Fibroblast growth factor 2 (Fgf2) is involved in several biological functions. Fgf2 requires glycosaminoglycans, like chondroitin and dermatan sulfates (hereafter denoted CS/DS) as co-receptors. CS/DS are linear polysaccharides composed of repeating disaccharide units [-4GlcUAβ1-3-GalNAc-β1-] and [-4IdoUAα1-3-GalNAc-β1-], which can be sulfated. Uronyl 2-O-sulfotransferase (Ust) introduces sulfation at the C2 of IdoUA and GlcUA resulting in over-sulfated units. Here, we investigated the role of Ust-mediated CS/DS 2-O sulfation in Fgf2-induced cell migration. We found that CHO-K1 cells overexpressing Ust contain significantly more CS/DS 2-O sulfated units, whereas Ust knockdown abolished CS/DS 2-O sulfation. These structural differences in CS/DS resulted in altered Fgf2 binding and increased phosphorylation of ERK1/2 (also known as MAPK3 and MAPK1, respectively). As a functional consequence of CS/DS 2-O sulfation and altered Fgf2 binding, cell migration and paxillin activation were increased. Inhibition of sulfation, knockdown of Ust and inhibition of FgfR resulted in reduced migration. Similarly, in 3T3 cells Fgf2 treatment increased migration, which was abolished by Ust knockdown. The proteoglycan controlling the CHO migration was syndecan 1. Knockdown of Sdc1 in CHO-K1 cells overexpressing Ust abolished cell migration. We conclude that the presence of distinctly sulfated CS/DS can tune the Fgf2 effect on cell migration.
Fibroblast growth factor 2 (Fgf2) is a member of the Fgf family. This family contains 22 members each with different binding affinities for Fgf receptors (FgfRs) (Ornitz and Itoh 2001; Itoh and Ornitz 2004). Depending on the cell type examined, Fgf2 can play important roles in proliferation, migration, differentiation and cell repair (Yun et al., 2010). In addition to FgfR binding, Fgfs require glycosaminoglycans (GAGs) like heparan sulfate (HS) or chondroitin and dermatan sulfates (hereafter denoted CS/DS) as low affinity receptor for their biological activity (Pellegrini et al., 2000, Taylor et al., 2005). As extracellular matrix (ECM) components these GAGs control multiple cellular functions, including cell–matrix, cell–cell and ligand–receptor interactions, including Fgf2 signaling (Nugent and Edelman 1992, Chua et al., 2004). However, there is growing evidence that CS/DS influence physiological and pathophysiological functions (Seidler et al., 2007; Seidler 2012; Thelin et al., 2012).
The biosynthesis of CS/DS begins in the Golgi with the addition of d-glucuronic acid (d-GlcUA) and d-N-acetylgalactosamine (d-GalNAc) on a GAG-specific carbohydrate–protein linkage region. On polymer level d-GlcUA is epimerized to l-iduronic acid (l-IdoUA) by the dermatan C5-epimerases (Dse and Dse-like) followed by O-sulfation at the C-2, C-4 or C-6 position by several sulfotransferases, thus introducing a micro-heterogeneity along the chain (Yamada and Sugahara, 2008, Malmström, 1984) (supplementary material Fig. S1). The enzyme introducing the sulfate group at position 2-O of d-GlcUA or l-IdoUA is uronyl 2-O sulfotransferase (Ust) (Kobayashi et al., 1999). The role of CS/DS, specifically the 2-O sulfation is not well understood.
Over-sulfated CS is involved in neuronal migration and axon regeneration (Kwok et al., 2012, Maeda and Noda, 1998) and Ust knockdown lead to an altered neurite outgrowth (Ishii and Maeda, 2008). Furthermore, the lack of CS/DS 2-O sulfation in skin of decorin-deficient (Dcn−/−) mice affects Fgf2 and Fgf7 binding and keratinocyte differentiation (Nikolovska et al., 2014). For HS-Fgf2 interaction, l-IdoUA 2-O sulfation plays a vital role and influences binding of Fgf2 to the receptor (Jemth et al., 2002; Ashikari-Hada et al., 2009). However, CHO cells (pgsD-677) deficient in HS still exhibit Fgf2 signaling (Ashikari-Hada et al., 2009) indicating that the remaining CS/DS could support Fgf2 signaling.
Fgf2 activates multiple signaling pathways and results in activation of ERK1 and ERK2 (ERK1/2; also known as MAPK3 and MAPK1, respectively), which in turn regulates proliferation, differentiation, survival and migration (for a review, see Turner and Grose, 2010). Studies have highlighted the importance of the Fgf2-triggered pathways in cell migration (Rieck et al., 2001; Horowitz et al., 2002) and focal adhesion formation (Woods and Couchman, 1992). Interestingly, CS/DS is known to be involved in migration as enzymatic digestion of CS/DS decreases motility of macrophages and osteosarcoma cells (Hayashi et al., 2001; Merle et al., 1999). Furthermore, a decreased CS/DS IdoUA content leads to the loss of directional migration in smooth muscle cells and reduces FAK activation (Bartolini et al., 2013).
The modification of CS/DS, specifically the 2-O sulfation provides important information for modulating the interaction of CS/DS with Fgf2. Our data demonstrate that cell surface CS/DS 2-O sulfation modulates the Fgf2 response by augmenting the induction of ERK1/2 thereby triggering CHO cells to migration.
Ust overexpression in CHO-K1 cells leads to an increase in CS/DS 2-O sulfation
Subcloning of Ust-transfected CHO-K1 cells (denoted CHO-K1/Ust) resulted in two clones with a 70% increase in Ust expression compared to mock-transfected cells (Fig. 1A). Two clones with 30% increase in Ust expression were also tested (data not shown). Immunoblotting showed a significant increase of 60±2% (mean±s.d.) in Ust protein in CHO-K1/Ust clone 2 (Fig. 1B) and clone 10 (data not shown) compared to CHO-K1 cells. As expected, we observed a higher uronyl 2-O sulfotransferase activity in CHO-K1/Ust (0.47±0.07 pmol/min/µg) compared to CHO-K1 cells (0.28±0.01 pmol/min/µg). Chondroitin 6-sulfate was used as substrate and is converted into chondroitin 2,6 sulfate by Ust. Following chondroitinase ABC digestion, the 2,6 sulfate disaccharides were detected by fluorophore-assisted carbohydrate electrophoresis (FACE) (supplementary material Fig. S2). The sulfotransferase activity in CHO-K1/Ust cells displayed a ∼60% increase, similar to the increase in Ust protein. Because there is endogenous Ust expression in CHO-K1, we also investigated CHO-K1 cells where Ust was knocked down by using small hairpin RNA (shRNA) (denoted CHO-K1/shUst). Two clones of CHO-K1/shUst were analyzed and both Ust mRNA and protein were reduced by ∼40% compared to CHO-K1 cells (Fig. 1A,B). The mock-transfected shRNA CHO-K1 cells showed no alterations in Ust expression (Fig. 1B).
To evaluate the changes in CS/DS 2-O sulfation, GAGs from transfected CHO cells were extracted as described previously (Zamfir et al., 2012). Total CS/DS extracts from the CHO-K1, CHO-K1/Ust and CHO-K1/shUst cell lines were similar (Fig. 1C). However, using high-pressure liquid chromatography (HPLC)-DEAE and a discontinuous NaCl gradient for further fractionation, we obtained less highly sulfated CS/DS, eluting at 0.68 M NaCl, in CHO-K1/Ust compared to CHO-K1 and CHO-K1/shUst cells (Fig. 1C). Analysis of the total and highly-sulfated CS/DS extracts revealed significantly higher sulfate content in CHO-K1/Ust and CHO-K1/shUst cells compared to CHO-K1 (data not shown). The disaccharide composition of highly sulfated CS/DS was determined by FACE analysis. CHO-K1 CS/DS and CHO-K1/Ust cells contained six, and CHO-K1/shUst only four different disaccharides units. The percentages of ΔHexUA-GalNAc (ΔDi0S), ΔHexUA-GalNAc(4S) (ΔDi4S) and ΔHexUA-GalNAc(4S,6S) (ΔDi4,6S) units were similar in all three cell lines (Fig. 1D). CHO-K1 cells contained 2±0.8% of ΔHexUA(2S)-GalNAc (ΔDi2S), which was increased by Ust overexpression to 11±3.7%. Furthermore, we detected traces of oversulfated ΔHexUA(2S)-GalNAc(4S or 6S) [ΔDi2,XS-units (X = 4 or 6)] in CHO-K1 cells, which was increased by a factor of >5% in Ust overexpressing cells. Knocking down Ust abolished both ΔHexUA(2S)-GalNAc (ΔDi2S) and ΔHexUA(2S)-GalNAc(4S or 6S) [ΔDi2,XS units (X = 4 or 6)]. Interestingly, increases in the ΔHexUA-GalNAc(6S) (ΔDi6S) units were also observed, by 45% in CHO-K1, 36% in CHO-K1/Ust and 60% in CHO-K1/shUst cells (Fig. 1D). The HS disaccharide structures are similar between the different clones (data not shown). To determine possible functional consequences of altered 2-O sulfation, the binding capacity of Fgf2 to CS/DS was analyzed.
Increased 2-O sulfation influences the binding of Fgf2 to CS/DS
Fgfs interact with cell surface GAGs, including CS/DS in a manner that is dependent on sulfation pattern (Taylor et al., 2005). A solid-phase binding assay has previously shown that there were different binding properties between the highly sulfated CS/DS extracted from wild-type and Dcn−/− mice towards digoxygenin (Dig)-conjugated Fgf2 (Nikolovska et al., 2014). BSA was used as negative control and decorin with ∼50% IdoUA content as positive control (Seidler at al., 2006). Dig–Fgf2 binding was significantly increased to 2-O over-sulfated CS/DS (CHO-K1/Ust) compared to control CS/DS (CHO-K1). In contrast, the loss of 2-O sulfation significantly decreased the binding of Dig–Fgf2 (Fig. 2A).
Using CHO cells, we determined the impact of FgfR and GAGs on binding of Dig–Fgf2 to the cell surface (Fig. 2B). A database analysis has shown that CHO-K1 cells express four different FgfRs (Hammond et al., 2012). Nonspecific binding of Dig–Fgf2 was analyzed with the Fab fragments of the anti-digoxigenin antibody. A similar binding of human Dig–FGF2 (data not shown) and murine Dig–Fgf2 to CHO-K1, CHO-K1/Ust and CHO-K1/shUst cell surface (Fig. 2B) was observed. In the following experiments we used murine Fgf2. HS digestion significantly reduced Dig–Fgf2 binding to all cell lines. However, after HS digestion, CHO-K1/Ust cells had a significantly higher Fgf2-binding capacity than did CHO-K1 and CHO-K1/shUst (Fig. 2C). Digestion of CS/DS (enzyme cleavage sites are as shown in supplementary material Fig. S1) reduced Fgf2 binding for the three cell lines. However, Fgf2 binding was significantly lower for CHO-K1/Ust compared to CHO-K1 and CHO-K1/shUst (Fig. 2C) indicating that 2-O sulfation of CS/DS is involved in this process. As expected, enzymatic digestion of cell surface HS and CS/DS decreased the Dig–Fgf2 binding to the baseline level (Fig. 2B). To determine whether CS or DS are involved in the Fgf2 binding, HS was first enzymatically digested with a heparitinase mix followed by chondroitin ACI lyase to remove CS or chondroitin B lyase to remove DS (supplementary material Fig. S1). Digestion of cell surface CS revealed that CHO-K1/Ust cells bound significantly more Fgf2 compared to CHO-K1 and CHO-K1/shUst. In contrast, the presence of CS, after chondroitin B lyase treatment, caused no difference in Fgf2 binding for any cell lines (Fig. 2B). These results suggest that DS is present on the cell surface and that 2-O sulfated DS might block Fgf2 binding. Presence of DS oligosaccharides on the cell surface of CHO cells was confirmed by gel permeation chromatography (Seidler et al., 2002) and further defined by fluorophore-assisted carbohydrate electrophoresis (FACE) disaccharide analysis (data not shown). Oligosaccharide analysis showed no differences in the amount of DS in CHO-K1 or CHO-K1/Ust cells (Table 1).
Chondroitin-ACI-lyase-digested highly sulfated CHO-K1 and CHO-K1/Ust CS/DS GAGs were fractioned on a Superdex Peptide column. The column was calibrated with cytochrome C to determine void volume (V0) and glycin to determine total column volume (Vt) (n = 2).
To confirm Fgf2 binding to CS/DS, we used cell line pgsD-677, which lacks endogenous HS (Fig. 2C) (Lidholt et al., 1992). pgsD-677 cells express 2–3 times more CS and substantially more Ust than does CHO-K1 cells (data not shown). To obtain similar Ust levels to those in CHO-K1 or CHO-K1/shUst cells, we knocked down Ust in pgsD-677 cells (denoted pgsD-677/shUst) and observed a significantly lower Fgf2-binding capacity than in pgsD-677 cells (Fig. 2C). This supports the hypothesis that CS/DS 2-O sulfation is required for Fgf2 binding. Enzymatic digestion of CS/DS in pgsD-677 and pgsD-677/shUst cells reduced Fgf2 binding to control levels. Sequential digestion of CS or DS blocks revealed no differences in the binding to pgsD-677 cells; however, lack of 2-O sulfation resulted in a significantly reduced Fgf2-binding capacity (Fig. 2C).
To demonstrate the importance of cell surface GAG sulfation, cells were pre-treated with 30 mM NaClO3. Chlorate treatment leads to an inhibition of 3′-phosphoadenosine-5′-phosphosulfate (PAPS) synthesis, a reduction in cell surface sulfate content (Keller et al., 1989) and reduced GAG sulfation. NaClO3 treatment of CHO-K1 and CHO-K1/Ust cells resulted in lower-sulfated CS/DS, eluting from an anion-exchanger at 0.275 M NaCl instead of 0.68 M NaCl (data not shown), and a significantly reduced Fgf2-binding capacity. Interestingly, NaClO3-treated CHO-K1/Ust cells bound significantly more Fgf2 than did CHO-K1 cells. This was abolished after CS/DS digestion. HS digestion after NaClO3 treatment did not influence the Fgf2-binding capacity of CHO-K1/Ust cells (Fig. 2D). These results indicate that CS/DS 2-O sulfation is important for Fgf2 binding to the cell surface of CHO cells.
Increased CS/DS 2-O sulfation in CHO cells affects proliferation
Next, we determined whether CS/DS 2-O sulfation affects CHO cell proliferation. All analyzed CHO cell lines proliferate even without exogenous stimuli (Fig. 3A). CHO-K1/Ust cells have a significantly higher proliferation rate (tdouble = 34.5 h) compared to CHO-K1 (tdouble = 45 h) and CHO-K1/shUst (tdouble = 49 h) cells. Using Fgf2 for 24 h as a stimulus, CHO-K1 and CHO-K1/shUst proliferated significantly more than CHO-K1/Ust cells (Fig. 3A). Surprisingly, CHO-K1/Ust Fgf2-treated cells had a significantly reduced proliferation compared to non-treated cells (Fig. 3A). Chlorate treatment showed similar results; however, CHO-K1/Ust cells proliferated more than cells containing less CS/DS 2-O sulfation (Fig. 3A). Fgf2 plus chlorate treatment significantly reduced proliferation of CHO-K1 and CHO-K1/shUst cells, indicating the importance of the GAG sulfation in this process. Surprisingly, CHO-K1/Ust cells treated with chlorate plus Fgf2 still proliferated (Fig. 3A). The role of CS/DS 2-O sulfation in Fgf2 function was further evaluated in pgsD-677 cells. These cells can respond to Fgf2 even in the absence of HS (Ashikari-Hada et al., 2009). In these cells, Ust knockdown led to a significant inhibition of proliferation compared to pgsD-677 cells. Neither pgsD-677 nor pgsD-677/shUst cells responded to Fgf2 treatment (data not shown).
By using the FgfR-specific inhibitor PD173074, we found that blocking FgfR diminished the proliferation differences between CHO-K1, CHO-K1/Ust and CHO-K1/shUst cells (Fig. 3B). To understand the difference in proliferation induced by 2-O sulfation, Fgf2 signaling was analyzed for presence of ERK1/2 activation. ERK1/2 was significantly activated by Fgf2 in CHO-K1/Ust. Interestingly, Fgf2 had no influence on ERK1/2 activation in CHO-K1/shUst cells lacking CS/DS 2-O sulfation (Fig. 3C,D). Fgf2 treatment of pgsD-677 cells lacking HS-induced ERK1/2 phosphorylation (data not shown). A slight activation of ERK1/2 has been reported previously for pgsD-677 cells when using human FGF2 (Ashikari-Hada et al., 2009). Therefore, we examined the effect of human FGF2 on CHO-K1 cells and also found low ERK1/2 activation (supplementary material Fig. S3), despite the absence of differences in the binding of the two Fgf species to CS/DS.
Our results indicate the importance of CS/DS on Fgf2 signaling in CHO cells. The fact that cell surface CS/DS sulfation is a prerequisite for Fgf2 signaling (Ling et al., 2006; Ramachandra et al., 2014) is supported by the finding that there was a reduced basal ERK phosphorylation after chlorate treatment for all tested cell lines (data not shown). Treatment with the inhibitor PD173074, specific for FgfR kinase, revealed the same results as for chlorate treatment (data not shown).
2-O sulfation of cell surface CS/DS affects cell migration
Given that Fgf2 induced ERK activation and did not affect CHO-K1/Ust and pgsD-677 cell proliferation, we used a wound scratch assay to test whether Fgf2 can induce migration in these cell lines. The presence of serum did not influence cell migration. The migration rate of CHO-K1/Ust cells under starving conditions in low-glucose medium was ∼40% higher than for CHO-K1 cells (Fig. 4A,B). Interestingly, pgsD-677 displayed an ∼50% to 70% higher number of migrating cells compared to CHO-K1/Ust and CHO-K1 cells (Fig. 4A,B). Although Ham's F12, a high-glucose medium, is recommended for culturing CHO-K1 cells (Lidholt et al., 1992), we used low-glucose minimal essential medium (MEM) to avoid high glucose levels (Wang and Hascall, 2009). We also tested Ham's F12 medium and observed an even faster migration of CHO-K1/Ust compared to CHO-K1 cells (supplementary material Fig. S4). After 28 h of Fgf2 treatment, the number of migrating CHO-K1/Ust and pgsD-677 cells was increased by ∼50%. No differences were observed in total cell numbers for CHO-K1 or CHO-K1/Ust cells (data not shown). Fgf2 did not induce changes in CHO-K1 cell migration (Fig. 4B). Knockdown of Ust in CHO-K1 and pgsD-677 cells completely blocked the cell migration, even in presence of Fgf2 (data not shown). This confirms our hypothesis that 2-O sulfation of CS/DS is important for Fgf2 signaling. Furthermore, increased CS/DS 2-O sulfation triggers cells to migrate but does not induce proliferation (Fig. 3A).
To demonstrate the importance of GAG sulfation for Fgf2-induced migration, a wound scratch assay was performed in the presence of chlorate. Analysis of the faster migrating cell lines CHO-K1/Ust and pgsD-677 showed a 30% reduction in migration upon chlorate treatment (Fig. 4C,D).
To analyze the impact of HS, CHO-K1 and CHO-K1/Ust cells were treated for 1 h with heparitinase mix (Hep-ase) and were analyzed by a wound scratch assay for 28 h. Fig. 4E,F show that the digestion of HS influences migration of both cell lines. CHO-K1 cells treated with Hep-ase migrated significantly faster in the presence of Fgf2. Migration of CHO-K1/Ust cells treated with Hep-ase was significantly increased without Fgf2 compared to CHO-K1 cells (Fig. 4E,F). Of note, although digestion of HS alters the proliferation of CHO-K1 and CHO-K1/Ust cells, no differences in cell number were observed between CHO-K1 with or without Fgf2. In contrast, digestion of HS of CHO-K1/Ust cells led to 10% increase in cell number followed Fgf2 treatment as compared to CHO-K1 cells. CS/DS digestion with chondroitinase ABC for 1 h followed by wound scratch assay with Fgf2 treatment blocked migration in both cell lines (data not shown). We also observed for all non-migrating cell lines that there was a 15% increase in cell number. Furthermore, the FgfR inhibitor PD173074 completely blocked Fgf2 induced migration in CHO-K1/Ust and pgsD-677 cells (Fig. 4G). These results indicate that CS/DS 2-O sulfation influences the presentation of Fgf2 to the respective cell surface receptor, inducing migration.
CS/DS 2-O sulfation and Fgf2 affects paxillin activation
Cytoskeleton organization and focal adhesion complex formation are known regulators of cell migration (Mitra et al., 2005). To further support the role of CS/DS 2-O sulfation in migration, we visualized F-actin and paxillin activation in these cells. Ust-transfected CHO-K1 cells adhered and spread differently compared to CHO-K1 cells. Ust overexpression induced morphological alterations in the F-actin compared to the normal phenotype of CHO-K1 cells (Fig. 5A) (Naik and Naik, 2011). For CHO-K1 cells, which proliferate in response to Fgf2, weak paxillin activation was observed. In contrast, Fgf2 treatment of CHO-K1/Ust and pgsD-677 cells promoted paxillin phosphorylation, corresponding to increased cell migration (Fig. 5A).
These results show that migration of CHO cells induced by Fgf2 can be correlated to the presence of CS/DS cell surface 2-O sulfation. Furthermore, this event is mediated by activation of ERK1/2 and paxillin. To strengthen our hypothesis that a minor modification like CS/DS 2-O sulfation is important for migration, we analyzed 3T3 fibroblasts. Previously, we have shown that 3T3 fibroblasts express Ust (Nikolovska et al., 2014). Therefore, we knocked down Ust in 3T3 fibroblasts which led to a ∼40±3% reduction in Ust protein (Fig. 5B,C). The functional consequence of the Ust knockdown was also evaluated by a wound scratch assay. Time course experiments showed that 3T3 and 3T3/mock cells close the gap within 40 h when cultured in MEM containing FBS (data not shown). Treatment with Fgf2, however, revealed that 3T3 and 3T3 mock cells closed the wound within 24 h, whereas this took ∼36 h for 3T3/shUst cells (Fig. 5D).
Lack of syndecan 1 affects migration of CHO cells
Epidermal cells that display a high degree of migration and proliferation express a high amount of the proteoglycan syndecan 1 (Sdc1), which is modified with HS and CS chains (Elenius et al., 1991; Kato et al., 1995). To confirm a role of 2-O sulfated CS/DS of syndecan 1 in migration we knocked down this hybrid proteoglycan by using shRNA (shSdc1) (Bernfield et al., 1992; Chen et al., 2009; Deepa et al., 2004). Syndecan 1 was significantly knocked down in CHO-K1/shSdc1 and CHO-K1/Ust/shSdc1 cells (Fig. 6A,B); the analysis of syndecan 1 expression revealed an ∼70% reduction for CHO-K1/shSdc1 ∼70% and ∼60% for CHO-K1/Ust/shSdc1 cells. Reduced syndecan 1 expression affected cell adhesion in both these cells line. Subsequently the cells were cultured on d-lysine-coated plates. Cell migration was inhibited for both CHO-K1/shSdc1 and CHO-K1/Ust/shSdc1 cells compared to their respective mock controls. Digestion of HS on syndecan 1 knockdown cells resulted in impaired adhesion (data not shown). Syndecan 1 knockdown of both cell lines abolished the pro-migratory effect of Fgf2 (Fig. 6C), as is supported by their lack of protrusions and paxillin activation (Fig. 6D). There was no difference in proliferation rates in Sdc1 knockdown CHO cells and Fgf2 did not affect cell proliferation (data not shown). These experiments support the hypothesis that cell surface CS/DS 2-O sulfation and syndecan 1 are involved in Fgf2 signaling in cell migration.
We previously demonstrated that the reduced amount of Ust and the lack of 2-O sulfated CS/DS in skin or fibroblasts of Dcn−/− mice leads to an altered Fgf7- and Fgf2-binding capacity (Nikolovska et al., 2014), suggesting they have a new and important role in modulating the growth factor response. In the present study, we show that the degree of 2-O sulfation of cell surface CS/DS can modulate Fgf2-induced migration through ERK1/2 activation. Furthermore, knockdown of the proteoglycan syndecan 1 abolished Fgf2-mediated migration.
The first step of Fgf2 signaling is the ligand binding to cell surface GAGs of proteoglycan, the low-affinity receptors, following which ligand presentation to FgfRs occurs. HS proteoglycans and heparin are well studied in this context (Pellegrini et al., 2000) as they promote cell proliferation and ERK1/2 activation (Guimond et al., 1993, Pye and Kumar, 1998, Delehedde et al., 2000). Recent studies have revealed distinct structural features of heparin that allow it to bind to Fgfs with a clear preference for the importance of 2-O sulfation of l-IdoUA (Xu et al., 2012; Jemth et al., 2002). The structural and functional aspects of HS and CS/DS in Fgf mitogenic activity are controversial, as both stimulation and inhibition have been reported (Guimond and Turnbull, 1999; Pye and Kumar, 1998; Fthenou et al., 2009; Yamada and Sugahara, 2008). This could be because the respective proteoglycans, such as syndecan 1 and 4 expressed by a mouse mammary gland epithelial cell line, can be linked with both CS and HS with different structures (Deepa et al., 2004).
Here, we found that genetic modulation of Ust influenced the Fgf2-binding capacity depending on the presence or absence of CS/DS 2-O sulfation. This observation is supported by analysis of pgsD-677 cells, which deficient in HS (Lidholt et al., 1992) but still exhibit Fgf2-induced ERK1/2 activation, although at a lower level than control cells (Ashikari-Hada et al., 2009). Notably, we observed increased Ust expression in pgsD-677 cells to a similar extent as CHO-K1/Ust (data not shown). The Ust knockdown in pgsD-677 resulted in reduced Fgf2 binding and loss of Fgf2-induced ERK1/2 activation. Inhibition of either GAG sulfation or FgfR also abolished ERK1/2 activation in CHO-K1/Ust or pgsD-677 cells exposed to Fgf2. Our model clearly shows that Fgf2 does not induce proliferation if CS/DS 2-O sulfation is increased on the cell surface despite ERK1/2 activation. Brittle star (class Ophiuroidea) 2-O sulfated CS/DS has a high amount of l-IdoA and is able to activate ERK1/2 in the presence of Fgf2 in CHO cells that lack any GAGs (Ramachandra et al., 2014). An explanation could be the domain organization of GAGs, as shown for HS oligosaccharides and their ability to modulate Fgf2-induced ERK1/2 activation (Jastrebova et al., 2010).
Although ERK1/2 activation contributes to proliferation (Talarmin et al., 1999; Chua et al., 2004), it is also implicated in cell migration. ECM proteins are also able to trigger epithelial or endothelial cell migration through ERK1/2 activation (Klemke et al., 1997; Webb et al., 2000). We visualized migration and showed activation of ERK1/2 and paxillin in CHO cells with increased amount of CS/DS 2-O sulfation. Paxillin activation regulates focal adhesion dynamics, thus regulating migration and cytoskeleton rearrangement (Parsons et al., 2000). GAGs bound to the cell membrane are also involved in cytoskeletal organization (Okina et al., 2012).
To evaluate the potential proteoglycan in cell migration, we knocked down Sdc1. Syndecan 1 is the main proteoglycan expressed in simple epithelial cells that bears HS and CS chains (Kato et al., 1995; Rapraeger et al., 1985), and it is known to influence Fgf2 mitogenic activity (Eriksson and Spillmann, 2012). Of note, syndecan 4, a cell surface proteoglycan also involved in migration, is not detected in CHO-K1 cells (Echtermeyer et al., 1999). In our experimental settings, CHO-K1 cells endogenously expressing Sdc1 did not migrate, and Sdc1 knockdown did not affect migration. However, some of our data support previously published observations concerning cell proliferation, adhesion, morphology and migration (Liu et al., 1998; Carey et al., 1994; Kato et al., 1995, Chen et al., 2009). First, knockdown of Sdc1 inhibited Fgf2-induced proliferation in CHO-K1 cells; second, adhesion was affected for both CHO-K1 and CHO-K1/Ust cells; and third, Fgf2 treated Sdc1 knockdown cells have a different F-actin distribution and morphology on plastic, independent of the level of Ust. Our results are further supported by the lack of activated paxillin in Fgf2-treated CHO-K1/shSdc1 and CHO-K1/Ust/shSdc1 cells. These results support our hypothesis that 2-O sulfation of CS/DS as a minor modification, probably on syndecan 1, affects Fgf2-induced migration. It has been shown that minor differences in disaccharide sequences result in remarkable changes in ligand specificity (Guimond et al., 1993; Nurcombe et al., 1993; Sanderson et al., 1994). The syndecan 1 ectodomain can also influence Fgf2 mitogenic activity and migration (Eriksson and Spillmann, 2012; Chen et al., 2009). Upon treating cells with heparin lyase to remove attached HS chains, the amount of the syndecan 1 ectodomain in conditioned medium is increased, suggesting that having less HS on the core protein will increase the shedding (Ramani et al., 2012). One might speculate that the specific sulfation pattern of syndecan 1 CS/DS, like the absence of 2-O sulfation, would have an impact on the sensitivity for proteolytic action, so that shedding by matrilysin (Endo et al., 2003) results in an increased ectodomain and signals back to the cells (Eriksson and Spillmann, 2012).
Consistent with our findings is the impact of over-sulfated CS in in vivo neuronal migration (Ishii and Maeda, 2008) and loss of directional migration of aortic smooth muscle cells with reduced cell surface l-IdoUA content (Bartolini et al., 2013). This underlies the importance of DS and the downstream 2-O sulfation in mediating cell migration and adhesion (Malmström et al., 2012). Notably, Dse1−/− mice display reduced levels of l-IdoUA blocks and ΔDi2,4S units and therefore might have an altered growth factor function (Maccarana et al., 2009). Interestingly, we detected DS units on the cell surface of the CHO cell lines. In vivo a reduction of DS and, therefore, a decrease in 2-O sulfation diminishes neuronal stem cell proliferation associated with an upregulation of growth factors (Bian et al., 2011). Analysis of epithelial syndecan 1 GAGs has shown that CS and HS can simultaneously bind Fgf2 suggesting that there is a regulatory mechanism in presenting the factors to the cell surface receptors (Deepa et al., 2004). Our results support this possibility, because Sdc1 knockdown abolished Fgf2-induced migration in Ust overexpressing CHO cells. A functional overlap of CS and HS for VEGF-induced angiogenesis has been previously discussed (Le Jan et al., 2012). Recently, the small non-coding microRNA149 has been linked to a new fine-tuning element of Fgf2-induced proliferation in endothelial cells by regulating the HS proteoglycan glypican (Chamorro-Jorganes et al., 2014). We have previously shown that the lack of decorin in skin led to the reduction of Ust and an impaired cell morphology in a 3D environment (Nikolovska et al., 2014; Jungmann et al., 2012), which could partially contribute to the delayed wound healing of the Dcn−/− mice (Järveläinen et al., 2006). We proved the importance of 2-O sulfation by showing that there was an impaired migration of 3T3 cells with a knockdown of Ust.
In summary, our data demonstrate a new role of cell surface CS/DS in Fgf2-induced migration. The results link specific GAG-dependent modification to ERK1/2 activation and increased migration. The proteoglycan likely to be involved in this process is syndecan 1 modified with HS and CS. These findings might explain further the altered wound healing in Dcn−/− mice as being caused by reduced fibroblast migration.
MATERIALS AND METHODS
Materials and cells
Primary antibodies against the following proteins were used: UST (D-20) and Sdc1 (H-174) (Santa Cruz Biotechnology), β-actin, p44/42 (ERK1/2) MAPK Rabbit mAb and phospho-p44/42 MAPK (Thr202/Tyr204) Rabbit mAb, phospho-paxillin (Tyr118) (1∶1000; all Cell Signaling; MA). Secondary antibody was horseradish peroxidase (HRP)-conjugated anti-rabbit-IgG secondary antibody (GE Healthcare, UK). F-actin cytoskeleton was visualized by Alexa-Fluor-488-conjugated phalloidin (1 unit/100 µl, Invitrogen). Recombinant mouse Fgf2 and human FGF2 were from Sigma Aldrich, chondroitin ABC lyase, chondroitin ACI lyase, chondroitin B lyase, heparitinase mix (heparinase II/III, 4∶1) were from Amsbio, UK, and heparin lyase I, II and III were from IBEX, Montreal, Canada.
Chinese hamster ovary cells (CHO-K1) and NIH/3T3 fibroblasts were obtained from American Type Culture Collection (Manassas, VA). Mutant CHO cells (pgsD-677) lacking heparan sulfate (Lidholt et al., 1992) were a gift from Jeffrey D. Esko (UCSD, CA, USA). Cells were maintained in α-minimum essential medium (MEM) with 5.6 mM d-Glucose, supplemented with 10% fetal bovine serum (FBS) (Biochrom AG, Berlin), 1% non-essential amino acids, 100 U/ml penicillin and 0.1 mg/ml streptomycin (PAA, Austria) under 5% CO2 at 37°C.
GAG sulfation was partially inhibited with NaClO3 in MEM medium containing sulfate. The concentration was determined by titration curves with 10, 20, 30, 50 and 100 mM of NaClO3. An MTT test (Sigma) was used to follow possible cytotoxic effects (data not shown). At 30 mM NaClO3 cells were viable.
Overexpression and knockdown of Ust and knockdown of Sdc1
The full length murine Ust (NCBI BC138155) cDNA clone IRCKp5014K0715Q (IMAGE: 8861174) obtained from imaGENES (Berlin, Germany), was subcloned by digesting with EcoRI and ligating into pcDNA3.1(+) (Invitrogen). CHO-K1 cells were transfected with pcDNA3-Ust using Polyfect (Qiagen; Hilden, Germany) according to the manufacturer's instructions, and selected with 300 µg/ml Geneticin (Life Technology, Paisley, UK) in MEM. After 4 weeks, cells were subcloned from a single cell by limiting dilution. As a control, CHO-K1 cells were mock-transfected with the vector alone.
CHO-K1, pgsD-677 and NIH/3T3 cells were stably transfected with shRNA-Ust(m) plasmid as a pool of three target-specific lentiviral vector plasmids each encoding 19–25 nt (plus hairpin) shRNAs designed to knockdown Ust gene expression (Santa Cruz Biotechnology) following the manufacturer's protocol. As a control, cells were mock-transfected with control shRNA plasmid-A. CHO-K1 and pgsD-677 shUst-transfected cells were selected with 10 µg/ml puromycin (Santa Cruz Biotechnology) for 2 weeks. Generated CHO-K1/shUst and pgsD-677/shUst cell lines were subcloned by limiting-dilution, whereas NIH/3T3 cells were selected with 2.5 µg/ml puromycin for 1 week.
Sdc1 knockdown cell lines were generated by transfecting CHO-K1 and CHO-K1/Ust cells with the shRNA-Syndecan1(m) plasmid as described for shUst. CHO-K1/shSdc1 and CHO-K1/Ust-shSdc1 were selected with 10 µg/ml puromycin for 2 weeks.
RNA extraction and quantitative real-time PCR
Total RNA, extracted with an RNAeasy kit, was transcribed into cDNA using the Omniscript RT Kit (both Qiagen, Hilden, Germany). cDNA (25 ng) was used as a template to monitor sulfotransferases expression levels by qRT-PCR using the MESA GREEN qPCR kit (Eurogentec, Köln, Germany). The following primers were used: uronyl-2-sulfotransferase-F, 5′-CTACCCTGCTGGTCTTCTGC-3′; uronyl-2-sulfotransferase-R, 5′-AAGCAAGACCACGGTACGAC-3′; β-actin-F, 5′-GGGTGTGATGGTGGGAATGG-3′; β-actin-R, 5′-TGGCTGGGGTGTTGAAGGTC-3′; GAPDH-F, 5′-ATTCAACGGCACAGTCAAG-3′; GAPDH-R, 5′-TTCACACCCATCACAAACAT-3′; ubiquitin-F, 5′-GCCCAGTGTTACCACCAAG-3′; and ubiquitin-R′ 5′-CACCAAAGAACAAGCACAAG-3′. The expression was normalized to β-actin, GAPDH and ubiquitin and determined by the ΔΔCt method (Livak and Schmittgen, 2001).
Detection of Ust and syndecan 1 in CHO cells and NIH/3T3 cells
Cell lysates from ∼1×106 cells were obtained using a lysis buffer [7 M Urea, 2 mM Thiourea, 40 mM Tris-HCl pH 8, 0.001% (w/v) Bromphenol Blue, 1% (w/v) ASB-14]. Cell lysates (50 µl) were separated under reduced conditions with a 4.5–15% gradient SDS–gel and transferred onto a PVDF membrane. Membranes were blocked with 5% skim milk in TBST for 1 h at room temperature and incubated overnight at 4°C with a Ust antibody (1∶500). Membranes washed with TBST were incubated for 1 h at room temperature with anti-rabbit-IgG HRP-conjugated secondary antibody and visualized by enhanced chemiluminescence (Perkin-Elmer Life Sciences, Boston, MA) and Fusion-SL 4.2 MP (PeqLab, Erlangen, Germany). Membranes were treated with stripping buffer (100 mM β-mercaptoethanol, 62.5 mM Tris-HCl pH 6.8 and 2% SDS) at 50°C for 30 min, reprobed for β-actin overnight at 4°C and developed as described for Ust. Band density was determined with ImageJ analysis software (rsb.info.nih.gov). Ust expression was normalized to β-actin. Detection of syndecan 1 in CHO cell lines was performed in ∼1×106 cells (shSdc1 and mock transfected) lysed as described for Ust. Syndecan 1 was detected with a polyclonal antibody.
Sulfotransferase activity in CHO-K1 and CHO-K1/Ust cells
Sulfotransferase reaction was carried out according to the manufacturer instructions using the universal sulfotransferase assay (R&D). To determine sulfotransferase activity protein lysates (25–200 µg) of CHO-K1 and CHO-K1/Ust cells were incubated with 10 mM chondroitin 6-sulfate as substrate, PAPS (R&D), and a coupling phosphatase and the respective controls. The color was developed by addition of 30 µl of Malachite reagent at room temperature for 20 min and monitored at 620 nm with an ELISA reader [optical density (OD)/µg]. A phosphate standard curve was used to determine the activity (OD/pmol). The specific activity was determined with the following equation: , where S is the slope of the line with the OD values of the sulfotransferase assay and CF the phosphate conversion factor (taken from the phosphate standard). To evaluate the Ust activity, the products were subjected to FACE analysis (supplementary material Fig. S2).
Purification of cell surface CS/DS proteoglycans from CHO-K1/Ust, CHO-K1/shUst and CHO-K1
GAGs were extracted as described previously (Zamfir et al., 2012). Briefly, ∼4×107 cells were lysed in extraction buffer (4 M guanidium hydrochloride, 50 mM sodium acetate, 1× protease-inhibitor-mix pH 6; Roche Applied Sciences, Mannheim, Germany) and stirred overnight at 4°C. Cell lysates were applied on a preparative DEAE-Ceramic HyperD F (Pall, Life Sciences, France) column and CS/DS was eluted as previously published (Nikolovska et al., 2014). Further CS/DS fractionation was performed using a DEAE-PW5 column (Tosoh, Stuttgart, Germany) as described previously (Seidler et al., 2002; Zamfir et al., 2009). Under-sulfated CS/DS were separated from highly sulfated CS/DS using different NaCl concentrations. In the following experiments, the fraction eluting with 0.68 M NaCl was defined as highly sulfated CS/DS.
Characterization of CS/DS
CS/DS were released by reductive β-elimination and purified over the DEAE column. The HexUA content was determined using an m-hydroxydiphenyl reaction (van den Hoogen, 1998). Uronic acid was hydrolyzed in 80% sulfuric acid containing tetraborate at 80°C, incubated with m-hydroxydiphenyl (Sigma Aldrich) at room temperature and measured at OD540 using heparin as standard. The total sulfate content was then evaluated (Nikolovska et al., 2014). Briefly, 100 µl of sample was hydrolyzed with 50 µl of 3 M HCl for 3 h at 110°C and then cooled down. Luviskol K90 [2% (w/v)] and BaCl2 [1% (w/v) in H2O] was added to the sample (100 µl). After 30 min of rigorous shaking at room temperature samples were measured at OD366 using (NH4)2SO4 as a standard.
Disaccharide analysis of glycosaminoglycans
10 µg CS/DS were digested with 10 mU of chondroitinase ABC, chondroitin ACI and chondroitin B lyase. The disaccharides were labeled with 2-aminoacridon (AMAC) and separated on 30% borate-polyacrylamide gel (Seidler et al., 2002). Chondroitinase-ABC-digested CS6S was used as a standard. The AMAC disaccharides were detected with UV-Transiluminator and monitored by Photosystem (Biometra). Signal intensities were analyzed using ImageJ software. Cell pellets of CHO-K1, CHO-K1/Ust and CHO-K1/shUst cells were prepared as described previously to analyze HS composition (Kumar et al., 2014). After enzymatic removal of CS/DS, the heparin lyase I-, II- and III-digested CHO-K1, CHO-K1/Ust and CHO-K1/shUst GAGs were fractioned by RPIP-HPLC. The peaks were identified by co-elution with standard HS disaccharides.
Binding of Dig-conjugated Fgf2 to CS/DS and the cell surface of CHO cells
The binding of Dig–Fgf2 to extracted CS/DS was carried out as described previously (Nikolovska et al., 2014). 50 ng proteoglycans were coated to Nunc MaxiSorp plates (Fisher Scientific, Germany) overnight at 4°C in coating buffer (50 mM NaCl, 100 mM Tris/HCl, pH 7.4). Plates were washed, blocked with 1% BSA in TBS and then incubated with Dig–Fgf2.
5×103 cells/well suspended in MEM were seeded in 96-well plates for 24 h, then serum-starved overnight and treated as described previously (Ashikari-Hada et al., 2009). The influence of different cell surface GAGs on Fgf2 binding was determined by their digestion with 4 mU (1) of heparitinase II and III; (2) chondroitinase ABC; (3) chondroitinase ABC and heparitinase II and III; (4) chondroitinase ACI and heparitinase II and III or (5) chondroitinase B and heparitinase II and III at 37°C for 15 min. Non-digested cells were used as a positive control. The influence of GAG sulfation was determined by treating the cells with 30 mM NaClO3 (Keller et al., 1989) for 6 h at 37°C prior to the enzyme digestion. Cells were washed with PBS and incubated for 2 h at 4°C with Dig–Fgf2 prepared as described previously (Ashikari-Hada et al., 2009; Nikolovska et al., 2014). Cells were washed, fixed for 30 min with 0.4% PFA in PBS and blocked with 10 mg/ml BSA in PBS for 1 h at 4°C. Alkaline-phosphatase-conjugated Fab fragments of the anti-digoxigenin antibody were incubated for 1 h at room temperature, washed and developed with 2 mg/ml p-nitrophenyl phosphate in 1 M diethanolamine/HCl, pH 9.8. The enzyme activity was measured at OD405.
Proliferation of different CHO cell lines stimulated with Fgf2
3×104 cells/cm2 cells were seeded in six-well plate, cultured for 24 h and serum starved for 16 h prior to the experiment. Experiments were performed in serum-free MEM supplemented with or without 10 ng/ml Fgf2 for 24 h. The same experiment was repeated in the presence of 30 mM NaClO3 with or without Fgf2, after 6 h pretreatment of the cells with 30 mM NaClO3. Cells were pretreated with the FgfR inhibitor PD173074 (20 nM) (Sigma Aldrich) for 6 h, followed by 24 h incubation with 20 nM PD173074 with or without Fgf2 (Welti et al., 2011). Working concentrations for Fgf2, NaClO3 and PD173074 were determined based on titration curves. Proliferation experiments were evaluated by cell counting or BrdU incorporation (Cell Proliferation ELISA, Roche Applied Science).
Wound scratch assay
CHO cell lines (CHO-K1, CHO-K1/Ust, CHO-K1/shUst, pgsD-677 and pgsD-677/shUst) and NIH/3T3 fibroblasts (NIH/3T3 and NIH/3T3/shUst) were seeded in six-well plates at a density of 1×105 cells/cm2 in MEM medium and starved overnight. 5×104 cells/cm2 CHO-K1/shSdc1, CHO-K1/Ust-shSdc1 and their respective mock controls were seeded on pre-coated d-lysine (50 µg/well) 24-well plates, cultured in MEM medium and starved overnight. An artificial wound was generated by a scratch and cells were incubated with serum-free MEM (control) or MEM supplemented with 10 ng/ml Fgf2 for 6, 12 and 28 h at 37°C (Liang et al., 2007). Cell migration was determined after 6 h pre-treatment of the cells with 30 mM NaClO3 or 20 nM PD173074 followed by 28 h incubation with or without 10 ng/ml Fgf2. To digest cell surface HS, CHO-K1 cell lines were pre-incubated with heparitinase II and III for 1 h, for the digestion of CS/DS, cells were pretreated with chondroitin ABC lyase. In the case of NIH/3T3, cells were starved for 2 h before scratching, followed by 24 h incubation with or without 10 ng/ml Fgf2. Images were captured, using a Zeiss Axiovert 100 microscope with AxioCam ICc1 camera, at 0, 24 and 28 h. Cell migration was measured by counting the number of migrating cells using ImageJ. First, the threshold was set for each grayscale image, thus converting them into a binary image. Overlapping cells were separated using the Watershed segmentation process. Selection frame with the gap size at time point 0 was generated and applied over the same position in images at 28 h, to calculate the number of cells migrated in the gap. For each well two to four pictures were acquired (n = 3 independent experiments).
Analysis of ERK1/2 signaling molecules
2.3×105 cells/cm2 CHO cells were seeded in a six-well plate, starved overnight then treated for 5, 15 and 30 min with or without 10 ng/ml Fgf2. Cells were harvested with lysis buffer [10 mM Tris-HCl pH 7.4, 150 mM NaCl, 1% Triton X-100, 1.5 mM EDTA, phosphatase inhibitor cocktail and protease inhibitor mixture (CompleteMini, Roche Applied Science)]. Immunoblotting was performed as described above using antibodies against ERK1/2, membranes were reprobed with ERK1/2 antibody. β-actin was used as loading control. Signals were visualized by enhanced chemiluminescence (Perkin-Elmer Life Sciences, Boston, MA) and Fusion-SL 4.2 MP (PeqLab, Erlangen, Germany) and quantified with ImageJ software. Data were normalized to β-actin and calculated as a ratio of phosphorylated to non-phosphorylated protein.
Cytoskeleton organization of CHO cells by phalloidin
1.2×104 cells/cm2 cells were seeded in 8-well slides (Zell-Kontakt, Nörten-Hardenberg, Germany) and incubated for 24 h. Cells were starved overnight followed by treatment with 10 ng/ml Fgf2 for 1 h. Cells were washed, fixed with 0.5% PFA in PBS for 20 min, permeabilized then blocked with 3% BSAin PBS for 30 min. Cells were incubated with phospho-paxillin in 1% BSA/PBS overnight. Phosphorylated paxillin and the actin cytoskeleton were co-visualized with 10 mU/µl phalloidin–Alexa-Fluor-488 in 1% BSAin PBS containing Alexa-Fluor-647-conjugated anti-rabbit-IgG and 1 µg/ml DAPI for 1 h at room temperature. Fluorescence was monitored by confocal microscope Zeiss AxioImager M2 with two to five pictures per well (n = 3).
Statistical evaluation was performed with GraphPad Prism4. If not mentioned, we used Student's t-test and considered P<0.05 as significant.
We thank Margret Bahl for her expert technical assistance and Ines Lippmann for cloning Ust-pcDNA3, Lydia Sorokin for the use of the confocal microscope and the IZKF qRT-PCR core unit of the Medical Faculty of University of Münster. We specifically thank Zerina Lokmic and Christian Stock for carefully reading and editing the manuscript.
D.G.S. and K.N. conceived and designed the experiments. K.N. and D.G.S. performed the experiments. K.N., D.G.S. and D.S. analyzed the data.
This work was financially supported by the GRK1549 International Research Training Group ‘Molecular and Cellular GlycoSciences’ to K.N..
The authors declare no competing or financial interests.