A Protein Kinase C/Ras/ERK Signaling Pathway Activates Myeloid Fibronectin Receptors by Altering β1 Integrin Sialylation*

Here we report that myeloid cells differentiating along the monocyte/macrophage lineage down-regulate the ST6Gal-I sialyltransferase via a protein kinase C/Ras/ERK signaling cascade. In consequence, the β1 integrin subunit becomes hyposialylated, which stimulates the ligand binding activity of α5β1 fibronectin receptors. Pharmacologic inhibitors of protein kinase C, Ras, and MEK, but not phosphoinositide 3-kinase, block ST6Gal-I down-regulation, integrin hyposialylation, and fibronectin binding. In contrast, constitutively active MEK stimulates these same events, indicating that ERK is both a necessary and sufficient activator of hyposialylation-dependent integrin activation. Consistent with the enhanced activity of hyposialylated cell surface integrins, purified α5β1 receptors bind fibronectin more strongly upon enzymatic desialylation, an effect completely reversed by resialylation of these integrins with recombinant ST6Gal-I. Finally, we have mapped the N-glycosylation sites on the β1 integrin to better understand the potential effects of differential sialylation on integrin structure/function. Notably, there are three N-glycosylated sites within the β1 I-like domain, a region that plays a crucial role in ligand binding. Our collective results suggest that variant sialylation, induced by a specific signaling cascade, mediates the sustained increase in cell adhesiveness associated with monocytic differentiation.

The U937 and THP-1 cell lines represent well accepted model systems for studying myeloid differentiation along the monocyte/macrophage lineage. Following treatment with phorbol myristate acetate (PMA), 4 these cells exhibit phenotypic changes that are characteristic of cell differentiation, including increased respiratory burst activity, enhanced phagocytotic capability, and markedly elevated cell adhesiveness to extracellular matrix ligands such as fibronectin. In vivo, the increased adhesiveness of monocytes/macrophages contributes to the extravasation of cells from the vasculature as well as tethering of cells within inflamed tissues.
Differentiating myeloid cells bind to fibronectin through the integrin family of cell adhesion receptors, including the ␣5␤1 integrin species. The molecular mechanisms underlying PMA-dependent cell adhesion have not been well defined, although it has been reported that PMA increases the synthesis of both ␣5 and ␤1 integrin subunits (1)(2)(3)(4). However, myeloid cells (U937 and THP-1) express an abundant amount of ␣5␤1 in the absence of PMA treatment, and yet these cells bind very poorly to fibronectin. This suggests that myeloid ␣5␤1 receptors are normally in an inactive state and that increased expression alone cannot account for the dramatically increased fibronectin binding induced by PMA.
In our prior study (5) we observed that PMA stimulated a rapid but transient increase in fibronectin binding that was likely due to the activation of integrins already present on the cell surface. However, following this initial transient event there was a second phase of elevated fibronectin binding that was sustained over many hours. The onset of this second phase of integrin activation was temporally correlated with the synthesis of a ␤1 integrin isoform that lacked ␣2-6-linked sialic acids, a sugar modification directed by the ST6Gal-I sialyltransferase. Our laboratory has previously determined that ␤1 integrins serve as a substrate for ST6Gal-I in several different cell types (5)(6)(7). In differentiating myeloid cells the expression of hyposialylated ␤1 integrins results from PMA-induced down-regulation of ST6Gal-I (5).
Given that PMA is a known activator of protein kinase C (PKC), our goal in this investigation was to identify the signaling molecules that direct ST6Gal-I down-regulation and, correspondingly, integrin hyposialylation. Other studies have suggested that extracellular signal-regulated kinase (ERK) signaling is required for monocytic differentiation (8 -11); however, the mechanism by which ERK regulates integrin function in differentiated cells has not been elucidated. Our current results suggest that a PKC/Ras/ERK signaling cascade mediates the sustained phase of fibronectin binding by inhibiting ␤1 integrin sialylation.

MATERIALS AND METHODS
Cell Culture-A U937 myeloid cell subclone selected for granulocyte/macrophage colony-stimulating factor sensitivity was obtained from Dr. Elizabeth Eklund (Northwestern University). The cells were maintained in Dulbecco's modified Eagle's medium with 4.5 mg ml Ϫ1 glucose, L-glutamine (Cellgro), 10% fetal bovine serum, and gentamicin. U937 cells expressing constitutively active MEK were generated by using Lipofectamine Plus (Invitrogen) to transfect cells with a hemagglutinin-tagged, activated MEK construct (available from Upstate Biotechnology). A pooled population of clones expressing activated MEK was obtained by selection in G418. Verification of MEK expression was accomplished by immunoblotting for the hemagglutinin tag.
Western Blotting-U937 cells were treated with or without 50 ng ml Ϫ1 PMA for 15 h. Cells were then lysed in 50 mM Tris-HCl buffer (pH 7.4) containing 1% Triton X-100, 0.5 mM phenylmethylsulfonyl fluoride, 20 g ml Ϫ1 leupeptin, 4 mM sodium fluoride, and 200 M sodium pervanadate. Protein concentrations in the lysates were determined using a modified Bradford assay (Sigma). Lysates were resolved by reducing SDS-PAGE, and ␤1 integrins were Western blotted using a monoclonal antibody from BD Transduction Laboratories. Western blot analysis of ST6Gal-I was accomplished using a polyclonal antibody generously provided by Dr. Karen Colley (University of Illinois, Chicago, IL).
Enzyme Inhibitor Studies-Cells were incubated for 20 min at 37°C with one of the following enzyme inhibitors: 10 M R031-8220 (Calbiochem), 30 M manumycin A (Sigma), 50 M PD98059 (Calbiochem), or 30 nM wortmannin (Sigma). PMA was then added to a final concentration of 50 ng ml Ϫ1 , and cells were incubated in the presence of both PMA and the selected inhibitor for an additional 15 h at 37°C.
Lectin Affinity Analyses-Cell lysates (600 g) were incubated for 3 h at 4°C with 4 g of the biotinylated lectins SNA or ECL (Vector Laboratories). Streptavidin-agarose (20 l, Sigma) was then added, and samples were incubated for an additional 2 h at 4°C with rotation. Lectinglycoprotein complexes were collected by brief centrifugation and then washed three times with lysis buffer followed by one wash with phosphate-buffered saline. Glycoproteins were released from the complexes by boiling in SDS-PAGE sample buffer, resolved by reducing SDS-PAGE, and immunoblotted to detect the ␤1 integrin.
Cell Attachment Assays-Cells were treated with or without 50 ng ml Ϫ1 of PMA and also with enzyme inhibitors in some trials and then seeded onto tissue culture dishes that had been precoated with 20 g ml Ϫ1 fibronectin. Adhesion was quantified as described previously using a crystal violet staining method (5).
Sialidase and ST6Gal-I Enzyme Treatment of Purified ␣5␤1 and Modified ELISA Integrin Binding Assay-As reported previously (5), purified ␣5␤1 integrins (Chemicon) were suspended in 50 mM Tris buffer containing 150 mM NaCl, 2 mM MgCl 2 , 0.1 mM CaCl 2 , and 0.1% Triton X-100 ("ELISA buffer") adjusted to pH 6.5. Agarose-conjugated Vibrio cholerae sialidase (200 milliunits; Calbiochem) was added to the integrin solution, and samples were incubated for 6 h at 37°C with rotation. An equal amount of the integrin solution was incubated for 6 h in buffer without sialidase as a control. Following this incubation, the agarose-conjugated sialidase was removed by centrifugation. Both control and sialidase-treated integrin solutions were subdivided into two tubes and then treated with or without 5 milliunits ml Ϫ1 of rat recombinant ST6Gal-I (Calbiochem) in the presence of 50 M CMP-Neu (the activated sugar donor substrate required by ST6Gal-I) for 4 h at 37°C. Following this incubation, the integrin solution buffer was adjusted to pH 7.4 by addition of pH 8.0 ELISA buffer. Integrin solutions were loaded onto 12-well tissue culture dishes precoated with fibronectin (100 g ml Ϫ1 ). A final amount of 350 ng of purified ␣5␤1 integrin was added to each well. Samples were also loaded onto wells that were precoated with denatured bovine serum albumin to control for nonspecific binding. Purified integrins were allowed to adhere for 1.5 h at 37°C; the wells were then washed three times with ELISA buffer (pH 7.4) and exposed for 1 h to the glycosylation-insensitive anti-␤1 integrin monoclonal antibody MAB2000 (Chemicon). After washing, wells were incubated with a horseradish peroxidase-coupled secondary antibody (Amersham Biosciences) followed by the colorimetric horseradish peroxidase substrate chromogen (BIOSOURCE). Integrin binding was quantified by absorbance spectroscopy at a 450-nm wavelength. Specific binding values were obtained by subtracting bovine serum albumin binding values ("nonspecific") from total fibronectin binding.
Transfection of N-Glycosylation Site Mutants-CHO-K1 cells were transfected with a V5-tagged ST6Gal-I construct (provided by Dr. Karen Colley, University of Illinois, Chicago, IL), and stable clones were generated by selection in G418. The pECE plasmid containing the human ␤1 integrin sequence (12) was obtained from Dr. Erkki Ruoshlati (The Burnham Institute), and site-directed mutagenesis was accomplished using the QuikChange site-directed mutagenesis kit (Stratagene). Reactions were performed in a thermal cycler with 40 ng of template DNA and complementary primers at 300 nM and with the following cycling steps repeated 16 times: 30s at 94°C, 60 s at 55°C, and 12 min at 68°C. Mutant clones were verified by DNA sequence analysis. The mutated constructs were transiently transfected into ST6Gal-Iexpressing CHO-K1 cells using Lipofectamine Plus according to the vendor protocol (Invitrogen). Forty-eight hours following transfection the cells were lysed, and detection of the constructs was accomplished by Western blotting with antibodies that recognized only the transfected form of the ␤1 integrin. Two antibodies were used, one from BD Transduction Laboratories (catalog number 610468) and one from Chemicon (catalog number MAB1965). Of note, the N74Q mutant (numbering begins with first amino acid following the signal sequence) was only detectable with the Chemicon antibody, possibly because this mutation alters epitope recognition by the BD Transduction Laboratories antibody.

Activation of PKC/Ras/ERK Induces Expression of Hyposialylated ␤1
Integrins-Previously we reported that ␤1 integrins expressed by PMAtreated U937 and THP-1 cells have a smaller apparent molecular mass when analyzed by SDS-PAGE, and it was subsequently shown this was due to the PMA-induced synthesis of integrins lacking ␣2-6 sialic acids (5). To evaluate whether the expression of hyposialylated integrins was regulated by PKC, we pretreated U937 cells with the PKC inhibitor R031-8220, stimulated cells with PMA (plus inhibitor), and then examined the electrophoretic mobility of ␤1 integrins. As shown (Fig. 1A), mature ␤1 integrins from PMA-treated cells migrated more rapidly than ␤1 integrins from control cells, reflecting the expression of the hyposialylated glycoform. Pretreatment with R031-8220 blocked the PMA-induced mobility shift, implicating PKC as a modulator of integrin sialylation. In contrast to mature ␤1, neither PMA nor R031-8220 had any effect on the mobility of precursor ␤1 integrins, a species that resides in the endoplasmic reticulum (13)(14)(15)(16) and is therefore not a substrate for sialyltransferases.
Other studies from our laboratory have shown that forced expression of oncogenic Ras in epithelial cells causes altered ␣2-6 sialylation of ␤1 integrins (7); we therefore speculated that Ras may act as a regulator of integrin sialylation in myeloid cells. To test this hypothesis, cells were treated with manumycin A, a compound that blocks Ras activation by preventing farnesylation (17,18). Similar to results with R031-8220, manumycin A prevented the PMA-induced electrophoretic mobility shift (Fig. 1A).
We next sought to identify downstream effectors of Ras that might be involved in regulating integrin sialylation. Ras can activate multiple signaling cascades; however, the phosphoinositide 3-kinase and Raf/MEK/ ERK signaling pathways are among the best characterized mediators of Ras-dependent cellular responses (19). Accordingly, we treated cells with an inhibitor (PD98059) of the ERK-activating kinase MEK as well as with an inhibitor of phosphoinositide 3-kinase (wortmannin). As shown in Fig. 1A, the MEK inhibitor blocked PMA-induced expression of the hyposialylated glycoform, whereas the phosphoinositide 3-kinase inhibitor was without effect. These data suggest that PKC regulates integrin sialylation by activating a Ras/Raf/MEK/ERK signaling cascade.
To more directly examine integrin sialylation, we performed a lectin affinity assay. Briefly, cell lysates were incubated with SNA, a lectin that binds specifically to ␣2-6-linked sialic acids. Sialylated proteins were precipitated and electrophoresed, and ␤1 integrins were subsequently detected by Western blotting. Consistent with results from mobility shift assays, SNA failed to precipitate ␤1 integrins from PMA-treated cells, indicating that these integrins are lacking ␣2-6 sialic acids (Fig.  1B). However, SNA reactivity could be restored when PMA-treated cells were preincubated with R031-8220, manumycin A, and PD98059, but not with wortmannin.
Down-regulation of ST6Gal-I Is Mediated by PKC/Ras/ERK-Given that PMA induces down-regulation of ST6Gal-I (5, 20), we examined the effects of pharmacologic inhibitors on ST6Gal-I protein levels. Western blots of ST6Gal-I revealed that the PMA-dependent downregulation in ST6Gal-I expression could by blocked by preincubating cells with R031-8220, manumycin A, and PD98059 but not with wortmannin (Fig. 1C).
Hyposialylated ␤1 Integrins Have Increased Levels of Galactose-terminated N-Glycans-ST6Gal-I directs the addition of sialic acid in an ␣2-6-linkage to the terminal galactose of N-linked polylactosamine chains. However, this terminal galactose is a potential substrate for other trans-Golgi glycosyltransferases including several ␣2-3-sialyltransferases, which are known to be active in U937 cells (21). It follows that in cells with down-regulated ST6Gal-I the terminal galactoses of ␤1 could either remain unmodified or, alternately, become capped with other types of sugars or sugar linkages. To establish whether ␤1 integrins become targeted by competing glycosyltransferases as a result of PMA-induced down-regulation of ST6Gal-I, we performed a lectin affinity analysis with ECL, a lectin specific for the unsubstituted terminal galactose of N-linked polylactosamine chains. As shown in Fig. 1D, ␤1 integrins from PMA-treated cells were much more reactive with ECL, suggesting that a substantial proportion of integrin polylactosamine chains remain uncapped in the absence of ST6Gal-I activity.
PMA-dependent Cell Binding to Fibronectin Is Mediated by PKC/Ras/ERK-Having determined that a PKC/Ras/ERK signaling cascade directs ST6Gal-I down-regulation and hyposialylated integrin expression, we anticipated that inhibitors of this pathway would block integrin-dependent cell adhesion to ␤1 substrates. Thus, cells were pretreated with inhibitors as before, stimulated with PMA, and then subjected to standard cell adhesion assays using fibronectin as a substrate. These assays showed that PMA-dependent fibronectin binding was blocked by R031-8220, manumycin A, and PD98059 but not by wortmannin (Fig. 1E).
Constitutively Active MEK Mimics the Effect of PMA on Integrin Sialylation and Function-To verify that integrin hyposialylation and function are regulated by an ERK-dependent signaling cascade, we generated cells that stably express constitutively active MEK. SNA analyses of integrins harvested from these cells revealed that activated MEK induced the expression of hyposialylated ␤1 integrins in tandem with down-regulation of ST6Gal-I (Fig. 2, A and B). We also found that MEK-dependent integrin hyposialylation was associated with enhanced cell adhesion to fibronectin (Fig. 2C). These data, combined with results from the pharmacologic inhibitor studies (Fig. 1), indicate that ERK is both a necessary and sufficient regulator of sialylation-dependent integrin activation.
␣2-6 Sialylation Directly Regulates ␣5␤1 Binding to Fibronectin-To further establish that ␣2-6-linked sialic acids play a causal role in regulating integrin function, we manipulated the sialylation of purified ␣5␤1 integrins and then monitored integrin binding to fibronectin using a modified ELISA. Consistent with our prior results (5), the enzymatic desialylation of purified ␣5␤1 integrins stimulated fibronectin binding (Fig. 3A). However, we now show that this increased fibronectin binding can be reversed by using recombinant ST6Gal-I to add ␣2-6 sialic acid residues back onto desialylated ␣5␤1 integrins. These data provide FIGURE 1. A PKC/Ras/ERK signaling cascade regulates integrin sialylation and function. A, cells were pretreated with R031-8220 (R031), manumycin A (man A), PD98059 (PD98), or wortmannin (wort) and then further incubated with the respective inhibitor plus PMA for 15 h. Cells were lysed, and the electrophoretic mobility of the ␤1 integrin was evaluated by Western blotting. PMA treatment induced increased mobility of the mature ␤1 integrin species, indicating reduced sialylation, whereas no alteration was noted in the endoplasmic reticulum-resident, the precursor ␤1 integrin isoform. Cont, control. B, cell lysates harvested from cells treated with PMA and inhibitors as described above were incubated with biotinylated SNA lectin followed by precipitation with streptavidin-coupled agarose beads. Lectin-glycoprotein complexes were resolved by SDS-PAGE and then Western-blotted for ␤1 integrins. Note that only the mature integrin species is precipitated by SNA, because precursor ␤1 isoforms are never sialylated. Loss of SNA reactivity in samples treated with PMA only or PMA plus wortmannin reflects the expression of mature integrins lacking ␣2-6 sialic acid. C, cells treated as above were subjected to Western blot analysis to determine levels of ST6Gal-I. D, cell lysates were incubated with biotinylated ECL lectin, and glycoproteins with terminal galactoses were precipitated using streptavidin-agarose. ␤1 integrins precipitated by ECL were detected by Western blot. E, cells were treated with inhibitors and PMA as described previously and then seeded onto fibronectin-coated tissue culture dishes. Cell adhesion was quantified using a standard crystal violet staining method. Values represent the means and S.E. for three independent experiments performed in duplicate.
To confirm the activity of both the sialidase and ST6Gal-I enzymes in our assays, treated integrins were precipitated with SNA and then Western-blotted for ␤1 integrin. As shown in Fig. 3B, sialidase treatment of ␣5␤1 led to significantly reduced SNA reactivity, suggesting that the sialidase was very effective in removing ␣2-6-linked sialic acids. The subsequent incubation of desialylated ␣5␤1 integrins with recom-binant ST6Gal-I restored SNA reactivity to base-line levels, indicating re-addition of ␣2-6-linked sialic acids. Treatment of control ␣5␤1 integrins, which are already heavily sialylated, with ST6Gal-I slightly increased ␣2-6 sialylation, although this did not appear to affect ligand binding activity (Fig. 3A) ␤1 Integrins Are Glycosylated on 10 of 12 of the Asparagine Residues That Have the Appropriate Consensus Sequence for N-Glycosylation, Including Three Sites within the Functionally Important I-like Domain-Our understanding of the role of glycosylation in regulating integrin structure/function has been limited by the lack of information concerning specific sites of N-glycosylation. To address this deficiency, we used a mutagenesis

. Constitutively active MEK induces loss of integrin sialylation and enhanced cell adhesion to fibronectin. A, cells stably expressing constitutively active MEK (ca MEK)
were generated using standard protocols. Lysates harvested from parental cells or cells expressing activated MEK were Western-blotted to detect ST6Gal I. Cont, control. B, SNA was used to precipitate ␣2-6 sialylated proteins; the precipitated proteins were resolved by reducing SDS-PAGE, and ␤1 integrins were detected by Western blot. C, control or constitutively activated MEK-expressing cells were seeded onto fibronectin-coated tissue culture dishes, and adhesion was quantified as described previously. Values represent the means and S.E. for three independent experiments performed in triplicate (*, p Ͻ 0.05). FIGURE 3. ␣2-6 sialylation of purified ␣5␤1 integrins modulates binding to fibronectin. A, purified ␣5␤1 integrins were treated with (ϩ) or without (Ϫ) sialidase to remove sialic acids. The integrins were then treated with or without recombinant ST6Gal-I to restore ␣2-6-linked sialic acids. The purified integrins were added to fibronectin-coated dishes, and binding was quantified using a modified ELISA assay. Values represent the means and S.E. for three independent experiments performed in duplicate (*, p Ͻ 0.05). B, the treated integrins from above were subjected to lectin precipitation using SNA. The precipitates were resolved by SDS-PAGE, and the ␤1 subunit was detected by Western blotting.
approach to identify the sites carrying N-linked glycans. Specifically, the asparagine residues within the NX(S/T) glycosylation consensus sequence were mutated to glutamine; these mutated cDNAs were then transfected into CHO-K1 cells, and expression of the constructs was detected by Western blotting using antibodies that recognize only the transfected ␤1 isoform. We anticipated that the loss of an N-glycan at a given site would result in a reduced apparent molecular mass when compared with wild type ␤1 integrins. As shown in Fig. 4, A and B, 10 of the 12 mutant constructs demonstrated increased electrophoretic mobility by SDS-PAGE, indicating decreased molecular mass (N461Q is shown in Fig. 4A as a representative example). These data suggest that the ␤1 integrin is typically N-glycosylated on 10 sites, including three sites within the ␤1 I-like domain, a region crucial for ligand binding. In contrast, the mobilities of the N564Q and N74Q constructs were identical to that of the wild type ␤1 isoform, suggesting that these sites do not normally carry N-linked glycans.
Interestingly, only one band was typically observed for the transfected ␤1 construct (Fig. 4A), and this band migrated to the expected position of the partially glycosylated precursor isoform. In a few blots a small amount of mature ␤1 was detected upon extended exposure of the films (not shown). The mature band, when observed, also showed an electrophoretic mobility shift as expected. It is currently unclear why the majority of the transfected ␤1 construct remains in the endoplasmic reticulum.

DISCUSSION
The regulation of Golgi glycosyltransferases by signaling mechanisms has been little studied, and even less is known about how such regulation affects the function of specific substrates targeted by these enzymes. We and others have shown that forced expression of oncogenic Ras alters ST6Gal-I expression in epithelial cells (7) and fibroblasts (22)(23)(24)(25). However, the current study describes regulation of ST6Gal-I by an endoge-nous multistep signaling cascade. These results are noteworthy because they indicate that Golgi enzymes such as ST6Gal-I can be dynamically regulated and further imply that variantly glycosylated substrates may be expressed in response to extracellular stimuli that activate appropriate signaling cascades.
Signaling through ERK has been suggested as an essential step in the differentiation of myeloid cells along the monocyte/macrophage lineage (8 -11), and the heterologous expression of constitutively active Ras or MEK stimulates monocytic cell behaviors including phagocytosis and adhesion to fibronectin (8,26,27). However, the molecular mechanisms linking ERK to specific changes in integrin structure/function have not been well defined. Our results indicating that expression of hyposialylated integrins is induced by PKC/Ras/ERK signaling, combined with our prior observation that the synthesis of hyposialylated integrins is temporally correlated with enhanced adhesion to fibronectin, strongly support variant ␤1 sialylation as a mechanism for activation of myeloid fibronectin receptors during monocytic differentiation.
Unequivocal evidence that ␣2-6 sialic acids play a direct role in regulating the activity of ␣5␤1 fibronectin receptors is provided in this report by purified integrin receptor/ligand binding assays. The enzymatic removal of sialic acids from purified ␣5␤1 integrins increases fibronectin binding, whereas the re-addition of these sugars by recombinant ST6Gal-I attenuates binding to basal levels. Importantly, the behavior of purified ␣5␤1 integrins in a cell-free assay system mimics the behavior of integrins on the cell surface; ␣5␤1 integrins lacking ␣2-6 sialic acids bind better to fibronectin. We postulate that the enhanced fibronectin binding activity of hyposialylated integrins contributes to the recruitment of leukocytes to sites of inflammation. This hypothesis is consistent with results from a recent animal study in which ST6Gal-I expression was reduced in selected tissues by mutating one of the multiple ST6Gal-I promoter sequences. In these ST6Gal-I-deficient mice the introduction of a bacterial pathogen stimulated enhanced recruitment of leukocytes into the peritoneum (28).
Elegant work from Luo et al. (29) suggests that glycosylation can affect integrin conformation. These investigators engineered an artificial N-glycosylation site into the ␤1 integrin by substituting an asparagine residue for proline at amino acid 333, which created the NXS consensus sequence for N-glycosylation. Upon transfection of this construct in CHO-K1 cells, the ␤1 subunit was N-glycosylated, and this variant glycoform was shown to have increased ligand binding activity. Site 333 is in the C-terminal end of the ␤1 I-like domain, and it was hypothesized that the addition of a glycan at this site caused an increase in the distance between the ␤1 head and stalk domains, thus inducing the integrin heterodimer to assume a more extended (activated) integrin conformation. Although P333N is not a naturally occurring mutation, these data are important because they provide proof of concept that changes in the glycan structure within key regions of the integrin molecule can alter integrin conformation and activity. Notably, the specific sites on ␤1 that normally carry N-linked glycans have not previously been identified, which is surprising given the fact that there is currently intense interest in delineating integrin structure. Our studies now suggest that 10 of the 12 consensus asparagine residues are elaborated with N-linked glycans. Three of these sites lie within the ␤1 I-like domain, a region critical for ligand binding, and seven other sites are distributed among the plexin-semaphorin-integrin, hybrid, integrinepidermal growth factor, and ␤-tail domains. We hypothesize that sialic acid, a negatively charged sugar, either alters the overall conformation of the integrin receptor or more directly regulates ligand binding. Negative charges within the ␤1 I-like domain are very important for integrin function, as negatively charged amino acids within this region are FIGURE 4. Expression of N-glycosylation site mutants. A, site-directed mutagenesis was used to generate ␤1 integrin constructs containing glutamine substitutions for asparagine residues lying within N-glycosylation consensus sequences. The constructs were transfected into CHO-K1 cells engineered to express ST6Gal-I, and expression of the constructs was verified by Western blotting. Ten of the twelve mutant constructs exhibited reduced molecular mass as compared with the wild-type ␤1 isoform (WT), reflecting the loss of an N-linked glycan. The construct containing the N461Q substitution is shown in the left section as a representative example. Mutants N564Q (right section) and N74Q (not shown) did not show mobility shifts, suggesting that these sites are not normally glycosylated. B, summary of results from transfections of N-glycosylation site mutants. Ten of twelve sites appear to carry N-linked glycans, including three sites within the I-like domain. PSI, plexin-semaphorin-integrin; I-EGF, integrin-epidermal growth factor. known to complex with divalent cations. In turn, the binding of divalent cations is a requisite event in integrin activation. We speculate that the addition or subtraction of sialic acids within this domain could either influence coordination of divalent cations or alter positioning of the ligand within the ligand binding surface.
Results generated from integrins with artificial N-glycosylation sites, combined with our ligand-binding studies using enzymatically manipulated purified integrins, provide much needed causal evidence that integrin function can be regulated by changes in glycosylation. However, it is important to note that there is an extensive literature showing that naturally occurring changes in integrin glycan structure are associated with dramatic alterations in cell behaviors such as adhesion, migration, and invasion. Variant ␤1 glycoforms have been observed in numerous cell types including fibroblasts, myeloid cells, keratinocytes, cytotrophoblasts, T lymphocytes, and several kinds of epithelial cells (reviewed in Refs. 30 and 31). Importantly, the altered ␤1 glycosylation described in most of these studies occurred in response to physiologic events or stimuli and not merely as a consequence of in vitro molecular manipulations of cell lines. Thus, differential glycosylation likely represents an important feature of the natural biology of integrins containing the ␤1 subunit. The current study adds to the prior body of literature by defining a specific endogenous signaling mechanism that regulates variant ␤1 glycosylation and by demonstrating that hyposialylation is the likely mechanism underlying the known involvement of ERK in regulating the increased cell adhesiveness associated with monocytic/macrophage differentiation.