The influence of an intramolecular hydrogen bond in differential recognition of inhibitory acceptor analogs by human ABO(H) blood group A and B glycosyltransferases.

Human ABO(H) blood group glycosyltransferases GTA and GTB catalyze the final monosaccharide addition in the biosynthesis of the human A and B blood group antigens. GTA and GTB utilize a common acceptor, the H antigen disaccharide alpha-l-Fucp-(1-->2)-beta-d-Galp-OR, but different donors, where GTA transfers GalNAc from UDP-GalNAc and GTB transfers Gal from UDP-Gal. GTA and GTB are two of the most homologous enzymes known to transfer different donors and differ in only 4 amino acid residues, but one in particular (Leu/Met-266) has been shown to dominate the selection between donor sugars. The structures of the A and B glycosyltransferases have been determined to high resolution in complex with two inhibitory acceptor analogs alpha-l-Fucp(1-->2)-beta-d-(3-deoxy)-Galp-OR and alpha-l-Fucp-(1-->2)-beta-d-(3-amino)-Galp-OR, in which the 3-hydroxyl moiety of the Gal ring has been replaced by hydrogen or an amino group, respectively. Remarkably, although the 3-deoxy inhibitor occupies the same conformation and position observed for the native H antigen in GTA and GTB, the 3-amino analog is recognized differently by the two enzymes. The 3-amino substitution introduces a novel intramolecular hydrogen bond between O2' on Fuc and N3' on Gal, which alters the minimum-energy conformation of the inhibitor. In the absence of UDP, the 3-amino analog can be accommodated by either GTA or GTB with the l-Fuc residue partially occupying the vacant UDP binding site. However, in the presence of UDP, the analog is forced to abandon the intramolecular hydrogen bond, and the l-Fuc residue is shifted to a less ordered conformation. Further, the residue Leu/Met-266 that was thought important only in distinguishing between donor substrates is observed to interact differently with the 3-amino acceptor analog in GTA and GTB. These observations explain why the 3-deoxy analog acts as a competitive inhibitor of the glycosyltransferase reaction, whereas the 3-amino analog displays complex modes of inhibition.

observations explain why the 3-deoxy analog acts as a competitive inhibitor of the glycosyltransferase reaction, whereas the 3-amino analog displays complex modes of inhibition.
The human blood group A and B oligosaccharide antigens are respectively formed by the transfer of GalNAc by glycosyltransferase GTA 1 or Gal by glycosyltransferase GTB to the common H-disaccharide ␣-L-Fucp-(132)-␤-D-Galp-OR, where R is a glycoprotein or glycolipid (1). Generally, humans with the gene for GTA have blood group A, those with GTB have blood group B, those with both genes have blood group AB, and those with neither have blood group O. Glycosyltransferases in general have been implicated as indicators of cancer progression, susceptibility to infectious diseases, glycoprotein activity, and heart and autoimmune diseases (for review, see Ref. 2), and the human ABO(H) blood group glycosyltransferases in particular are viewed as a model system for the study of action and specificity of this class of enzyme.
When the primary structures of GTA and GTB were determined, it was found that they differ in only four amino acid residues which, given that they share a common acceptor, were assumed to confer their ability to distinguish between the donor substrate molecules (3). X-ray crystallographic studies of the catalytic domains of the two enzymes revealed that their structures are almost identical outside of these four residues, that only two of these four residues served to distinguish between donor substrates, and that the acceptor substrate binding sites were nearly superimposable (4).
To assist in further elucidating the mechanisms of these glycosyltransferases, specific analogs of the acceptor molecules were made and characterized kinetically by their ability to inhibit the enzyme reaction. The targeting of the acceptor moiety at the 3-OH linkage site (the point at which the monosaccharide is transferred to the acceptor) of the Gal ring via modification to 3-deoxy and 3-amino analogs (5, 6) produced potent inhibitors of glycosyltransferase activity (Fig. 1). The 3-deoxy analog was found to be a competitive inhibitor of both GTA and GTB, with K i ranging from 14 to 68 M, and was shown to inhibit GTA in cell culture such that the expression of surface A antigen was significantly reduced (7). The K i of the 3-amino analog could not be determined, as the mode of inhi- bition for both GTA and GTB was observed not to fit standard models of inhibition. However, it was estimated that K i for the 3-amino analog is in the 200-nM range for GTA (5,6).
To understand the different behaviors of the two inhibitors and, specifically, the mode of binding of the 3-amino analog, we determined structures of both GTA and GTB in complex with the 3-deoxy and 3-amino analogs both in the absence and presence of UDP.

MATERIALS AND METHODS
Protein Production-Protein production was carried out as described in Ref. 8.
Protein Crystallization-Crystals of native GTA and GTB were grown as reported previously (4). Crystals were soaked with various combinations of UDP-GalNAc, UDP-Gal, UDP, and acceptor analogs. Soaking solution contained 7.5% polyethylene glycol 4000 (Sigma), 15% glycerol (BDH), 75 mM ADA buffer (Sigma), pH 7.5, 10 mM MnCl 2 (Fisher), and 10 mM acceptor for 3-4 days. UDP, UDP-GalNAc, or UDP-Gal (Sigma) was added to the soaking solution at a concentration of 10 -15 mM for 20 -25 h. At the end of the soaking period, crystals were frozen in liquid propane using magnetic crystal caps (Hampton Research), and the caps were stored in liquid nitrogen for transport to the beamline.
Data Collection and Structure Determination-Data was collected at beamline X8C at the National Synchrotron Light Source at Brookhaven National Laboratories at a wavelength of 1.15 Å. Two data sets FIG. 1. Chemical structure of the H-antigen acceptor analogs used to generate complexes with GTA and GTB, where R ‫؍‬ H is the 3-deoxy acceptor analog, and R ‫؍‬ amino is the 3-amino acceptor analog. RЈ is an aliphatic group used in the purification of the analogs. The native acceptor is given by R ϭ OH.   (GTAϩDI and GTAϩAI) were collected on a MAR 300 mounted on a Rigaku RU300 generator at Queen's University (Kingston, Ontario, Canada). All data were collected at low temperature using a Cryostream 600 cooler. Data was reduced and scaled with HKL2000 software (9). Initial rigid body refinement in CNS (10) was carried out by using the native GTA and GTB structures with and without H-antigen and UDP bound (PDB codes 1LZ0, 1LZ7, 1LZI, and 1LZJ). This procedure was followed by overall structural refinement using CNS. Leastsquared overlaps of structures were calculated by using LSQKAB within the CCP4 program suite (11). All overlaps shown are based on protein-protein overlaps to the GTB structure (PDB code 1LZJ). Diagrams were made using ChemSketch and SETOR (12).

RESULTS
Data Collection and Structure Refinement-Details of the data collection and structure refinement are shown in Tables I  and II. Data were collected to a maximum of 2.09 -1.55 Å resolution, with R and R free for the final models from ϳ0.188 -0.208 and from 0.209 -0.227, respectively. All structures show excellent electron density over the course of the polypeptide chain, with the exception of the disordered loop between residues 177-195 and the final 10 residues of the C terminus, which were also absent in the native structures (4). These data sets (see Table I) are GTA ϩ 3-deoxy inhibitor (GTAϩDI), GTB ϩ 3-deoxy inhibitor (GTBϩDI), GTA ϩ 3-amino inhibitor (GTAϩAI), and GTB ϩ 3-amino inhibitor (GTBϩAI). UDP was present in the soaking solution of four crystals, and UDP appears clearly in the corresponding electron density maps. These data sets (see Table II) are GTA ϩ 3-amino inhibitor ϩ UDP (GTAϩAIϩUDP), and GTB ϩ 3-amino inhibitor ϩ UDP (GTBϩAIϩUDP). GTA and GTB were also co-crystallized with 3-amino inhibitors in the presence of UDP-GalNAc and UDP-Gal, respectively, (GTAϩAIϩUDP-GalNAc and GTBϩAIϩ UDP-Gal). The structures have been deposited with the Protein Data Bank with accession codes 1R7T, 1R7U, 1R7V, 1R7X, 1R7Y, 1R80, 1R81, and 1R82, respectively.

Structural Analysis of 3-Deoxy Acceptor
Analog-The 3-deoxy acceptor analog of the H-disaccharide bound to both GTA and GTB in the absence of UDP (Fig. 2, a and b), which was surprising given that UDP is known to form an integral part of the acceptor-binding site (4). This 3-deoxy inhibitor has the same overall conformation and binding interactions observed for the GTA and GTB structures containing H-antigen and UDP ( Fig. 3a; Ref. 4). No significant changes in polypeptide structures were observed. The acceptor analog displayed the same position and orientation in the binding site as the native acceptor ( Fig. 4a; Ref. 4). As in the structures of the native enzymes, the aliphatic tail of the acceptor analog occupies a different specific location in GTA and GTB due to the presence of Ser-235 in GTB compared with the Gly-235 in GTA (Fig. 2, a   FIG. 2. Wire- This analog differs from the H-disaccharide acceptor by the lack of the 3-OH group on the Gal residue and so prevents the transfer of saccharide from donor; it was hoped that this difference would allow the co-crystallization of enzyme with acceptor analog and donor. However, experiments to co-crystallize UDP-GalNAc and UDP-Gal with GTA and GTB, respectively, did not yield any electron density corresponding to the donor sugar. (Electron density consistent with partial occupancy by UDP was observed similar to that described for the 3-amino inhibitor, below.) Structures of 3-Amino Acceptor Analog-Unlike the 3-deoxy inhibitor, the kinetic data regarding the 3-amino analog indicated that there was a complex mode of binding involved in its inhibition of both GTA and GTB (6). Fig. 2 shows the electron density surrounding the acceptor analogs in GTA and GTB complexed to the 3-amino analog crystallized in the absence (Fig. 2, c and d) and presence (Fig. 2, e and f) of UDP. In the absence of UDP, the analog is seen to adopt a conformation that is significantly different from that observed for either the native acceptor or 3-deoxy analog, which is induced by the presence of the amino group. This group would be protonated at physiological pH and is observed in the crystal structures of the complexes to form a hydrogen bond with hydroxyl on O-2Ј of the fucose residue. This new conformation has the Gal residue in approximately the same location as the native acceptor, with the maintenance of the O-4Ј and O-6Ј hydrogen bonds observed in other structures (Fig. 3b), with the result that the L-Fuc residue moves sharply toward the surface of the protein pocket and into the space normally occupied by UDP. This new position is further stabilized by the formation of a new hydrogen bond from Fuc O-2Ј through a bridging water molecule to potential nucleophilic residue Glu-303 in both GTA and GTB (Figs. 3b and 4b).
In the presence of UDP, the Gal moiety of the 3-amino analog remains in approximately the same position observed for the native acceptor and deoxy analog. The fucose residue is thus displaced, causing the 3-amino analog to abandon the intramolecular hydrogen bond and the hydrogen bond to the bridging water molecule. The fucose ring takes up a less ordered position, with no apparent contact with any part of the enzyme (Fig. 4c), and displays significantly higher temperature factors than the Gal residue.
As with the 3-deoxy analog, attempts made to co-crystallize the 3-amino analog with GTA and GTB in the presence of UDP-GalNAc and UDP-Gal did not reveal a density corresponding to either donor; however, as was found for the deoxy analog, the enzyme was able to hydrolyze significant amounts of donor over the crystallization experiments such that partial occupancy for UDP was observed. In these crystal structures (GTAϩAIϩUDP-GalNAc and GTBϩAIϩUDP-Gal), excellent density was observed for the Gal residue, but the L-Fuc residue was disordered, especially in GTA (possibly over the two basic conformations observed), to the extent that the electron density corresponding to this residue was poorly defined (Fig. 5, a  and b). The repositioning of the 3-amino acceptor analog is not restricted to the influence of UDP alone, and Fig. 4b shows that, in the absence of UDP, the analog occupies distinctly different positions in GTA and GTB, which correspond to changes in the identity of amino acid residue 266. Of the four amino acid differences between GTA and GTB, residue 266 has been shown to dominate the selection between UDP sugar donors (4), with GTA having Leu and GTB having Met. Small differences between the binding of H-disaccharide acceptor to GTA and GTB have been noted in the past (8) and attributed to Leu/Met-266; however, the positions of both the native (4) and now the 3-deoxy acceptor analogs in GTA and GTB are nearly superimposable. The large difference in position observed for the 3-amino analog emphasizes the influence that Leu/Met-266 can have on acceptor binding.
Finally, although the aliphatic aglycones are placed to aid in the purification of the acceptor analogs, they are attached at the same position as the natural glycoprotein or glycolipid substrate of the A and B antigens, and their conformations give some insight into the differential recognition of these substrates by GTA and GTB (4). Fig. 4a shows the positions of the aliphatic groups observed for the deoxy acceptors, where the difference in conformation is similar to that observed in the native acceptors and can be attributed to one of the four critical residue differences between GTA and GTB (Gly/Ser-235). The observed movement of the Fuc residue to form a hydrogen bond with the 3-amino Gal in the absence of UDP opens a pocket in the binding site into which the aliphatic tail moves (Fig. 4b).
The presence of UDP in the binding site causes the Fuc ring on the inhibitor to become disordered but still permits the aliphatic aglycone to assume the same general conformation.

CONCLUSIONS
The x-ray crystal structures of the 3-deoxy and 3-amino analogs in complex with GTA and GTB show that the binding of these modified H-antigen acceptor analogs are consistent with their observed activity as inhibitors of acceptor binding in GTA and GTB. The binding of the 3-amino analog takes a marked departure from that displayed by the H-disaccharide acceptor and 3-deoxy analog, where this binding changes in the presence or absence of UDP. Remarkably given that GTA and GTB share a common acceptor, the binding of acceptor analog is significantly affected by whether the enzyme is GTA or GTB. The competition of the 3-amino acceptor with UDP for the same binding site in the protein is consistent with the observed complex mode of inhibition.