Changing the Donor Cofactor of Bovine α1,3-Galactosyltransferase by Fusion with UDP-galactose 4-Epimerase

Two fusion enzymes consisting of uridine diphosphogalactose 4-epimerase (UDP-galactose 4-epimerase, EC 5.1.3.2) and α1,3-galactosyltransferase (EC 2.4.1.151) with an N-terminal His6 tag and an intervening three-glycine linker were constructed by in-frame fusion of the Escherichia coli galEgene either to the 3′ terminus (f1) or to the 5′ terminus (f2) of a truncated bovine α1,3-galactosyltransferase gene, respectively. Both fusion proteins were expressed in cell lysate as active, soluble forms as well as in inclusion bodies as improperly folded proteins. Both f1 and f2 were determined to be homodimers, based on a single band observed at about 67 kDa in SDS-polyacrylamide gel electrophoresis and on a single peak with a molecular mass around 140 kDa determined by gel filtration chromatography for each of the enzymes. Without altering the acceptor specificity of the transferase, the fusion with the epimerase changed the donor requirement of α1,3-galactosyltransferase from UDP-galactose to UDP-glucose and decreased the cost for the synthesis of biomedically important Galα1,3Gal-terminated oligosaccharides by more than 40-fold. For enzymatic synthesis of Galα1,3Galβ1,4Glc from UDP-glucose and lactose, the genetically fused enzymes f1 and f2 exhibited kinetic advantages with overall reaction rates that were 300 and 50%, respectively, higher than that of the system containing equal amounts of epimerase and galactosyltransferase. These results indicated that the active sites of the epimerase and the transferase in fusion enzymes were in proximity. The kinetic parameters suggested a random mechanism for the substrate binding of the α1,3-galactosyltransferase. This work demonstrated a general approach that fusion of a glycosyltransferase with an epimerase can change the required but expensive sugar nucleotide to a less expensive one.

Oligosaccharides are attractive targets for the development of new pharmaceuticals because of their important roles in cell recognition, cell signaling, and other biological processes (1,2). ␣-Gal 1 epitopes (Gal␣1,3Gal-terminated oligosaccharide sequences including di-, tri-, and pentasaccharides) have drawn increasing attention since it was discovered that the interaction of preexisting natural antibodies in human serum with this specific xenoactive oligosaccharide sequence on animal cells is the main cause of hyperacute rejection in xenotransplantation (3). ␣-Gal epitopes exist as glycolipids or glycoproteins on the cell surface of mammals other than humans, apes, and Old World Monkeys (4,5). The unique enzyme responsible for the formation of the terminal glycoside bond in nature is UDP-Gal:Gal␤1,4GlcNHAc ␣1,3-galactosyltransferase (␣1,3GalT), a protein that is absent in humans due to mutational inactivation of the gene (6,7). In contrast, humans produce a large amount of anti-Gal antibodies including IgG, IgM, and IgA isotypes (8).
The discovery of the interaction of anti-Gal and ␣-Gal epitopes has led to experimental attempts to overcome hyperacute rejection by either depleting the recipient's anti-Gal through ␣-Gal-immobilized affinity columns or antagonizing anti-Gal by infusing the recipient's body with soluble synthetic ␣-Gal oligosaccharides (9,10). However, such procedures require access to a substantial amount of ␣-Gal oligosaccharides as well as synthetically derived ␣-Gal analogs and mimetics. Due to the high cost associated with multiple protection and deprotection steps with a tedious separation procedure at each step in traditional chemical synthesis of oligosaccharides (11)(12)(13), the most practical production of ␣-Gal oligosaccharides is by glycosyltransferase-catalyzed enzymatic synthesis (14). Nevertheless, since sugar nucleotides required by glycosyltransferases are exceptionally expensive, much work has been focused on in situ sugar nucleotide regeneration through multiple-enzyme systems (15)(16)(17)(18) and on enzymatic transformation to inexpensive sugar derivatives. We have demonstrated that a number of ␣-Gal oligosaccharides can be synthesized by glycosyltransferase-catalyzed reactions with regeneration of UDP-Gal through a five-enzyme system (19). Such a multipleenzyme system undoubtedly increased the complexity of the reaction. We also showed that a simpler alternative was a two-enzyme system in which UDP-galactose 4-epimerase (GalE) converted relatively inexpensive sugar nucleotide UDP-Glc to UDP-Gal, and ␣1,3-galactosyltransferase carried out the subsequent glycosylation reaction (19). In order to further reduce the cost of ␣-Gal synthesis as well as to avoid multiple fermentation for enzyme preparations, in this work two bifunctional fusion proteins containing both GalE and ␣1,3GalT were constructed. GalE carries out the interconversion of UDP-Glc to UDP-Gal, and ␣1,3GalT catalyzes the transfer of galactose from UDP-Gal to N-acetyllactosamine or its derivatives (Fig.  1A). The overall function of the fusion proteins is the transfer of galactose from UDP-Glc to the acceptor (Fig. 1B). These fusion enzymes change the donor requirement of the ␣1,3GalT from UDP-Gal to UDP-Glc. Furthermore, the fused enzyme system that catalyzes a sequential reaction may also have a kinetic advantage over the mixture of two separated enzymes, since the product of one enzyme travels a shorter distance to be captured by the next enzyme (20,21).
Glycosyltransferases can be classified mechanistically into two major families, inverting glycosyltransferases and retaining glycosyltransferases, depending on the anomeric configuration at the reaction center. ␣1,3GalT belongs to the retaining family, since it transfers a galactose residue from an ␣-linked nucleotide diphospho sugar (UDP-galactose) to the acceptor, forming an ␣-linked product via the retention of the anomeric configuration. The reaction mechanism was proposed to consist of two nucleophilic substitutions at the sugar anomeric carbon including the transient formation of a glycosyl enzyme intermediate (22)(23)(24). However, the details of the catalytic mechanism of ␣1,3GalT remain unknown due to the unavailability of the three-dimensional structure of any retaining glycosyltransferase. The comparison of the kinetic parameters of two fusion proteins with reverse sequence as well as of the native ␣1,3GalT could provide insights into the mechanism of the ␣1,3GalT-catalyzed reaction.
Construction of Plasmids pET15b-f1 and pET15b-f2-Chromosomal DNA of E. coli strain K-12 was purified using the QIAamp tissue kit. All of the polymerase chain reactions were performed in a total 50-l reaction volume containing 5 l (0.2 g) of template DNA, a 1 M concentration of each of two corresponding primers, MgCl 2 (2.5 mM), 5 l of 10ϫ buffer B (100 mM Tris-HCl, pH 8.3, at 25°C, 500 mM KCl), 1 mM dNTPs, and 2.5 units of Taq polymerase. The reaction mixture was covered with 50 l of mineral oil and subjected to 30 cycles of amplification with an annealing temperature of 55°C in a Thermolyne Amplitron I thermal cycler (Barnstead Thermolyne Corp., Dubque, IA). Restriction digestions and DNA ligations were performed as directed by the enzyme manufacturers. The resulting plasmids were transformed into E. coli cloning strain DH5␣ and then expression strain BL21 (DE3). Selected clones were characterized by restriction mapping and DNA sequencing. The strategy for constructing plasmids for fusion enzymes is described in "Results." Overexpression and Purification of the Fusion Enzymes-The expression and purification of the fusion enzymes from the cell lysate were as described before (19). Briefly, the overexpression of the enzymes was induced by 0.4 mM isopropyl-1-thio-␤-D-galactopyranoside for 3 h at 37°C in a C25 incubator shaker (New Brunswick Scientific Co., Inc., Edison, NJ). The cell lysate and inclusion bodies were separated by centrifugation at 12,000 rpm for 20 min.
From cell lysate, the active enzymes were purified using a Ni 2ϩ -NTAagarose affinity column. After elution, the fractions containing the purified enzyme (detected by UV-visible spectrometry) were combined, and ammonium sulfate was slowly added with stirring to 80% saturation. The suspension was kept at 4°C for an additional 30 min and then centrifuged. The pellet was dissolved in 20 mM Tris-HCl, pH 7.9, and applied to a Superdex 200 preparation FPLC column preequilibrated with Tris-HCl buffer (50 mM, pH 8.0) containing 2-mercaptoethanol (10 mM), glycerol (10%), and NaCl (0.25 M). The column was then eluted with the same buffer, and fractions containing fusion protein activity were pooled and concentrated in a Centricon 3 concentrator (Millipore Corp., Bedford, MA).
To obtain the active fusion proteins from the inclusion bodies, the purification and refolding procedures were carried out as described in Ref. 25 except that a calculated E 280 nm 0.1% value of 1.6 (26) was used for the fusion enzymes for the determination of enzyme concentration.
SDS-PAGE-SDS-PAGE was performed in a 10% gel in a Mini Protein III cell gel electrophoresis unit (Bio-Rad) at DC 200 V. High range (40 -212 kDa) SDS-PAGE standards (Promega) were used as molecular weight standards, and the gel was stained with Coomassie Blue.
Assay for Fusion Enzymes-Enzyme assays for fusion proteins were performed at 37°C for 30 min in a final volume of 100 l containing Tris-HCl (10 mM, pH 7.0), MnCl 2 (10 mM), bovine serum albumin (0.1%), UDP-D-[6-3 H]galactose (0.3 mM) or UDP-D-[6-3 H]glucose (final specific activity of 1000 cpm/nmol), fusion enzyme (6.3 pmol), and acceptor (50 mM, lactose or lactose derivatives). Acceptor was omitted for blank. The reaction was stopped by adding 100 l of ice-cold EDTA (0.1 M). Dowex 1 ϫ 8 -200 chloride anion exchange resin was then added in a water suspension (0.8 ml, 1:1 (v/v)). After centrifugation, supernatant (0.5 ml) was collected in a 20-ml plastic vial, and ScintiVerse BD (5 ml) was added. The vial was vortexed thoroughly before the radioactivity of the mixture was counted in a liquid scintillation counter (Beckmann LS-3801 counter). One unit of fusion enzyme activity is defined as the amount of enzyme that catalyzes the transfer of 1 mol of galactose from UDP-Glc to lactose/min at 37°C.

RESULTS
Construction of pET15b-f1 and pET15b-f2-The bifunctional enzyme f1 was prepared by in-frame linking of the galE gene downstream to the truncated ␣1,3GalT gene, whose translational stop signal has been removed, through an in-frame linker (GGTGGAGGC) coding for three glycine residues. To obtain the plasmid pET15b-f1, the plasmids pET15b-␣1,3GalT (f1) encoding individual ␣1,3GalT and pET15b-galE (f1) encoding GalE were constructed first ( Fig. 2A). The gene for the ␣1,3GalT was amplified with two primers ␣1,3GalT-F-f1 (forward primer, containing restriction site NdeI) and ␣1,3GalT-R-f1 (reverse primer, containing restriction sites SpeI, SalI, and BamHI and the codons of the three-glycine linker). NdeI and BamHI restriction sites facilitated the insertion of the ␣1,3GalT gene into the multiple restriction sites in vector pET15b, while SpeI and SalI were for the introduction of the galE gene. Similarly, the galE gene was amplified by polymerase chain reaction from the chromosomal DNA of E. coli strain K-12 with primers galE-F-f1 (forward primer, containing NdeI and SpeI restriction sites) and galE-R-f1 (reverse primer, containing SalI and BamHI restriction sites). Plasmids pET15b-␣1,3GalT (f1) and pET15b-galE (f1) were obtained, respectively, by the digestion with NdeI and BamHI and consequent ligation with the pET15b vector. After confirmation by restriction mapping, both plasmids pET15b-␣1,3GalT (f1) and pET15b-galE (f1) were digested with SpeI and SalI. The 6.7-kb digested fragment of pET15b-␣1,3GalT (f1) and the 1.1-kb fragment of pET15b-galE (f1) were purified and ligated by T4 DNA ligase to form plasmid pET15b-f1.
Another way of fusing the ␣1,3GalT gene and galE gene together was to join the structural gene of ␣1,3GalT to the 3Ј-end of the galE gene. In this case, the translational stop signal of the galE gene was removed. A similar strategy was applied to obtain plasmid pET15b-f2 for fusion protein f2 (Fig.  2B). The primers in the construction of fusion proteins are listed in Table I. All of the plasmids contained a T7 promoter, a T7 terminator, and an ampicillin-resistant gene. The His 6 tag fused to the N terminus of the fusion proteins simplified the purification of enzymes by Ni 2ϩ -NTA affinity chromatography. The expressions of both f1 and f2 were under the control of the T7 lac promoter and induced by the addition of isopropyl-1thio-␤-D-galactopyranoside.
The plasmid pET15b-f1 or pET15b-f2 was transformed into E. coli cloning host DH5␣, and positive recombinants were transformed into E. coli expression host BL21 (DE3). Selected colonies were confirmed by restriction mapping and DNA sequencing. When pET15b-f1 was digested with NdeI and BamHI, two fragments with sizes of 5.7 kb (pET15b vector) and 2 kb (␣1,3GalT gene plus galE) were observed in agarose gel electrophoresis. When pET15b-f1 was digested with SpeI and SalI, two fragments with sizes of 6.6 kb (pET15b vector plus ␣1,3GalT gene) and 1100 bp (galE) were observed (Fig. 3A). Similarly, when pET15b-f2 was digested with NdeI and BamHI, two fragments with sizes of 5.7 and 2 kb were shown. If f2 was digested with SalI and BamHI, two fragments with sizes of 6.8 and 900 bp (␣1,3GalT gene) were shown (Fig. 3B).
Expression and Purification of Fusion Enzymes-Both fusion proteins f1 and f2 were efficiently expressed in E. coli expression host BL21 (DE3) and appeared as predominant bands corresponding to a molecular mass of 67 kDa, which were absent in the BL21 (DE3) transformed with the pET15b vector plasmid only (Fig. 4).  pET15b-f2 (B). The galE gene for UDP-Gal 4-epimerase was cloned directly from the genome of E. coli K-12 and constructed into pET15b vector. The gene of ␣1,3-galactosyltransferase was amplified from a previously constructed plasmid and ligated into the same pET15b vector. A, pET15b-f1 was produced by the insertion of galE to the 3Ј terminus of the ␣1,3galactosyltransferase gene. B, pET15b-f2 was produced by the insertion of the ␣1,3galactosyltransferase gene into the 3Ј-terminus of the galE gene.
failed. The successful refolding was achieved in the same buffer except that the dithiothreitol was substituted by a redox system containing 10 mM 2-mercaptoethanol and 1 mM 2-hydroxyethyl disulfide. The necessity of a redox system for generating active enzymes indicated the importance of disulfide bond formation during refolding of the fusion proteins.
A better expression level was achieved by lowering the temperature to 30°C and elongating the incubation time to 20 h after the addition of isopropyl-1-thio-␤-D-galactopyranoside. However, as the case under the normal expression conditions (37°C, 3 h), the amount of the proteins obtained from 1 liter of bacteria culture for f1 remained similar to that for f2, as de-termined by SDS-PAGE and the activity assay. These results indicated that the relative location of the transferase and the epimerase in the fusion enzymes was not critical for the expression level in the current overexpression system.
Purification of the fusion proteins was achieved by a nickel affinity chromatography and a subsequent FPLC HL Superdex 200 (16/60) gel filtration chromatography. After the affinity chromatography, the fusion proteins were purified to about 90% purity. Further purification was achieved for both f1 and f2 by FPLC gel filtration chromatography, which gave a single peak containing fusion protein activity in FPLC elution and a clean single band of about 67 kDa in SDS-PAGE (Fig. 4).
Molecular Mass Determination-The molecular weights of the fusion enzymes were estimated by gel filtration chromatography on a Superdex 200 column. Both hybrid enzymes eluted as single peaks corresponding to a molecular mass of 140 kDa, as determined by comparison with the protein standards performed under the same conditions. The molecular mass for either f1 or f2 determined by SDS-PAGE was estimated to 67 kDa (Fig. 4). The monomers of both fusion proteins have theoretical calculated molecular masses of 74 kDa. These results suggest that both f1 and f2 are homodimers with total molecular mass of about 140 kDa under native conditions. Apparent Kinetic Parameters for the Fusion Enzymes-The measured kinetic parameters in Table II reveal some properties of the fusion enzymes. First, the k cat values of the ␣1,3GalT moieties (UDP-Gal 3 Gal␣1,3Lac reaction) in f1 and f2 are similar to that of the native ␣1,3GalT. However, the K m value of f1 for lactose (8.5 mM) is decreased to about a half of that of the native ␣1,3GalT (15 mM). Therefore, the catalytic efficiency of the ␣1,3GalT moiety in f1 is higher (k cat /K m ϭ 0.053 s Ϫ1 mM Ϫ1 ) compared with that of f2 or of the native ␣1,3GalT. It is important to point out that the epimerase part of the fusion enzymes participated in the conversion of UDP-Gal to UDP-Glc, which decreased UDP-Gal concentration and slowed down the reaction, during the measurement of the ␣1,3GalT activity for f1 and f2. Therefore, the kinetic parameters shown in Table  II are only approximate determinations. Second, it is found that the catalytic efficiencies of the epimerase moieties (UDP-Glc 3 UDP-Gal reaction) in f1 and f2 are greatly reduced (ϳ25-fold) in comparison with that of the native GalE. Dissecting to individual parameters, the k cat value of f1 is similar to that of the native GalE, and the efficiency loss of the GalE in f2 mainly comes from the poorer binding to UDP-Glc. For f2, the decrease of the efficiency comes from both a lower k cat and a higher K m . Third, when considering the overall fusion enzyme activity in the coupled reaction (UDP-Glc 3 Gal␣1,3Lac reaction), the catalytic efficiency of f1 is about 8 times higher than that of f2. This is mainly due to the fact that f1 has a lower K m .
Proximity Effect-UDP-Gal 4-epimerase catalyzes the interconversion of UDP-Glc to UDP-Gal (Reaction 1). ␣1,3-galactosyltransferase catalyzes the addition of a galactose from UDP-Gal to its acceptor lactose (Reaction 2). The fusion enzymes contain both epimerase and transferase activities and will catalyze the transformation of galactose from UDP-Glc (instead of

5Ј-GGAATTCCATATGAGAGTTCTGGTTACC-3Ј
GalE-R-f2 from UDP-Gal) to lactose (Reaction 3). The net result is the change of donor requirement of ␣1,3GalT from UDP-Gal to UDP-Glc, which decreases the cost for the synthesis of ␣-Gal epitope and its derivatives more than 40-fold.
Due to their proximity effects, the fusion proteins may also have kinetic advantages over the system containing two separate enzymes in terms of catalytic efficiency. To test this hypothesis, the rates of the coupled reaction catalyzed by the same amount of f1, f2, or a combination of the two native enzymes were measured and compared under same conditions (Fig. 5). Significantly, the initial rates of catalysis by the fusion enzyme f1 and f2 were about 300 and 50% higher, respectively, than those of the native enzymes at all substrate concentrations. The activity improvements of the fusion enzymes suggested that proximity effects occurred in the fusion enzymes to increase the catalytic activities for the overall reaction. UDP-Gal produced by the epimerase should have traveled a shorter distance to be captured by the ␣1,3-galactosyltransferase active sites in the fusion proteins compared with that of the two-enzyme system.
Acceptor Specificity-The natural acceptor for native ␣1,3galactosyltransferase is N-acetyllactosamine. The recombinant ␣1,3-galactosyltransferase was shown to be able to accept a wide range of substrates including lactose and galactose derivatives but not glucose derivatives (29). In order to determine whether the acceptor specificity of the ␣1,3-galactosyltransferase had been changed after the fusion with epimerase, acceptor specificities were measured for fusion protein f2. Selected monosaccharides and disaccharides were used as the acceptors, and either UDP-Gal or UDP-Glc was used as the sugar nucleotide donor for the ␣1,3GalT activity and the overall coupled reaction.
As shown in Table III, the acceptor specificity of the ␣1,3galactosyltransferase moiety (UDP-Gal 3 Gal␣1,3Lac reaction) in f2 and that of the overall fusion enzyme f2 (UDP-Glc 3 Gal␣1,3Lac reaction) are similar to that obtained for the individual recombinant ␣1,3-galactosyltransferase. For example, lactose derivatives are acceptors for the hybrid protein f2 and the ␣1,3-galactosyltransferase moiety in f2 with the activities corresponding to the substitution of aglycones as N 3 Ͼ OH Ͼ SPh. The presence of axial configuration of hydroxyl group at the C-4 position of the Gal residue is important. Lactose and galactose derivatives are acceptors; however, glucose derivatives are not accepted by the enzyme. As in the case of the recombinant ␣1,3GalT, ␣or ␤-linkage of the Gal terminus at the nonreducing end was confirmed to be an important but not crucial factor for f2. All of the Gal ␤-linked compounds (entries  The assays are carried out in duplicate at 37°C for 30 min with UDP-Gal (0.3 mM) or UDP-Glc (0.3 mM) and fixed acceptor concentration (50 mM) in the presence of 10 mM MnCl 2 in a total volume of 0.1 ml.

Entry
Donor candidates ␣1,3GalT Activity a f2 Activity b a Assay using UDP-Gal as the donor, the acceptor specificity of the ␣1,3GalT moiety in f2.
UDP-Gal 4-epimerase is one of the key enzymes of the Leloir pathway for galactose metabolism (32). The epimerase from E. coli is a homodimer with an overall molecular mass of 79 kDa (33). The epimerase was cloned from E. coli K-12 into pET15b expression vector (27). FPLC gel filtration and SDS-PAGE data indicate that the recombinant epimerase is a homodimer. It is interesting to know that both fusion proteins f1 and f2 have dimeric configurations as that for the native epimerase. The relative position of the transferase and the epimerase does not alter this property. Also, the His 6 tag at the N terminus of protein does not prevent dimerization. Since the recombinant ␣1,3GalT is a monomer, the recombinant epimerase is a homodimer, it is reasonable to assume that, the interactions between the two subunits of the fusion proteins result from the interactions of the epimerase moieties in the fusion enzymes (Fig. 6). The same conclusion was made by Bulow and Mosbach (34) that a dimeric hybrid polypeptide ␣-␤␤-␣ was usually obtained by the fusion of a dimeric protein (␤␤) with a monomeric one (␣).
A previous report about a fusion protein composed of UDP-Gal 4-epimerase and galactose-1-phosphate uridylyltransferase (GalT) with an intervening Ala 3 linker was shown to exist in three forms: a monomer, dimer, or tetramer (20). Another fusion protein of ␤-galactosidase and galactose dehydrogenase was also shown to exist as two major forms, consisting of four or six subunits, as well as other forms (21). Unlike these reports, due to the monomeric nature of the ␣1,3GalT, both fusion proteins f1 and f2 in the present study exist as homodimers.
It was reported that each subunit of the dimeric epimerase contained one irreversibly, noncovalently bound NAD ϩ (33), and no additional NAD ϩ was required for the epimerase-catalyzed reaction (32). This is also the case for the fusion enzymes. No external NAD ϩ is necessary for either the coupled reaction or the activity of the epimerase moiety in the fusion enzymes. This indicates that NAD ϩ is also tightly bound to the epimerase during the expression of fusion proteins in E. coli.
Our kinetic studies indicate that the fusion proteins have kinetic advantages over the system of individual enzymes for the overall coupled reaction. Significantly, the reaction rates in producing Gal␣1,3Lac from UDP-Glc and lactose are increased 4-and 1.5-fold by f1 and f2, respectively, in comparison with the native enzymes. This result can be explained by the sub-strate channeling effect (21), indicating that the UDP-Gal produced by the epimerase is captured faster by the transferase moiety in either f1 or f2 than that in the system of native enzymes. Furthermore, the higher activity of f1 over f2 suggests that the active sites for the epimerase and ␣1,3-galactosyltransferase part are in a more optimal configuration in f1.
It is noticed that the overall fusion enzyme activities (UDP-Glc 3 Gal␣1,3Lac reaction) are higher than the activity of the corresponding transferase moieties (UDP-Gal 3 Gal␣1,3Lac reaction) in both f1 and f2. This provided us with an insight into the mechanism of ␣1,3-galactosyltransferase. The steady state kinetic properties of ␣1,3-galactosyltransferase indicate a sequential mechanism in which metal cofactor, donor, and acceptor bind enzyme prior to catalysis. The low level of UDP-Gal hydrolase activity found in ␣1,3-galactosyltransferase excluded the possibility of an ordered sequential mechanism with binding of acceptor prior to donor. Previous studies were not able to determine whether an ordered mechanism with the binding of donor prior to acceptor or a random mechanism applies for ␣1,3-galactosyltransferase. Our results on the kinetic parameters of coupled reaction and the native transferase reaction in both fusion enzymes suggest that a random mechanism should be a more plausible explanation. For instance, if ␣1,3-galactosyltransferase obeys an ordered mechanism with the binding of donor prior to acceptor, then in the absence of UDP-Gal (as in the coupled reaction), UDP-Glc has to be captured by the epimerase and be converted to UDP-Gal prior to the function of transferase. Opposite to the results we obtained, the initial rate in the coupled reaction should be lower than that in the activity assay for the ␣1,3-galactosyltransferase moieties in fusion enzymes, in which UDP-Gal is present and able to be captured and utilized by the transferase right away.
Beside N-acetyllactosamine, lactose and lactose derivatives are found to be good acceptors for the recombinant ␣1,3-galactosyltransferase. Our results indicate that the in-frame fusing with epimerase does not change the acceptor specificity of the transferase. This suggests that without changing many important properties of individual native enzymes, fusing two enzymes together has the advantages of simplicity, economy, and efficiency. In our laboratory, the fusion enzymes f1 and f2 have been successfully applied in the synthesis of ␣-Gal epitope and its derivatives (35,36). This novel approach reduces the cost for the synthesis of oligosaccharide by over 40-fold and provides an easy and economic access to a wide spectrum of ␣-galactosyl epitopes and their derivatives to support the continuous studies on xenotransplantation as well as other pharmaceutical research (37,38). The methodology can also be applied for other galactosyltransferases that require UDP-Gal as the donor, such as ␣1,4-galactosyltransferase, ␣1,6-galactosyltransferase, ␤1,4-galactosyltransferase, ␤1,3-galactosyltransferase, human blood type B galactosyltransferase, etc. From a broader point of view, depending on which sugar nucleotide is the most economical source, the fusion of a glycosyltransferase and an epimerase can change a required, but more expensive, sugar nucleotide to a less expensive one. For instance, broader application of such an approach may include the combination of a variety of glycosyltransferases with different epimerases, such as N-acetylgalactosaminyltransferase and UDP-N-acetylglucosamine 4-epimerase, galactosaminyltransferase and UDP-N-glucosamine 4-epimerase, or galacturonosyltransferase and UDP-glucuronate 4-epimerase.