XT-II, the Second Isoform of Human Peptide-O-xylosyltransferase, Displays Enzymatic Activity*

Peptide O-xylosyltransferase (EC 2.4.2.26) is the first enzyme required for the generation of chondroitin and heparan sulfate glycosaminoglycan chains of proteoglycans. Cloning of cDNAs has previously shown that, whereas invertebrates generally have a single xylosyltransferase gene, vertebrate genomes encode two similar proteins, xylosyltransferase I and II (XT-I and XT-II). To date, enzymatic activity has only been demonstrated for the human XT-I, Caenorhabditis SQV-6, and Drosophila OXT isoforms. In the present study, we demonstrate that a soluble form of human XT-II expressed in the xylosyltransferase-deficient pgsA-745 (S745) Chinese hamster ovary cell line is indeed capable of catalyzing the transfer of xylose to a variety of peptide substrates; its enzyme activity was also proven using a Pichia-expressed form of XT-II. Its pH, temperature, and cation dependences are similar to those of XT-I expressed in either mammalian cells or yeast. Our data suggest that XT-I and XT-II are, at least in vitro, functionally identical.

Morphogenesis is a highly complex process in animal development and requires the creation of growth factor gradients for its control. Central to the correct formation of such gradients and to the associated signal transduction pathways are the glycosaminoglycan chains, both chondroitin and heparan sulfates, of various proteoglycans (1)(2)(3)(4). In the past few years, a number of mutations affecting glycosaminoglycan biosynthesis have shown the key role of these macromolecules. Particularly striking is the series of Caenorhabditis elegans squashed vulva (sqv) mutants (5,6); each of these mutants is defective in a different gene involved in the formation of chondroitin sulfate. In mammals, the EXT1 and EXT2 genes involved in heparan sulfate biosynthesis are implicated in brain morphogenesis (7) and in the bone disease hereditary multiple exostoses (8,9), whereas a form of heparan sulfate, heparin, acts as an anticoagulant by activating antithrombin (10).
The initial step in the biosynthesis of both chondroitin and heparan sulfates begins with the transfer of xylose to serine residues of a proteoglycan core protein (11). Although this enzyme activity was first described in the 1960s (12)(13)(14), it was first purified some 35 years later (15), and the first cDNAs were cloned from mammalian species (16). Subsequently, homologues were identified in Drosophila, encoded by the oxt gene (17), and in Caenorhabditis, encoded by the sqv-6 gene (18,19). Both of these invertebrate species only possess a single xylosyltransferase gene; however, vertebrates, starting with fish, appear to have two genes (XYLT1 and XYLT2) encoding highly homologous enzymes named XT-I 3 and XT-II. Of vertebrate forms, enzymatic activity has been demonstrated for human XT-I (16).
Recently, it has been reported that polymorphisms exist within both human xylosyltransferase genes. A variety of these have been correlated with blood pressure levels in diabetic patients, with non-diabetic nephropathy, with osteoarthritis, and with symptoms in pseudoxanthoma elasticum (20 -22). Both genes are, as judged by reverse transcription-PCR with mRNA from eight tissues, widely expressed (15), and green fluorescent protein-tagged forms of both enzymes were found to be targeted to the Golgi (23), a location consistent with some earlier cell fractionation studies (24). However, the enzymatic function of XT-II has remained elusive. Preliminary data in this laboratory suggested a rather minor activity in the culture supernatant of Pichia pastoris transformed with an XT-II construct; thus, as an alternative, we decided to attempt the expression in a mammalian cell line lacking xylosyltransferase activity (25). Indeed, for the first time, we report that human XT-II is an active enzyme with properties not significantly different from those of XT-I.

EXPERIMENTAL PROCEDURES
Vector Construction and Transfection-Partial reading frames of human XT-I and XT-II were isolated by PCR using the forward primers HsXTI/1/FseI (5Ј-cccacgaaggccggcctaagtgtgacatctcag-3Ј) or HsXTII/1/FseI (5Ј-cccacgaaggccggcccaagtgcgagatcgt-3Ј) and the reverse primers HsXTI/2/NotI (5Ј-ttttccttttgcggccgcactcctcggtgcccag-3Ј) or HsXTII/2/NotI (5Јttttccttttgcggccgctggggccctgctac-3Ј) with pGEM-T vectors previously prepared using fragments generated by PCR of human embyronic kidney (HEK-293) cell line cDNA (19) as templates and KOD polymerase (Novagen). Purified PCR fragments were then cut with FseI and NotI and ligated into the pPA-TEV vector (26), which was the kind gift of Dr. James Rini, University of Toronto. This vector, derived from pIRESpuro3 (Clontech), is intended for the expression of secreted protein A fusion proteins under control of the cytomegalovirus promoter.
The inserts of selected clones were sequenced in their entirety using the BigDye kit (Applera). Verified clones encoding XT-I and XT-II, as well as an empty vector control, were transiently transfected into the pgsA-745 (S745) Chinese hamster ovary cell line (kindly supplied by Dr. Jeffrey Esko, University of California, San Diego, CA; Ref. 25) using Lipofectamine 2000 reagent (Invitrogen). Independent transfections of XT-II were performed on separate days. Cells were cultivated for 24 h, and the culture supernatants were collected, stored at 4°C and, as required, concentrated using Ultrafree microcentrifuge devices with an M r 30,000 cut-off (Millipore). Protein content in the supernatants was determined using the MicroBCA protein assay kit (Pierce).
The construction of the pPICZ␣C vector with the XT-II insert, transformation of P. pastoris strain GS115, and induction of expression with methanol has been described previously (19). Supernatants were concentrated 10-fold using Ultrafree microcentrifuge devices, diluted with 10 mM HEPES, pH 8, containing 0.01% (w/v) sodium azide and 1 mM phenylmethylsulfonyl fluoride, and again concentrated prior to storage at 4°C.
Western Blotting-Aliquots of concentrated supernatants, equivalent to 12.5 l of original culture medium, were subject to SDS-PAGE (10% gel) and Western blotting using rabbit antiprotein A (Sigma, 1:2000) and alkaline phosphatase-conjugated goat anti-rabbit (Vector Laboratories, 1:2000). SigmaFAST tablets were used to prepare the chromogenic substrate for detection. Aliquots of the supernatants were also digested overnight with PNGase F (1 unit, Roche Applied Science) at 37°C in 10 mM HEPES, pH 8, prior to SDS-PAGE.
Assay of Xylosyltransferase Activity-Culture supernatants were assayed using the unmodified syndecan peptide (DDDS-IEGSGGR), the bikunin peptide (QEEEGSGGGGQR), or a dansylated form of the perlecan peptide (DSISGDDLGSGDLGSG-DFQR) as substrates. Dansylation was performed as follows: 3 mg of peptide were dissolved in 7 ml of Li 2 CO 3 (40 mM, pH 9.5); thereafter, 40 mg of dansyl chloride in 16 ml of acetone were  added, and the mixture was incubated in a water bath for 60 min at 37°C. After the addition of 14 l of ethylamine (70% (w/v)), the sample was subject to rotary evaporation and purified by gel filtration (Sephadex G25 fine, 1 ϫ 50 cm in 25% methanol) and RP-HPLC. Detection of the dansylated product was performed using fluorescence at 315/550 nm. The syndecan (1 mM final concentration), bikunin (1 mM) or dansylated perlecan (30 M) peptides were then incubated with supernatants of xylosyltransferase-expressing cells in the presence or absence of 1 mM UDP-Xyl, and the products were analyzed by MALDI-TOF MS (Perspective Biosystems Voyager-DE STR work station) as described previously (17,19) or by RP-HPLC at a flow rate of 1.5 ml/min (Hypersil ODS). In the case of syndecan, the RP-HPLC gradient was of 11.4 -13.3% acetonitrile from 5 to 13 min with UV detection at 214 nm (19), whereas for dansylated perlecan, a gradient of 34.2-41.8% acetonitrile was applied from 6 to 22 min with fluorescence detection at 315/550 nm.
Further Characterization of XT-I and XT-II-The enzymatic parameters of both human xylosyltransferases expressed in pgsA-745 cells were examined using the aforementioned RP-HPLC method with the syndecan peptide as substrate. For the pH dependence, HEPES and AMPD buffers of various pH values (6.5-9.0) were used. Various cations were tested at a final concentration of 10 mM; the effects of temperature during the reaction and of substrate concentration were examined using incubations containing 40 mM HEPES, pH 8, and 10 mM MnCl 2 . All assays were performed in duplicate and within the linear range of product formation with respect to time. Relative activities (for pH, temperature, and cation dependence) were determined based on HPLC peak areas (for the product peak and the sum of product and substrate peaks); the percentage conversion of product was then converted to an activity relative to that under the condition giving the highest rate of conversion to product. Specific activities were also calculated based upon the amount of substrate, as determined by the relative HPLC peak areas for product and substrate, converted within the linear time range, in the presence of Mn(II) at 37°C, and are expressed in terms of milliunits/mg, where 1 milliunits/mg corresponds to 1 nmol substrate turned over per minute per mg of protein in the culture supernatant.

RESULTS AND DISCUSSION
Expression of Xylosyltransferases-In previous studies, we have found that forms of Caenorhabditis SQV-6, Drosophila OXT, and human XT-I could be expressed using P. pastoris in an active form (17,19). These forms were truncated so as to delete the cytosolic, transmembrane and stem regions but retained the Pro-Xaa-Cys-(Asp/Glu) motif noted as being the most N-terminal conserved region in invertebrate and verte- brate peptide O-xylosyltransferases (11); the two human xylosyltransferase sequences display 61% identity (75% similarity) in the region beginning with this motif. However, although the expression in Pichia was easy to perform, the HPLC-and MSbased assays using the culture supernatants were sometimes complicated by the peptones in the medium and, furthermore, a barely detectable activity was found in preliminary trials with human XT-II. Considering that there was no published record of XT-II displaying activity, we decided that another system should be used to determine whether this protein is indeed enzymatically active. Previously, human XT-I has been expressed in an active soluble form in wild-type Chinese hamster ovary cells (15) and insect cells (27), as well as in a full-length form in HEK-293 cells (23). On the other hand, others have used the Caenorhabditis enzyme to rescue the defect in the pgsA-745 xylosyltransferase-deficient Chinese hamster ovary cell line (18). Specifically, an unknown genetic defect in this cell line results in a xylosyltransferase activity in crude cellular extracts at least 15-fold lower than in the parental line and, as a result, in an incorporation of sulfate into glycosaminoglycan that is 2% of normal levels (25). Therefore, as an alternative to Pichia while still using a xylosyltransferase-deficient host, we generated mammalian expression plasmids containing the same insert sequences as the previously used yeast vectors and then employed these to transfect pgsA-745 cells.
Upon transient expression, Western blotting of culture media indicated that for both XT-I and XT-II, bands of M r ϳ105,000 were detected after PNGase F digestion. This is consistent with the expected size of the tagged soluble enzymes (Fig. 1A); the tag contains residues 53-306 of protein A corresponding to M r ϳ30,000, and the soluble form of either enzyme theoretically confers M r ϳ80,000 to the overall size. The reduction in apparent M r and the narrowing of the bands after PNGase F would be consistent with the presence of predicted N-glycosylation sites within the sequences of both the xylosyltransferase and the protein A domains; both the XT-I and the XT-II constructs contain the same number of predicted N-glycosylation sites. However, due to the low percentage change in molecular mass at this high range, no conclusion can be made as to the exact numbers of occupied N-glycosylation sites.
As expected, transfection with the vectors encoding either human XT-I (Fig. 1B) or Drosophila OXT (data not shown) resulted in the appearance of xylosyltransferase activity in the media as judged by MALDI-TOF MS using the syndecan peptide; this activity was absent from the empty vector control (Fig.  1D). Furthermore, transfection of human XT-II into these cells also resulted in a distinct xylosyltransferase activity (Fig. 1C). Transfer of xylose to the peptide was also observed by RP-HPLC (Fig. 2); based on the latter assays, the specific activities in the supernatants were estimated to be 0.55 and 0.23 milliunits/mg for XT-I and XT-II, respectively. The lower specific activity of XT-II is consistent with the somewhat lower intensity of the XT-II band in the Western blot.
Enzymatic Characteristics of XT-I and XT-II-Using the HPLC-based method with the syndecan peptide, we examined the various characteristics of XT-II and compared them with those of XT-I. As shown in Fig. 3A, both XT-I and XT-II displayed significant activity in the range 16 -37°C; the optimum of 37°C is the same as that of the Pichia-expressed XT-I (19) and is as expected for a mammalian enzyme. The pH optima of both enzymes with both buffers employed was fairly broad but with a peak around pH 7.5-8.0 in both cases (Fig. 3B); both In all cases, a secondary reaction resulting in a shift in m/z of Ϫ17 (species of 1173 and 1305) was also observed, as described previously (19), and is presumed to be due to the conversion of the N-terminal glutamine residue of both the acceptor and the product to pyroglutamate. This modification has no obvious effect on the xylosyltransferase reaction. enzymes displayed significant activity at pH 6.5, which contrasts with the apparent lack of activity of the Pichia-expressed XT-I using AMPD at pH 6.5 (19). There is no difference, however, between the actual optima of XT-I expressed in the different systems. These values, however, are still somewhat higher than that which may be expected for a Golgi enzyme, considering that the pH inside this organelle has been estimated in one study as being pH 6.2 Ϯ 0.4 (28). Others have found pH optima for natural forms of mammalian xylosyltransferases of between pH 6.5 and 7.5 (29 -31), although it is not known whether these native forms were mixtures of both XT-I and XT-II or not.
The cation dependence was also broadly comparable for XT-I expressed in both the Pichia and the mammalian systems (Fig. 3C); the highest activities were again observed with Ca(II), Co(II), Mg(II), and Mn(II). For XT-II, the pattern was similar except that the activity in the presence of Co(II) was lower than that in the presence of EDTA and that XT-II, in contrast to the XT-I expressed in the mammalian system, is almost completely inhibited by Ni(II). Both enzymes were still active in the absence of any added cation and in the presence of EDTA. Recently, this lack of a necessity for a divalent cation has been structurally explained for an enzyme, core 2 ␤1,6-N-acetylglucosaminyltransferase, from the same glycosyltransferase family (CAZy family 14); indeed, all members of this enzyme family lack the degenerate DXD sequence motif involved in metal ion binding and catalysis in other so-called GT-A glycosyltransferases (32). Although mammalian xylosyltransferase sequences actually contain an (Asp/Glu)-Trp-Asp sequence necessary for the activity of XT-I (33), it does not correspond to a canonical DXD motif. It is located in the unique C-terminal region, which displays no homology to other glycosyltransferases, and we assume that it does not play a direct role in the catalytic mechanism or in metal ion binding.
Finally, the substrate concentration dependence of XT-I and XT-II was examined (Fig. 3, D and E). The apparent K m values for UDP-xylose were determined to be in the region of 250 M; these values are somewhat higher than the value of 100 M found with the Drosophila enzyme using the same method (17) but similar to the 300 M estimated for the Pichia-expressed form of XT-I (data not shown). In comparison, values of between 10 and 180 M were found with various native mammalian enzymes (31, 34 -36), whereas ϳ1 M was found by others for a form of XT-I expressed in insect cells and assayed using a protein acceptor (33). For syndecan, the K m values were in the region of 1 mM, which are comparable with those found for the same substrate with Caenorhabditis and Drosophila xylosyltransferases expressed in Pichia (17,19). Considering that the various parameters, including K m values, for XT-I and XT-II are approximately the same, there is no obvious enzymological reason for the previous lack of success in detecting the activity of the latter; we can only assume that some aspect of the vector construction, exact insert sequence, promoter, medium or cell line employed plays a role.
Use of Alternative Substrates by XT-II-Previously, we have assayed recombinant xylosyltransferases from vertebrate (XT-I) and invertebrate (OXT and SQV-6) sources with a variety of peptide substrates: i.e. those based on portions of Drosophila syndecan, human bikunin, or human perlecan (17,19). As shown above, an initial hypothesis that XT-II may require some particular unique substrate (19) was disproved in the course of the present study since the syndecan peptide was a substrate when this enzyme was expressed in pgsA-745 cells. However, it was still of interest to test XT-II with the bikunin and perlecan substrates. Indeed, as for XT-I, transfer of xylose by XT-II was observed with the bikunin peptide as judged by the appearance of a peak with m/z 1322 (Fig. 4A). Furthermore, encouraged by the results with XT-II in the mammalian expression system, we reattempted the expression of XT-II in Pichia and, this time, a significant transfer of xylose to the bikunin peptide was also observed (Fig. 4B). In controls (empty vector or no donor), no species with m/z 1322 was present (Fig. 4, C and D). This result is an independent indication that XT-II is indeed truly active since yeasts are not known to possess an endogenous peptide O-xylosyltransferase.
In the case of the perlecan substrate, we decided to develop an HPLC-based assay. In this case, the peptide was modified by dansylation to yield a fluorescently tagged form, whose detection was not expected to be subject to interference by other substances in the culture medium. Initial tests with recombinant Drosophila OXT showed that the dansylated perlecan peptide was a suitable substrate for sensitive detection of xylosyltransferase activity (data not shown). Thus, we also tested this substrate with supernatants of cells transfected with the XT-I and XT-II constructs. In both cases, a number of products were observed by RP-HPLC (Fig. 5, A and C), which were not present in incubations with the control supernatants (Fig. 5, E and F) or in the absence of the donor substrate (Fig. 5, B and D). The various fractions were collected and analyzed by MALDI-TOF MS either separately (see annotations on Fig. 5, A and C) or as a pool (as shown for the incubation with XT-II; Fig. 5G).
In contrast to the transfer of only two xylose residues to the non-dansylated form of the perlecan peptide in incubations of either the mammalian-expressed XT-I and XT-II (data not shown) or the Pichia-expressed XT-I and OXT (19), up to three xylose residues were transferred by both XT-I and XT-II to the dansylated perlecan peptide; this corresponds to the number of Ser-Gly motifs in this peptide. We assume that the modification of the N terminus by the dansyl group facilitates transfer to the first Ser-Gly motif (this serine residue being the fourth residue of the peptide); in the case of the non-dansylated form xylosylated with Pichia-expressed enzymes, this residue does not, as shown by tandem MS, serve as a xylosylation acceptor site (19). These data may be compatible with the effects of peptide length on the degree of xylosylation previously noted by others (29,37) and show that relatively small changes in a peptide substrate can have significant effects on the results obtained.
Conclusions-In this study, we have shown that both isoforms of human xylosyltransferase, XT-I and XT-II, are indeed active enzymes when expressed in the Chinese hamster ovary mutant pgsA-745 cell line, thereby showing activity of the latter for the first time. The demonstration of XT-II activity also in recombinant Pichia confirms this result and suggests that the activity of the supernatants of the transformed pgsA-745 cells is indeed due to XT-II activity and not due to some indirect rescue of the endogenous defective xylosyltransferase gene(s). Thus, now knowing that both xylosyltransferase isoforms are active and assuming their relatively ubiquitous expression (15), it may be that many or most mammalian cells have two enzymes capable of initiating chrondroitin and heparan sulfate biosynthesis, although the situation in, for example, wild-type Chinese hamster ovary cells has been unknown. On the other hand, there is only one isoform of the next enzyme in the glycosaminoglycan biosynthetic pathway (38), ␤4Gal-TVII, and it is interesting to note that some cases of Ehlers-Danlos syndrome, a connective tissue disorder, are associated with mutations in the B4GALT7 gene, which result in reduced enzymatic activity and defective proteoglycan glycosylation (39). This raises the issue as to whether mutations in one xylosyltransferase gene will result in strong phenotypes or whether only ablation of both will result in lethality in mammals. This is in contrast to the situation in invertebrates, which generally only have a single xylosyltransferase gene, but may be akin to the situation with another Golgi enzyme pair, mannosidase II and mannosidase II x . In the case of the latter, although the single knock-outs do show either immunological or reproduction defects, both mannosidase genes must be ablated for either in utero or postnatal lethality to occur in the mouse (40). Perhaps the possession of two xylosyltransferases with similar activities is an evolutionary accident or is indeed a reflection of an essential role for proteoglycans in many developmental and physiological processes in mammals.