Biosynthesis of Dermatan Sulfate

We identified the gene encoding chondroitin-glucuronate C5-epimerase (EC 5.1.3.19) that converts d-glucuronic acid to l-iduronic acid residues in dermatan sulfate biosynthesis. The enzyme was solubilized from bovine spleen, and an ∼43,000-fold purified preparation containing a major 89-kDa candidate component was subjected to mass spectrometry analysis of tryptic peptides. SART2 (squamous cell carcinoma antigen recognized by T cell 2), a protein with unknown function highly expressed in cancer cells and tissues, was identified by 18 peptides covering 26% of the sequence. Transient expression of cDNA resulted in a 22-fold increase in epimerase activity in 293HEK cell lysate. Moreover, overexpressing cells produced dermatan sulfate chains with 20% of iduronic acid-containing disaccharide units, as compared with 5% for mock-transfected cells. The iduronic acid residues were preferentially clustered in blocks, as in naturally occurring dermatan sulfate. Given the discovered identity, we propose to rename SART2 (Nakao, M., Shichijo, S., Imaizumi, T., Inoue, Y., Matsunaga, K., Yamada, A., Kikuchi, M., Tsuda, N., Ohta, K., Takamori, S., Yamana, H., Fujita, H., and Itoh, K. (2000) J. Immunol. 164, 2565–2574) with a functional designation, chondroitin-glucuronate C5-epimerase (or DS epimerase). DS epimerase activity is ubiquitously present in normal tissues, although with marked quantitative differences. It is highly homologous to part of the NCAG1 protein, encoded by the C18orf4 gene, genetically linked to bipolar disorder. NCAG1 also contains a putative chondroitin sulfate sulfotransferase domain and thus may be involved in dermatan sulfate biosynthesis. The functional relation between dermatan sulfate and cancer is unknown but may involve known iduronic acid-dependent interactions with growth factors, selectins, cytokines, or coagulation inhibitors.

Proteoglycans consist of glycosaminoglycan (GAG) 5 chains covalently linked to core proteins. The GAGs play important roles in mediating bio-logical functions of proteoglycans, mainly due to their ability to interact with a variety of proteins (1)(2)(3)(4)(5). Chondroitin sulfate (CS)/dermatan sulfate (DS) proteoglycans carry GAGs composed of alternating units of GalNAc and GlcA, or IdoUA in the case of DS. The chondroitin/CS/DS family has been ascribed a variety of physiological/developmental effects that range from control of basic cellular processes such as cell division in Caenorhabditis elegans, scaffold functions in various types of connective tissue, to highly cell type-specific effects, as exemplified by the neurite outgrowthpromoting activity mediated by rare structures in cerebral CS/DS (6 -9). Protein ligands targeted by CS/DS include growth factors, modulators of blood coagulation, selectins, and chemokines. It is important to define the CS/DS structures involved in selective protein binding and to understand how they are generated. The IdoUA-containing domains of DS chains are of particular significance in this regard, because IdoUA residues endow conformational flexibility to the polymer, which is believed to facilitate protein interactions (10).
Biosynthesis of CS/DS involves initial formation of a precursor polysaccharide composed of alternating GlcA and GalNAc residues, which subsequently undergoes a series of modification reactions (11). Our previous studies established that the generation of IdoUA units in DS (12,13), as well as in heparin/heparan sulfate (14), occurs by C5-epimerization of a portion of the GlcA residues previously incorporated into the polysaccharide chain. Moreover, CS/DS polysaccharides are O-sulfated at C2 of GlcA and IdoUA and C4 and/or C6 of GalNAc (15). Notably, IdoUA units are consistently found adjacent to 4-O-sulfated GalNAc residues. The extent of these modifications varies between tissues and seems to be influenced by the core protein structure (16). In addition, both epimerization and sulfation can be affected by growth factors (17).
The enzymes involved in the biosynthesis of heparin/heparan sulfate and CS/DS have been cloned, all except the chondroitin-glucuronate C5-epimerase required for IdoUA formation in DS (in the following denoted DS epimerase). In this study, we therefore purified this enzyme ϳ43,000-fold from bovine spleen microsomes and identified by mass spectrometry a candidate protein, SART2 (squamous cell carcinoma antigen recognized by T cells 2) that had previously been cloned but not assigned any specific function. Transgenic expression of this protein in mammalian cell lines yielded a product with DS epimerase activity, capable of inducing IdoUA formation in exogenous chondroitin substrate. Moreover, cells transfected with DS epimerase synthesized dermatan sulfate chains with increased IdoUA content compared with mock-transfected control cells.

Analytical Methods
DS epimerase was assayed by its ability to release 3 H-labeled water from chondroitin substrate containing [ 3 H]GlcA residues, essentially as described (23), but with some modifications based on preliminary kinetics analysis of semipurified enzyme (data not shown). Enzyme samples, containing 3 mg of BSA as carrier unless otherwise stated, were desalted at 4°C on Sephadex G-25 columns (0.7 ϫ 3 cm), equilibrated with desalting buffer (20 mM MES (pH 5.5 at 37°C), 10% glycerol, 0.5 mM EDTA, 0.1% Triton X-100, 1 mM DTT, protease inhibitors). Incubations were performed in a 100-l final volume of 0.8ϫ desalting buffer, 0.5 mg of BSA, 2 mM MnCl 2 , 0.5% Nonidet P-40, and 30,000 dpm [5-3 H]dK4 or [1-14 C]dK4 (ϳ200 M HexA for both substrates). Incubations were done at 37°C for 2-14 h and stopped by boiling for 5 min, and samples were centrifuged. For quantitation of [ 3 H]water, 90 l of the incubations were transferred to a distillation tube containing 200 l of (unlabeled) water. After distillation (24), 200 l of the distillate was analyzed by scintillation counting. The assay is linear up to 3000 dpm of released 3 H.
Protein was estimated by the Bradford assay (Bio-Rad), using BSA as standard. HexA content was determined by the carbazole reaction (25).

Purification of DS Epimerase
Bovine spleen was obtained fresh from the local slaughterhouse and was processed immediately. All procedures were carried out at 4°C, and all buffers contained 1 mM DTT as reducing agent.
Step 1. Microsomal Preparation and Extraction-On day 1, three spleens were freed from surrounding fat tissue, cut into 1-2-cm cubes, washed twice with cold distilled water, and placed in homogenization buffer (20 mM MES, pH 6.5, 250 mM sucrose, 5 mM EDTA, protease inhibitors). Batches of 350 ϫ g of tissue were first homogenized without any added buffer in a food processor and were then rehomogenized three times after step additions of 350, 500, and 500 ml of buffer. The homogenate was centrifuged for 15 min at 8,000 ϫ g. The resulting supernatant was centrifuged at 38,400 ϫ g for 45 min. The 38,400 ϫ g pellet was extracted with 90 ml of solubilization buffer (20 mM MES, pH 6.5, 150 mM NaCl, 1 mM EDTA, 1% glycerol, 1% Triton X-100, protease inhibitors). The combined extracted material from 11 such preparations, in all corresponding to 3.9 kg of tissue, was subjected to two strokes with a 200-ml Potter device and centrifuged at 38,400 ϫ g for 45 min. The supernatant was collected. On day 2, the final supernatant from day 1 was further centrifuged at 125,000 ϫ g for 40 min. Approximately half of the resulting supernatant was clear and was collected, whereas the other half was turbid and was diluted 1:2 with solubilization buffer and recentrifuged, and the resulting clear supernatant was collected.
Step 4. dK4-Sepharose-The eluate from Step 3 was applied to a dK4-Sepharose column (2.6 ϫ 3 cm) at 2 ml/min. The column was washed with 10 bed volumes of Buffer A, 20 mM NaCl, 10 mM CHAPS, further washed with 10 bed volumes of Buffer A, 20 mM NaCl, and finally eluted with Buffer A, 150 mM NaCl. Fractions with epimerase activity were pooled, dialyzed versus Buffer A, 10 mM NaCl, and concentrated by applying them to a Mono-Q 5/5 column, run at 0.5 ml/min. After application, the column was inverted and eluted at a flow rate of 0.1 ml/min with Buffer A, 1 M NaCl.
Step 5. Superose 12-The 1-ml concentrated material from Step 4 was injected to a Superose 12 column, which was subsequently eluted at 0.05 ml/min with Buffer A, 150 mM NaCl. Fractions of ϳ0.3 ml were collected ( Fig. 2A) and analyzed for epimerase activity. Active fractions were pooled, diluted with Buffer A to a final NaCl concentration of 20 mM, and concentrated on a Mono-Q PC 1.6/5 column, operated as described above.
Step 6. Superose 12-The 0.2-ml concentrated material from Step 5 was injected to a second Superose 12 column, operated as above.
Step 7. Red-Sepharose-The most active fraction from the previous step was diluted with Buffer A to 100 mM NaCl; CHAPS was added to final 10 mM concentration, and the sample was batch-incubated with 50 l of fresh Red-Sepharose gel. After incubation for 2 h, the gel was washed five times with 500 l of Buffer A, 100 mM NaCl, 10 mM CHAPS, further washed five times with 1 ml of the same solution without CHAPS, and finally eluted with 10 ϫ 150 l of Buffer A, 2 M NaCl.
The ϳ0.6 million-fold purified analytical sample shown in Fig. 2B, lane 2, was prepared as above, with one modification; an initial elution step with Buffer A, 1 M NaCl preceded the final elution with Buffer A, 2 M NaCl. The final eluate contained ϳ10% of the epimerase activity incubated with the Red-Sepharose gel.

Tryptic Peptide Preparation
Material from the last purification step was precipitated with trichloroacetic acid, and the pellet was resuspended in reducing Laemmli buffer. SDS-PAGE was carried out on NuPAGE pre-made 10% acrylamide gels (Invitrogen) that were stained with Brilliant Blue G Colloidal staining solution (Sigma). Gel bands of interest were cut out, and the proteins were digested with trypsin (Promega) as described (26).
LC-MS/MS was performed using an LTQ ion trap (Thermofinnigan, San Jose, CA) with an electrospray voltage of 2 kV. The instrument was set up to perform one MS scan (400 -1600 Da) followed by three MS/MS analyses in a data-dependent mode with an intensity threshold of 15,000 counts. The repeat count was set to 3, the repeat duration to 30 s, and the exclusion duration was 60 s; and the exclusion list size was 100.
Data base search was carried out using SEQUEST (28), and the search results were further analyzed by peptide and protein prophet as described previously (29,30). The identified peptides and corresponding proteins were stored and analyzed in a MySQL data model, which was used to accommodate further analysis of the data (31).

Expression of SART2/DS Epimerase
SART2 human cDNA clone (IMAGE number 5272885) was obtained from RZPD, Germany, and subcloned into pcDNA 3.1/myc-His vector (Invitrogen) using XhoI and AgeI restriction sites (underlined in the primers) introduced by PCR using primers 5Ј-GATCCTCGAGATGAG-GACTCACACACGGGG-3Ј and 5Ј-GATCACCGGTACACTGTGAT-TGGGAACAAGA-3Ј, respectively. The insert was confirmed by sequencing. Ligation into the expression vector resulted in a construct with SART2 in-frame with a C-terminal His 6 tag. CHO-K1 cells, maintained in F12-K medium, 10% FBS, HFL-1 cells and 293HEK cells, both maintained in minimum Eagle's medium, 10% FBS, were grown in 6-well plates and transiently transfected with pcDNA-His or pcDNA SART2-His plasmid, using Lipofectamine 2000 reagent (Invitrogen), according to the manufacturer's instruction. After 48 h, cells were washed with phosphate-buffered saline and lysed in 20 mM MES, pH 6.5, 150 mM NaCl, 10% glycerol, 2 mM DTT, 1 mM EDTA, 1% Triton X-100, protease inhibitors. After 30 min at 4°C, cell lysate was centrifuged for 1 h at 20,000 ϫ g, and 200 l of the supernatant was desalted (without carrier BSA added prior to desalting), followed by determination of protein content and enzyme activity.

Preparation of Rat Tissue Lysates
DA strain rats, 6 weeks old, were sacrificed, and the organs were put in ϳ3-fold excess (v/w) lysis buffer (see above for buffer composition and lysate preparation). The predominantly muscular tissues heart, uterus, and skeletal muscle were ground and further homogenized by three Potter strokes. For the remaining soft tissues the use of the Potter homogenizer alone was adequate.

Product Analysis of [ 14 C]dK4 Incubated with DS Epimerase
Two strategies were adopted to analyze the reaction products. In one experimental set, samples of 15,000 dpm of [1-14 C]dK4, recovered after incubation with cellular lysate as described in Fig. 4A, were digested with chondroitinase AC-I, and the split products were analyzed by gel chromatography, as described below. Alternatively, the overall composition of enzyme-incubated [1-14 C]-labeled dK4 was determined by analysis of labeled HexA-aTal R disaccharides, generated by N-deacetylation followed by deaminative cleavage and reduction of the products essentially as described (20). Briefly, samples were subjected to hydrazinolysis (64% hydrazine, 36% H 2 O, 1% hydrazine sulfate) at 100°C for 16 h, and the product was reisolated on a PD-10 column eluted with water and then treated with HNO 2 at pH 3.9. The resultant disaccharides, representing 64% of the total radioactivity, were reduced with NaBH 4 and recovered by gel chromatography on a Superdex Peptide column, eluted at 0.3 ml/min with 0.2 M NH 4 HCO 3 . The labeled disaccharides were separated by paper chromatography (20).

Analysis of Cellular CS/DS
Transient transfection of 293HEK cells, grown in 75-cm 2 flasks was performed as described above, with pcDNA-His or pcDNA SART2-His plasmid. After transfection, cell were grown for 24 h in ordinary minimum Eagle's medium, washed once in sulfur-deprived Dulbecco's modified Eagle's medium, maintained for 2 h in 10 ml of sulfur-deprived Dulbecco's modified Eagle's medium, 10% FBS, and finally 100 Ci/ml 35 SO 4 was added. After an additional 24-h culture period, medium was collected and applied to a 2-ml DE52 column, equilibrated with 50 mM acetate, pH 5.5, 0.1% Triton X-100, 6 M urea. The column was washed with 30 bed volumes of equilibration buffer and then with 5 volumes of water, 5 volumes of 0.2 M NH 4 HCO 3 , and finally eluted with 10 volumes of 2 M NH 4 HCO 3 . Ten g of cold DS were added, and the samples were lyophilized and then subjected to alkaline ␤-elimination (50 mM KOH, 1 M NaBH 4 , 45°C for 16 h). GAG chains were reisolated using the DE52 column, operated as described above. The GAG chains were deaminated at pH 1.5 (32), and intact CS/DS chains were recovered after gel filtration. 35

RESULTS
Purification of DS Epimerase-Initial data base searches for clues to the DS epimerase coding sequence based on similarity with the HS or alginate C5-epimerases were unsuccessful, and we therefore decided to isolate the DS epimerase protein. Screening various rat tissues for enzyme activity pointed to spleen as the richest source of DS epimerase (Fig. 1). A survey of bovine tissues gave similar results (data not shown), and bovine spleen was therefore selected as starting material for purification of DS epimerase. The overall purification process (see "Experimental Procedures") is summarized in Table 1. Solubilized microsomes were first applied to Red-Sepharose gel, yielding a 10-fold purification with excellent recovery. A ConA-Sepharose column efficiently removed contaminants but allowed only 30 -40% recovery of epimerase activity. dK4-Sepharose provided consistent 3-fold purification, with high recovery. After concentration, the partially purified enzyme was applied to a Superose 12 column. Most of the protein emerged as high molecular weight complexes, but the epimerase fraction was more retarded and relatively homogeneous in size, being eluted 4-fold and purified with the peak of activity at the position of a 67-kDa marker ( Fig. 2A). Size fractionation was refined by reapplication of the active pool to the same column, thus yielding a sharp peak with 80% of the eluted activity in two effluent fractions, again with a 4-fold purification. The most active fraction was incubated, batchwise, with a small amount of Red-Sepharose gel. Most of the activity was recovered and 14-fold further purified. The complexity of this sample, purified altogether ϳ43,000-fold, was assessed on SDS-PAGE (Fig. 2B, lane 1). An 89-kDa band was considered of particular interest, because the (Colloidal Blue) staining intensity of this band in the eluted fractions from Superose 12 and Red-Sepharose correlated with epimerase activity (data not shown). Furthermore, by modifying the elution conditions of the last purification on Red-Sepharose (see "Experimental Procedures"), ϳ0.6 millionfold purified active material was obtained, at the expense of recovery. Silver staining after SDS-PAGE showed the 89-kDa band as a major component (Fig. 2B, lane 2).
Gene Identification-Seven SDS-PAGE Colloidal Blue-stained bands (Fig. 2B, lane 1) were trypsinized and subjected to electrospray ionization-based LC-MS/MS analysis. Data base search of the results against the NCBI nonredundant (NRP.nci.fasta.20041115) data base yielded 56 proteins identified with more than one peptide matched and with a protein prophet probability of 1.0 (29,30). Several proteins were identified in more than one band, decreasing the number of unique proteins identified to 33. The following four proteins were identified from the 89-kDa gel band, believed to contain the DS epimerase: human SART2 (UPTR:Q9UL01; 6 peptides), lactotransferrin (gi.7428768; 11 peptides), hu-k4 (Q92853; 2 peptides), and pld3 protein (O35405; 2 peptides). The seven LC-MS/MS runs were then searched against a data base generated from the merged output from two genome-wide gene-finder programs, GenScan and GeneId and protein sequences from Bos taurus at NCBI (www.ncbi.nlm.nih.gov). Again, the 89-kDa band yielded four proteins with more than one peptide and a protein prophet score of 1.0: SCAFFOLD270174.1 (11 peptides; SART2), gi.7428768 (12 peptides; lactotransferrin), SCAFFOLD796601.1 (2 peptides; ATPase), and SCAFFOLD172907.1 (2 peptides; prolylcarboxypeptidase). The identities of these scaffolds (indicated above in parentheses) were assessed by blast searches. In all, 18 peptides (14 unique peptides; see Fig. 2, B, lane 1, and C) with a peptide prophet probability score of Ն0.93 for all but one, and ranging from 8 to 24 amino acid residues, matched SART2 or the corresponding scaffold and are shown in Table 2 and supplemental data. The resulting protein coverage was 8% for human SART2 and 26% for the corresponding scaffold. Of the 33 unique proteins identified, SART2 was the only one without a known function.
SART2/DS Epimerase Expression-Full-length human SART2 cDNA ( Fig. 2C for amino acid sequence) was subcloned into an expression vector in-frame with a C-terminal His tag and transiently transfected into three different cell lines. SART2/DS epimerase transfection increased the DS epimerase-specific catalytic activity 2-, 3-, and 22-fold in cellular lysates of HFL-1, CHO, and 293HEK cells, respectively, compared with mock transfections (Fig. 3A). Assays of culture media showed essentially background levels of DS epimerase activity for SART2/DS epimerase-transfected HFL-1 and CHO cells, whereas medium from 293HEK cells displayed activity ϳ15-fold higher than that from mock-transfected cells (Fig. 3B). The transfection data unambiguously indicated that SART2 DNA encodes a protein that can release a volatile radiolabeled component from the C5-3 H-labeled dK4 polysaccharide substrate. Gel chromatography of similar incubations containing equal amounts of cell lysates from either mock-transfected or SART2/DS epimerase-transfected 293HEK cells showed that 3 and 46%, respectively, of the total 3 H emerged at the elution position of 3 H 2 O (Fig.  4A). Because the 3 H label was about equally distributed between the C-5 positions of the GlcA and GalNAc units (20), this result indicates that essentially all GlcA residues had been targeted by the transfected epimerase.
Effect of DS Epimerase on dK4 Incubated in Vitro-The 14 C-labeled polysaccharide products were recovered following incubation with lysates of either mock-transfected or SART2/DS epimerase-trans-  fected cells, and their HexA composition was analyzed in two different ways. One approach involved depolymerization of the polysaccharides by chondroitinase AC-I, which cleaves N-acetylgalactosaminidic linkages to GlcA but not to IdoUA, and the digests were analyzed by gel chromatography (Fig. 4B). The AC-I products of dK4 incubated with mock-transfected 293HEK cell lysate consisted of 99% disaccharides, ϳ1% tetrasaccharides, and Յ0.2% hexasaccharides. The corresponding proportions for a sample incubated with epimerase-transfected cell lysate were 85% disaccharides, 12% tetrasaccharides, and 2.5% hexasaccharides (Fig. 4B). The IdoUA contents calculated from these analyses were 0.6% of total HexA in a sample incubated with mock-transfected lysate, and 8% after incubation with epimerase-overexpressing cell lysate. In a second strategy, GalNAc residues in dK4 polysaccharide, incubated with mock-transfected or epimerase-transfected cell lysates, were N-deacetylated by hydrazinolysis and were then deaminated with nitrous acid (pH 3.9 procedure) to cleave the chains at the sites of N-unsubstituted GalN units. Disaccharides, recovered by gel chromatography, were separated into GlcA-anhydro-D-talitol and IdoUA-anhydro-D-talitol species by paper chromatography (designated GT and IT, respectively, in Fig. 4C). The IdoUA-containing disaccharide barely exceeded background in samples that had been treated with endogenous epimerase (mock-transfected) but increased to 9% in a sample incubated with overexpressed epimerase.
To exclude that the DS epimerase could have a depolymerizing activity on the substrate, 3 H-labeled dK4 was incubated with ϳ200-fold purified enzyme and its size verified on a Superose 6 column. The elution position of dK4 was unchanged after incubation with the enzyme (data not shown).  Table 1), was applied to a Superose 12 column, and effluent fractions were assayed for epimerase activity and protein. Arrows indicate elution positions of molecular weight markers. Active fractions were pooled as indicated and further purified. B, SDS-PAGE analysis of purified DS epimerase. Lane 1, the ϳ43,000-fold most purified preparative sample (see Table 1

Effects of DS Epimerase Overexpression on DS Biosynthesis-293HEK
cells were grown in the presence of inorganic [ 35 S]sulfate, and the isolated GAGs were treated with nitrous acid to eliminate any heparan sulfate present (see "Experimental Procedures"). The GAG resistant to cleavage, ϳ60% of initial labeled polysaccharide, was quantitatively converted into disaccharides upon digestion with chondroitinase ABC (data not shown) and thus was identified as galactosaminoglycan (CS/ DS). This material was digested with chondroitinase B lyase, which cleaves -GalNAc-IdoUA-sequences in DS chains but leaves -GalNAc-GlcA-sequences intact. DS chains isolated from DS epimerase-transfected cells yielded a conspicuous disaccharide peak that amounted to 18% of the total label, with smaller but significant proportions of tetraand hexasaccharides (Fig. 5A). Together, these fractions accounted for 23% of the total label of the digest. The corresponding sample derived from mock-transfected cells was more resistant to chondroitinase B digestion, the disaccharide peak representing 4% and the sum of the di-, tetra-, and hexasaccharides amounting to 7% of the total label. Disregarding the possible occurrence of nonsulfated disaccharide units, lacking radiolabel, calculations based on the di-, tetra-, and hexasaccharides resulted in IdoUA contents of 5% in CS/DS from control cells and 20% in CS/DS from epimerase-overexpressing cells. The large proportion of disaccharides compared with tetra-and hexasaccharides generated by digestion with chondroitinase B (Fig. 5A) suggested that the IdoUA residues were preferentially introduced in blockwise fashion by the transgenic DS epimerase. This conclusion was corroborated by digestion of the 35 S-labeled polysaccharide with chondroitinase AC-I that cleaves -GalNAc-GlcA-but not -GalNAc-IdoUA-sequences. Enzyme-resistant blocks, corresponding to the general structure, -GlcA-GalNAc-(IdoUA-GalNAc-) Ն3 in the parental chains, constituted 5% of the labeled CS/DS chains modified by endogenous DS epimerase and 22% of the chains produced by overexpressing cells (Fig. 5B). Notably, the ratio of tetrasaccharides (indicative of isolated IdoUA residues in intact polysaccharide) to larger oligosaccharides (Ն3 consecutive IdoUA-containing disaccharide units, i.e. block sequences) was reversed between the two samples. Chondroitinase AC-I-generated oligosaccharides larger than octasaccharides were quantitatively converted to disaccharides by treatment with chondroitinase B (data not shown), confirming their IdoUA content. In summary, analysis of DS chains produced by transfected cells conclusively established that the SART2 gene encodes a DS epimerase that is active also in the context of the biosynthetic cellular machinery.

DISCUSSION
DS epimerase was purified from bovine spleen, using the previously established radiochemical assay procedure to monitor the process. MS analysis of peptides derived from partially purified enzyme identified SART2 as a candidate protein. Overexpression of human SART2 cDNA in 293HEK cells resulted in a Ͼ20and Ͼ10-fold increase of DS epimerasespecific activity in cellular lysate and culture medium, respectively, compared with mock-transfected control cells. The product of the reaction catalyzed by the recombinant enzyme, acting on a chondroitin substrate,  was confirmed to be IdoUA residues by specific enzymatic digestion and identification of chemical degradation products. Based on these findings we conclude that the protein encoded by the SART2 gene is a DS epimerase. The designation SART2 relates to the previous observation that this protein is recognized by a subset of cytotoxic T lymphocytes in certain tumors (33) but does not define any function. We therefore propose to rename SART2 with the functional term chondroitin-glucuronate C5-epimerase (EC 5.1.3.19) or, in short, DS epimerase. Three classes of epimerases are known to act on HexA residues in polysaccharides as follows: HS epimerase, alginate epimerases, and DS epimerase. DS epimerase shows no significant homology with either one of the other two types of epimerases. In addition, it does not display the "consensus sequence" of six spaced amino acid residues revealed by hydrophobic cluster analysis in alginate epimerases and in the HS epimerase (34). It therefore appears likely that the DS and HS epimerases have evolved to similar functions by convergent evolution. The human DS epimerase gene (located at 6q22) includes six exons, and the HS epimerase gene (15q23) includes three exons (33,35). The DS epimerase Fractions were collected every minute and analyzed for radioactivity. 14 C-dK4 non-incubated control (OE) was incubated with mock-transfected (E) or DS epimerase-overexpressing (F) cell lysate. Elution positions of di-, tetra-, and hexasaccharides derived from dK4 are indicated by arrows. C, paper chromatography of disaccharides. 14 C-Labeled dK4, reisolated after incubation as in A, was extensively deacetylated, followed by deaminative cleavage at pH 3.9. The resulting disaccharides were recovered by gel chromatography and analyzed by paper chromatography. 14 C-dK4 was incubated with mock-transfected (E) or DS epimerase-overexpressing (F) cell lysate. Positions of GlcA-aTal R (GT) and IdoUA-aTal R (IT) are indicated by arrows.
is conserved between species; compared with the human sequence, the bovine (GenPept XP_591812.2, predicted sequence) protein shows 91% amino acid identity, mouse (GenPept NP_766096.1) 93%, chicken (Gen-Pept XP_419777.1; predicted sequence) 86%, and Zebrafish (GenPept CAI20604.1; predicted sequence) 65%. The 958-amino acid residue human DS epimerase has one putative transmembrane region at its N terminus (amino acids 8 -30) and two at its C terminus (amino acids 901-923 and 931-952; see Fig. 2C). An unexpected and highly interesting observation links the human DS epimerase with NCAG1 (SwissProt Q8IZU8), the protein product of the C18orf4 gene cloned in 2003 (36) and genetically linked to bipolar disorder. The DS epimerase sequence from amino acid 43 to 673 thus shares 50% identity and 66% similarity with NCAG1. Moreover, NCAG1 is recognized by protein family data bases, e.g. Pfam, as a chondroitin sulfate sulfotransferase with conserved 5Ј-and 3Ј-PAPS-binding sites. These homology observations suggest that that NCAG1 may be an enzyme with dual epimerase and O-sulfotransferase activities involved in DS biosynthesis. We are currently attempting to express the C18orf4 gene.
DS epimerase shares significant homology (21% identity and 36% similarity between amino acids 24 and 584) with a second protein, a bacterial putative alginate oligo-lyase (SwissProt Q8UBJ1). We have shown in this report that DS epimerase has no lyase, i.e. depolymerizing, activity. However, comparison of the two proteins might provide clues to the mode of action of DS epimerase. In fact, lyase and epimerase reactions are mechanistically related and are even expressed by a single protein, the alginate AlgE7, a single amino acid residue being essential for both activities (37).
The quantitative comparison of DS epimerase-specific activities in different organs (Fig. 1) showed that spleen was the richest site. Stomach, uterus, and ovary had approximately one-third of the specific activity of spleen; and kidney, lung, and liver had approximately one-tenth. Skeletal muscle, heart, and brain contained low but detectable activity, whereas serum contained none. Specific activities of the enzyme in lysates from different rat organs generally correlated with the ubiquitous but variable expression of a 4.0-kb SART2/DS epimerase human mRNA transcript (33).
Exhaustive treatment of chondroitin (dK4) substrate with recombinant DS epimerase resulted in conversion of only 8 -9% of the GlcA to IdoUA residues, largely as isolated units (Fig. 4). This result agrees with previous studies of solubilized microsomal enzyme, which showed a freely reversible reaction with favored retention of D-gluco configuration (20). Similar findings applied to solubilized HS epimerase (21), which, however, was found to act in essentially irreversible mode in the intact cell (22). The intracellular epimerase reactions are presumably closely associated with O-sulfation steps (12, 38 -40) that preclude "back epimerization" and thus promote generation of IdoUA-containing structures. Accordingly, transfection of DS epimerase into cultured cells resulted in formation of DS with sulfated IdoUA-containing disaccharide units in blockwise arrangement (Fig. 5).
SART2 was originally identified through its association with certain cancer forms. Western blot analysis thus showed expression of SART2 as a 100-kDa protein in squamous cell carcinoma (33), glioma (33), gynecological cancer (41), pancreatic cancer (42), colorectal carcinoma (43), and prostate cancer (44) but not in breast cancer cell lines or in any normal tissues (33). We assume that the apparent lack of the protein in normal tissues reflected the relative insensitivity of the Western blot method, such that only tissues with overexpressed SART2/DS epimerase would be detected. Nakao et al. (33) identified two peptides in the N terminus and one in the C terminus of SART2, which elicited a cytotoxic reaction in a HLA-A24-restricted cytotoxic T lymphocyte cell line. The authors hypothesized that SART2 may be targeted by cancer immunotherapy. Indeed, the three SART2 peptides originally identified were tested in phase I clinical trials in patients with hormone-refractory prostate cancer (44).
The overexpression of DS epimerase in various cancers raises questions about the pathophysiological role of DS chains (although we cannot exclude effects of the DS epimerase protein as such, which are independent of its function in DS biosynthesis). A tumor may be regarded as a specialized organ, in which growing cancer cells are embedded in a supportive stroma composed of cellular components and of a specialized extracellular matrix (45). Various DS-carrying core proteins have been reported overexpressed in different human cancers (46 -48), and several DS-dependent processes may be hypothetically related to tumor biology. Potentially relevant candidate protein ligands include heparin cofactor II (49), several growth factors (e.g. HGF (50), FGF-2 and FGF-7 (51)), and chemokines/cytokines (e.g. RANTES (regulated on activation normal T cell expressed and secreted) and interferon-␥ (9)). If DS epimerase can indeed be functionally linked to tumor biology, it might become a target for therapy.