Extended core 1 and core 2 branched O-glycans differentially modulate sialyl Lewis X-type L-selectin ligand activity.

It has been established that sialyl Lewis x in core 2 branched O-glycans serves as an E- and P-selectin ligand. Recently, it was discovered that 6-sulfosialyl Lewis x in extended core 1 O-glycans, NeuNAcalpha2-->3Galbeta1-->4(Fucalpha1-->3(sulfo-->6))GlcNAcbeta1--> 3Galbeta1-->3GalNAcalpha1-->Ser/Thr, functions as an L-selectin ligand in high endothelial venules. Extended core 1 O-glycans can be synthesized when a core 1 extension enzyme is present. In this study, we first show that beta1,3-N-acetylglucosaminyltransferase-3 (beta3GlcNAcT-3) is almost exclusively responsible for core 1 extension among seven different beta3GlcNAcTs and thus acts on core 1 O-glycans attached to PSGL-1. We found that transcripts encoding beta3GlcNAcT-3 were expressed in human neutrophils and lymphocytes but that their levels were lower than those of transcripts encoding core 2 beta1,6-N-acetylglucosaminyltransferase I (Core2GlcNAcT-I). Neutrophils also expressed transcripts encoding fucosyltransferase VII (FucT-VII) and Core2GlcNAcT-I, whereas lymphocytes expressed only small amounts of transcripts encoding FucT-VII. To determine the roles of sialyl Lewis x in extended core 1 O-glycans, Chinese hamster ovary (CHO) cells were stably transfected to express PSGL-1, FucT-VII, and either beta3GlcNAcT-3 or Core2GlcNAcT-I. Glycan structural analyses disclosed that PSGL-1 expressed in these transfected cells carried comparable amounts of sialyl Lewis x in extended core 1 and core 2 branched O-glycans. In a rolling assay, CHO cells expressing sialyl Lewis x in extended core 1 O-glycans supported a significant degree of shear-dependent tethering and rolling of neutrophils and lymphocytes, although less than CHO cells expressing sialyl Lewis x in core 2 branched O-glycans. These results indicate that sialyl Lewis x in extended core 1 O-glycans can function as an L-selectin ligand and is potentially involved in neutrophil adhesion on neutrophils bound to activated endothelial cells.

The roles of sialyl Lewis x in core 2 branched O-glycans have been demonstrated by analyzing mutant mice with deficient Core2GlcNAcT-I, obtained through homologous recombination (17). Leukocytes derived from null mutant mice display significantly reduced adhesion to L-, P-, and E-selectins, demonstrating that ligands for these selectins are mainly carried by core 2 branched O-glycans. By contrast, in these mice, lymphocyte adhesion to HEV in lymph nodes is only marginally impaired, and MECA-79 antibody staining that decorates HEV is not reduced (17). Recent studies reveal that L-selectin ligand activity remaining after abrogation of Core2GlcNAcT-I is due to the activity of 6-sulfosialyl Lewis x in extended core 1 O-glycans, NeuNAc␣233Gal␤134[Fuc␣133(sulfo36)]-GlcNAc␤13 3Gal␤133GalNAc␣13 Ser/Thr (18). Moreover, a minimum MECA-79 epitope was found to be a 6-sulfo structure in the extended core 1 O-glycan, and MECA-79 antibody binds efficiently to 6-sulfosialyl Lewis x-containing extended core 1 O-glycans (18). These findings are consistent with previous findings that MECA-79 antibody inhibits lymphocyte adhesion to HEV without removal of sialic acid (19) or fucose and that MECA-79 staining remains after expression of fucose is abrogated by inactivation of fucosyltransferase VII (FucT-VII) (20).
␤3GlcNAcT-3 transcripts are highly expressed in the small intestine, colon, and placenta and are moderately expressed in various tissues, including the liver, kidney, pancreas, and prostate (18). It is not known whether blood cells express ␤3GlcNAcT-3.
In this study, we address the function and expression pattern of ␤3GlcNAcT-3. First, we show that ␤3GlcNAcT-3 was the only enzyme that significantly formed extended core 1 among highly related ␤3GlcNAcTs. We then show that ␤3GlcNAcT-3 transcripts were present in both human neutrophils and lym-phocytes, but that these cells lacked LSST. Transfection studies using FucT-VII and ␤3GlcNAcT-3 or Core2GlcNAcT-I showed that extended core 1 could be synthesized in Chinese hamster ovary (CHO) cells and that extended core 1 structure was fucosylated by FucT-VII more efficiently than core 2 branches. Finally, we show that sialyl Lewis x synthesized in extended core 1 served as an L-selectin ligand, although it is apparently less potent than sialyl Lewis x in core 2 branches.
Expression of i Antigen in HeLa Cells by Different ␤3GlcNAcTs-To determine whether all of ␤3GlcNAcTs direct the synthesis of poly-Nacetyllactosamine synthesis, HeLa cells were transiently transfected with one of the pcDNA3.1(N)-␤3GlcNAcTs or pcDNA1.1-␤3GlcNAcT-3. Thirty-six hours after transfection, cells were dissociated into monodispersed cells using an enzyme-free cell dissociation solution (Hanks' balanced saline solution-based) purchased from Cell and Molecular Technologies. Monodispersed cells were incubated with human anti-i serum (Dench) (31), followed by affinity-purified fluorescein isothiocyanate (FITC)-conjugated goat anti-human IgM antibodies (Pierce). The stained cells were subjected to FACS analysis using a FACScan (BD Biosciences) as described previously (32).
HeLa cells were chosen as recipient cells for transfection because the molecular mass of lysosome-associated membrane protein-1, a major carrier of poly-N-acetyllactosamines (33), is the smallest among Namalwa, HL-60, CHO, HepG2, and HeLa cells. The results indicated that HeLa cells express minimum amounts of ␤3GlcNAcTs because poly-N-acetyllactosaminylated lysosome-associated membrane protein-1 displays a higher molecular mass than lysosome-associated membrane protein-1 containing minimum amounts of poly-N-acetyllactosamine (33).
Expression of MECA-79 in Lec2 Cells by LSST and ␤3GlcNAcT-To determine which ␤3GlcNAcT directs expression of MECA-79 antigen, Lec2 cells were transiently transfected with pcDNA1-LSST and one of the pcDNA3.1(N)-␤3GlcNAcTs or pcDNA1.1-␤3GlcNAcT-3. Thirty-six hours after transfection, cells were dissociated into monodispersed cells using the cell dissociation solution as described above. Monodispersed cells were incubated with MECA-79 antibody (BD Biosciences) (19), followed by affinity-purified FITC-conjugated goat anti-rat IgM antibody (ICN Biochemicals). The stained cells were subjected to FACS analysis as described above. CHO mutant Lec2 cells lack a functional Golgi CMP-sialic acid transporter; and therefore, sialylation is absent in Lec2 cells (34). The absence of sialylation facilitates core 1 extension because core 1 extension and sialylation compete with each other for the same acceptor, Gal␤133GalNAc␣13 R.
Core 1 Extension in PSGL-1 by ␤3GlcNAcTs-To determine which ␤3GlcNAcT can add ␤1,3-N-acetylglucosamine to core 1, Gal␤133GalNAc␣13 R, Lec2 cells were transiently transfected with pZeoSV-PSGL-1 (kindly provided by Dr. Richard Cummings) and vectors encoding ␤3GlcNAcT using LipofectAMINE as described previously (12). The ratio of pZeoSV-PSGL-1 and ␤3GlcNAcT cDNA was 1:5 (w/w) to achieve efficient modification of PSGL-1 by ␤3GlcNAcT. Fortyeight hours after transfection, cells were harvested in phosphatebuffered saline with a cell scraper. The cells were subjected to three cycles of freezing and thawing to disrupt the plasma membrane. The membrane fraction was collected by centrifugation at 12,000 ϫ g for 10 min. The resultant membrane fraction was first resuspended in 10 mM Tris-HCl and 1 mM EDTA (pH 8.0), and then 10% Triton X-100 was added to a final concentration of 1%. After gentle rocking at 4°C for 15 min, the Triton X-100-soluble membrane fraction, containing PSGL-1, was obtained by centrifugation at 12,000 ϫ g for 10 min.
The membrane fraction was then lysed in sample buffer and subjected to SDS-PAGE. After blotting onto a polyvinylidene difluoride membrane filter, the blot was reacted with anti-PSGL-1 antibody (KPL-1, BD Biosciences), followed by secondary antibody; and immunoreactive PSGL-1 was visualized using enhanced Luminol reagent (PerkinElmer Life Sciences).
As a control, CHO-PSGL-1 cells were stably transfected with pCDM8-FucT-VII and pcDNA3 and cultured in the presence of Zeocin and Geneticin. Cells expressing FucT-VII were selected after staining with CSELX-1 antibody, resulting in CHO-PSGL-1/F7 cells.

Measurement of L-selectin-mediated Rolling in CHO Cells Expressing Sialyl Lewis x in Extended Core 1 or Core 2 Branched O-Glycans-CHO
cells stably expressing PSGL-1, FucT-VII, and ␤3GlcNAcT-3 or Core2GlcNAcT-I were established as described above. These cells maintained similar amounts of sialyl Lewis x and PSGL-1 as assessed by FACS analysis using anti-PSGL-1 antibody KPL-1 and anti-sialyl Lewis x antibody CSELX-1 (see also ''Structural Analysis of PSGL-1 O-glycans Synthesized in the Presence of ␤3GlcNAcT-3 or Core2GlcNAcT-I''). These stably transfected cells seeded on dishes were used as the bottom plate of a parallel flow chamber as described previously (18).
Neutrophils or lymphocytes were initially introduced into the flow chamber at a wall sheer stress of 5 dynes/cm 2 for 15 s, followed by the termination of flow to allow the cells to adhere under static conditions (17). Flow rate was then initiated at different shear forces. Image analysis was performed and analyzed as described previously (17).
Because Gal␤134(sulfo36)GlcNAc␤133Gal␤133GalNAc-␣13 R is a minimum epitope for MECA-79 antigen, the formation of extended core 1 can be detected by immunostaining with MECA-79 antibody when LSST is also expressed.
First, we tested whether all of the cloned ␤3GlcNAcTs are active in synthesizing i antigen, Gal␤134GlcNAc␤133Gal-␤134GlcNAc3 R (21,31). The synthesis of i antigen is dependent on ␤3GlcNAcT, which adds ␤1,3-linked GlcNAc to Nacetyllactosamine. HeLa cells were thus transiently transfected with one of the ␤3GlcNAcTs in mammalian expression vectors, and the transfected cells were stained with anti-i antibody, followed by FITC-conjugated secondary antibody. Fig. 2 shows that the cells transfected with any of the ␤3GlcNAcTs tested displayed increased amounts of i antigen compared with mock-transfected cells. The results also indicate that the expression efficiency of different ␤3GlcNAcTs is essentially invariable because similar amounts of i antigen were detected in HeLa cells transfected with different ␤3GlcNAcTs.
As shown previously, CHO cells and the CHO mutant Lec2 cell line do not synthesize core 2 O-glycans (43)  formation of MECA-79 antigen by sialylation of Gal␤133GalNAc␣13 R. Lec2 cells were thus transiently transfected with LSST and one of the ␤3GlcNAcTs. As shown in Fig. 3, Lec2 cells transfected with ␤3GlcNAcT-3, also called core 1-␤3GlcNAcT, expressed significant amounts of MECA-79 antigen, but none of the other enzymes formed MECA-79 antigen. Because all of the ␤3GlcNAcTs tested were expressed in similar amounts in HeLa cells, it is reasonable to assume that these ␤3GlcNAcTs were expressed in similar amounts in Lec2 cells as well (see also Fig. 4). These results indicate that only ␤3GlcNAcT-3 can form extended core 1 structure.

Core 1 Extension Enzyme (␤3GlcNAcT-3) Is Also Expressed in Neutrophils and
Lymphocytes-Previously, we showed that ␤3GlcNAcT-3 is expressed in HEV in peripheral lymph nodes and forms L-selectin ligand critical for lymphocyte homing (18). To determine whether ␤3GlcNAcT-3 is also expressed in human neutrophils and lymphocytes, neutrophils and lymphocytes were isolated from the peripheral blood, and RT-PCR was used to assay for ␤3GlcNAcT-3 transcripts. The results shown in Fig. 5 demonstrate that the ␤3GlcNAcT-3 transcript was expressed in neutrophils and lymphocytes, although the amount of the transcript was much less than that of the Core2GlcNAcT-I transcript. In addition, the level of Core2GlcNAcT-I transcripts was less than that of PSGL-1. Notably, neutrophils, but not lymphocytes, contained a significant amount of FucT-VII transcripts. To confirm that these transcripts are derived from the proper corresponding Glc-NAcT locus, the RT-PCR products were digested with restriction enzymes and analyzed by agarose gel electrophoresis. The results shown in Fig. 5B demonstrate that the transcripts from neutrophils and lymphocytes produced the same restriction digest products as those derived from plasmids encoding those proteins, supporting our conclusion that the transcripts derived from these cells represent ␤3GlcNAcT-3, Core2Glc-NAcT-I, FucT-VII, and PSGL-1, respectively.
On the other hand, LSST transcripts were barely detected in neutrophils or lymphocytes, indicating that LSST is expressed, if at all, in very low quantities in these cells (Fig. 5). These results indicate that neutrophils express each of the enzymes necessary to form sialyl Lewis x in extended core 1 structure, A, cDNAs were synthesized by reverse transcriptase using total RNA isolated from human neutrophils and lymphocytes. PCR was carried out using the cDNAs as templates and oligonucleotide primers specific to each transcript (ϩ). Control reactions employed total RNA that was not reverse-transcribed (Ϫ). The products were separated by electrophoresis on 1% agarose gels. Transcripts encoding glyceraldehyde-3-phosphate dehydrogenase (G3PDH) served as a positive control. B, PCR products derived from neutrophils (N) and lymphocytes (L) were digested with the indicated restriction enzymes and separated by electrophoresis on 2% agarose gel. Positive controls employed PCR using plasmids (P) containing respective cDNAs as templates. The migration positions of molecular mass markers are shown on the left. whereas lymphocytes express trace amounts of the sialyl Lewis x moiety due to negligible expression of FucT-VII.

Structural Analysis of PSGL-1 O-Glycans Synthesized in the Presence of ␤3GlcNAcT-3 or Core2GlcNAcT-I-Because
PSGL-1 is expressed in neutrophils and lymphocytes as indicated, the above results suggest that ␤3GlcNAcT-3 can form extended core 1 structure in PSGL-1, a counter-receptor for L-, P-, and E-selectins in neutrophils. Such expression led to the formation of sialyl Lewis x structure in extended core 1 Oglycans when FucT-VII was also present. Fig. 6 illustrates that either ␤3GlcNAcT-3 (C1) or Core2GlcNAcT-I (C2) could convert PSGL-1/IgG chimeric protein into a polydisperse high molecular mass glycoform, which could be metabolically labeled with [ 3 H]glucosamine (second and third lanes). By contrast, PSGL-1 chimeric protein migrated as sharper bands when isolated from control CHO cells expressing only FucT-VII (first lane).
These samples were treated with alkaline borohydride to release O-glycans and subjected to Sephadex G-50 gel filtration. Released O-glycans (Fig. 7, A and E, horizontal bars) were subjected then to Bio-Gel P-4 gel filtration. As shown in Fig. 7  (B and F), PSGL-1 O-glycans derived from CHO cells expressing FucT-VII and ␤3GlcNAcT-3 or Core2GlcNAcT-I mainly produced two or three peaks (I, IЈ, and II). After desialylation and Bio-Gel P-4 gel filtration, peak I produced peaks IA and IB (Fig. 7, C and G). As shown in Fig. 7 (C and G), peak C2-IA from Core2GlcNAcT-I-expressing CHO cells eluted slightly later than peak C1-IA from ␤3GlcNAcT-3-expressing CHO cells.
After desialylation, peaks C1-I and C2-I (Fig. 7, B and F) also produced peaks C1-IB and C2-IB, which eluted at the same positions as Gal␤133GalNAcOH and sialic acid monomer (Fig.  7, C and G). These results indicate that peaks C1-I and C2-I and C2-IA (C) were subjected to HPLC using an amino-bonded column (Asahipak NH 2 P50-4E). The sample was eluted isocratically in the first 40 min and then eluted with a linear gradient by decreasing acetonitrile concentration. In B and D, the peaks that eluted at 64 -67 ml in A and at 68 -71 ml in C were separately digested with ␣1,3-fucosidase and subjected to the same HPLC step. In a separate experiment, a mixture of peaks C1-IA and C2-IA was subjected to the same HPLC step, demonstrating that peaks from C1-IA and C2-IA can be separated under these elution conditions (data not shown). The main peaks in B and D correspond to Gal␤134GlcNAc␤133Gal␤133GalNAcOH and Gal␤134GlcNAc␤13 6(Gal␤133)GalNAcOH, respectively. ( Fig. 7, B and F) also contained NeuNAc␣236(NeuNAc-␣233Gal␤133)GalNAcOH, which was recovered in peaks C1-IB and C2-IB after desialylation. After desialylation, peak C1-II (monosialylated fraction in Fig. 7B) produced peak C1-IIB and a small amount of peak C1-IIA (Fig. 7D). Peak C1-IIA did not change its elution position after desialylation and corresponds to Gal␤134(ϮFuc␣133)GlcNAc␤133Gal␤133-GalNAcOH. Peak C1-IIB was, on the other hand, found to be a mixture of Gal␤133GalNAcOH and sialic acid. Similarly, peak C2-II (Fig. 7F) produced sialic acid and Gal␤133GalNAcOH (data not shown). The amount of Gal␤133GalNAcOH in peaks IB and IIB was determined after sialic acid was removed by QAE-Sephadex A-25 column chromatography.
These combined results indicate that extended core 1 Oglycans can be fucosylated more efficiently than core 2 branched O-glycans (Fig. 8, compare A and C; and Table I).
However, the conversion of core 1 structure to extended core 1 is less efficient than the conversion of core 1 structure to core 2 branched structure (Fig. 7, B versus F, compare peaks I and II; and Table I). These results as a whole indicate that sialyl Lewis x in core 2 branched O-glycans is expressed at levels equivalent to sialyl Lewis x in extended core 1 O-glycans (Table I).
Sialyl Lewis x in Extended Core 1 Functions as an L-selectin Ligand-To determine the role of sialyl Lewis x in extended core 1 structure, tethering and rolling of neutrophils and lymphocytes were assayed in CHO cells expressing PSGL-1 and sialyl Lewis x in extended core 1 O-glycans or core 2 branched O-glycans. Because CHO cells lack endogenous ␤3GlcNAcT-3 (18) or Core2GlcNAcT (4,39), CHO cells expressing only PSGL-1 and FucT-VII lack sialyl Lewis x in mucin-type Oglycans. As shown in Fig. 9A, neutrophil tethering and rolling were not detected in this CHO cell line lacking both ␤3Glc-NAcT-3 and Core2GlcNAcT-I (open circles). By contrast, CHO cells expressing sialyl Lewis x in extended core 1 (closed circles) or core 2 (open triangles) oligosaccharides supported neutrophil tethering and rolling under shear. Essentially identical results were obtained when lymphocyte tethering and rolling were assayed in the same transfected CHO cells (Fig. 9B). These combined results indicate that non-sulfated sialyl Lewis x in extended core 1 or core 2 oligosaccharides functions as an L-selectin ligand, although sialyl Lewis x in core 2 branched O-glycans functions as a more efficient L-selectin ligand than does sialyl Lewis x in extended core 1 O-glycans (Fig. 9, compare open triangles and closed circles).
This study unexpectedly demonstrated that extended core 1 is most likely present in neutrophils and lymphocytes, although the amount of extended core 1 structure is less than that of core 2 branched structure. Interestingly, LSST, which is required to form 6-sulfo-GlcNAc structure, is apparently not expressed in neutrophils and lymphocytes. This finding is consistent with the fact that neutrophils and lymphocytes are negative for MECA-79 antigen. ␤3GlcNAcT-3 belongs to the ␤3GlcNAcT gene family, which consists of at least eight different ␤3GlcNAcTs. The members of this gene family include ␤3GlcNAcTs encoded by fringe and brainiac (44 -47). Fringe was identified as a protein that regulates Notch signaling by adding ␤1,3-linked GlcNAc to ␣-fucose attached to the extracellular domain of Notch (44,45). By contrast, Brainiac apparently acts on glycolipids and thereby modulates Notch activity by another mechanism (46,47). Although ␤3GlcNAcT-1 does not have discernible homology to the ␤3GlcNAcT gene family, recent studies show that one protein predicted by DNA sequence within the human or mouse Large locus has some homology to ␤3GlcNAcT-1 (48). Premature translation termination of this putative glycosyltransferase within the Large locus in mice results in myodystrophy (48), whereas the LARGE locus is deleted in human patients with meningioma, a tumor of the meninges of the central nervous system (49). These result suggest that ␤3GlcNAcT-1 may play an important role in development and cancer because ␤3GlcNAcT-1 is ubiquitously expressed (21).
This study demonstrated that, in transfected CHO cells, fucosylation of N-acetyllactosamine in extended core 1 structure takes place more efficiently than does fucosylation in core 2 branched O-glycans. This finding is consistent with the previous finding that a significant portion of extended core 1 structure is fucosylated in HEV (18). Our results show that extended core 1 structure is efficiently fucosylated once core 1 structure is formed. On the other hand, the amount of extended core 1 O-glycans is less than that of core 2 branched O-glycans (Table I). As an aggregate, extended core 1 and core 2 branched O-glycans contain similar amounts of sialyl Lewis x. Synthesis of both extended core 1 and core 2 branches competes with ␣2,3-sialylation of core 1, which is catalyzed by ␤-galactoside ␣2,3-sialyltransferase I (50). It is thus possible that both the levels of expressed ␤3GlcNAcT-3 and its catalytic activity are not as great as those of expressed Core2GlcNAcT-I in transfected CHO cells.
It is also possible that localization of ␤3GlcNAcT-3, Core2GlcNAcT-I, and ␤-galactoside ␣2,3-sialyltransferase I in different Golgi compartments is a key factor in determining the amount of oligosaccharides synthesized by ␤3GlcNAcT-3 or Core2GlcNAcT-I (51, 52). Previously, we have shown that Core2GlcNAcT-I resides in the cis-to medial-Golgi, whereas the majority of N-and O-glycan sialyltransferases are thought to reside in the medial-to trans-Golgi (51). This difference allows Core2GlcNAcT-I to add core 2 branch before core 1 oligosaccharide is sialylated. If core 1 oligosaccharide is sialylated first, it becomes unavailable for Core2GlcNAcT-I action (see Fig. 1 in Ref. 2). It is tempting to speculate that ␤3Glc-NAcT-3 resides in later compartments of the Golgi than does Core2GlcNAcT-I, thus directly competing with ␤-galactoside ␣2,3-sialyltransferase I for the same acceptor, Gal␤13 3GalNAc␣3 Ser/Thr. Such direct competition, if it occurs, should lead to moderate synthesis of extended core 1 structure.
Previously, we showed that 6-sulfosialyl Lewis x in extended core 1 O-glycans serves as an L-selectin ligand as efficiently as 6-sulfosialyl Lewis x in core 2 branched O-glycans (18). In this study, we extended this finding by showing that sialyl Lewis x in extended core 1 O-glycans also functions as an L-selectin ligand, although it is not as efficient as sialyl Lewis x in core 2 branched O-glycans. Our results are also consistent with previous reports showing that sialyl Lewis x functions as an Lselectin ligand, although extended core 1 structure was not evaluated in that study (53). In our previous study, we found that the sialyl Lewis x structure is present in extended core 1 O-glycans of HEV-derived GlyCAM-1, which were converted to neutral oligosaccharides after desialylation (18). These results combined indicate that sialyl Lewis x in extended core 1 serves as an L-selectin ligand in HEV. L-selectin-mediated neutrophil rolling was shown to take place in adherent neutrophils bound to activated endothelial cells (10,11). Very recently, we obtained mice heterozygous for ␤3GlcNAcT-3 deficiency and knock-in of green fluorescent protein under the control of the ␤3GlcNAcT-3 promoter. Analysis of these mice indicated that ␤3GlcNAcT-3 is expressed in neutrophils and T lymphocytes because neutrophils and T lymphocytes stained with markers CD11b (Mac-1) and CD3, respectively, were also positive for green fluorescent protein as determined by FACS analysis. 2 These studies did not, however, inform at to how much sialyl Lewis x in extended core 1 structure is present in neutrophils. Further studies on the knockout mice are important to determine the degree to which sialyl Lewis x moieties in extended core 1 structure contribute to L-selectin-mediated adhesion in HEV and neutrophil-neutrophil interaction.