Identification and Modification of the Uridine-binding Site of the UDP-GalNAc (GlcNAc) Pyrophosphorylase*

UDP-GalNAc pyrophosphorylase (UDP-GalNAcPP; AGX1) catalyzes the synthesis of UDP-GalNAc from UTP and GalNAc-1-P. The 475-amino acid protein (57 kDa protein) also synthesizes UDP-GlcNAc at about 25% the rate of UDP-GalNAc. The cDNA for this enzyme, termed AGX1, was cloned in Escherichia coli, and expressed as an active enzyme that cross-reacted with antiserum against the original pig liver UDP-HexNAcPP. In the present study, we incubated recombinant AGX1 with N3-UDP-[32P]GlcNAc and N3-UDP-[32P]GalNAc probes to label the nucleotide-binding site. Proteolytic digestions of the labeled enzyme and analysis of the resulting peptides indicated that both photoprobes cross-linked to one 24-amino acid peptide located between residues Val216 and Glu240. Four amino acids in this peptide were found to be highly conserved among closely related enzymes, and each of these was individually modified to alanine. Mutation of Gly222 to Ala in the peptide almost completely eliminated UDP-GlcNAc and UDP-GalNAc synthesis, while mutation of Gly224 to Ala, almost completely eliminated UDP-GalNAc synthesis, but UDP-GlcNAc was only diminished by 50%. Both of these mutations also resulted in almost complete loss of the ability of the mutated proteins to cross-link N3-UDP-[32P]GlcNAc or N3-UDP-[32P]GalNAc. On the other hand, mutations of either Pro220 or Tyr227 to Ala did not greatly affect enzymatic activity, although there was some reduction in the ability of these proteins to cross-link the photoaffinity probes. We also mutated Gly111 to Ala since this amino acid was reported to be necessary for catalysis (Mio, T., Yabe, T., Arisawa, M., and Yamada-Okabe, H. (1998) J. Biol. Chem. 273, 14392–14397). The Gly111 to Ala mutant lost all enzymatic activity, but interestingly enough, this mutant protein still cross-linked the radioactive N3-UDP-GlcNAc although not nearly as well as the wild type. On the other hand, mutation of Arg115 to Ala had no affect on enzymatic activity although it also reduced the amount of cross-linking of N3-UDP-[32P]GlcNAc. These studies help to define essential amino acids at or near the nucleotide-binding site and the catalytic site, as well as peptides involved in binding and catalysis.

UDP-GlcNAc pyrophosphorylase (UDP-GlcNAcPP) 1 is a key enzyme in the biosynthesis of complex carbohydrates, since it is the enzyme that produces the activated form of GlcNAc, i.e. UDP-GlcNAc, for polymerization. This enzyme catalyzes the reversible reaction: UTP ϩ GlcNAc-1-P º UDP-GlcNAc ϩ PP i . The enzyme was initially purified about 100-fold from Staphylococcus aureus, and that enzyme preparation utilized UDP-GalNAc (in the reverse direction) at about 2.8% the rate of UDP-GlcNAc (1). The enzyme was also partially purified from calf brain, and in that case, the enzyme also utilized UDPglucose at about 36% the rate of UDP-GlcNAc (2). However, it was not clear from that report whether the calf brain enzyme had any activity with UDP-GalNAc, or whether the activity with UDP-glucose was due to the same enzyme, or to a contaminating pyrophosphorylase.
We purified the UDP-GlcNAcPP from pig liver cytosol to apparent homogeneity, and found that the purified enzyme contained two proteins with molecular masses of about 64 and 57 kDa. Interestingly enough, this purified enzyme also catalyzed the synthesis of UDP-GalNAc from UTP and GalNAc-1-P, and at saturating concentrations of GalNAc-1-P, the rate of synthesis of UDP-GalNAc was equal to that of UDP-GlcNAc (3). We sequenced three peptides from the 64-kDa protein and two from the 57-kDa protein and found that all of these sequences showed 100% identity to AGX1, a protein of unknown function from the sperm of infertile males (4). A second human protein of wider tissue distribution, but also of unknown function and called AGX2, has the identical sequence to AGX1 except that AGX2 has an additional 17-amino acid insert near the carboxyl terminus (5).
We cloned the cDNAs for AGX1 and AGX2 and expressed the active proteins in Escherichia coli. AGX1 migrated with the 57-kDa pig liver enzyme on SDS gels, and had about three times as much UDP-GalNAc synthetic activity as UDP-GlcNAc activity. On the other hand, AGX2 migrated with the 64-kDa pig liver protein and also had both UDP-GalNAc and UDP-GlcNAc activity, but at a ratio of about 1:8 (5). Thus, addition of the 17-amino acid insert changes the activity of the enzyme from synthesizing UDP-GalNAc to synthesizing UDP-GlcNAc (5).
In the current study, we have identified a 24-amino acid peptide involved in binding the uridine portion of the substrate, i.e. either UDP-GalNAc or UDP-GlcNAc. To do this, we synthesized N 3 -UDP-[ 32 P]GlcNAc and N 3 -UDP-[ 32 P]GalNAc (6,7), and incubated both of these photoprobes with the recombinant AGX1 (i.e. UDPGalNAcPP). Both photoprobes cross-linked specifically to amino acids in the active site, and the resulting radiolabeled peptide was identified and located in the protein sequence of AGX1. We then used this information to do site-directed mutagenesis of four highly conserved amino acids, and determined the effect of amino acid changes on enzymatic activity and substrate binding. These studies indicated that two glycines in that peptide (at positions 222 and 224) are important in cross-linking the uridine portion of the substrate, as well as in enzyme catalysis. These glycines may play a direct role in binding and catalysis, or they may be important in protein conformation.
Another study had previously reported the identity between AGX1 and the UDP-GlcNAc pyrophosphorylase (8). That study indicated that Gly 112 , Arg 116 , and Lys 123 were important for catalytic activity of this enzyme. We also confirmed that changing Gly 111 (position 111 in our sequence) to Ala resulted in loss of catalytic activity, but importantly, this modification did not eliminate the ability of the mutant protein to cross-link the photoprobes. These studies begin to identify important amino acid residues and peptide sequences involved in substrate binding and catalysis.
Photoaffinity Labeling of AGX1-Photoaffinity labeling was done using previously described methods (6). Briefly, incubations for labeling the proteins were as follows: 2-10 g of purified wild type or mutant recombinant AGX1, in a final volume of 50 l of 50 mM Tris buffer, pH 7.5, plus 5 mM MgCl 2 , were incubated for 30 s, on ice, with either N 3 -UDP-[ 32 P]GlcNAc or N 3 -UDP-[ 32 P]GalNAc. The reaction mixture was then exposed to short-wave UV light for 90 s, with agitation, to activate the azido group and covalently cross-link the probe to the protein. One ml of methanol, cooled to Ϫ80°C, was added to stop the reaction, and the precipitated protein was isolated by centrifugation in a microcentrifuge at 10,000 ϫ g for 10 min. The supernatant was discarded and the pellet was suspended in 20 l of SDS gel-loading buffer, and subjected to SDS-PAGE on a 10% gel. Location of radioactive proteins was determined by autoradiography (9,10). To determine the specificity of the labeling, some incubations contained the radioactive probe, plus various amounts of unlabeled UDP-GlcNAc or UDP-GalNAc. These reaction mixtures were processed as described above for the control incubations (5,6).
Enzymatic Digestions of Radiolabeled AGX1 and Isolation of Peptides-The protein pellets from incubations of AGX1 and the photoprobes were isolated by centrifugation of the methanolic extracts, and suspended in 40 l of digestion buffer (100 mM Tris buffer, pH 8.0, containing 0.1% SDS), rather than the SDS gel loading buffer. The sample mixture was divided into four equal aliquots, each of which was subjected to digestion at 37°C for various times with 1 unit of one of the following proteases: endoproteinase Lys-C, V8 proteinase, endoproteinase Asp-N, or both endo-V8 and Asp-N.
The protease digests were subjected to SDS-PAGE on precast SDS/ Tricine gels in 10 -20% polyacrylamide. Protein and peptide standards (MARK 12, Novex) were run on these gels to determine the sizes of the radiolabeled peptides obtained from each protease digestion. The location of the radiolabeled peptide was determined by subjecting the gels to autoradiography, while the standards were detected by Coomassie Blue staining. The predicted molecular masses of proteolytic peptides from AGX1 were determined using a Wisconsin Package Program.
Site-directed Mutagenesis of Amino Acid Residues of AGX1-Sitedirected mutagenesis was performed on a pGEX-4T-2 plasmid with full-length AGX1 cDNA using a Quickchange Site-directed Mutagenesis Kit (Stratagene), according to the manufacturer's instructions. Codons for Pro 220 , Gly 222 , Gly 224 , Tyr 227 , Gly 111 , and Arg 115 were mutated to alanine by using the appropriate oligonucleotides.
Western Blot Analysis-Western blot analysis was done using stand-ard technology. Wild type and mutant AGX1s were purified as described previously (5). Equal amounts of the various proteins were subjected to SDS-PAGE on 10% gels. The proteins from these gels were transferred electrophoretically to nitrocellulose membranes, and sequentially blotted.
Measurements of Enzyme Activities and Determination of K m Values-Assays for UDP-GlcNAcPP and UDP-GalNAcPP activities were usually done in the forward direction (formation of UDP-HexNAc from HexNAc-1-P and UTP). These assays were done with the various mutant proteins and the activities compared with that of purified wild type AGX1. The incubation mixtures contained the following components in a final volume of 100 l: 50 mM Tris buffer, pH 7.5, 5 mM MgCl 2 , 2 mM [ 3 H]UTP (55,000 cpm), 5 mM GlcNAc-1-P or GalNAc-1-P, and various amounts of the protein to be tested. After an incubation of 10 min at 37°C, the incubations were terminated by heating at 100°C for 30 s, and the mixture was applied to a column of DE-52. The UDP-HexNAcs were eluted with 70 mM (NH 4 )HCO 3 , and could be readily separated from UTP which required much higher concentrations of bicarbonate for elution. Thus, the amount of radioactivity in the 70 mM elution was determined as a measure of activity.
Enzymatic activity could also be measured in the reverse direction. In this case, the incubation mixtures contained the following components in a final volume of 100 l: 50 mM Tris buffer, pH 8.5, 5 mM MgCl 2 , 5 mM sodium pyrophosphate, 2 mM UDP-[ 3 H]GlcNAc (or Gal-NAc) (35,000 cpm), and various amounts of the proteins to be tested. After an incubation of 10 min at 37°C, the reactions were stopped by the addition of 0.5 ml of 5% trichloroacetic acid, and 0.5 ml of acid washed charcoal was added to absorb the nucleotides. The charcoal was removed by centrifugation, and the amount of radioactivity in the supernatant liquid was determined as a measure of the amount of HexNAc-1-P produced. Kinetic parameters, i.e. K m and V max /K m , were determined from Lineweaver-Burk plots obtained by varying the concentration of UDP-GlcNAc from 0.1 to 1 mM, and the concentration of UDP-GalNAc from 0.5 to 5 mM. In these experiments, the concentration of sodium pyrophosphate was kept constant at 5 mM.

FIG. 1. Photoaffinity labeling of AGX1 with either N 3 -UDP-[ 32 P]GlcNAc (A) or N 3 -UDP-[ 32 P]GalNAc (B).
Recombinant AGX1 (5 g of purified protein) was incubated as described under "Experimental Procedures" in a final volume of 50 l, but with increasing amounts of radioactive probe, for 30 s and then exposed to UV for 90 s with shaking. The protein was then precipitated with methanol, resuspended in a SDS gel loading buffer, and subjected to SDS-PAGE. Radioactive proteins were detected by autoradiography. Lanes are as follows: lane 1, AGX1 ϩ 10 l of probe, but no UV; lanes 2-6, AGX1 ϩ 1, 3, 6, 10, and 15 l of probe ϩ UV. In B, lanes are essentially the same, but contained 2, 5, 10, and 20 l of the UDP-GalNAc probe.

RESULTS AND DISCUSSION
Identification of a Peptide Involved in Binding UDP-Hex-NAc-Recombinant AGX1 was incubated with either N 3 -UDP-[ 32 P]GlcNAc or N 3 -UDP-[ 32 P]GalNAc for 30 s and the mixtures were exposed to short wave UV light to activate the azido group. The protein was isolated and separated on SDS gels, and the amount of label was determined by exposure to film. The data in Fig. 1 indicate that the photo cross-linking of either of these probes (see Fig. 1, A and B) to the recombinant AGX1 increased with increasing concentrations of probe, and furthermore, this incorporation was saturable (not shown in this figure).
As shown in Fig. 2, the photo cross-linking of N 3 -UDP-[ 32 P]GlcNAc to AGX1 (lane 2) was inhibited in a concentrationdependent manner by adding increasing amounts of unlabeled UDP-GlcNAc (lanes [3][4][5] to the incubation mixtures. Photoincorporation of the N 3 -UDP-[ 32 P]GlcNAc probe was also inhibited by increasing amounts of unlabeled UDP-GalNAc (lanes 6 -8). However, the amount of UDP-GalNAc required for inhibition of cross-linking was about 10 times greater then the amount of UDP-GlcNAc. These data are in agreement with the K m values for UDP-GlcNAc and UDP-GalNAc, and they strongly indicate that these photoaffinity probes are recognized as substrate analogs by the active UDP-GalNAcPP (AGX1). Similar results on cross-linking specificity were observed when the N 3 -UDP-[ 32 P]GalNAc probe was tested (data not shown). As expected since it is also a substrate, cross-linking of the azido probe was also inhibited by increasing amounts of UTP (lanes 9 -11).
Since these photoprobes cross-linked in a specific manner to AGX1, it seemed likely that they would be useful as reagents to specifically label the substrate-binding site. Thus, large scale incubations were done with AGX1 and either N 3 -UDP-[ 32 P]Gl-cNAc or N 3 -UDP-[ 32 P]GalNAc, and after exposure to UV light, the two labeled proteins were isolated by centrifugation. Each protein was divided into four equal aliquots and each aliquot was digested with a specific proteases. Fig. 3A shows the results of an overnight digestion of UDP-GlcNAc-labeled enzyme  3. A, autoradiography of peptides produced by protease digestion of AGX1 labeled with N 3 -UDP-[ 32 P]GlcNAc. AGX1 was incubated with probe as described in the legend to Fig. 1 and protein was isolated by precipitation with methanol. The labeled AGX1 prepared from each of the probes was divided into four equal aliquots, and each was digested with a different protease as follows: lane 1, Lys-C; lane 2, V8; lane 3, Asp-N; lane 4, Asp-N ϩ V8. After digestion for 24 h, peptides were separated on Tricine-SDS gels and radiolabeled peptides were detected by autoradiography. Peptides of known size were run as standards as shown by the numbers on the left. B, diagram showing cleavage sites of the various proteases (Lys-C, V8, Asp-N) and estimated sizes of peptides released. In the case of Lys-C, cleavage at positions 207 (Lys 207 ) and 278 (Lys 278 ) gave a radiolabeled peptide of 7.9 kDa, and further cleavage of this peptide at Lys 215 and Lys 271 gave rise to a labeled peptide of about 6.3 kDa, which was cleaved again at Lys 258 to give a 5.0-kDa radioactive peptide, encompassing residues Asn 214 to Lys 258 . Since the 6.3-kDa Lys-C peptide was predicted to encompass residues Val 216 to Lys 271 , the smallest Lys-C peptide involved in binding would be from Val 216 to Lys 258 . This peptide was further defined by V8 digestion which resulted in a radiolabeled peptide of about 5.7 kDa encompassing Asn 190 to Glu 240 . Confirmation that the peptide from Val 216 to Glu 240 was the correct and smallest binding region for the uridine portion of the substrate was the isolation of a radiolabeled peptide by Asp-N digestion of about 5.4 kDa representing Asp 205 to Val 252 . with the various proteases (i.e. Lys-C, Asp-N, or V8). After a long incubation with each protease, most of the protein was digested to smaller peptides (as evidenced by Coomassie Blue staining), and only one or a few of these peptides were radiolabeled as shown in the figure. The resulting peptides were separated on SDS/Tricine gels and their migration was compared with various peptide standards.
Since the site of action of each of the different proteases is known, and the amino acid sequence of AGX1 is also known, we can predict the probable peptide sequence that is cross-linked, based on the size of the resulting radiolabeled peptides in each incubation and the overlapping sequences of the peptides produced by each protease (9,10). The rationale for identifying the substrate-binding region by photolabeling is outlined in Fig. 3B and was used to identify overlapping radiolabeled peptides released from AGX1 by different proteases. The primary amino acid sequence of AGX1 was used to predict the molecular masses of peptides that would be released by each protease.
For example, the UDP-GlcNAc cross-linked AGX1 was digested overnight with Lys-C to give labeled peptides of 7.9, 6.3, and 5.0 kDa. Using the amino acid sequence of AGX1 to derive predicted peptide masses produced by Lys-C, the only peptide of 7.9 that could give rise to peptides of 6.3 and 5.0 kDa was in the region of the protein from amino acids 208 -278. Likewise, the only peptide of 6.3 kDa that would yield the 5.0-kDa peptide was in the region from amino acid 216 to 271, by loss of a 1.49-kDa peptide by Lys-C action. The critical region could be further refined using V8 and Asp-N proteases. Thus Asp-N gave labeled peptides of about 11.2, 8.0, 6.5, and 5.5 kDa, with the 5.5-kDa peptide being the major species. The only peptide of 8.0 kDa that could give rise to a peptide of about 6.5 and 5.5 was in the region from amino acid 205 to 273. The 5.5-kDa peptide resulting from this digestion would be between amino acids 205 and 252. Finally, by the same analysis with V8 protease, it was deduced that the 5.7-kDa peptide must be between amino acids 190 and 240. Thus, the smallest peptide that could carry the labeled photoprobe and be compatable with the peptides produced by these three proteases would have to be between residues 216 and 240. The amino acid sequence of this peptide is presented in Fig. 4, and is from Val 216 to Glu 240 . Essentially similar results, and the same peptide sequence, were obtained when labeling was done with N 3 -UDP-[ 32 P]Gal-NAc (data not shown). Fig. 4 also compares this peptide sequence with sequences from other related UDP-sugar pyrophosphorylases. This data indicates that there are a number of residues that are highly conserved, and thus it was important to determine whether modification of any of the conserved amino acids would affect the activity of the mutant protein to synthesize UDP-GlcNAc and UDP-GalNAc. We were also interested in knowing whether such changes would affect the ability of the protein to bind the UDP-HexNAc substrates.
Mutant proteins were produced by site-directed mutagenesis in which selected amino acids were modified to alanines. Alanine was chosen since it does not prevent or disrupt formation of ␣-helix or ␤-pleated sheet secondary structures (11). Thus, we chose to modify Pro 220 , Gly 222 , Gly 224 , and Tyr 227 . The results of these alterations on the ability of these proteins to catalyze the synthesis of either UDP-GlcNAc or UDP-GalNAc are shown in Fig. 5 (A and C). In these incubations, each mutant protein was incubated with either GlcNAc-1-P (Fig. 5A) or GalNAc-1-P (Fig. 5C) and [ 3 H]UTP, and the formation of UDP-GlcNAc or UDP-GalNAc was determined. It can be seen that mutation of Gly 222 resulted in an almost complete loss in the ability of that mutant protein to synthesize the corresponding UDP-HexNAc, whereas change of Gly 224 resulted in considerable, but not complete loss in activity. On the other hand, mutation of Pro 220 to alanine resulted in only a slight decrease in activity, while change of Tyr 227 to alanine did not greatly affect the ability of that mutant protein to synthesize the UDP-HexNAcs. It should be noted that all of these comparisons of enzymatic activity were on the basis of specific activity measurements with the purified mutant AGX1s.
Similar results were obtained when these various mutant UDP-GalNAcPPs were assayed in the reverse direction. These results are presented in Fig. 5 (B and D). In this case, the Gly 224 mutant appeared to retain more activity with UDP-GlcNAc (Fig. 5B) than with UDP-GalNAc. However, the results were generally similar with either the forward or the reverse assay.
In order to determine whether these changes in activity were due to differences in the ability of the mutant enzymes to bind FIG. 4. Predicted sequence of peptide of AGX1 labeled by affinity probes and comparison with other known sequences. The AGX1 sequence was aligned with the sequence from UDP-GlcNAcPP of Candida albicans (8), as well as sequences from UDP-GlcPP of human (14) and bovine liver (15) and of yeast (16). The AGX1 sequence begins with Val 216 and ends at Glu 240 .
FIG. 5. Enzymatic activity of various mutant AGX1 preparations made by site-directed mutagenesis of conserved amino acids. Recombinant proteins were prepared and pyrophosphorylase activity was tested in the forward direction with GlcNAc-1-P (A) or GalNAc-1-P (C) ϩ UTP, and in the reverse direction with UDP-GlcNAc (B) or UDP-Gal-NAc (D) ϩ inorganic pyrophosphate, as described under "Experimental Procedures." Mutations are proline to alanine (P220A), glycine to alanine (G222A or G224A), and tyrosine to alanine (Y227A). the UDP-HexNAc substrates, the K m values of the mutant proteins for UDP-GlcNAc were determined, and are presented in Table I. Mutants P220A and Y227A had K m values that were essentially the same as those of the wild type AGX1 (approximately 0.8 mM), whereas G224A showed a K m of about 2.2 mM and G222A had such a low binding affinity that it was not possible to obtain an accurate K m value. Because the determination of K m values reflect both substrate binding and catalysis, the apparent decreases in UDP-GlcNAc affinity could actually be due to changes in catalytic activity, binding affinity, or both.
In order to be certain that the mutated amino acid changes were really affecting binding rather than catalysis, we sought to separate these two steps. One way to determine relative binding affinities without involving the catalytic step is to use the photoaffinity substrates (i.e. N 3 -UDP-[ 32 P]GlcNAc and N 3 -UDP-[ 32 P]GalNAc), and to evaluate their ability to label the mutant proteins. However, prior to doing the cross-linking studies, it was important to be certain that each recombinant mutant protein was produced at the same rate in the various E. coli strains. Fig. 6 demonstrates, by SDS-gel electrophoresis and Western blotting, that each of the AGX1 recombinant proteins reacted equally well with the polyclonal antibody prepared against the purified pig liver UDP-HexNAcPPs suggesting that the single amino acid changes did not grossly affect protein conformation. In addition, each of the E. coli strains, with the exception of the G224A strain, grew at about the same rate, and produced approximately the same amount of recombinant AGX1 (data not shown). These findings indicate that changes in Pro 220 , Gly 222 , or Tyr 227 do not affect synthesis, targeting, or turnover of the corresponding AGX1s. In the case of mutation G224A, that E. coli strain grew significantly slower and had only about 25% of the amount of recombinant protein found in the other strains. We do not know at this time why this particular amino acid substitution affects production of the protein, but it is clear from other data presented here that substitution of glycine 224 with alanine also greatly affects the binding of UDP-GlcNAc and UDP-GalNAc. Fig. 7 demonstrates that the results of labeling with the photoprobes corresponded well with the activity measurements shown in Fig. 5, as well as with the K m measurements presented in Table I. That is, very little binding of the UDP-GlcNAc probe (Fig. 7A), or of the UDP-GalNAc probe (Fig. 7B), was detected with the G222A or the G224A mutant proteins (lanes 3 and 4). On the other hand, the other two mutant proteins, P220A (lane 2) and Y227A (lane 5), did bind the probe, although labeling was significantly lower then with the wild type AGX1 (lane 1).
We also examined the binding of N 3 -UDP-[ 32 P]GlcNAc to two other mutant proteins in which amino acids at the proposed catalytic site had been altered. Mio et al. (8) had previously reported that modification of Gly 112 to alanine resulted in loss of catalytic activity. We confirmed that modification of Gly 111 to alanine in the AGX1 sequence did cause almost complete loss of catalytic activity, and the data in Table I  . Each recombinant AGX1 (5 g of protein) was incubated with probe, exposed to UV, and subjected to SDS-PAGE. Gels were exposed to film to visualize radioactive proteins. was incubated with the radioactive probe, exposed to UV light, and subjected to SDS-PAGE. Gels were exposed to film to visualize radioactive proteins. Lane O, AGX1 without exposure to UV; lane 1, AGX1 with exposure to UV. In all other lanes, the incubations were exposed to