Cloning, Golgi Localization, and Enzyme Activity of the Full-length Heparin/Heparan Sulfate-Glucuronic Acid C5-epimerase*

While studying the cellular localization and activity of enzymes involved in heparan sulfate biosynthesis, we discovered that the published sequence for the glucu-ronic acid C5-epimerase responsible for the intercon-version of D -glucuronic acid and L -iduronic acid resi- dues encodes a truncated protein. Genome analysis and 5 * -rapid amplification of cDNA ends was used to clone the full-length cDNA from a mouse mastocytoma cell line. The extended cDNA encodes for an additional 174 amino acids at the amino terminus of the protein. The murine sequence is 95% identical to the human epimerase identified from genomic sequences and fits with the general size and structure of the gene from Drosophila melanogaster and Caenorhabditis elegans . Full-length epimerase is predicted to have a type II transmembrane topology with a 17-amino acid transmembrane domain and an 11-amino acid cytoplasmic tail. An assay with increased sensitivity was devised that detects enzyme activity in extracts prepared from cultured cells and in recombinant proteins. Unlike other enzymes involved in glycosaminoglycan biosynthesis, the addition of a c- myc tag or green fluorescent protein to the highly conserved COOH-terminal portion of the protein inhibits its activity. The amino-terminally truncated epimerase does not localize to any cellular compartment, whereas the full-length enzyme is in the Golgi, where heparan dIII inserting the extended 5 9 end amplified by PCR. The GFP-tagged full-length epimerase amplifying the full coding region from pCDNA3.1 with the primers 5EpiXhoI and 3 9 Epi-GFPBam and by cloning into the Xho I and Bam HI sites of pEGFP-N1. The coding sequence was verified by directly sequencing PCR products from three independent amplifications from the MST cDNA library. All PCR amplifications were done using Vent DNA polymerase (New England Bio- labs), and clones were sequenced on an ABI 373 DNA sequencer using dye terminator cycle sequencing. Generation of Full-length Murine Epimerase— 5 9 -Rapid amplification of cDNA ends was performed according to the instruc- tions (CLONTECH) with mRNA isolated from MST cells. Two gene-specific primers were used based on the murine sequence for the epimerase 3EpiBam (5 9 -GGATCCGAGATTCCATGC-CGCGCTCGTACAAG-3 and BC15 (5 9 -ACATTGGTGGATCTAGACT- T-3 9 Analysis of five independent clones, each with the same 5 9 end, consensus sequence. the 3 9 end of murine 3 9 end MST cDNA cleotides

Heparan sulfate proteoglycans are located on the cell surface and in the extracellular matrix, where they play important roles in cell adhesion, differentiation, and growth in vitro and in vivo (1)(2)(3). To a large extent, these biological activities depend on the heparan sulfate chains attached to the core protein. Heparan sulfate, a type of glycosaminoglycan, initially assembles by the copolymerization of N-acetyl-D-glucosamine (GlcNAc) and D-glucuronic acid (GlcA). The backbone then undergoes extensive modification initiated by the N-deacetylation and N-sulfation of subsets of GlcNAc residues. Subsequently, D-GlcA residues adjacent to the N-sulfated sugars are converted to L-IdoUA 1 by a C5-epimerase and are sulfated at C-2 by a specific sulfotransferase. The glucosamine units also can be sulfated at C-6 and to a lesser extent at C-3. The blocklike arrangement of the modified residues confers specific binding properties to the chains for protein ligands, which in turn facilitate various biological activities.
Many of the enzymes involved in heparan sulfate and heparin formation seem to be members of multienzyme gene families. Two exceptions are the C5-epimerase that interconverts D-GlcA and L-IdoUA and the 2-O-sulfotransferase that adds sulfate to C-2 of IdoUA residues and to a lesser extent GlcA residues. The C5-epimerase has been partially purified from mouse mastocytoma (4) and purified to homogeneity from bovine liver (5). A bovine cDNA for the epimerase has been cloned as well (6). Kinetic studies have clarified the substrate specificity of the epimerase, showing that the enzyme will react with both D-GlcA (forward reaction) and L-IdoUA (reverse reaction) when these residues are located toward the reducing side of N-sulfated glucosamine residues, but it will not react with uronic acids that are O-sulfated or that are adjacent to Osulfated glucosamine residues (7,8). This specificity is consistent with the overall order of modification, suggesting that epimerization begins to occur after GlcNAc N-deacetylation and N-sulfation but before glucosamine residues undergo 6-Osulfation and 3-O-sulfation (7,9). The fact that the epimerase seems to be represented only once in vertebrate and invertebrate genomes suggests that the extent of uronic acid epimerization depends on the level of enzyme expression and production of the N-sulfated tracts.
In an attempt to study the cellular localization and potential interaction of the epimerase with other enzymes in the pathway, we discovered that the published bovine sequence encodes a truncated protein. 2 This report provides the full-length sequence from mouse and human, an improved set of conditions for assaying the epimerase in cell extracts, and a demonstration that the enzyme is localized to the Golgi in vertebrate cells.

EXPERIMENTAL PROCEDURES
Cell Culture-Chinese hamster ovary cells (CHO-K1) were obtained from the American Type Culture Collection (CCL-61, Manassa, VA). MST cells were derived from the Furth murine mastocytoma (10). CHO cells were grown in Ham's F-12 medium (Life Technologies, Inc.), and MST cells were grown in RPMI 1640 medium. Both media were supplemented with 10% (v/v) fetal bovine serum (Hyclone Laboratories), 100 g/ml streptomycin sulfate, and 100 units/ml penicillin G. The cells were cultured at 37°C under an atmosphere of 5% CO 2 in air at 100% relative humidity.
Cloning the Murine C5-Epimerase-A murine epimerase cDNA frag-* This work was supported by National Institutes of Health Grant R37GM33063 (to J. D. E.), a fellowship from the PEW Latin American Fellows Program in the Biomedical Sciences (to M. A. S. P.), and National Institutes of Health Training Grants CA67754 (to B. E. C.) and GM08666 (to S. K. O.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed: Universidade Federal De Sao Paulo, Vila Clementino, CEP 04044020, Sao Paulo, Brazil. E-mail: maspinhal.bioq@epm.br. 1 The abbreviations used are: IdoUA, L-iduronic acid; CHO, Chinese hamster ovary; GFP, green fluorescent protein; PCR, polymerase chain ment corresponding to the published bovine sequence (GenBank TM accession number AF003927) was cloned from an MST cDNA library using PCR, the forward primer 5Ј-ATGTCCTTTGAAGGCTACAAT-GTGG-3Ј, and the reverse primer 5Ј-CTAGTTGTGCTTTGCCCGGCT-GCC-3Ј, which anneal with the first and last 24 bases of the partial bovine sequence (6). The PCR product was blunt-end cloned into pGEM (Promega) for sequencing. Subsequently, the primers XhoEpi5Ј (5Ј-CCCCGGCTCGAGGCCGCCATGTCCTTTGAAGCCTACAATG-3Ј) and BamEpi3Ј (5Ј-CTGGATCCTAGTTGTGCTTTGCCCGG-3Ј) were used to amplify the cDNA from the pGEM epimerase clone for transfer into pCDNA3.1 (Invitrogen) using the XhoI and BamHI sites. To generate the YFP-tagged truncated epimerase, the primers XhoEpi5Ј and 3ЈEpi-GFPBam (5Ј-CTGGATCCCCGTTGTGCTTTGCCCGG-3Ј) were used to amplify the epimerase cDNA, which was cloned into the XhoI and BamHI sites of pEYFP-N1.
The cDNA containing the full-length epimerase was cloned using a primer designed to the proposed 5Ј end deduced from the human genomic DNA sequence (GenBank TM accession number AC026992). This primer, 5EpiXhoI (5Ј-CTCGAGCCATGCGTTGCTTGGCAGCTCG-G-3Ј), was used with an internal reverse primer, 3EpiBam (5Ј-GGAT-CCGAGATTCCATGCCGCGCTCGTACAAG-3Ј), to amplify the 5Ј 900 base pairs of the cDNA from the murine MST cDNA library. The cDNA encoding the full-length epimerase was then constructed by digesting the truncated epimerase in pCDNA3.1 with XhoI and HindIII and then by inserting the extended 5Ј end amplified by PCR. The GFP-tagged full-length epimerase was generated by amplifying the full coding region from pCDNA3.1 with the primers 5EpiXhoI and 3ЈEpi-GFPBam and by cloning into the XhoI and BamHI sites of pEGFP-N1. The coding sequence was verified by directly sequencing PCR products from three independent amplifications from the MST cDNA library. All PCR amplifications were done using Vent DNA polymerase (New England Biolabs), and clones were sequenced on an ABI 373 DNA sequencer using dye terminator cycle sequencing.
Generation of Full-length Murine Epimerase-5Ј-Rapid amplification of cDNA ends was performed according to the manufacturer's instructions (CLONTECH) with mRNA isolated from MST cells. Two genespecific primers were used based on the murine sequence for the epimerase described above, 3EpiBam (5Ј-GGATCCGAGATTCCATGC-CGCGCTCGTACAAG-3Ј) and BC15 (5Ј-ACATTGGTGGATCTAGACT-T-3Ј). Analysis of five independent clones, each with the same 5Ј end, yielded a consensus sequence.
The sequence of the 3Ј end of the murine coding sequence was determined by amplifying the 3Ј end from an MST cDNA library using BC11 (5Ј-GGAGACCACAGAAAAGAATC-3Ј) and BC42 (5Ј-GGAGAC-CACAGAAAAGAATC-3Ј). BC11 was designed to anneal between nucleotides 1164 and 1183 of the mouse epimerase cDNA, whereas BC42 was designed to anneal to the 3Ј-untranslated region (UTR) and was designed based on nucleotides 1832-1855 of the partial human sequence (GenBank TM accession number AB020643). The PCR product was cloned, and three independent isolates were sequenced. All three clones contained two silent changes from the published human sequence. The GenBank TM accession numbers for the murine cDNA and encoded protein are AF325532 and AAG42004, respectively.
Enzyme Localization-Epimerase constructs were generated with COOH-terminal c-myc and GFP tags by subcloning into pCDNA3.1 containing myc/His 6 and pEGFP (CLONTECH), respectively. An amino-terminal c-myc-tagged clone was generated by annealing the oligonucleotides BC46 (5Ј-ATGTCTAGAGAACAAAAACTCATCTCAG-AAGAGGATCTGTCTAGAGCA-3Ј) and BC47 (5Ј-TGCTCTAGACAGA-TCCTCTTCTGAGATGAGTTTTTGTTCTCTAGACAT-3Ј), which codes for a myc tag. The oligonucleotides were boiled for 1 min, cooled on ice, phosphorylated with polynucleotide kinase (New England Biolabs), and cloned in-frame into the EcoRV site in the polylinker region of pcDNA3.1 containing the full-length epimerase. The clone used in these experiments actually contained two Myc tags in a tandem repeat at the amino terminus.
CHO cells were transiently transfected with 2 g of plasmid DNA using LipofectAMINE according to the manufacturer's directions (Life Technologies, Inc.). Cells were grown on 24-well glass microscope slides and were processed for enzyme localization studies 24 -36 h after transfection. After the cells were fixed for 1 h with 2% paraformaldehyde in 75 mM phosphate buffer, pH 7.5, they were rinsed several times with phosphate-buffered saline (PBS) (11). Cells were then permeabilized with 1% Triton X-100 (v/v) and 0.1% bovine serum albumin (BSA) (w/v) in PBS. The primary antibody, mouse anti-Myc (Invitrogen) monoclonal antibody, and rabbit polyclonal anti-␣-mannosidase II antiserum (a gift from Marilyn G. Farquhar, University of California, San Diego) were diluted 1:400 in PBS with 1% BSA and incubated for 1 h with the fixed cells. To remove the unbound primary antibody, the cells were washed several times for 30 min with PBS containing 0.1% BSA. The samples were then incubated for 1 h with the secondary antibody, anti-rabbit Cy5 (Qccuvate Chemicals, NY) or anti-mouse-TRITC (Sigma), diluted 1:200 in PBS containing 1% BSA. After several washes, the cells were mounted with Vectashield containing 4Ј,6-diamidino-2-phenylindole for nuclear staining (Vector Laboratories). The photomicrographs shown in Fig. 5, A-D, were captured with a Photometrics charge-coupled device mounted on a Nikon microscope adapted to a DeltaVision (Applied Precision, Inc.) deconvolution imaging system. The data sets were deconvolved and analyzed using SoftWorx software (Applied Precision, Inc.) on a Silicon Graphics Octane work station. The photomicrograph shown in Fig. 5E was captured with a Hamamatsu C5810 three-color chilled charge-coupled device camera mounted on a Zeiss Axiophot (ϫ100 lens) microscope.
Epimerase Assay-Normal and transfected cells were washed twice with cold PBS and once with cold 0.25 M sucrose in 20 mM Tris, pH 7.4, and were then scraped with a rubber policeman into 100 l of cold buffer containing 0.25 M sucrose, 20 mM Tris-HCl, pH 7.4, 20 M phenylmethylsulfonyl fluoride, 0.5 g/ml leupeptin, and 0.5 g/ml pepstatin. Cells were lysed by sonication with a microtip sonicator, and the protein concentration was quantitated with the Bradford assay (Bio-Rad) using BSA as the standard. The extracts were stable when stored at Ϫ20°C.
The epimerase substrate consisted of modified N-acetylheparosan and was prepared as described (12). Briefly, Escherichia coli K5 capsular polysaccharide was labeled in vivo with D-[5-3 H]glucose (PerkinElmer Life Sciences) and purified from the growth medium. The GlcNAc residues in the labeled polysaccharide were N-deacetylated to near completion with anhydrous hydrazine and hydrazine sulfate (Sigma) and were N-sulfated with trimethylamine sulfur trioxide complex (Sigma). The concentration of N-acetylheparosan was determined by a carbazol assay for uronic acids (13), which yielded a radiospecific activity of 76 cpm/pmol GlcA (43 Ci/mol).
Detection of epimerase activity was based on the release of 3 H from [5-3 H]GlcA units in the polysaccharide and recovery as 3 H 2 O (12). Initial assays were set up according to the published reaction conditions ("original"), which contained 50 mM HEPES, 15 mM EDTA, 100 mM KCl, and 0.015% Triton X-100, pH 7.4 (12). Protein, substrate, and various ancillary factors were adjusted to maximize the activity detected in normal and transfected cells. The "revised" assay consisted of 25 mM HEPES, pH 7.0, 0.1% Triton X-100, 300 pmol of 3 H-sulfated heparosan substrate, and 2 g of cell protein in a total volume of 20 l. Some assays contained 40 mM CaCl 2 , but divalent cations were later found not to be required. The reactions were incubated for 2 h at 37°C and halted by the addition of 50 l of cold 50 mM sodium acetate buffer, pH 4.0, containing 50 mM LiCl. The sample and a 100-l rinse of the tube with buffer (25 mM HEPES, pH 7.0, and 0.1% Triton X-100) were transferred to a 0.4-ml column of DEAE-Sephacel (Amersham Pharmacia Biotech) that was equilibrated with the same buffer. The column was washed with 0.9 ml of assay buffer, and the 3 H 2 O recovered in the flow-through fractions was counted by liquid scintillation spectrometry using Ultima Gold (Packard Instrument Co.). A reagent blank containing everything except a source of enzyme was included as a control. This yielded values of ϳ200 cpm, which were subtracted from the experimental values that ranged from 300 to 3000 counts. All assays were done in duplicate with comparable results from three or more independent experiments.
Western Blotting-Cells were harvested, and 25 g of protein for each sample was analyzed by SDS-polyacrylamide gel electrophoresis on a 10% gel. The samples were transferred to a nitrocellulose membrane using the Bio-Rad Mini Protean II system. The membrane was blocked at 4°C overnight with 4% BSA in Tris-buffered saline (10 mM Tris, pH 8.0, and 150 mM NaCl) with 0.1% Tween 20 (TBST). Mouse anti-GFP (CLONTECH) and mouse anti-Myc (Invitrogen) were diluted 1:1000 and 1:5000 in TBST, respectively, and incubated for 1 h with the membrane at room temperature with shaking. The membrane was washed three times with TBST before the application of the secondary antibody, goat anti-mouse horseradish peroxidase (Bio-Rad) diluted 1:3000 in TBST. The membrane was washed six times with TBST and developed with SuperSignal West Pico chemiluminescent substrate (Pierce).

RESULTS AND DISCUSSION
The GlcA C5-epimerase was previously purified from murine mastocytoma and bovine liver (4, 5). Sequencing the purified protein yielded proline as the amino-terminal amino acid, sug-gesting that the protein had been proteolytically cleaved during the purification process (6). Several internal peptides also were generated by controlled proteolysis, and the peptide sequences were used to design oligonucleotides for screening a bovine lung cDNA library. A cDNA sequence was obtained that was 3085 base pairs and contained an open reading frame corresponding to a 444-amino acid protein from the first in-frame ATG codon. The deduced amino acid sequence predicted a 49,905-Da protein with a potential transmembrane domain located near the amino terminus. Expression of this clone in the baculovirus system yielded the expected activity (6).
Further analysis of the epimerase sequence suggested that it was incomplete. First, computer-aided analysis using PSORT did not indicate the predicted transmembrane domain and in fact suggested that the protein was most likely soluble (14). As described below, this was confirmed in localization studies with GFP-tagged constructs. Second, alignment of the published bovine epimerase sequence with other orthologs in the Gen-Bank TM data base suggested that the bovine sequence was potentially missing a large domain from the amino terminus (Fig. 1). A BLAST search revealed a human genomic DNA clone (GenBank TM accession number AC026992) containing an extended open reading frame that more closely matched the size of the epimerase orthologs found in Caenorhabditis elegans and Drosophila melanogaster. In addition, a partial human cDNA also was identified in the sequence data bases (GenBank TM accession number AB020643).
Based on this information, we cloned an extended cDNA from mouse mastocytoma mRNA using primers based on the human sequence. This fragment was further extended using 5Ј-rapid amplification of cDNA ends. The additional sequence added 753 nucleotides including 231 nucleotides in a 5Ј-UTR and 522 coding nucleotides. The 5Ј-UTR contains an in-frame termination codon (TGA) 21 base pairs upstream of the new initiation codon, suggesting that the cDNA encodes the fulllength epimerase (Fig. 2). The context of the new start codon conforms to an "adequate" Kozak sequence (AATatgC, consensus RNNatgY, where R ϭ A or G and Y ϭ T or C) (15), whereas the previously suggested start codon lacks A or G at Ϫ3 (tttatgt). Previous studies reported a 3Ј-UTR sequence of ϳ1.6 kilobases (6). Because the mRNA was found to be ϳ5 kilobases, an additional untranslated sequence of ϳ1.4 kilobases apparently exists, but its location is unknown.
The revised sequence adds 174 amino acids to the amino terminus of the previous sequence of the bovine epimerase for a total of 618 amino acids. The full-length epimerase predicts a protein of 70,099 Da with a relatively basic isoelectric point (pI ϭ 8.95). The protein contains a stretch of 17 hydrophobic residues located 11 amino acids from the amino terminus. As expected, PSORT predicts this as a transmembrane domain, and therefore the protein would most likely have a type II transmembrane topology like other enzymes involved in polysaccharide biosynthesis. The full-length clone contains three potential N-linked glycosylation sites at residues 225, 304, and 394, consistent with previous studies suggesting that the protein contained one or more Asn-linked chains. Using the extended sequence to perform a BLAST search of the GenBank TM data bases did not reveal additional homologs, suggesting that there may be only one heparin/heparan sulfate C5-epimerase (16).
A comparison of C. elegans, D. melanogaster, mouse, bovine, and human epimerase sequences indicated weak homology in the amino-terminal domain (residues 1-171) followed by a region of high identity (62%, residues 172-223). However, neither of these regions seems critical for catalytic activity because the purified protein from bovine liver was truncated at residue 248 (6). The COOH-terminal domain was also highly conserved across phylogeny (60% identity, residues 497-618), suggesting that this may represent an important functional part of the protein.
Expression studies of the truncated and full-length epimerase were undertaken to determine whether the additional 174 amino acids had any effect on catalysis. Initial attempts to assay the basal enzyme activity in cell extracts prepared from CHO and MST cells met with limited success. Expression of the truncated or full-length enzyme gave variable results, suggesting that the assay originally described for the bovine enzyme might not be optimal for cultured cells or recombinant enzymes expressed in cultured cells (4,5). Variation of each component improved the activity additively. Monovalent salts were inhibitory in contrast to previous findings (Fig. 3A) (5, 7). Divalent The triangle indicates the putative initiating Met residue in the original bovine sequence (6). The asterisks indicate conserved lysine residues in the COOH terminus that may be involved in catalysis (17).
cations, such as Ca 2ϩ or Mg 2ϩ , and EDTA (up to 40 mM) had no effect (4,7). The activity was highly dependent on detergent, even in sonicated extracts, with maximal effects obtained with 0.1% Triton X-100 (Fig. 3B). However, other detergents inhibited the reaction, suggesting that the effect was not merely because of solubilization of the protein from membranes. The pH optimum was ϳ7.0, which is in general agreement with previous findings (Fig. 3C) (4, 5, 7), but the activity showed marked sensitivity to the type of buffer (Fig. 3D). HEPES was found to be optimal. Under the revised conditions, the reaction was proportional with time for over 2 h and with protein concentration in the range of 1-30 g. The K m of the enzyme for the N-deacetylated/N-sulfated heparosan was estimated to be 25 M GlcA equivalents (ϳ500 pmol of GlcA/assay). With 300 pmol of GlcA/assay, a 4.5-fold increase in epimerase activity in MST cell extracts and a 6.5-fold enhancement in CHO cell extracts were observed compared with the original conditions (Fig. 3E). At lower concentrations of substrate, the difference was even more dramatic (data not shown).
Transient transfection of CHO cells revealed 4-fold greater activity associated with the full-length protein compared with the truncated enzyme in the revised assay (Fig. 4). Increasing the substrate 10-fold did not enhance the rate of reaction for either recombinant enzyme (data not shown). These findings indicated that the natural amino terminus was not a prerequisite to detect activity, which is consistent with previous find -FIG. 2. Full-length cDNA of murine C5-epimerase. The isolation of the cDNA encoding the full-length epimerase is described in the text. The revised sequence adds another 522 bases of coding nucleotides that translate into 174 additional amino acids on the amino terminus. In addition, the cDNA contains a 231-base 5Ј-UTR. The transmembrane domain identified by PSORT is enclosed by a box extending from residues 12-28. The initiating methionine residue indicated in the previously published bovine cDNA is enclosed by a circle (residue 175), and the amino-terminal proline residue of the purified bovine epimerase is enclosed by a box at residue 248. Three potential Asn-linked glycans may be attached at residues 225, 304, and 394 (shaded boxes). The GenBank TM accession number for the murine cDNA is AF325532.
ings showing that the truncated protein purified from liver and mastocytoma had substantial activity (4,5,12,17). Extracts prepared from cells transfected with epimerase containing a COOH-terminal GFP or YFP tag were analyzed by Western blotting with an anti-GFP monoclonal antibody. As shown in the inset of Fig. 4, the tagged full-length protein was present at higher levels than the truncated enzyme. Both forms were engineered into a near perfect Kozak sequence in the expression vector, suggesting that their expression was similar. Thus, we believe that the lower amount of the truncated enzyme was caused by decreased stability. As shown below, the truncated enzyme was also mislocalized, which may add to its instability. Thus, the amino-terminal domain does not seem to enhance the intrinsic activity of the enzyme.
Fusing c-myc or GFP to the COOH terminus resulted in a dramatic reduction of enzyme activity (Ͻ1 pmol/min/mg), but when a c-myc tag was placed at the amino terminus, enzyme activity was normal (23 pmol/min/mg versus 26 pmol/min/mg, respectively). These findings suggested that the highly con-  3. Effects of salt, detergent, pH, and buffer on epimerase activity. Cell extracts prepared from MST cells (A-D) were assayed for served COOH terminus plays an important role in binding, conformation, or catalysis. Recent investigations into the catalytic mechanism of the C5-epimerase implicated two polyprotic bases in the proton exchanges at C-5 (17) that are possibly mediated by two lysine residues. Interestingly, two lysine residues (amino acids 547 and 616) in the COOH-terminal domain of the epimerase are highly conserved across phylogeny (Fig. 1, asterisks). Adding GFP to the COOH terminus of other enzymes involved in heparin/heparan sulfate biosynthesis does not result in loss of activity (18). 2 The Full-length Epimerase Localizes to the Golgi-To study the intracellular localization of epimerase, cDNAs encoding the truncated and full-length enzymes were fused to GFP or c-myc and expressed in CHO cells. Full-length epimerase was located in a juxtanuclear position, co-localizing with the Golgi marker, ␣-mannosidase II (Fig. 5, A-C). This localization was observed with tags on either the C or amino terminus, indicating that the location of the tag did not interfere with subcellular localization signals in the protein (Fig. 5E). When the truncated epimerase was expressed, it behaved as a soluble protein exhibiting diffuse cytoplasmic staining (Fig. 5D). This is not an unexpected result given that the protein lacks a signal peptide.
The mislocalization of the truncated enzyme may act to destabilize its structure and activity (Fig. 4, inset).
Future studies of the epimerase will be greatly expedited by having the full-length sequence. Interestingly, very little information is available about the function of IdoUA in the biological activity of heparin and heparan sulfate. In general, it is assumed that the greater conformational flexibility of IdoUA will enhance the binding opportunities for heparin and heparan sulfate (19). The best studied example is the interaction of antithrombin with a heparin pentasaccharide, in which a critical IdoUA residue located to the reducing side of a central 3-O-sulfated glucosomine unit confers high affinity binding to antithrombin (20). Fibroblast growth factor-2 also apparently requires at least one IdoUA unit for binding and activation (21,22). In the former case, the addition of the 2-O-sulfate group to the IdoUA residue seems to be dispensable (20,23), whereas in the latter it is essential for binding (24 -26). These findings suggest that in some cases the IdoUA may play a direct role in binding to the ligand, whereas in others it may simply serve as a scaffold for placement of a critical sulfate residue. In both cases, the epimerase plays an essential role in creating the preferred binding site for the ligand. With full-length recombinant enzyme now available, it should be possible to engineer binding sites in isolated oligosaccharides and to explore the function of epimerase in vivo by creating mutants in cells and model organisms.