COS-7 cells, but it could be targeted to the nucleus as an SV40 NLS-GST fusion protein (Fig. 5, A and B).
To determine whether the amino-terminal 30 amino acids of HCDA contains a functional NLS that can confer nuclear import upon a heterologous cytoplasmic protein, it was fused to the amino terminus of GST and the intracellular localization of the chimeric protein examined in COS-7 cells by indirect immunofluorescence. The fusion protein was localized to the nucleus (Fig. 5, A and B).
Hence, the amino-terminal HCDA sequence behaves as a classical NLS. It is necessary for nuclear localization of the parent protein (HCDA) and sufficient to localize a non-nuclear heterologous protein (GST) to the nucleus.
Importin ␣ (Karyopherin ␣) Interacts with HCDA-The import of nuclear proteins containing a classical NLS involves binding to a heterodimeric cytoplasmic receptor complex (importin ␣-␤/karyopherin ␣-␤) that translocates to the nuclear pore (reviewed in Ref. 25). A 60-kDa subunit of this complex, importin ␣, recognizes and binds proteins containing an NLS (41,42). In order to test whether HCDA might enter the nucleus by the pathway of classical NLS-bearing proteins, we asked whether HCDA could bind to the importin nuclear transport machinery. This was addressed by testing the interaction between HCDA and importin ␣ in the yeast two-hybrid system. Importin ␣ was cloned into the vector pJG4 -5 and HCDA cloned into pSB202 (see under "Experimental Procedures"). Empty pJG4 -5 and pSB202 plasmids, respectively, were used as controls in the experiment (Fig. 6). Importin ␤ and transportin (import receptor of M9-like NLSs (43)) were also cloned into pJG4 -5, and interaction with HCDA was assessed (data not shown). An interaction between importin ␣ and HCDA was observed. There was no interaction between either importin ␤ or transportin and HCDA or with empty plasmid controls. This suggests that HCDA interacts with importin ␣ or an importin ␣-like protein and enters the nucleus by the importin ␣/␤ transport machinery or an analogous pathway, consistent with that of a classical nuclear localization signal.
A Bipartite NLS May Exist in Other Cytidine Deaminases-Sequence comparison of deaminases reveals amino acid sequence homologies between the enzymes and conservation of the HCDA bipartite NLS at the amino terminus of human deoxycytidylate deaminases (HdCMPDs) and in deoxycytidylate deaminases from yeast (YdCMPD) (Fig. 7). Therefore, these other deaminases may also be nuclear proteins that enter the nucleus by recognition of a bipartite NLS.
Structural Characterization of the HCDA Gene-To characterize the HCDA gene, we screened a phage genomic library with HCDA cDNA. Two positive clones were isolated. These were characterized by DNA sequence analysis. The 3Ј flanking region of the last exon and the 5Ј flanking region of the first exon encoding the HCDA mRNA and all exon boundaries were sequenced. The organization of the HCDA gene is presented in Fig. 8. The cDNA and genomic coding sequences were identical. The HCDA gene is contained within 28 kilobases. Intron-exon boundaries corresponding to four coding exons were identified (Table II). These exons vary in size from 58 base pair (exon 3) to over 450 base pair (exon 4). The sequences at the intron-exon junctions all followed the characteristic GT-AG rule. The coding sequence is interrupted by three introns at codon positions Gly-52, Ser-89, and Glu-108. The first protein-encoding exon is exon 1 and contains amino acid residues 1-52, including the NLS. The second exon encodes residues 53-89, the third has residues 90 -108, and the fourth and final coding exon has residues 109 -147. The catalytic site core domain with the zinc coordinating amino acid residues, proton shuttling glutamine, and the proline residue that encompass the active site is split The results have shown that the intron/exon organization of the HCDA, HdCMPD (44), and APOBEC-1 genes are distinctly different (Fig. 8). In both HdCMPD and APOBEC-1, the catalytic center is contained within exon 3, and the centers consist of 4 and 5 exons, respectively. In APOBEC-1, Arg-15 is the first amino acid of the putative bipartite NLS and the last amino acid of the third exon (second coding exon) of the gene (45). The fourth exon (third coding exon) of APOBEC-1 contains the remainder of the putative NLS. The putative NLS of HdCMPD is contained within the first coding exon of the gene, as for HCDA.
Comparison of the yeast dCMPD cDNA and genomic sequences from sequence data bases indicated that the yeast dCMPD gene does not contain introns. Yeast dCMPD does, however, appear to have a putative bipartite NLS (Fig. 7). DISCUSSION In the present study, we show for the first time that HCDA is present in the nucleus. We have also shown the presence of a classical amino-terminal NLS on HCDA and binding to the nuclear transport receptor, importin ␣. Similar NLS sequences are present on HdCDA, HdCPMD, and YdCMPD. The intracellular distribution of HCDA had not been previously defined, but it had been considered to be a cytosolic enzyme (17).
Many nuclear proteins contain nuclear localization signals, which often consist of a series of basic residues (21). On analysis of the HCDA amino acid sequence, we identified an aminoterminal sequence that resembles a classical bipartite NLS. Disruption of this sequence abolished nuclear accumulation of HCDA. The putative sequence is a functional NLS, which is sufficient to target a non-nuclear reporter protein to the nucleus. Therefore, the amino-terminal region of HCDA is neces-sary for nuclear localization of the enzyme and also sufficient to target a non-nuclear protein to the nucleus, meeting the criteria of a classical NLS.
We report an interaction between HCDA and the nuclear import receptor, importin ␣, in the yeast two-hybrid system for testing protein-protein interaction. This suggests that HCDA enters the nucleus by a pathway similar to that of the importin ␣␤-mediated pathway for nuclear import. Recent x-ray crystallographic studies of importin ␣ by Conti et al. (26) focused on the region involved in NLS recognition and binding. Analysis revealed two binding sites, and hence a possible mechanism for the binding of bipartite NLSs. The larger binding site can accommodate up to five lysine or arginine residues, whereas the smaller binding site recognizes specifically only two amino acid residues. The structure also allowed a spacer length of 10 residues between the two basic clusters of a bipartite NLS, with potential for a longer spacer sequence, but not a shorter one. The HCDA NLS consists of one cluster bearing a lysine and an arginine residue, a spacer region of 19 amino acids, and a second cluster of two lysine residues. The presence of a proline residue in an NLS indicates the presence of a turn in the polypeptide backbone. In the case of the nucleoplasmin NLS, the equivalent proline residue following the first cluster of basic amino acid residues could be mutated with no effect on nuclear localization (46). Although the presence of a proline residue is not always essential to NLS function, in HCDA the proline residue following the first cluster may play a role in compensating for its longer than average spacer region, adjusting the NLS into the required conformation so that the two basic clusters are able to fit the binding sites of the receptor groove. It will be interesting to establish whether this is the case.
We have also identified similar bipartite NLS sequences in  other cytidine/deoxycytidine deaminases. Interestingly, the long spacer region discussed above is also a feature of these putative bipartite NLSs. These enzymes may also be nuclear proteins that enter the nucleus by a bipartite NLS and a pathway similar to that of the importin ␣␤-mediated pathway for nuclear import. Deoxycytidine kinase (dCK), which catalyzes the rate-limiting step of the deoxynucleoside salvage pathway in mammalian cells, is located mainly in the nucleus of cells transfected with a dCK-green fluorescent protein fusion protein (47). However, in a recent study of the intracellular localization of dCK, native dCK was found to be located mainly in the cytoplasm in several cell types (48). In contrast, when dCK was overexpressed in the cells, it was mainly localized in the nucleus, as with previous studies by Johansson et al. (47). The nuclear localization was therefore attributed to the overexpression of the protein. It was suggested that although dCK has the ability to enter the nucleus, the endogenous protein is kept in the cytoplasm by a cytoplasmic retention mechanism. We have found both endogenous and overexpressed HCDA to be present in the nucleus of a number of different cell types, suggesting that the nuclear localization observed was not solely a consequence of overexpression. Although the data presented shows a predominantly nuclear staining pattern in cells in which HCDA is overexpressed, compared with a nuclear and diffuse cytoplasmic staining pattern of endogenous HCDA, some transfected cells also showed a signal in the cytoplasm (data not shown). It is possible that, as suggested for dCK (48), HCDA may have a regulated cytoplasmic retention mechanism. However, it has been reported that vectors bearing the CMV promoter show variable levels of activity between cells according to serum levels and stage in the cell cycle (49). Therefore, the difference in the staining pattern observed between transfected cells may be attributed to overexpression. Additional studies using a vector containing an inducible promoter may elucidate this.
The data presented in this study, together with the presence of putative NLSs in cytidine and deoxycytidine deaminases, suggest that the pyrimidine nucleotide salvage pathway may operate in the nucleus. dCMPD activity is high in rapidly dividing cells and in the S phase of mitosis (10 -13). dCMPD also provides dUMP for thymidylate synthase (14,15). By analogy, CDA may be important in providing substrate for RNA biosynthesis. It will be of interest to test whether the putative NLSs in other members of this gene family encode proteins that have functional NLSs. The intracellular localization of these enzymes and their proximity to the site of utilization of their products are likely to be of importance in the regulation of nucleoside and nucleotide metabolism and nucleic acid biosynthesis.