Nuclear localization signal of murine CMP-Neu5Ac synthetase includes residues required for both nuclear targeting and enzymatic activity.

5-N-Acetylneuraminic acid (Neu5Ac) is the major sialic acid derivative found in animal cells. As a component of cell surface glycoconjugates, Neu5Ac is pivotal to numerous cellular recognition and communication processes including host-parasite interactions. A prerequisite for the synthesis of sialylated glycoconjugates is the activation of Neu5Ac to cytidine-monophosphate N-acetylneuraminic acid (CMP-Neu5Ac). The reaction is catalyzed by CMP-Neu5Ac-synthetase (syn), which, for unknown reasons, resides in the nucleus. Sequence analysis of the cloned murine CMP-Neu5Ac synthetase identified three clusters of basic amino acids (BC1-BC3) that might function as nuclear localization signals (NLS). In the present study chimeric protein and mutagenesis strategies were used to show that BC1 and BC2 are active NLS sequences when attached to the green fluorescent protein (enhanced GFP), but only BC2 is necessary and sufficient to mediate the nuclear import of CMP-Neu5Ac synthetase. Site-directed mutations identified the residues K(198)RXR to be essential for nuclear transport and Arg(202) to be necessary to complete the transport process. Cytoplasmic forms of CMP-Neu5Ac synthetase generated by single site mutations in BC2 demonstrated that (i) enzyme activity is independent of nuclear localization, and (ii) Arg(199) and Arg(202) are involved in both nuclear transport and synthetase activity. Comparison of all known and predicted CMP-sialic acid synthetases reveals Arg(202) and Gln(203) as highly conserved in evolution and critically important for optimal synthetase activity but not for nuclear localization. Combined, the data demonstrate that nuclear transport and enzyme activity are independent functions that share some common amino acid requirements in CMP-Neu5Ac synthetase.

Sialic acids acids are a family of negatively charged, 9-carbon sugars that form terminal residues on cell surface glycoproteins and glycolipids and provide the bulk of the negative charge, which is characteristic for animal cell surfaces (for review, see Ref. 1). More than 40 sialic acid derivatives have been identified in biological systems. The most abundant form in higher vertebrates is 5-N-acetylneuraminic acid (Neu5Ac). 1 Sialic acids play important roles in cell-cell communication and recognition processes and mediate immune responses by binding to selectins and other members of the SIGLEC family (for review, see Refs. 2 and 3). Moreover, vital processes such as fertilization (for review, see Ref. 4), neural cell growth, differentiation and plasticity (for review, see Refs. 5 and 6; Ref. 7), biological aging (8), as well as development (6,9) and progression of malignancies (10 -12) are accompanied by alterations in the cellular sialylation pattern. In bacteria sialic acids are found as components of capsules and lipooligosaccharides and often are important virulence factors, mediating resistance to host defense mechanisms (reviewed in Refs. [13][14][15]. For example, Neisseria meningitidis serogroup B (NmB), the major cause of meningitis outbreaks in the western hemisphere, expresses a capsular polysaccaride consisting of ␣2,8-linked sialic acid residues, exclusively. The polysialic acid (polySia) of the NmB capsule is identical to host expressed polySia, which represents a specific posttranslational modification of the neural cell adhesion molecule (for review, see Ref. 16). This classical example of antigenic mimicry explains the low serum response caused by NmB in infected individuals (17).
A prerequisite for the incorporation of sialic acids into glycoconjugates is their activation as cytidine-monophosphate diester (CMP-Neu5Ac). This reaction is catalyzed by the CMP-N-acetylneuraminic acid synthetase (CMP-Neu5Ac-syn, EC 2.7.7.43; for review, see Ref. 18). The nucleotide sequences of six bacterial (19 -24) and two vertebrate CMP-sialic acid-syn (25,26) are available. The alignment of the deduced protein sequences revealed five highly conserved motifs indicating a common ancestor (25,26). Thus the CMP-sialic acid synthetases provide the first example for an evolutionary conservation from bacteria to vertebrates among the sialic acid metabolizing enzymes.
The first vertebrate CMP-Neu5Ac-syn cDNA was isolated by complementation cloning in the Chinese hamster ovary (CHO) mutant LEC29.Lec32, which due to the lec32 mutation does not contain an active CMP-Neu5Ac-syn. These cells cannot transfer the non-activated sugar onto glycoconjugates and therefore lack sialic acid and polySia on the cell surface. Expression cloning with polySia detection as an assay of complementation led to the isolation of a mouse CMP-Neu5Ac-syn cDNA (25). Analysis of the intracellular localization of the recombinant murine CMP-Neu5Ac synthetase confirmed earlier studies that showed the majority of CMP-Neu5Ac-syn activity to be localized in the nuclear compartment (for review, see Ref. 18). This is in clear contrast to all other nucleotide sugar synthetases, which are restricted to the cytoplasm. The unusual localization of the CMP-Neu5Ac-syn has been a focus of scientific interest for more than 30 years but still remains an enigma (for review, see Refs. 18 and 25).
While small molecules and proteins are able to enter the nucleus by passive diffusion, the import of larger proteins through the nuclear pore complex is mediated by active transport. The nuclear import machinery is complex and various pathways are used simultaneously to achieve the import of many different substrates (for review, see Refs. [27][28][29]. Most karyophilic proteins contain one or more nuclear localization signals (NLS) that are recognized by soluble factors of the nuclear transport system (for review, see Ref. 30). NLS motifs are non-cleaved signal sequences within a protein that do not fit a tight consensus motif. The so called classical or canonical NLS typically consists of one or more clusters of basic amino acids often preceded by a proline residue (for review, see Refs. 27, 31, and 32). One of the best defined NLSs is located in the simian virus SV40 large T (SV40-T) antigen (P 126 KKKRKV; Ref. 33). In the murine CMP-Neu5Ac-syn three potential NLSs termed basic clusters (BCs) 1, 2, and 3 are present, and their involvement in nuclear transport has been proposed earlier (25).
In the present study, wild-type and mutant sequences of BC1, BC2, and BC3 were tested for their ability to transport eGFP chimeras and CMP-Neu5Ac-syn to the nuclear compartment. In addition, the enzymatic activity of the generated CMP-Neu5Ac-syn mutants was analyzed. The data provide clear evidence that (i) functional activity of CMP-Neu5Ac-syn is not dependent on nuclear localization, and (ii) some mutations that prevent nuclear localization also reduce or eliminate synthetase activity.
Plasmids for eukaryotic expression of C-terminally extended eGFP fusion proteins were generated by the use of the eukaryotic expression vector peGFPCI (CLONTECH). Sense and antisense oligonucleotides that encode BC1, BC2, and BC3 of murine CMP-Neu5Ac-syn or the SV40-T NLS were synthesized. Oligonucleotides were generated with overhanging sequences to allow for the directed cloning into BglII and EcoRI restriction sites as shown in Table I. Matching oligonucleotide pairs were annealed and ligated into the BglII and EcoRI sites of peGFPCI, resulting in 3Ј-extended eGFP cDNAs. The integrity of all plasmids was confirmed by sequencing.
Construction of Mutants-Site-directed mutations were introduced using the QuikChange TM site-directed mutagenesis kit (Stratagene) according to the supplier's instructions, using Pfu polymerase (Stratagene). Mutation primers were designed to either delete nucleotide triplets encoding selected amino acids or to replace natural triplets by GCN, which encodes alanine. Sequences of the mutation primers together with the plasmid names are given in Table II. The integrity of all N-terminally FLAG/Myc-tagged mutants was confirmed by automated sequencing both, before and after subcloning.
Expression of CMP-Neu5Ac-Syn-The functionality of wild-type and mutant murine CMP-Neu5Ac-syn was analyzed in complementation studies using CHO LEC29.Lec32 (36) and E. coli EV5 (37). For eukaryotic expression 1.8 ϫ 10 6 LEC29.Lec32 cells were grown overnight on 60-mm tissue culture dishes, rinsed twice with PBS (10 mM sodium phosphate, pH 7.4, 150 mM NaCl), and transiently transfected with 2 g of plasmid DNA and 12 l of LipofectAMINE (Invitrogen) in 2.5 ml of Opti-MEM (Invitrogen). After 7 h transfections were stopped by adding 5 ml of medium containing 10% fetal calf serum. After another 17 h the medium was exchanged, and cells were harvested 48 h later. After washing with PBS, transfected cells of one 60-mm-plate were subdivided into two aliquots, and both aliquots were incubated for 30 min at 37°C in the absence or presence of 100 ng of endoNE to remove polySia. Subsequently, the cells of both aliquots were lysed in 100 l of ice-cold lysis buffer (50 mM Tris-HCl, pH 8.0, 1 mM MnCl 2 , 1% Nonidet P-40, 200 units of aprotinin and 1 mM phenylmethylsulfonyl fluoride), sonicated (Branson Sonifier, 50% duty cycle, output control 5, 4°C, 18 ϫ 5 s), and centrifuged at 16,000 ϫ g. The protein concentration of the supernatant was determined using BCA protein assay reagent (Pierce). Lysates diluted to a final concentration of 2.5 mg/ml in lysis buffer were analyzed by SDS-PAGE (see below).
For prokaryotic expression E. coli EV5 were transformed with 10 ng of plasmid DNA. Bacteria were grown at 37°C to an A 600 0.4, and protein expression was induced with 0.2 mM isopropyl-␤-D-thiogalactoside for 2 h. Thereafter, bacteria were harvested and washed with PBS. An aliquot was resuspended in sonication buffer (PBS containing 200 units of aprotinin and 1 mM phenylmethylsulfonyl fluoride), and cells were sonicated as described previously. After removal of cell debris, the protein concentrations were determined using BCA protein assay reagent (Pierce). Two parallel samples (bacteria corresponding to 200 g of protein each) were incubated for 30 min at 37°C in the absence or presence of 100 ng of endoNE. The bacteria from these parallel samples were pelleted, resuspended in 50 l of PBS, mixed 1:1 with 2ϫ Laemmli sample buffer, sonicated as described, and further analyzed by SDS-PAGE as described below.
SDS-PAGE and Western Blot Analysis-SDS-PAGE was performed according to Laemmli (38). Samples were reduced with 2.5% (v/v) ␤-mercaptoethanol and heated at 65°C for 20 min. After electrophoretic separation proteins were blotted onto nitrocellulose membranes (Schleicher & Schuell), and Western blots were developed using 5 g/ml of primary antibody followed by anti-mouse alkaline phosphatase conjugate (Dianova). Nitro blue tetrazolium and 5-bromo-4-chloro-3-indoyl phosphate were used as substrates for alkaline phosphatase.
Transfection of NIH 3T3 Cells and Immunofluorescence-Twentyfour h before transfection, 3 ϫ 10 4 cells per well were seeded in 12-well plates containing glass coverslips. Transient transfections were performed using the Superfect transfection kit (Qiagen). In 20 l of Dulbecco's modified Eagle's medium, 0.4 g of DNA were mixed with 2 l of Superfect, incubated at room temperature for 10 min, mixed with 500 l complete medium, and then added to the washed cells. Cells were carefully washed with PBS 24 h after transfection, fixed in 4% paraformaldehyde for 20 min, and washed twice with PBS. For indirect immunofluorescence staining, transfected cells were permeabilized with 0.2% Triton X-100 for 10 min at room temperature, washed with PBS, and incubated with the anti-FLAG mAb M5 (3.5 g/ml in 20% horse serum in PBS) for 1 h at 37°C. After washing with PBS, cells were incubated with sheep anti-mouse IgG-Cy3 (Sigma; 1:300 in PBS containing 20% horse serum) for 1 h at 37°C. After three additional washes with PBS, nuclei were stained with Hoechst 33258 (Hoechst Pharmaceuticals) at a concentration of 500 ng/ml in PBS for 4 min at room temperature. Cells transfected with the peGFPCI constructs (see Table I) were used for direct immunofluorescence 24 h after transfection. Cells were fixed in 4% paraformaldehyde and stained with Hoechst 33258 as described previously. All coverslips were mounted in Moviol and analyzed under a Zeiss Axiophot fluorescence microscope (ϫ400).
Computational Analysis-PSORTII was performed for the prediction of protein localization sites in cells (psort.nibb.ac.jp; Refs. 39 and 40), and the method of Cokol (Ref. 41; cubic.bioc.columbia.edu/predictNLS) was applied to search for putative NLS.

Identification of Potential Nuclear Localization Signals-
Murine CMP-Neu5Ac synthetase has been shown to reside in the cell nucleus, and three clusters of basic amino acids have been identified that may function as NLS sequences (25). The locations and amino acid sequences of the basic clusters (BC1, BC2, and BC3) are shown in Fig. 1. NLS sequences are often preceded by helix-breaking proline residues (for review, see Ref. 32). Therefore, the sequences K 198 RPRR 2 (BC2) and P 196 AKRPRR (BC2 PA ) are both potential NLS motifs and were investigated. The gray boxes in Fig. 1 indicate sequence motifs that are highly conserved between bacterial and vertebrate CMP-Neu5Ac-syn, suggesting their importance for enzyme structure and/or function. It should be noted that BC2 is surrounded by highly conserved domains. Computational analyses, performed with different algorithms as described under "Experimental Procedures," identified only BC2 as a potential NLS (program PSORTII; Refs. 39 and 40). BC1 and BC3 were determined by eye.
Initial experiments investigated the ability of BC sequences to direct a non-nuclear protein to the nuclear compartment. BC1, BC2, BC2 PA , and BC3 were fused to eGFP, which encodes a soluble protein that distributes evenly throughout the cell (Ref. 42; Table I). The NLS (P 126 KKKRKV) of simian virus SV40 large T antigen (SV40-T) was fused to eGFP as a positive control (Ref. 33; Table I). eGFP and C-terminal-extended eGFP-fusion proteins were transiently expressed in NIH 3T3 cells and visualized by direct immunofluorescence microscopy ( Fig. 2, panel 1). DNA staining with Hoechst 33258 was used to detect cell nuclei (Fig. 2, panel 2). The merged images shown in Fig. 2, panel 3, clearly demonstrate that eGFP without any fusion peptide is homogenously distributed throughout the cell, while the eGFP-SV40-T chimera localizes to the cell nucleus. Like eGFP-SV40-T the fusion proteins eGFP-BC1, eGFP-BC2, and eGFP-BC2 PA were found in the nuclear compartment. In contrast, eGFP-BC3 mirrored the staining obtained with eGFP. Therefore, BC1 and BC2/BC2 PA , but not BC3, are sufficient to target a non-nuclear protein to the cell nucleus.
BC2 Is Necessary and Sufficient to Target Murine CMP-Neu5Ac-syn to the Nucleus-To determine the importance of BC1, BC2, and BC3 in nuclear transport of murine CMP-Neu5Ac-syn itself, each motif was individually deleted in an N-terminally FLAG/Myc-tagged enzyme giving the constructs ⌬BC1, ⌬BC2, ⌬BC2 PA , and ⌬BC3 (Table IIA). Deletion mutants and wild-type murine CMP-Neu5Ac-syn were expressed in NIH 3T3 cells, and the intracellular localization of these proteins was analyzed by indirect immunofluorescence microscopy using anti-FLAG mAb M5 and a Cy3-labeled secondary antibody (Fig. 3, panel 1). Cell nuclei were stained with Hoechst 33258 (Fig. 3, panel 2), and the merged photos are shown in panel 3. The data show that all mutant proteins were expressed. However, only ⌬BC1 and ⌬BC3 were efficiently transported to the cell nucleus. Nuclear transport was abolished in mutants ⌬BC2 and ⌬BC2 PA . Moreover, an insertionmutant Ins-BC2, in which K 198 RPRR is replaced by Q 198 DWDGNLNPA, by chance found in mutagenesis experiments and used as an additional control, was unable to enter the cell nucleus. These results confirm BC2 as being the only essential NLS in CMP-Neu5Ac-syn. BC1, which functions as an NLS for eGFP (Fig. 2), had no visible effect on the nuclear transport of murine CMP-Neu5Ac-syn under the conditions used in this study.
All Basic Residues in BC2 Are Important for Nuclear Targeting-Next, the contribution of single amino acid residues in BC2 for its function as an NLS was investigated. The sequence stretch P 196 AKRPRRQD containing BC2 PA plus two additional downstream residues was systematically modified by site-directed mutagenesis. Amino acids were replaced by alanine, either in pairs (mutants K198A/R201A and R199A/R202A) or individually (all other mutants, Table IIB). After transient expression in NIH 3T3 cells, the subcellular localization of mutant proteins was analyzed by indirect immunofluorescence microscopy using the anti-FLAG mAb M5 and a Cy3-labeled secondary antibody (Fig. 4, panel 1). Cell nuclei were stained with Hoechst 33258 (Fig. 4, panel 2), and images were merged in panel 3. Simultaneous mutation of two basic amino acids trapped the proteins K198A/R201A (Fig. 4A) and R199A/ R202A (Fig. 4B) in the cytoplasm. Moreover, individual replacement of the above mentioned residues by alanine led to the expression of cytoplasmic proteins in the case of K198A, R199A, and R201A (Fig. 4, D-F). Only mutant R202A (Fig. 4G) was targeted to the nucleus, although a weak cytoplasmic staining was visible for R202A in a subsequent set of experiments. In contrast, replacement of Pro 200 (P200A) and of the neighboring residues Pro 196 , Gln 203 , and Asp 204 by alanine (P196A, Q203A, D204A) had no effect on nuclear transport (Fig. 4, H-K).
Cytoplasmic Forms of Murine CMP-Neu5Ac-syn Are Catalytically Active-To determine whether nuclear localization is a prerequisite for CMP-Neu5Ac synthetase activity in mammalian cells, the enzymatic activity of nuclear-and cytoplasmiclocalized CMP-Neu5Ac-syn mutants used in Figs. 3 and 4 was analyzed in vivo by complementation of two cellular systems lacking active CMP-Neu5Ac-syn: CHO cells of the complementation group LEC29.Lec32 (36) and E. coli strain EV5, a derivative of E. coli K1 (37). In both mutants a genetic defect inactivates endogenous CMP-Neu5Ac-syn, leading to an asialophenotype (36,43). One sialic acid epitope missing in both cellular systems is polySia, and cell lysates did not bind the polySia-specific mAb 735 (see first lane (Ø) in Fig. 5). However, after transfection with wild-type murine CMP-Neu5Ac-syn a return to the polySia-positive phenotype was observed in LEC29.Lec32 cells as well as in E. coli EV5 (Fig. 5). Complementation of the polySia-negative phenotype was used to test the activity of CMP-Neu5Ac-syn variants. The bacterial mutant served as a control, because defects influencing nuclear transport of CMP-Neu5Ac-syn would not be functionally relevant in the prokaryotic system. EndoNE treatment of lysates to remove polySia was used to show that mAb 735 is specific for the polySia epitope (34). Parallel samples of the cell lysates were analyzed by Western blot before (Ϫ) and after (ϩ) treatment with endoNE (Fig. 5). PolySia, which is bound to the 2 The position of the amino acid in the primary sequence is indicated as a superscript in the top right hand corner (e.g. K 198 ). neural cell adhesion molecule in mammalian cells (16), migrates as a diffuse band above 200 kDa (Fig. 5A). In E. coli K1 polySia is bound to a lipid anchor of yet unidentified structure (for review, see Ref. 44) and migrates as a typical smear between 30 and 100 kDa (Fig. 5C). Mutant forms of N-terminally FLAG/Myc-tagged murine CMP-Neu5Ac-syn (see Fig. 5) were transiently expressed in LEC29.Lec32 and in E. coli EV5 cells, and re-expression of polySia was monitored as described. Deletion mutants ⌬BC1 and ⌬BC3, which are both targeted to the cell nucleus (Fig. 3), restored polySia expression in both the mammalian and the bacterial cell system (Fig. 5, A and C). The low polySia signal observed in cells transfected with ⌬BC3 corresponds to low expression of the protein (Fig. 5, B and D). While ⌬BC1 is expressed at the level of the wild-type enzyme, the expression level of ⌬BC3 is drastically reduced, in LEC29.Lec32 (Fig. 5B) and E. coli (Fig. 5D) cells, indicating that deletion of BC3 interferes with expression and/or stability of the mutant protein. In contrast, cells transfected with the deletion mutants ⌬BC2 and ⌬BC2 PA or the insertion mutant Ins-BC2 remained polySia-negative, despite stable expression of the proteins in E. coli and, with the exception of Ins-BC2, also in LEC29.Lec32. Because loss of activity in E. coli cannot be solely due to the deletion of an NLS (see Fig. 3), this data suggest that BC2 contains elements that are required for enzymatic activity.
To test for this possibility mutants harboring double and single amino acid exchanges were analyzed for enzymatic activity as described. As shown in Fig. 5, B and D, all protein variants were stably expressed in mammalian and bacterial cells. Simultaneous mutation of two basic amino acids, which kept CMP-Neu5Ac-syn in the cytoplasm (see K198A/R201A and R199A/R202A in Fig. 4) led to a loss of enzymatic activity only in the mutant R199A/R202A (Fig. 5, A and C). The double mutant K198A/R201A was fully active and consistent with this observation, and the single site mutants K198A and R201A were, likewise, cytoplasmic but active. In contrast, the cytoplasmic mutant R199A, as well as the predominantly nuclear mutant R202A, showed drastically reduced synthetase activity. Whereas all basic residues of BC2 are important for nuclear localization, Arg 199 and Arg 202 in addition are crucial for enzymatic activity.
Helix-breaking proline residues found N-terminal (Pro 196 ) and in the middle (Pro 200 ) of BC2 are not involved in either nuclear transport (Fig. 4, H and I) nor catalytic function of CMP-Neu5Ac-syn (Fig. 5, A and C). Similarly, replacement of Asp 204 with alanine had no deleterious effect on either function (see mutant D204A), while mutant Q203A, although correctly located in the nucleus, showed drastically reduced activity in both mammalian and bacterial cells. These findings suggest that Gln 203 is involved in the catalytic function of CMP-Neu5Ac-syn. The combined data demonstrate that synthetase activity and nuclear transport are independent properties of the murine CMP-Neu5Ac-syn, but they are located in the same domain, and thus they may both be inactivated by changes in certain amino acids they share in common.

TABLE I eGFP Fusion proteins
Plasmids for eukaryotic expression of eGFP fusion proteins were generated by ligating hybridized oligonucleotides to the 3Ј-end of the eGFP cDNA in the vector peGFPCI. This table shows the names of the resulting plasmids (column 1) as well as the amino acid sequences of the basic clusters (column 2) encoded by the corresponding oligonucleotides (column 3). The oligonucleotides were generated with overhanging sequences to allow for the directed cloning into BglII and EcoRI sites of the vector (BglII and EcoRI extensions are italicized). Oligonucleotides were hybridized and ligated into the corresponding restriction sites in peGFPCI. The NLS of the large T antigen of simian virus SV40 (SV40-T) served as a positive control.

FIG. 2. Intracellular localization of eGFP fusion proteins.
BCs of murine CMP-Neu5Ac-syn or SV40-T antigen NLS were fused individually to the C terminus of eGFP (Table I). NIH 3T3 cells were transiently transfected with cDNAs of either eGFP alone or eGFP fusion proteins. Intracellular localization was analyzed by direct immunofluorescence at ϫ400 magnification (panel 1). Cell nuclei were stained with Hoechst 33258 (panel 2). The photos are merged in panel 3.

Arg 202 and Gln 203 Are Evolutionary Conserved Amino Acid
Residues-The functional importance of Gln 203 , which is not part of BC2, prompted us to analyze amino acids at related positions in other CMP-Neu5Ac-syn sequences. The alignment in Fig. 6A shows a comparison of bacterial CMP-sialic acid synthetases (19 -24), murine CMP-Neu5Ac-syn (25), and the recently cloned rainbow trout CMP-Kdn-syn, which exhibits a similar affinity for Neu5Ac and Kdn (26). Additionally, the sequences of putative CMP-sialic acid synthetases that have been identified based on homology were aligned in Fig. 6B. Amino acid residues that comprise BC2 in the murine sequence are shaded in gray. The alignment reveals that Arg 202 (part of BC2) and Gln 203 represent highly conserved positions (open boxes), although a conservative substitution of glutamine by asparagine occurs in E. coli and Legionella pneumoniae, and cysteine substitutes arginine in the L. pneumoniae enzyme.
Evolutionary Conserved Residues in BC2 Are Important for Enzymatic Activity-Our finding that Arg 202 is part of the active site in murine CMP-Neu5Ac-syn is in excellent agreement with crystal structure data recently obtained for the NmB CMP-Neu5Ac-syn (45). In the NmB enzyme Arg 165 , which corresponds to Arg 202 in the murine enzyme, participates in Neu5Ac binding, providing an explanation for the dramatic drop in activity seen for murine mutant R202A (Fig. 5, A and  C). However, the structural data do not contain information about the functional relevance of the neighboring glutamine residue (Gln 166 in NmB; Gln 203 in mouse). To investigate this in more detail, both Arg 165 and Gln 166 in the NmB enzyme BCs in the murine CMP-Neu5Ac-syn were deleted using the primers given in A. In mutant Ins-BC2 the basic cluster 2 is replaced by a nonsense insertion that occurred by chance during mutagenesis. Single amino acids exchanges were introduced by site-directed mutagenesis using the primers given in B for the murine enzyme and in C for the NmB enzyme. The nucleotide triplet used for the selected amino acid was changed to alanine (GCN) by the indicated nucleotide exchanges (bold letters in the primer sequences). Deleted or exchanged nucleotides are given in second column.
were mutated to alanine, resulting in the mutants NmB R165A and NmB Q166A (Table IIC), respectively. Activity was tested by complementation in E. coli EV5 as described (Fig. 7). As expected from the crystal structure, no activity was found for NmB R165A (Fig. 7A). The respective murine mutant R202A exhibited low but detectable activity in EV5 (see Fig. 5C). Moreover, the activity in mutant NmB Q166A, like the corresponding murine mutant Q203A (Fig. 5C), was strongly reduced (Fig. 7A). Both proteins were expressed at the level of the wild-type enzyme (Fig. 7B), indicating that catalytic function, but not protein stability, is impaired in the mutant proteins. These data confirm the essential function of this conserved amino acid pair and strongly suggest that Arg 202 in the murine enzyme, which is analogous to Arg 165 in the NmB enzyme, participates in substrate binding. The functional role of the conserved glutamine residue is not clear, but the structural data obtained for the NmB enzyme (45) suggest that this conserved glutamine residue is involved in the quaternary organization of CMP-Neu5Ac-syn. DISCUSSION Isolation of the murine CMP-Neu5Ac-syn cDNA and visualization of the epitope-tagged recombinant protein in mammalian cells (25) corroborated earlier reports that CMP-Neu5Acsyn is a nuclear protein in different vertebrate cells (for review, see Refs. 18;46). Translocation in the nucleus often depends on the existence of an NLS sequence, which in the case of canonical NLSs are short stretches of basic amino acids that mediate nuclear import of mature proteins. Apart from the well defined NLSs found in SV40 large T antigen and nucleoplasmin (33, 47), a variety of NLS have been identified (for review, see Refs. 27; 31). Although no strictly conserved consensus sequence has been delineated so far, variations of the four-pattern motif K-K/R-X-K are available for data base searches (PSORTII; Refs. 39, 40, and 48). However, the existence of a putative NLS must still be verified experimentally.
Three NLS-like motifs previously identified in the primary

FIG. 3. Intracellular localization of murine CMP-Neu5Ac-syn mutants lacking individual basic clusters (⌬BC).
Wild-type and mutant constructs were transiently expressed in NIH 3T3 cells. The mutant Ins-BC2 harbors a ten-amino acid nonsense insertion (Q 198 DWDGNLNPA) instead of wild-type sequence (K 198 RPRR) at BC2. N-terminally FLAG/Myc-tagged CMP-Neu5Ac-syn and the different mutant forms were localized by immunostaining with anti-FLAG mAb M5 and a Cy3-conjugated secondary antibody (panel 1). Nuclear staining was performed with Hoechst 33258 (panel 2). The photos are merged in panel 3 (ϫ400 magnification).

FIG. 4. Intracellular localization of murine CMP-Neu5Ac-syn mutants.
Wild-type (A) and mutant constructs (B-K) were transiently expressed in NIH 3T3 cells. To localize the FLAG/Myc-tagged proteins, permeabilized cells were stained with anti-FLAG mAb M5. Bound primary antibodies were visualized with anti-mouse IgG-Cy3 (panel 1). Nuclear staining was performed with Hoechst 33258 (panel 2). The photos are merged in panel 3 (ϫ400 magnification). All analyzed amino acid exchanges are located in the nine-amino acid stretch given below. The sequence of BC2 is highlighted with a gray box. Amino acids essential for nuclear transport are marked with an asterisk.
sequence of the murine CMP-Neu5Ac-syn were investigated for function in this study (25). Initially the capability of each basic cluster to target eGFP into the cell nucleus was analyzed (Fig.  2). This technique has been employed successfully to study NLS sequences from other nuclear-localized proteins (e.g. Stat5b (49), diacylglycerol kinase-(50), and Ring3 (42)). Furthermore, the importance of individual basic clusters for the nuclear import of CMP-Neu5Ac-syn was tested with deletion mutants ⌬BC1-⌬BC3 (Fig. 3).
The most C-terminal basic cluster BC3 was unable to signif- icantly alter the localization of eGFP (Fig. 2), and deletion of BC3 had no deleterious effect on nuclear import of murine CMP-Neu5Ac-syn (Fig. 3). Both findings demonstrate that BC3 is not a functional NLS. In addition, BC3 seems not to contain elements that are required for catalytic activity. The low expression levels of this mutant protein in transfection experiments, however, suggest that this amino acid stretch may be important for the folding and/or stability of the CMP-Neu5Ac-syn.
The most N-terminal basic cluster BC1 is sufficient to direct eGFP to the nuclear compartment (Fig. 2), but it is not necessary for nuclear import of the parent protein under the experimental conditions used in this study (Fig. 3). It may be that BC1 supports nuclear transport physiologically, for example by improving the kinetics of the transport process. There are examples where an enhancement in nuclear transport is observed if two or more NLS are present in a given protein (51,52). Alternatively, BC1 could be masked in the native CMP-Neu5Ac-syn by inter-or intramolecular interactions and thus not be accessible to transport factors (53). Secondary structure predictions argue against this latter hypothesis and suggest BC1 to be a flexible hydrophilic region, most likely located at the surface of the protein. In line with this we found that the stability of a bacterial expressed recombinant CMP-Neu5Acsyn is enhanced if 38 amino acids are deleted from the N terminus. This truncation, which includes BC1, does not affect the activity of the recombinant protein. 3 So far, our data are not sufficient to elucidate the role of BC1 in the native protein.
BC2 is sufficient to target eGFP to the nuclear compartment and, in addition, is the dominant NLS in murine CMP-Neu5Acsyn. Deletion of the five amino acid residues that form BC2 (K 198 RPRR) was sufficient to retain the enzyme in the cytoplasm. Using site-directed mutagenesis the basic amino acids Lys 198 , Arg 199 , and Arg 201 were shown to be essential for the formation of the functional NLS. Isolated replacement of these positions by alanine completely abolished nuclear transport (see Fig. 4, D-F). Arg 202 is important for optimal NLS function. The effect on transport observed in the mutant R202A was small but clearly visible in repeated experiments (see Fig. 4G). Pro 200 , despite being part of BC2, is not required for the formation of a functional NLS. Additional experiments are required to determine whether Pro 200 is of importance for the spatial geometry of the NLS. At this point we are also unable to judge the importance of Pro 196 (see BC2 PA ). Helix-breaking proline residues often precede an NLS and augment activity (for review, see Refs. 27 and 31). Such an effect was not observed in this study, but may be of importance under physiological conditions. Finally, variations in the highly conserved C-terminal-flanking amino acid residues of BC2 (mutants Q203A and D204A) are irrelevant for the nuclear transport. Because neither the N-nor the C-terminal ends of BC2 are involved in nuclear transport of murine CMP-Neu5Ac-syn, the active NLS was determined to be K 198 RXRR. This motif fits well with the four-residue motif K-K/R-X-K used in data base searches (PSORT II; Refs. 39 and 40) and originally suggested by Chelsky (K-R/K-X-R/K; Ref. 48). Interacting partners recognizing BC2 and the transport pathway for nuclear import of the protein remain to be elucidated.
Expression of active murine CMP-Neu5Ac-syn in bacteria has already suggested that CMP-Neu5Ac-syn activity from animal cells is not dependent on nuclear localization (Ref. 25 and this paper). To determine connections between nuclear localization and synthetase activity, mutants generated in this study were tested for their ability to complement two CMP-Neu5Ac-syn-deficient mutants, CHO LEC29.Lec32 cells (36) and E. coli EV5 (43). Enzymatic activity and protein expression were found to correlate directly in bacteria and mammalian cells for the wild-type and mutant CMP-Neu5Ac-syn. Importantly, activity in eukaryotic cells seems not to be dependent on nuclear localization, since the cytoplasmic mutants K198A and R201A restored polySia biosynthesis. Furthermore, the fact that Arg 199 and Arg 202 are involved in both nuclear transport and catalytic activity of CMP-Neu5Ac-syn suggests that a functional NLS is also required for optimal catalysis. BC2 may contain part of the active site of the enzyme or, perhaps more likely, is important for maintaining the structure of the active site. The results are supported by the comparison of various known CMP-Neu5Ac-syn sequences (Fig. 6). The alignment revealed no significant conservation for BC2 but identified Arg 202 and Gln 203 as highly conserved residues. This is in line with crystal structure data recently obtained for the NmB enzyme (45), which show that Arg 165 (corresponding to Arg 202 in the murine enzyme) is part of the CMP-Neu5Ac binding pocket and of the dimerization domain. As expected from this central function, the exchange R165A completely abolished catalytic activity in the NmB enzyme. It is most likely that Arg 202 has an identical function in the murine enzyme. It was, however, surprising to find that the negatively charged Glu 162 , also part of the Neu5Ac binding pocket in the NmB enzyme, corresponds to the positively charged Arg 199 in the murine enzyme.
Mutation of the second highly conserved residue, Gln 203 in mouse and Gln 166 in NmB, to alanine was followed by a drastically reduced functional activity (Figs. 5 and 7). According to the existing three-dimensional data this position does not participate in substrate binding but seems to be required for the quaternary organization of the protein (45). A lack of quaternary organization in mutants may worsen the overall kinetics of their reactions thus leading to the decreased enzyme activity.
Several suggestions have been made to explain the unusual subcellular localization of CMP-Neu5Ac-syn. (i) The nuclear environment may be a prerequisite for enzymatic activity. (ii) The CTP concentration may be higher in the cell nucleus. (iii) Production of CMP-Neu5Ac in the nucleus may protect the nucleotide sugar from subsequent modifications by cytoplasmic enzymes, such as the CMP-Neu5Ac hydroxylase (54,55). (iv) CMP-Neu5Ac is an allosteric inhibitor of UDP-GlcNAc 2-epimerase/kinase. This enzyme, which is localized in the cytoplasm, initiates the synthesis of sialic acids. Sequestration of CMP-Neu5Ac may be necessary to prevent early inactivation of UDP-GlcNAc 2-epimerase/kinase. (v) CMP-Neu5Ac may possibly be required for sialyltransferases localized in the nucleus. Sialyltransferase activity associated with rat liver nuclei has been reported that may be critical for the regular function of various nuclear proteins, which never transit the Golgi apparatus (56,57). Our data show that cytoplasmic forms of the murine CMP-Neu5Ac-syn are active in mammalian cells, thus nuclear localization cannot be a prerequisite for enzymatic activity and the CTP concentration in the cytoplasm is at least sufficient for the functionality of the enzyme. Moreover, trials to discriminate between cytosolic and nuclear CTP pools failed (58). On the other hand, Vionnet et al. (46) demonstrated DNAbinding for the purified bovine CMP-Neu5Ac-syn. The data presented in this study, as well as earlier data describing the unusual localization of CMP-Neu5Ac-syn in the nuclear compartment, allow us to speculate that CMP-Neu5Ac-syn or its product may have a second cellular function. Initial experiments have been carried out in our laboratory toward understanding the physiological relevance of the nuclear localization, but at the cell culture level no noticeable differences were found between LEC29.Lec32 cells transfected with either nuclear or cytoplasmic forms of CMP-Neu5Ac-syn. We are currently investigating whether the ratio between N-acetyl-and N-glycolylneuraminic acid varies with the subcellular localization of the CMP-Neu5Ac-syn. Most important, a transgenic mouse model shall soon be available, in which cytoplasmic forms substitute endogenous CMP-Neu5Ac-syn. This model should provide a new basis for studies aimed at enlightening the functional role of CMP-Neu5Ac-syn in the cell nucleus.