Poly-N-acetyllactosamine Synthesis in BranchedN-Glycans Is Controlled by Complemental Branch Specificity of i-Extension Enzyme and β1,4-Galactosyltransferase I*

Poly-N-acetyllactosamine is a unique carbohydrate that can carry various functional oligosaccharides, such as sialyl Lewis X. It has been shown that the amount of poly-N-acetyllactosamine is increased inN-glycans, when they contain Galβ1→4GlcNAcβ1→6(Galβ1→4GlcNAcβ1→2)Manα1→6 branched structure. To determine how this increased synthesis of poly-N-acetyllactosamines takes place, the branched acceptor was incubated with a mixture of i-extension enzyme (iGnT) and β1,4galactosyltransferase I (β4Gal-TI). First,N-acetyllactosamine repeats were more readily added to the branched acceptor than the summation of poly-N-acetyllactosamines formed individually on each unbranched acceptor. Surprisingly, poly-N-acetyllactosamine was more efficiently formed on Galβ1→4GlcNAcβ1→2Manα→R side chain than in Galβ1→4GlcNAcβ1→6Manα→R, due to preferential action of iGnT on Galβ1→4GlcNAcβ1→2Manα→R side chain. On the other hand, galactosylation was much more efficient on β1,6-linked GlcNAc than β1,2-linked GlcNAc, preferentially forming Galβ1→4GlcNAcβ1→6(GlcNAcβ1→2)Manα1→6Manβ →R. Starting with this preformed acceptor,N-acetyllactosamine repeats were added almost equally to Galβ1→4GlcNAcβ1→6Manα→R and Galβ1→4GlcNAcβ1→2Manα→R side chains. Taken together, these results indicate that the complemental branch specificity of iGnT and β4Gal-TI leads to efficient and equal addition ofN-acetyllactosamine repeats on both side chains of GlcNAcβ1→6(GlcNAcβ1→2)Manα1→6Manβ→R structure, which is consistent with the structures found in nature. The results also suggest that the addition of Galβ1→4GlcNAcβ1→6 side chain on Galβ1→4GlcNAcβ1→2Man→R side chain converts the acceptor to one that is much more favorable for iGnT and β4Gal-TI.

Poly-N-acetyllactosamine is a unique carbohydrate that can carry various functional oligosaccharides, such as sialyl Lewis X. It has been shown that the amount of poly-N-acetyllactosamine is increased in N-glycans, when they contain Gal␤134GlcNAc␤136(Gal␤134Glc-NAc␤132)Man␣136 branched structure. To determine how this increased synthesis of poly-N-acetyllactosamines takes place, the branched acceptor was incubated with a mixture of i-extension enzyme (iGnT) and ␤1,4galactosyltransferase I (␤4Gal-TI). First, N-acetyllactosamine repeats were more readily added to the branched acceptor than the summation of poly-N-acetyllactosamines formed individually on each unbranched acceptor. Surprisingly, poly-N-acetyllactosamine was more efficiently formed on Gal␤134GlcNAc␤132Ma-n␣3 R side chain than in Gal␤134GlcNAc␤136Ma-n␣3 R, due to preferential action of iGnT on Gal␤134G-lcNAc␤132Man␣3 R side chain. On the other hand, galactosylation was much more efficient on ␤1,6-linked GlcNAc than ␤1,2-linked GlcNAc, preferentially forming Gal␤134GlcNAc␤136(GlcNAc␤132)Man␣136Man␤ 3 R. Starting with this preformed acceptor, N-acetyllactosamine repeats were added almost equally to Gal␤134GlcNAc␤136Man␣3 R and Gal␤134GlcNAc-␤132Man␣3 R side chains. Taken together, these results indicate that the complemental branch specificity of iGnT and ␤4Gal-TI leads to efficient and equal addition of N-acetyllactosamine repeats on both side chains of GlcNAc␤136(GlcNAc␤132)Man␣136Man␤3 R structure, which is consistent with the structures found in nature. The results also suggest that the addition of Gal␤134GlcNAc␤136 side chain on Gal␤134GlcNAc-␤132Man3 R side chain converts the acceptor to one that is much more favorable for iGnT and ␤4Gal-TI.
The increased expression of sialyl Le x apparently takes place in those tumor cells when ␣1,3-fucosyltransferase is also present. In fact, highly metastatic colonic carcinoma cells express more sialyl Le x in poly-N-acetyllactosamines than poorly metastatic counterparts (27). Moreover, our recent studies demonstrated that B16 mouse melanoma cells produced many more lung tumor foci after the cells were transfected with ␣1,3fucosyltransferase to form sialyl Le x in poly-N-acetyllactosamines (29). These results, as a whole, indicate that the formation of poly-N-acetyllactosamine plays a critical role for carbohydrate recognition in cell-cell interaction.
Previously, we have demonstrated that poly-N-acetyllactosamines in core 2 branched O-glycans are synthesized through iGnT and a newly discovered member of ␤1,4-galactosyltransferase, ␤4Gal-TIV (30). We have also found that ␤4Gal-TI, iGnT, and I-branching enzyme are involved in the synthesis of I-branched poly-N-acetyllactosamines (31). In the same study, we found that the addition of N-acetyllactosamine repeats to linear poly-N-acetyllactosamines is preferred over the extension of N-acetyllactosamine to I-branches, mainly because the first galactosylation of a GlcNAc␤136 branch is inefficient. These studies thus demonstrate intricate interaction between a specific glycosyltransferase and an acceptor molecule, which largely contributes to the control of poly-N-acetyllactosamine synthesis. In these studies, we utilized molecular tools that have recently become available, such as cDNAs encoding iextension enzyme (32), novel members of ␤4Gal-T (33), and I-branching enzyme (34).
These results prompted us in the present study to determine how poly-N-acetyllactosamine is formed in branched N-glycans. Our results demonstrated that the addition of GlcNAc␤136Man␣3 R side chain renders the resultant GlcNAc␤136(GlcNAc␤132)Man␣136Man␤3 R an extremely efficient acceptor for ␤4Gal-TI and i-extension enzyme. Moreover, we found that the i-extension enzyme prefers Gal␤134GlcNAc␤132Man␣3 R side chain over Gal␤134 GlcNAc␤136Man␣3 R side chain, whereas the opposite is true for ␤4Gal-TI. Such complementary effects apparently result in forming similar size and abundance of poly-N-acetyllactosamines in both chains of GlcNAc␤136 (GlcNAc␤132)-Man␣3 R structures.

EXPERIMENTAL PROCEDURES
Isolation and Expression of cDNA Encoding iGnT-cDNA encoding iGnT was cloned into pcDNA3.1, resulting in pcDNA3.1-iGnT as described previously (32). pcDNAI-A, harboring cDNA encoding a signal sequence and an IgG binding domain of Staphylococcus aureus protein A, was constructed as described before (35). The catalytic domain of iGnT was cloned into this vector, resulting in pcDNAI-A⅐iGnT (32).
pcDNAI-A and pcDNAI-A⅐iGnT were separately transfected with LipofectAMINE Plus (Life Technologies, Inc.) into COS-1 cells as described previously (36). The chimeric enzyme released into serum-free OPTI-MEM was used after adsorbing the protein A chimeric enzyme to IgG-Sepharose 6FF (Amersham Pharmacia Biotech) as described previously (36). Alternatively, the culture medium was concentrated 100fold or 1000-fold by a Centricon 10 concentrator (Amicon) and directly used as an enzyme source. In most of the studies, the concentrated culture medium was used for iGnT because IgG-Sepharose bound enzymes had a low activity, as seen for other glycosyltransferases (37,38). Typically, the activities of iGnT in the incubation mixture was 38.0 nmol/h/ml (for addition of one GlcNAc) or 380 nmol/h/ml (for poly-Nacetyllactosamine synthesis), using 0.5 mM Gal␤134Glc␤3p-nitrophenol (Toronto Research Chemicals) as an acceptor. The medium from mock-transfected COS-1 cells contained less than one-fifth of iGnT activity compared with that derived from pcDNAI-A⅐iGnT-transfected COS-1 cells, as described (32).
When [ 3 H]Gal␤134GlcNAc␤136(Gal␤134GlcNAc␤132)Man␣1 36Man␤13octyl (compound 11, see above) and Gal␤134GlcNAc-␤136([ 3 H]Gal␤134GlcNAc␤132)Man␣136Man␤13octyl (compound 12, see above) were used as acceptors, nonradioactive donor substrates were used. The products were purified by HPLC using the same NH 2 -bonded silica column as described above. A peak containing one N-acetyllactosamine repeat was digested with diplococcal ␤-galactosidase (41), and the digest was purified by a Sep-Pak column and then analyzed by HPLC as described above. A peak containing two N-acetyllactosamine repeats was sequentially digested with diplococcal ␤-galactosidase, jack bean ␤-N-acetylglucosaminidase, and diplococcal ␤-galactosidase. After each digestion, the digest was analyzed by HPLC using the same NH 2 -bonded column as described above. Aliquots were taken for determining radioactivity, obtaining the ratio of the radioactivity in the product and starting material.
Analysis of Products by Endo-␤-galactosidase Digestion-Products were digested with Escherichia freundii endo-␤-galactosidase for 18 h at 37°C (43). The digestion conditions used allowed the cleavage of galactose linkage, where no ␤1,6-linked N-acetylglucosamine is attached (2,43). The digests were subjected to HPLC using AX-5 column.

Addition of N-Acetylglucosamine to N-Glycan Acceptor by i-Extension
Enzyme-To determine whether iGnT has a preference for the addition of N-acetylglucosamine to one of Nglycan acceptors, Gal␤134GlcNAc␤136Man␣3 R, Gal␤1 34GlcNAc␤132Man␣3 R, and Gal␤134GlcNAc␤136 (Gal␤134GlcNAc␤132)Man␣3 R were incubated with iGnT. First, iGnT acted in almost identical efficiency on Gal␤13 4GlcNAc␤136Man␣3 R and Gal␤134GlcNAc␤13 2Man␣3 R (Figs. 1A and 2, A and B), indicating that the enzyme does not prefer one acceptor over the other. Second, the branched acceptor has a better affinity and higher V max for addition of one N-acetylglucosamine ( Fig. 1A and Table I). When the products from the 10-h incubation of the branched acceptor was analyzed, the majority of the products contained only one N-acetylglucosamine residue, and a mere 5.7% (in molar ratio) of the whole products contained two N-acetylglucosamine residues (Fig. 2C).
Poly-N-acetyllactosamine Synthesis on Unbranched or Branched Acceptor-As shown previously (30), among different members of ␤4Gal-Ts ␤4Gal-TI was most efficient in adding galactose to GlcNAc␤136Man␣136Man␤13octyl. Similarly, we found in the present study that ␤4Gal-TI is the most efficient in adding a galactose to GlcNAc␤132Man␣136 Man␤13octyl ( Fig. 1B; see also Table II).
To determine how N-acetyllactosamine repeats are added to two different side chains, Gal␤134GlcNAc␤136Man␣3 R or Gal␤134GlcNAc␤132Man␣3 R was incubated with iGnT, ␤4Gal-TI, UDP-[ 3 H]Gal and UDP-[ 3 H]GlcNAc. Fig. 4, A and B, illustrates that poly-N-acetyllactosamines were almost equally formed from these two acceptors. We then incubated a branched acceptor, Gal␤134GlcNAc␤136(Gal␤134GlcNAc-␤132)Man␣136Man␤13octyl with i-GnT, ␤4Gal-TI, and radioactive donor substrates. First, the branched acceptor incorporated 8.4 times more [ 3 H]Gal and [ 3 H]GlcNAc than the unbranched acceptors when incubated under the same conditions (Fig. 4C). The results also demonstrated that the branched acceptor contained more N-acetyllactosamine units, and the product containing three N-acetyllactosamine units constituted the major product (Fig. 4C). In contrast, the major product obtained from the unbranched acceptors contained one N-acetyllactosamine unit (Fig. 4, A and B). These results indicate that the addition of Gal␤134GlcNAc␤136 branch to Gal␤134GlcNAc␤132Man␣3 R in the acceptor has a synergistic effect on the branched acceptor, forming many more N-acetyllactosamine repeats than a summation of N-acetyllactosamine repeats formed when each branch was individually assayed. Moreover, such an effect extends to the formation of longer poly-N-acetyllactosamines.
Poly-N-acetyllactosamine Is Preferentially Added to Gal␤134GlcNAc␤132 Side Chain on Gal␤134GlcNAc-␤132(Gal␤134GlcNAc␣136)Man␣3 R Acceptor-To determine how each side chain of the branched acceptor was elongated with N-acetyllactosamine repeats, [ 3 H]Gal␤134-GlcNAc␤136(Gal␤134GlcNAc␤132)Man␣3 R, synthesized as described above, was incubated with iGnT, ␤4Gal-TI, and nonradioactive donor substrates. The products obtained exhibited an elution profile similar to that obtained in the above experiments (Fig. 4D). The difference between Fig. 4C and Fig. 4D was due to the fact that larger products were more visible in Fig. 4C
These results, taken together, indicate that preferential addition of N-acetylglucosamine by iGnT to Gal␤134GlcNAc-␤132Man␣3 R side chain leads into preferential formation of poly-N-acetyllactosamine in Gal␤134GlcNAc␤132Man␣3 R branch.
yield after endo-␤-galactosidase digestion. The results indicate that the molar ratio between Gal␤134GlcNAc␤136([ 3 H]Gal-␤134GlcNAc␤132)Man3 R and the whole products was 1.0:3.1, because each initial product contained three radioactive sugars (Fig. 8A; see also Fig. 9A). This can be translated into the conclusion that the N-acetyllactosamine repeat was formed on Gal␤134GlcNAc␤136Man␣3 R and GlcNAc-␤132Man␣vR in the acceptor in a ratio of 1:2.1.
When the same acceptor was incubated with increased amount of iGnT, and the ratio of iGnT and ␤4Gal-TI was 1:2, 1.4/7.2 of the radioactivity was recovered as GlcNAc-␤136([ 3 H]Gal␤134GlcNAc␤132)Man␣3 R after endo-␤-galactosidase digestion (Fig. 8B). The results indicate that Nacetyllactosamine repeat was added to Gal␤134GlcNAc-␤136Man branch 1.4 times more than to GlcNAc␤132Man branch. These results indicate that both branches were almost equally elongated by N-acetyllactosamine repeats when iGnT and ␤4Gal-TI was in a ratio of 1:2 under the incubation conditions employed (Fig. 9A).
As shown above, galactosylation takes place preferentially on ␤1,6-linked GlcNAc in GlcNAc␤136(GlcNAc␤132)Man-␣136Man␤13octyl. These combined results thus indicate that poly-N-acetyllactosamine extension almost equally takes place on both GlcNAc␤136Man and GlcNAc␤132Man side chains after the formation of Gal␤134GlcNAc␤136(GlcNAc␤1 32)Man␣136Man␤3 R. DISCUSSION The present study demonstrates that Gal␤134GlcNAc-␤132Man3 R antenna is extended by N-acetyllactosamine repeats as much as does Gal␤134GlcNAc␤136Man3 R antenna ( Figs. 1 and 2). When both galactosylated side chains are present in a branched acceptor, Gal␤134GlcNAc␤132Man-3 R is utilized more efficiently to add N-acetyllactosamine repeats than Gal␤134GlcNAc␤136Man3 R (Figs. 4 -6). This is mainly because iGnT preferentially acts on Gal␤134GlcNA-c␤132Man side chain (Figs. 2 and 3). Similarly, it was reported that iGnT from Novikoff hepatoma acts 1.6 times more efficiently on Gal␤134GlcNAc␤132Man␣3 R side chain than on Gal␤134GlcNAc␤136Man␣3 R side chain in a branched acceptor (44). The results suggest that iGnT present in Novikoff hepatoma is probably the same as the cloned iGnT used in the present study.
More importantly, the amount of poly-N-acetyllactosamine formed in the branched acceptor was much more than the summation of two separate reactions using either one of these side chains (Fig. 4). These results strongly suggest that addition of Gal␤134GlcNAc␤136 side chain on Gal␤134GlcNAc ␤132Man␣3 R side chain, or vice versa, must change the conformation of Gal␤134GlcNAc␤132Man␣3 R and possibly that of Gal␤134GlcNAc␤136Man␣3 R. As a result, both side chains become more favorable for iGnT and ␤4Gal-TI to act. FIG. 5. Analysis of the poly-N-acetyllactosaminyl products derived from asymmetrically labeled acceptors. A, ␤-galactosidase digest of peak 1 in Fig. 4D. B-D, sequential digestion of peak 2 in Fig.  4D by ␤-galactosidase (B), ␤-N-acetylhexosaminidase (C), and ␤-galactosidase (D). E-H, the products derived from Gal␤134GlcNAc-␤136([ 3 H]Gal␤134GlcNAc␤132)Man␣3 R were separated by HPLC and those containing one N-acetyllactosamine repeat (E) and two Nacetyllactosamine repeats (F-H) were digested with ␤-galactosidase (E) or sequentially digested with ␤-galactosidase (F), followed by ␤-N-acetylhexosaminidase (G) and ␤-galactosidase (H), and subjected to HPLC. Numbers indicate the relative ratio of the radioactivity of the digested product (solid line) and the starting material (dotted line) (see Fig. 6). This enhancement can be observed even when first galactosylation takes place on GlcNAc␤136(GlcNAc␤132)Man ␣136Man␤13octyl (see Tables II and III). In this context, it is noteworthy that GnTV adds N-acetylglucosamine only when Man␣136Man is in a gauche-gauche conformation (45). This result indicates that the action of GnTV restricts the conformation of resultant branched oligosaccharide, GlcNAc␤136(Glc-NAc␤132)Man␣136Man␤3 R. This conclusion is consistent with the results in the recent report indicating that Gal␤134GlcNAc␤136 side chain may be extended toward the proximal region of N-glycans (46). Addition of GlcNAc␤136 by chemical synthesis may also bring some conformational change in both side chains, which is favorable for the actions by iGnT and ␤4Gal-TI.
It is rather striking that Gal␤134GlcNAc␤132Man ␣136Man␤3 R is as good as Gal␤134GlcNAc␤136Man ␣136Man␤3 R as an acceptor for iGnT ( Figs. 1 and 2 and Table I). The results obtained in the present study are, however, consistent with the structural data on human red cell band 3 (18,47). Human red cell band 3 does not contain Gal␤134GlcNAc␤136Man␣136Man␤3 R side chain yet

TABLE III
Kinetic properties of ␤4Gal-TI Incubation conditions were exactly the same as Table II. Arrows indicate where galactose is added (see Fig. 7A). 33 197 a Relative values. V max of ␤4-Gal-TI is compared to the V max (118.3 nmol/h/ml) obtained using GlcNAc␤1-6Man␣1-6Man␤-octyl as an acceptor.
b Only one galactose was added under the incubation conditions used. However, distribution between these two side chains was not determined because Gal␤1-4GlcNAc␤1-2Man␣1-6Man␤-octyl and Gal␤1-4GlcNAc␤1-6Man␣1-6Man␤-octyl, which were obtained after ␤-Nacetylglucosaminidase treatment of products, could not be separated in various chromatographic conditions. contains very extended poly-N-acetyllactosamines. It is also noteworthy that a side chain extending from Man␣136Man-␤3 R contains more poly-N-acetyllactosamines than that extending from Man␣133Man␤3 R in human red cell band 3 (18,47).
The present study demonstrated that poly-N-acetyllactosamine extension takes place more efficiently on Gal␤134-GlcNAc␤132Man␣3 R side chain than Gal␤134GlcNAc-␤136Man␣3 R side chain when a fully galactosylated branch ed acceptor was utilized (Figs. 4 and 5). In contrast, poly-Nacetyllactosamine extension took place almost equally between two branches when increased amount of iGnT and Gal␤134 GlcNAc␤136(GlcNAc␤132)Man␣3 R as an acceptor were used (Figs. 8 and 9). The results obtained in the latter experiments are consistent with those obtained on the structural analysis of glycoproteins containing poly-N-acetyllactosamines from granulocytes (6, 7) and glycoproteins, such as human erythropoietin produced in Chinese hamster ovary cells (50 -54). In particular, NMR studies of the N-glycans isolated from the recombinant erythropoietin demonstrated that Nacetyllactosamine extension takes place almost equally in Gal␤134GlcNAc␤136Man␣3 R and Gal␤134GlcNAc-␤132Man␣3 R side chains (52)(53)(54). These results, as a whole, strongly suggest that the concentration of iGnT relative to that of ␤4Gal-TI may be more than that estimated from the activities in total cell lysates. Further studies will be of significance to determine whether iGnT is more enriched in narrower compartments of the Golgi than ␤4Gal-TI (see also Ref. 55).
The above results were obtained most likely due to the following reasons. First, it is almost certain that Gal␤134Glc-NAc␤136(GlcNAc␤132)Man␣3 R is formed before the formation of GlcNAc␤136(Gal␤134GlcNAc␤132) Man␣3 R, considering that ␤4Gal-TI greatly prefers ␤1,6linked GlcNAc over ␤1,2-linked GlcNAc residue (Fig. 10B). This allows poly-N-acetyllactosamine extension in Gal␤134-GlcNAc␤136Man side chain before initiation of poly-N-acetyllactosamine synthesis in GlcNAc␤132Man side chain (Fig.  10, B--E). In contrast, iGnT preferentially acts on Gal␤134-GlcNAc␤132Man␣3 R side chain once this side chain is galactosylated. Such a branch specificity of iGnT compensates the inefficient galactosylation of GlcNAc␤132Man␣3 R branch, leading into N-acetyllactosamine extension in Gal␤134Glc-NAc␤132Man␣3 R branch (Fig. 10I). These results indicate that branch specificity of ␤4Gal-TI and iGnT has the opposite effect on poly-N-acetyllactosamine extension in the GlcNAc-␤136Man␣3 R branch versus the GlcNAc␤132Man␣3 R branch. Such complemental branch specificities of ␤4Gal-TI and iGnT is a likely cause for almost equal distribution of poly-N-acetyllactosamine extension in two side chains in nature. Recently, a new member of iGnT was reported (56). However, no studies on branch specificity of this enzyme was carried out, and it is not known whether this additional iGnT may be responsible for poly-N-acetyllactosamine formation in certain cells.
In summary, the present study demonstrated that GlcNAc␤136Man␣3 R side chain itself is not a preferential site for poly-N-acetyllactosamine formation. Rather, addition of this side chain on GlcNAc␤132Man␣3 R side chain by GnTV, forming GlcNAc␤136(GlcNAc␤132)Man␣13 Man␤3 R, converts the acceptor extremely favorable for poly-N-acetyllactosamine formation (Fig. 4). Moreover, we found that the branched acceptor formed is first galactosylated at GlcNAc-␤136Man side, which is a key step to add N-acetyllactosamine extension equally in both GlcNAc␤136Man3 R and GlcNAc ␤132Man3 R side chains (Fig. 10). The present study, however, did not address why poly-N-acetyllactosamines are added more readily on membrane proteins than secretory proteins (57). Because the iGnT has a unique transmembrane domain (32), further studies will be of significance to address this point in relation to the actions of membrane-bound iGnT.