Molecular Basis for Protein-specific Transfer of N-Acetylgalactosamine to N-Linked Glycans by the Glycosyltransferases β1,4-N-Acetylgalactosaminyl Transferase 3 (β4GalNAc-T3) and β4GalNAc-T4*

Background: Luteinizing hormone and carbonic anhydrase-6 bear unique N-linked oligosaccharides terminating with LacdiNAc (GalNAcβ1,4GlcNAc). Results: Two related β1,4-N-acetylgalactosaminyltransferases have distinct specificities for peptide motifs on luteinizing hormone and carbonic anhydrase-6. Conclusion: Affinity for the peptide motif determines the efficiency of GalNAc transfer to oligosaccharide acceptors. Significance: Kinetic parameters define the basis for protein-specific synthesis of carbohydrate structures in vivo. Two closely related β1,4-N-acetylgalactosaminyltransferases, β4GalNAc-T3 and β4GalNAc-T4, are thought to account for the protein-specific addition of β1,4-linked GalNAc to Asn-linked oligosaccharides on a number of glycoproteins including the glycoprotein hormone luteinizing hormone and carbonic anhydrase-6 (CA6). We have utilized soluble, secreted forms of β4GalNAc-T3 and β4GalNAc-T4 to define the basis for protein-specific GalNAc transfer in vitro to chimeric substrates consisting of Gaussia luciferase followed by a glycoprotein substrate. Transfer of GalNAc by β4GalNAc-T3 and β4GalNAc-T4 to terminal GlcNAc is divalent cation-dependent. Transfer of GalNAc to glycoprotein acceptors that contain a peptide recognition determinant is maximal between 0.5 and 1.0 mm MnCl2; however, transfer is increasingly inhibited by concentrations of MnCl2 above 1 mm and by anion concentrations above 15 mm. In contrast, transfer of GalNAc to the simple sugar acceptor N-acetylglucosamine-β-p-nitrophenol (GlcNAcβ-pNP) is not inhibited by concentrations of MnCl2 or anions that would inhibit transfer to glycoprotein acceptors by >90%. This finding indicates that interaction with the peptide recognition determinant in the substrate is sensitive to the anion concentration. β4GalNAc-T3 and β4GalNAc-T4 have similar but distinct specificities, resulting in a 42-fold difference in the IC50 for transfer of GalNAc to chimeric glycoprotein substrates by agalacto human chorionic gonadotropin, comprising 29 nm for β4GalNAc-T3 and 1.2 μm for β4GalNAc-T4. Our in vitro analysis indicates that enzymatic recognition of the peptide determinant and the oligosaccharide acceptor are independent events.

dant than ␤4GalNAc-T4 transcripts. 3 A key question is whether ␤4GalNAc-T3 and ␤4GalNAc-T4 are simply redundant or display differences in specificity or catalytic properties. Why are two closely related transferases produced? Do ␤4GalNAc-T3 and ␤4GalNAc-T4 modify different substrates in vivo? Examination of these questions requires in vitro analysis of the specificity and properties of ␤4GalNAc-T3 and ␤4GalNAc-T4.
We have now examined protein-specific transfer of GalNAc by ␤4GalNAc-T3 and ␤4GalNAc-T4 to N-linked oligosaccharides on various chimeric glycoprotein substrates in vitro. We previously defined the peptide recognition determinant itself in in vivo studies in CHO cells expressing ␤4GalNAc-T3 and ␤4GalNAc-T4. Our present studies in the in vitro system complement the in vivo findings, revealing that ␤4GalNAc-T3 and ␤4GalNAc-T4 display similar but distinct specificities. In vitro analysis has further allowed us to separate enzymatic recognition of the peptide determinant and the oligosaccharide acceptor as two independent events.
Preparation of Substrates-Chimeric forms of the glycoprotein hormone ␣ subunit and Trf (see Fig. 1) were prepared as described (8). pCMV-GLuc was obtained from New England Biolabs. The Gaussia luciferase (GLuc) constructs in each case consist of GLuc followed by the ␣ subunit or Trf and the epitope Myc-His. GLuc chimeras were expressed in Lec8/CHO cells (10) grown in serum-free Ultra CHO medium (Lonza 12-724Q) using Lipofectamine and Lipofectamine Plus (Invitrogen) per the manufacturer's transfection protocol. Media containing the chimeric proteins were harvested and purified using nickelnitrilotriacetic acid-agarose (Qiagen). The bound proteins were eluted using 250 mM imidazole, 50 mM monobasic sodium phosphate, 300 mM sodium chloride, 0.05% Tween 20, pH 8.0. The purified substrates were exchanged into 25 mM HEPES buffer, pH 7.5, containing 0.1% BSA and stored at Ϫ80°C after the addition of glycerol to a final concentration of 10%.
Quantitation of GalNAc Incorporation into Gaussia Luciferase Containing Chimeric Glycoproteins-In vitro transfer of GalNAc to GLuc chimeric glycoproteins was carried out in a 50-l reaction containing: 25 mM Hepes buffer, pH 7.5, 10% glycerol, 0.05% Triton X-100, 5 mM KCl, 0.5 or 1.0 mM MnCl 2 , 1 mg/ml BSA, 0.24 mM UDP-GalNAc, and the affinity-purified GLuc substrate. The affinity-purified substrate was diluted sufficiently to maintain the final concentration of the imidazole below 5 mM. Following the addition of either ␤4GalNAc-T3-F or ␤4GalNAc-T4-F, transfer of GalNAc was allowed to proceed for 20 min at 25°C. Transfer was terminated by the addition of 450 l of ice-cold 2 mM EDTA, pH 7.5.
The amount of GalNAc transferred to GLuc substrates was determined using a microplate assay. B&W Isoplate-96 HB (PerkinElmer Life Sciences 6005580) 96-well plates were coated with Wisteria floribunda agglutinin (WFA) by adding 100 l of 25 mM sodium carbonate buffer, pH 8.5, containing 1.25 g of WFA (Sigma L8258) per well and incubating overnight at 4°C. Plates were washed with 0.1% BSA/PBS using a Bio-Rad Immunowash microplate washer and blocked with 5% BSA/PBS. The light units of GLuc activity present following GalNAc transfer were determined for each in vitro assay by adding 5-8 ϫ 10 5 light units of each transferase reaction in 100 l of 15.6 g/ml histone, 0.1% BSA, PBS (positive signal) or in 100 l of 15.6 g/ml histone, 0.1% BSA/PBS, containing 50 mM GalNAc (background signal) to WFA-coated wells. After incubation overnight at 4°C, unbound GLuc substrate was removed by washing, and the amount of bound GLuc substrate was determined using a Wallac Victor2 luminometer using the Renilla luciferase assay system (Promega E2820). Fifty mM GalNAc inhibits binding to immobilized WFA. The difference between the light units of GLuc substrate bound in the absence and presence of GalNAc reflects the amount of GLuc substrate that is modified with ␤1,4-linked GalNAc. In the absence of added ␤4GalNAc-T3 or ␤4GalNAc-T4, the amount of GLuc substrate bound in the absence and presence of GalNAc was identical. All assays were preformed in quadruplicate.
Transfer of UDP-[ 3 H]GalNAc to N-Acetylglucosamine-␤ p-Nitrophenol-Transfer of GalNAc to ␤-linked GlcNAc in the absence of peptide recognition was assayed using N-acetylglucosamine-␤-p-nitrophenol (GlcNAc␤-pNP) as a substrate. Unless otherwise indicated, the assay was carried out in a final volume of 50 l containing: 100 mM HEPES buffer, pH 7.5, 0.2% Triton X-100, 20 mM MnCl 2 , 2 mM ATP,1 mM GlcNAc␤-pNP (Sigma N9376), and 0.4 mM UDP-GalNAc plus 150,000 dpm of UDP-[ 3 H]GalNAc. Transfer was initiated by the addition of either purified ␤4GalNAc-T3-F or purified ␤4GalNAc-T4-F and incubating at 37°C for 2 h. The reaction was terminated by the addition of 950 l of ice-cold 2 mM EDTA, pH 7.0. The reaction was loaded onto a 500-mg Sep-Pak C-18 cartridge (Waters) previously activated with 2 ml of ethanol and equilibrated with 10

RESULTS
Secreted Forms of ␤4GalNAc-T3 and ␤4GalNAc-T4 Mediate Protein-specific Transfer of GalNAc in Vitro-Substrate constructs encoding chimeric glycoproteins consisting of GLuc, a secreted form of luciferase (11,12), followed by either the glycoprotein hormone ␣ subunit or Trf, were generated as described (8). The chimeric substrate GLuc glycoproteins shown in schematic form in Fig. 1 were epitope-tagged at their carboxyl terminus with sequences for Myc, and 6 residues of His that were utilized for purification by chelate affinity chromatography. The GLuc chimeras were expressed in Lec8/CHO cells, which are not able to transport UDP-Gal into the Golgi (10, 13) and thus do not add Gal to N-linked oligosaccharides on secreted glycoproteins. The fusion protein expressed in Lec8/CHO cells bears N-linked complex type oligosaccharides terminating with ␤-linked GlcNAc that acts as the substrate for GalNAc addition by either ␤4GalNAc-T3 or ␤4GalNAc-T4 in vitro. Equal light units of purified GLuc substrates were incubated with either affinity-purified ␤4GalNAc-T3-F or ␤4GalNAc-T4-F (secreted FLAG-tagged forms, see "Experimental Procedures") and UDP-GalNAc. Following termination of the transferase reaction, equal amounts of each GLuc substrate were incubated in 96-well plates coated with the Gal-NAc-specific lectin WFA (14). After removing unbound substrate, the amount of GLuc substrate bound by the WFA was then determined by injection of coelenterazine.
Conditions were established to ensure that both the transferase assay itself and the plate assay for bound GalNAcmodified GLuc substrates remained in a linear range. The incorporation of GalNAc into the N-linked oligosaccharides of GLuc␣(PLRSKK) and GLuc␣(PLRSKK)CA1-19 increased linearly for at least 15 min for both ␤4GalNAc-T3-F and ␤4GalNAc-T4-F (Fig. 2). The incorporation of GalNAc was proportional to the amount of ␤4GalNAc-T3-F and ␤4GalNAc-T4-F added (not shown).
Anion and Cation Requirements for Optimal Transfer of Gal-NAc by ␤4GalNAc-T3-F and ␤4GalNAc-T4-F in Vitro-Optimization of enzymatic reaction conditions revealed that transfer of GalNAc to GLuc␣(PLRSKK) by either ␤4GalNAc-T3-F or ␤4GalNAc-T4-F was progressively inhibited by concentrations of NaCl and KCl in excess of 10 -15 mM (Fig. 3). NaPO 4 reduced transfer of GalNAc by 80 -90% at a concentration of 10 mM. The addition of a divalent cation such as Mn 2ϩ was essential for transfer; maximal rates of transfer were achieved for both ␤4GalNAc-T3-F and ␤4GalNAc-T4-F at 0.5 mM Mn 2ϩ (Fig. 4). The apparent k d for Mn 2ϩ binding based on GalNAc transfer was 0.25 mM for ␤4GalNAc-T3-F and 0.06 mM for ␤4GalNAc-T4-F. Concentrations of Mn 2ϩ greater than 1 mM inhibited GalNAc transfer (Fig. 4), likely reflecting the free Clconcentrations. Thus, maintenance of a low anion concentration was essential for optimal transfer of GalNAc to glycoprotein substrates.
The reduced rate of GalNAc transfer by both ␤4GalNAc-T3 and ␤4GalNAc-T4 in the presence of increasing concentrations of anions could reflect an impact on peptide recognition, saccharide recognition, or both. We therefore determined the rate of [ 3 H]GalNAc transfer to GlcNAc␤-pNP, a small substrate that is devoid of peptide components, by ␤4GalNAc-T3-F under different conditions (Table 1). In the presence of 0.2 mM FIGURE 1. Schematic representation of chimeric glycoprotein substrates. Plasmids encoding GLuc followed by the glycoprotein hormone ␣ subunit (Alpha) or Trf were prepared by ribocloning (8). Locations of Asn glycosylation sites are indicated by the OE. The location of the sequence PLRSKK that was mutated to PLESEE is indicated for the ␣ subunit. The carboxyl-terminal 19 amino acids from CA6 (CA1-19) were added to the carboxyl terminus of ␣ and the carboxyl terminus of Trf and are indicated by the filled oval. Sequences of CA1-19 and of peptides with partial sequences are shown. All chimeric glycoprotein substrates contained a Myc epitope and His 6 residues. The sequence of the synthetic peptide from the ␣ subunit that contains the key residues required for recognition, ␣32-46, is also shown. for the times indicated. The amount of GalNAc incorporated into GLuc␣(PLRSKK)CA1-19 was determined by capturing GalNAc-modified GLuc␣(PLRSKK)CA1-19 onto 96-well plates coated with the lectin WFA that binds terminal ␤1,4-linked GalNAc. The input of luciferase activity was identical for each well. The amount of bound luciferase activity was determined using the Renilla luciferase assay (Promega) as described. The S.E. is shown for all data points, which were performed in triplicate, but is obscured by the data points. Incorporation of GalNAc increased linearly for both ␤4GalNAc-T3-F and ␤4GalNAc-T4-F for a minimum of 20 min. LU, light units.
Mn 2ϩ , the same amount of GalNAc was transferred to GlcNAc␤-pNP in 25 mM HEPES versus 100 mM HEPES. Furthermore, transfer of GalNAc increased 3-fold in the presence of 20 mM Mn 2ϩ and 100 mM HEPES, conditions that would have markedly inhibited transfer to the peptide substrate GLuc␣(PLRSKK) bearing a recognition motif. Thus, it is peptide recognition rather than saccharide recognition that is sensitive to inhibition by anions. In addition, the apparent K m for UDP-GalNAc is not sensitive to anion concentrations used for these assays (not shown).
␤4GalNAc-T3-F and ␤4GalNAc-T4-F Are Protein-specific-We previously demonstrated that hCG, but not Trf, that has had terminal Sia and Gal removed from its N-linked oligosaccharide structures is specifically modified with GalNAc by one or more ␤1,4GalNAc transferases present in the pituitary (2). We established an in vitro competition-type assay to examine the impact of different peptide recognition motifs on transfer of GalNAc by ␤4GalNAc-T3-F and ␤4GalNAc-T4-F. The addition of increasing amounts of agalacto-hCG to the transferase assay reduced the incorporation of GalNAc into GLuc␣(PLRSKK) by both ␤4GalNAc-T3-F and ␤4GalNAc-T4-F (Fig. 6). In contrast to the observed competition with agalacto-hCG, the highest concentration of agalacto-Trf examined, 1.05 M, did not reduce the incorporation of GalNAc into GLuc␣(PLRSKK) by either ␤4GalNAc-T3-F or ␤4GalNAc-T4-F (Fig. 6). The IC 50 for agalacto-hCG inhibition of GalNAc incorporation into GLuc␣(PLRSKK)CA1-19 was 29 nM for ␤4GalNAc-T3-F and 1.23 M for ␤4GalNAc-T4-F. The 42-fold difference in the IC 50 for agalacto-hCG competition indicates differences in the peptide specificities of ␤4GalNAc-T3-F as compared with ␤4GalNAc-T4-F.
The Presence of a Peptide Recognition Motif Is Essential for Efficient Protein-specific Transfer of GalNAc by ␤4GalNAc-T3-F and ␤4GalNAc-T4-F-We next examined GalNAc addition in vitro by ␤4GalNAc-T3-F and ␤4GalNAc-T4-F to GLuc substrates that did or did not contain the sequences that we had proposed are critical for recognition in vivo. The constructs we examined are shown in Fig. 1. Mutation of the basic residues in the sequence PLRSKK of ␣ subunit to acidic residues, i.e. PLESEE, reduced GalNAc transfer to GLuc␣ 42-fold by ␤4GalNAc-T3-F and 19-fold by ␤4GalNAc-T4-F (Fig. 8, A and B), confirming that the basic residues within the sequence PLRSKK are essential for substrate recognition by both ␤4GalNAc-T3-F and ␤4GalNAc-T4-F. The minimal amount of GalNAc transferred to GLuc␣(PLESEE) is similar to that transferred to GLucTrf. Trf is a glycoprotein that we have previously shown is not recognized by the pituitary ␤1,4GalNAc transferase (2). Adding the CA1-19 peptide sequence to the carboxyl terminus of GLucTrf increased the amount of GalNAc transferred to this substrate 9-fold for ␤4GalNAc-T3-F and 135-fold for ␤4GalNAc-T4-F (Fig. 8E), thus converting GLucTrf from a poor to a highly effective substrate that is recognized by both ␤4GalNAc-T3-F and ␤4GalNAc-T4-F. Similarly adding the CA1-19 peptide sequence to the carboxyl terminus of GLuc␣(PLESEE) restored the acceptor activity of GLuc␣(PLESEE)CA1-19 for both ␤4GalNAc-T3-F and ␤4GalNAc-T4-F. In fact GLuc␣(PLESEE)CA1-19 was modified to a greater extent than GLuc␣(PLRSKK) by both ␤4GalNAc-T3-F and ␤4GalNAc-T4-F. Adding CA1-19 to GLuc␣(PLRSKK) enhanced GalNAc transfer to the ␣ subunit oligosaccharides 4.2-fold for ␤4GalNAc-T3-F and 23-fold for ␤4GalNAc-T4-F, consistent with our observation above that ␤4GalNAc-T4-F has a stronger affinity for CA1-19 than for ␣32-46.
Minimal Requirements for Recognition of the Peptide Motif by ␤4GalNAc-T3-F and ␤4GalNAc-T4-F-The effectiveness of the CA1-19 recognition motif at promoting GalNAc transfer by ␤4GalNAc-T3-F and ␤4GalNAc-T4-F when this motif was added to the carboxyl terminus of multiple glycoprotein substrates suggested that the context of the sequence was not critical and that the in vitro assay could be used to determine the minimal requirements for recognition by ␤4GalNAc-T3-F and ␤4GalNAc-T4-F. We therefore appended different regions of the CA1-19 peptide sequence to the carboxyl terminus of GLuc␣(PLESEE) and compared these constructs as substrates for ␤4GalNAc-T3-F and ␤4GalNAc-T4-F (Fig. 8). The fulllength sequence CA1-19 was the best substrate for both ␤4GalNAc-T3-F (panel C) and ␤4GalNAc-T4-F (panel D).
Removing the carboxyl-terminal 3 amino acids to form CA1-16 reduced the extent of modification by 50% for both enzymes. The smallest region that was recognized by both ␤4GalNAc-T3-F and ␤4GalNAc-T4-F was CA7-16, which has the sequence QKITKRKKEK. The carboxyl-terminal region consisting of CA11-19 was recognized, whereas the aminoterminal half of the sequence consisting of CA1-10 was not. Thus, residues outside of the critical region CA7-16 enhance recognition but are not themselves essential.
Western blot analysis of ␤4GalNAc-T3-F and ␤4GalNAc-T4-F under nonreducing (Fig. 9A) and reducing (Fig. 9B) conditions using an antibody directed at the FLAG epitope revealed proteolytic cleavage of the major fraction of both transferases. The two fragments produced were covalently associated via one or more disulfide bonds because in the absence of reducing agents, both ␤4GalNAc-T3-F and ␤4GalNAc-T4-F migrated as doublets with M r 140,000 -160,000. Equal amounts of enzyme activity based on transfer of GalNAc to GLuc␣(PLRSKK) and to GLuc␣(PLRSKK)CA1-19, respectively, were subjected to quantitation by Western blot following separation by SDS-PAGE using an antibody directed at the FLAG epitope (Fig. 9C).  Alpha(CA1-19) lane in panel C). Thus, the specific activity for transfer of GalNAc associated with ␤4GalNAc-T3-F and ␤4GalNAc-T4-F is similar for transfer to GLuc␣(PLRSKK)CA1-19, whereas the specific activity associated with ␤4GalNAc-T4-F for transfer of GalNAc to GLuc␣(PLRSKK) is significantly lower than for transfer to GLuc␣(PLRSKK)CA1-19.
Clearly more ␤4GalNAc-T4-F was required to transfer the same amount of GalNAc to GLuc␣(PLRSKK) than to GLuc␣(PLRSKK)CA1-19. Furthermore, the specific activities of ␤4GalNAc-T3-F and ␤4GalNAc-T4-F were similar for transfer of GalNAc when GLuc␣(PLRSKK)CA1-19 was used as the substrate. Thus, the difference in the affinity of ␤4GalNAc-T4-F for the recognition motif found in the ␣ subunit as compared with the motif in CA6 accounts for this apparent difference in activity.

DISCUSSION
Unique oligosaccharide structures occur on a number of glycoproteins, serving as information-bearing signposts whose recognition is utilized for intracellular trafficking and/or extracellular clearance of these molecular species. For example, ␤1,4-linked GalNAc-4-SO 4 and Sia␣2,6GalNAc on glycoprotein hormones (16 -18) are recognized by receptors that mediate clearance of these hormones from the blood, and Man-6-PO 4 is recognized by receptors that direct enzymes and other glycoproteins bearing these structures to lysosomes (19). However, much remains to be learned about how these specific oligosaccharide structures are added to individual glycoproteins and, in some instances, how individual glycosylation sites are chosen. We recently described a system that provides unique opportunities to identify those parameters that are critical for synthesis of unique carbohydrate structures in vivo (8). This strategy, based on chimeric proteins consisting of GLuc followed by the glycoprotein of interest and an epitope tag, has defined the requirements for the protein-specific addition of   ␤1,4-linked GalNAc to N-linked oligosaccharides using CHO cells expressing ␤4GalNAc-T3 or ␤4GalNAc-T4 (8). In the current studies, we have used the identical GLuc chimeras to examine kinetic parameters for GalNAc addition by ␤4GalNAc-T3 and ␤4GalNAc-T4 to glycoprotein substrates in vitro.
Our in vitro studies demonstrate that the conditions optimal for transfer of GalNAc to glycoprotein substrates containing a particular recognition determinant surprisingly are not typical of the conditions traditionally utilized for in vitro characterization of glycosyltransferase properties. Transfer of GalNAc to glycoprotein acceptors by both ␤4GalNAc-T3-F and ␤4GalNAc-T4-F is progressively inhibited by increasing Cl Ϫ concentrations above 10 mM. The actual concentration of ions to which these transferases would be exposed in the cellular milieu is thus an interesting question. The concentration of Mn 2ϩ has not been directly measured in the Golgi. However, culturing cells in medium containing 100 M Mn 2ϩ induces transport of a cis-Golgi localized protein GPP130 to multivesicular bodies (20). Furthermore, culturing cells in the presence of 1 mM Mn 2ϩ reduces cell viability (21). Such studies suggest that the actual concentration of Mn 2ϩ is likely less than 200 M in the Golgi where ␤4GalNAc-T3 and ␤4GalNAc-T4 reside and carry out their functional activities; this would be consistent with their respective affinities for Mn 2ϩ of 0.25 and 0.06 mM.
The behavior of GalNAc transfer to a small-molecule acceptor that does not contain a peptide recognition determinant stands in sharp contrast to the transfer of GalNAc to GLuc␣(PLRSKK)CA1-19. In the former case, the transfer of GalNAc to GlcNAc␤-pNP is not inhibited by 20 mM MnCl 2 . Furthermore, the IC 50 values for inhibition of GalNAc addition to GLuc␣(PLRSKK)CA1-19 by agalacto-hCG and the peptide CA1-19 are shifted to higher concentrations in the presence of 5 mM MnCl 2 , suggesting that it is peptide recognition that is sensitive to the concentration of salt rather than the catalytic transfer of GalNAc from UDP-GalNAc to the oligosaccharide acceptor.
Both ␤4GalNAc-T3-F and ␤4GalNAc-T4-F require that a divalent cation be present for transfer of GalNAc from UDP-GalNAc to the glycoprotein oligosaccharide. They do, however, differ in the order of their preference for divalent cations, sug-  gesting that although the sequences of the catalytic domains are 67% identical, there are structural features that differentiate their catalytic properties. These same differences may also account for the difference in the affinity for Mn 2ϩ .
In previous studies using detergent extracts of bovine pituitary, we reported that GalNAc is transferred to the synthetic intermediate GlcNAc 2 Man 3 GlcNAc 2 Asn on hCG with an apparent K m of 4.4 M and to glycopeptides bearing the same synthetic intermediate with an apparent K m of 1.2-2.6 mM (22). The Ͼ270-fold difference in the K m for hCG versus glycopeptides can account for the ability of the ␤4GalNAc transferase present in pituitary extracts to add GalNAc to oligosaccharides on substrate glycoproteins such as hCG that contain a recognition determinant but not to the same oligosaccharide structures on glycoproteins such as Trf that do not contain a recognition determinant (2). hCG is a dimeric glycoprotein consisting of two noncovalently associated subunits, each of which has two Asn-linked oligosaccharides. The ␣ and ␤ subunits each contain a recognition determinant that can mediate the selective addition of GalNAc to the separated subunits. We used an in vitro assay to identify the basic residues in the sequence PLRSKK as being essential components of the glycoprotein hormone ␣ subunit recognition determinant (23). Both ␤4GalNAc-T3 and ␤4GalNAc-T4 display the predicted specificity for the glycoprotein hormone ␣ subunit versus Trf when expressed in CHO cells (8). Among those glycoproteins identified as bearing N-linked oligosaccharides with ␤1,4-linked Gal-NAc, we have demonstrated that regions with a recognition determinant of clustered basic residues are critical for the protein-specific addition of GalNAc by ␤4GalNAc-T3 and ␤4GalNAc-T4 to the ␣ subunit, to CA6 (8,15), and to SorLA/ LR11 (5).
Our in vitro studies reveal significant differences in the specificity of ␤4GalNAc-T3 versus ␤4GalNAc-T4 for the recognition motifs in the ␣ subunit and CA6. Transfer of GalNAc to GLuc␣(PLRSKK)CA1-19, which contains both recognition motifs, is inhibited by agalacto-hCG but not by agalacto-Trf, indicating that the peptide recognition motif is required for inhibition of transfer of GalNAc. However, inhibition of Gal-NAc transfer to GLuc␣(PLRSKK)CA1-19 by ␤4GalNAc-T4-F requires a 42-fold higher concentration of agalacto-hCG (IC 50 ϭ 1.23 M) than does inhibition of GalNAc transfer by ␤4GalNAc-T3-F (IC 50 ϭ 28.8 nM). The difference in IC 50 indicates that ␤4GalNAc-T3-F has a significantly stronger affinity for the recognition motif in agalacto-hCG than does ␤4GalNAc-T4-F.
Further support for the role of peptide recognition was obtained from inhibition studies. Peptides containing a recognition motif fully inhibited transfer of GalNAc to glycoprotein substrates containing a recognition motif but had no effect on transfer to GlcNAc␤-pNP. Peptide recognition and transfer of GalNAc appear to be independent events. The IC 50 values of 0.38 and 4.59 M obtained for ␣32-46 inhibition of GalNAc transfer by ␤4GalNAc-T3-F and ␤4GalNAc-T4-F, respectively, indicate that binding of the peptide recognition motif accounts for the majority of the affinity for glycoprotein substrates and that ␤4GalNAc-T3-F has a stronger affinity for the ␣ subunit recognition motif than does ␤4GalNAc-T4-F.
High levels of ␤4GalNAc-T3 and ␤4GalNAc-T4 expression in stably transformed CHO cells had prevented a definitive assessment of the degree to which the presence of a recognition determinant could increase modification of N-linked oligosaccharides with GalNAc in vivo (8). Our in vitro analysis demonstrated that mutation of the basic residues within the recognition motif of the ␣ subunit from PLRSKK to PLESEE reduced the amount of GalNAc transferred to GLuc␣ by ␤4GalNAc-T3-F and ␤4GalNAc-T4-F by 42-and 19-fold, respectively. Adding the CA1-19 sequence to GLucTrf increased the amount of GalNAc transferred to GLucTrf(CA1-19) 9-and 135-fold for ␤4GalNAc-T3-F and ␤4GalNAc-T4-F, respectively. Adding the CA1-19 sequence to the carboxyl terminus of GLuc␣(PLESEE) yielded a substrate, GLuc␣(PLESEE)CA1-19, that was more efficiently modified with GalNAc than GLuc␣(PLRSKK) by both ␤4GalNAc-T3-F and ␤4GalNAc-T4-F. Thus, the CA1-19 sequence was recognized more efficiently than the PLRSKK sequence. Furthermore, the location of the recognition determinant within the linear sequence of FIGURE 9. Western blot analysis of affinity-purified ␤4GalNAc-T3-F and ␤4GalNAc-T4-F. ␤4GalNAc-T3-F and ␤4GalNAc-T4-F were affinity-purified from medium using anti-FLAG-agarose and elution with FLAG peptide. The affinity-purified enzymes were examined by SDS-PAGE in the absence (A) or presence (B and C) reducing agent. Following electrophoresis, the proteins were electrophoretically transferred to a PVDF membrane, and proteins containing the FLAG epitope were visualized using anti-FLAG. In C, identical light units of activity using either GLuc␣(PLRSKK), indicated as Alpha, or GLuc␣(PLRSKK)CA1-19, indicated as Alpha(CA1- 19), were loaded onto the gel. In the absence of reducing agents, ␤4GalNAc-T3-F and ␤4GalNAc-T4-F both migrate as closely spaced doublets reflecting the presence of the uncleaved form and the disulfide-linked cleaved form.
the peptide was not critical because the PLRSKK motif precedes the two glycosylated Asn residues in the ␣ subunit, whereas the CA1-19 sequence has been added to the carboxyl terminus of the ␣ subunit. However, in spatial terms, accessibility of the peptide motif within the three-dimensional structure of the substrate relative to the oligosaccharide structure is likely critical. Based on the crystal structure of Trf (24), the CA1-19 added to the carboxyl terminus of Trf is predicted to be in close proximity to both N-linked oligosaccharides on Trf. Deletion of either glycosylation site on GLucTrfCA1-19 did not prevent modification with GalNAc (not shown), indicating that both sites can be modified in the presence of CA1-19.
The entire sequence of CA1-19 was not required for recognition by either ␤4GalNAc-T3-F or ␤4GalNAc-T4-F. Residues 7-16, QKITKRKKEK, were essential for recognition, whereas residues 1-6 and 17-19 appeared not to be essential but did enhance recognition. Given our previous insights on the importance of clustered basic residues (5,8,23), the bolded residues QKITKRKKEK within CA1-19 and PTPLRSKK within the ␣32-46 sequence may be the most critical for recognition by ␤4GalNAc-T3 and ␤4GalNAc-T4, whereas the other residues contribute to presentation of these basic residues to the transferase binding site.
The difference in substrate preference exhibited by ␤4GalNAc-T3 versus ␤4GalNAc-T4 was confirmed at the protein level. Western blot analysis with anti-FLAG demonstrated that a much greater quantity of ␤4GalNAc-T4-F than ␤4GalNAc-T3-F was required to provide the same light units of transferase activity based on GalNAc transfer to GLuc␣-(PLRSKK). In contrast, when equal light units of transferase activity based on GalNAc transfer to GLuc␣(PLRSKK)-CA1-19 were examined, the amounts of ␤4GalNAc-T3-F and ␤4GalNAc-T4-F present were similar. Thus, ␤4GalNAc-T3-F and ␤4GalNAc-T4-F have similar specific activities when using GLuc␣(PLRSKK)CA1-19 as the substrate.
The in vitro approach here expands our understanding of the contribution of peptide recognition determinants to the protein-specific transfer of GalNAc by ␤4GalNAc-T3 and ␤4GalNAc-T4. ␤4GalNAc-T3 and ␤4GalNAc-T4 are closely related structurally; however, they differ in their substrate preferences with ␤4GalNAc-T4 showing a strong preference for the recognition determinant represented by CA1-19 as compared with that present in the ␣ subunit. There is a single peptide binding site in both ␤4GalNAc-T3 and ␤4GalNAc-T4 because the peptides CA1-19 and ␣32-46 are both able to completely inhibit transfer of GalNAc to GLuc␣(PLRSKK)CA1-19. The IC 50 values for CA1-19 and ␣32-46 fall in the range of 0.38 -4.59 M, whereas the apparent K m for transfer to a glycopeptide acceptor devoid of any recognition determinant is 1.2-2.6 mM. Thus, as is seen with the phosphotransferase that adds GlcNAc-PO 4 to oligomannose structures on lysosomal enzymes (25), the affinity for the peptide recognition determinant is much stronger than for the oligosaccharide acceptor.
The kinetics for GalNAc transfer by ␤4GalNAc-T3-F and ␤4GalNAc-T4-F in vitro indicate that both transferases are highly protein-specific. Multiple lines of evidence indicate that ␤4GalNAc-T3 has stronger affinity for the recognition determinant in the hormone ␣ subunit than does ␤4GalNAc-T4 and, as a result, ␤4GalNAc-T3-F transfers GalNAc to the ␣ subunit oligosaccharides more efficiently than ␤4GalNAc-T4-F under the conditions of the in vitro assay. Nonetheless, ␤4GalNAc-T3 and ␤4GalNAc-T4 have similar specificities, and either or both could contribute to the modification of the reproductive glycoprotein hormones in vivo. Supporting this conclusion is the marked decrease in GalNAc transfer by both ␤4GalNAc-T3-F and ␤4GalNAc-T4-F in vitro when the PLRSKK sequence in GLuc␣ is mutated to PLESEE. Gotoh et al. (7) have reported that ␤4GalNAc-T4 transcripts are abundantly expressed in mouse brain, whereas ␤4GalNAc-T3 transcripts are not. In contrast, we have found that both ␤4GalNAc-T3 and ␤4GalNAc-T4 transcripts are present in pituitaries from mice. 3 Thus, in the pituitary, either or both transferases may contribute to the modification of oligosaccharides on the glycoprotein hormones with GalNAc. It is possible that ␤4GalNAc-T3 and ␤4GalNAc-T4 are both required for efficient modification of the oligosaccharides on hormones such as luteinizing hormone.
Our in vitro studies indicate that it is now possible to generate glycoprotein substrates that differ in their affinity for ␤4GalNAc-T3 and ␤4GalNAc-T4. Expression of substrates with different affinities for ␤4GalNAc-T3 and ␤4GalNAc-T4 in cells that express one or both transferases offers the opportunity to examine how differences in location, affinity for the substrate, and relative levels of transferases determine the structural outcome in vivo. Our ability to use the identical substrates for in vitro and in vivo studies provides a unique opportunity to examine how different parameters affect the pattern and extent of glycosylation within the milieu of the Golgi.