Glycobiology Advance Access originally published online on April 6, 2005
Glycobiology 2005 15(8):776-790; doi:10.1093/glycob/cwi060
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Mammalian-like nonsialyl complex-type N-glycosylation of equine gonadotropins in MimicTM insect cells
Unité de Physiologie de la Reproduction et des Comportements, Institut National de la Recherche Agronomique (INRA), Centre National de la Recherche Scientifique (CNRS) et Université François Rabelais de Tours, 37 380 Nouzilly, France
1 To whom correspondence should be addressed; e-mail: cahoreau{at}tours.inra.fr
Received on November 3, 2004; revised on March 29, 2005; accepted on March 29, 2005
 |
Abstract |
|---|
 Top Abstract IntroductionResultsDiscussionMaterials and methodsReferences |
|---|
Recombinant equine luteinizing hormone/chorionic gonadotropin (eLH/CG) was expressed in MimicTM insect cells, that are commercial stably transformed Spodoptera frugiperda (Sf9) cells expressing five mammalian genes encoding glycosyltransferases involved in the synthesis of complex-type monosialylated N-glycans. We previously showed that it exhibited no in vivo bioactivity although expressing full in vitro bioactivity, and it was suspected that this was because of insufficient sialylation of eLH/CG N-glycans. Lectin binding analyses were performed with recombinant dimeric eLH/CG or its alpha subunit, secreted in the serum-containing supernatant of infected Sf9 and MimicTM cells. Two types of specific lectin affinity assays (blot analyses and enzyme-linked immunosorbent assay) were used to compare the ability or inability of natural and recombinant gonadotropins to bind to various lectins. In natural equine chorionic gonadotropin (eCG), complex-type N-glycans terminating with both Sia
2,3Gal (based on Maackia amurensis agglutinin [MAA] binding) and Sia2,6Gal (based on Sambucus nigra agglutinin [SNA] binding) were found, but in the alpha subunit dissociated from natural eCG, we only detected Sia2–6Gal. In eLH/CG and its alpha subunit produced by Sf9 cells, N-glycans were found to be terminated by mannosyl residues (based on Galanthus nivalis agglutinin [GNA] binding), whereas those produced in MimicTM cells were terminated by galactoses (based on binding to Ricinus communis agglutinin I [RCA I] , but not to SNA or MAA). This is in agreement with the fact that the nucleotide donor substrate of sialic acid is not naturally synthesized in insect cells. On the basis of binding to Arachis Hypogaea agglutinin [PNA], O-glycans exhibited the Galß1–3GalNAc structure in recombinant-free alpha and eLH/CG from both Sf9 and MimicTM cell lines. Both N- and O-linked carbohydrate side chains synthesized in MimicTM cells should thus be amenable to further acellular sialylation. Key words: baculovirus / gonadotropin / glycosylation / insect cells / lectin
|
Introduction |
|---|
|
TopAbstractIntroductionResultsDiscussionMaterials and methodsReferences |
|---|
The glycosylation of circulating proteins in mammals plays a pivotal role in their maintenance in blood. In the case of glycoprotein hormones such as equine chorionic gonadotropin (eCG), removal of the terminal sialic acid residues from its carbohydrate chains has been known for a long time to dramatically diminish its in vivo activity (Morell et al., 1971
), because its half-life is reduced from 6 days to 
1–2 hours (Martinuk et al., 1991). Glycoprotein hormones including luteinizing hormone (LH), follicle-stimulating hormone (FSH), chorionic gonadotropin (CG), and thyroid-stimulating hormone (TSH) are composed of a common alpha subunit, that is noncovalently bound to a specific beta subunit (Pierce and Parsons, 1981; Combarnous, 1992). As eCG and eLH alpha and beta subunits are encoded by the same and ß genes expressed in placenta and pituitary respectively, the corresponding recombinant hormone has been denominated by eLH/CG.
Natural pituitary eLH and placental eCG subunits exhibit identical polypeptide structures and bear carbohydrate side chains at identical positions (Bousfield and Butnev, 2001), but these carbohydrate chains exhibit differences in their structures. Equine LH alpha subunit is composed of two N-linked carbohydrate chains terminating with sulphated N-acetylgalactosamines (SO4-4-GalNAc), whereas eCG alpha subunit carries two biantennary N-linked carbohydrate chains terminated with 2,6-linked sialic acids (Damm et al., 1990), possibly extended with lactosamine repeats (Bousfield et al. 2004, Gotschall and Bousfield, 1996). Equine LH and CG ß-subunits are composed of 149 residues (Sugino et al., 1987) with a single N-carbohydrate chain located at Asn 13 terminating with sulfated GalNAc in eLH beta subunit and sialylated 2,3Gal in eCG beta subunit (Smith et al., 1993; Matsui et al., 1994). Both beta subunits possess twelve O-glycans extended with sialylated (2,3) poly(N-acetyllactosamine) and located on serine and threonine of the C-terminal extension called carboxy terminal peptide (CTP; 122–149) (Bousfield and Butnev, 2001, Hokke et al., 1994). As they have identical polypeptide structures, the differences in carbohydrate structures between eLH and eCG are responsible for their differences in biological properties: equine CG exhibit much longer half-life in circulation but weaker affinity towards LH and FSH receptors than pituitary eLH, but the two of them exhibited high in vivo potencies.
In contrast, eLH/CG expressed in Spodoptera frugiperda (Sf9) cells or MimicTM cells were both found to be devoid of in vivo activity in spite of their full in vitro potency (Legardinier et al. 2005). It was thus important to further study the carbohydrate structure of eLH/CG expressed in Sf9 cells and MimicTM cells to design further modifications needed to convey in vivo biopotency to the recombinant hormone.
The baculovirus–insect cell system is very efficient in the production of proteins (Miller, 1988; Luckow and Summers, 1989) and glycoproteins (Altmann et al., 1999; Matsuura et al., 1987; Farrell et al., 1998). Insect cells generally produce glycoproteins with mannose-rich carbohydrate side chains whereas mammalian cells produce complex-type carbohydrates. To cope with the inability of insect cells to produce sialylated glycoproteins (Jarvis, 2003; Marchal, et al., 2001; Tomiya et al., 2003, 2004), baculoviruses or insect cells were genetically engineered to stably express lacking mammalian glycosyltransferases genes (Aumiller et al., 2003; Chang et al., 2003; Hollister, et al., 2003, 2002; Hollister and Jarvis, 2001).
Commercial MimicTM cells, referenced as SfSWT-1 cells (Hollister et al., 2002), are Sf9 cells stably transformed with mammalian genes encoding N-acetylglucosaminyltransferase-I (GlcNAcT-I), N-acetylglucosaminyltransferase-II (GlcNAcT-II), ß1–4 galactosyltransferase (ß4GalT), 2–3 sialyltransferase (ST3), and 2–6 sialyltransferase (ST6) and were reported to produce mammalian-like N-glycans up to the terminal 2–6 linked sialic acid residues in serum-containing medium (Hollister et al., 2003, 2002). Previous studies strongly suggest that sialic acid comes from serum (Hollister et al., 2003) because MimicTM cells in serum-free medium are unable to sialylate glycoproteins (Aumiller et al., 2003). Moreover, MimicTM cells like their parental Sf9 cell line are unable to generate the nucleotide monophospho-sialic acid (CMP-SA) sugar donor (Jarvis, 2003; Lawrence et al., 2001; Tomiya et al., 2003).
These insect cells have thus been used in our laboratory to attempt the production of recombinant eLH/CG with complex mammalian-like carbohydrates (Legardinier et al., 2005). To our knowledge, it was the first report of the production of a glycoprotein in the commercially available MimicTM cell line (Invitrogen, Cergy Pontoise, France). In this article, lectin binding analyses were performed with recombinant dimeric eLH/CG or its alpha subunit, secreted in the serum-containing supernatant of infected Sf9 and MimicTM cells. Two types of specific lectin affinity assays (blot analyses and enzyme-linked immunosorbent assay [ELISA]) were used to compare the ability or inability of natural and recombinant gonadotropins to bind to various lectins. The results are consistent with the conclusion that mammalian-like complex-type N-glycans with terminal galactoses are indeed synthesized on eLH/CG and alpha subunit in MimicTM cells without any detectable terminal sialic acid residue. In addition, short nonsialylated O-glycans with the Galß1–3GalNAc structure were also found in recombinant eLH/CG and free -subunit produced in Sf9 and MimicTM cell lines. Little is known about O-glycosidic glycosylation of proteins in insect cells. The presence of N-acetylgalactosaminyl O-glycans on glycoproteins isolated from insect cells has been reported using GalNAc-specific lectins or radiolabelled monosaccharides (Butters et al., 1981) and also very low amounts of Galß1–3GalNAc structures (Lopez et al., 1999; Thomsen et al., 1990).
The absence of sialic acid residues explains the lack of in vivo bioactivity of the hormone despite its full in vitro activity (Legardinier et al., 2005). Nevertheless, the presence of a terminal Gal residues makes it a potential good substrate for further in vitro sialylation in an acellular system (Legaigneur et al., 2001).
|
Results |
|---|
|
TopAbstractIntroductionResultsDiscussionMaterials and methodsReferences |
|---|
Expression of alpha-His6 and beta-His6 subunits in the supernatant of infected Sf9 and MimicTM cells
To express free alpha-His6 and beta-His6 subunits in the baculovirus insect cell system, we seeded Sf9 or MimicTM cells in 25-cm2 flasks and infected them with recombinant baculoviruses SLP10-
His6 (P10) and SLP10-ßHis6 (P10), respectively (See His6 vector in Figure 1). After five days incubation at 28°C, supernatants containing serum from infected cells were collected and analysed by western blotting using specific antibodies.
|
Natural and recombinant alpha subunits were detected using the anti- monoclonal antibody (mAb) 89A2 under nonreducing conditions. Protein doublets were detected at 22 kDa (Figure 2A, lanes 3–4) and 26 kDa (Figure 2A, lanes 5–6) for recombinant alpha-His6 subunit secreted in the serum-containing supernatant of infected Sf9 and MimicTM cells, respectively. The alpha-His6 subunit from MimicTM cells exhibited the same apparent molecular weight (MW) as that of alpha subunit coming from heat dissociated natural eCG (Figure 2A, lane 1), when the cells were used before 30 passages (Figure 2A, lanes 5–7) and lower apparent MW after 90 passages (Figure 2A, lane 8) suggesting that repeated subculturing of MimicTM cells is not recommended.
|
Natural and recombinant beta subunits were detected using the rabbit ß1–9 Ab under reducing conditions. Recombinant beta-His6 subunits appeared as doublet protein at 26 kDa in the serum-containing supernatant of infected Sf9 cells (Figure 2B, lanes 3–4) and as 26 and 28 kDa proteins for beta-His6 subunit secreted by MimicTM cells, at <30 passages (Figure 2B, lanes 5–6). The beta subunit from heat dissociated natural eCG exhibited an apparent MW between 45 and 66 kDa (Figure 2B, lane 1) resulting of the heterogeneity of its O-glycosylated C-terminal end (Butnev et al., 1996; Legardinier et al., 2005).
We previously showed that untagged alpha and beta subunits exhibited a similar profile in western blot using the same antibodies (Legardinier et al., 2005). The presence of the His6 tag at the C-terminus did not prevent His-tagged subunits from immunological recognition.
Time course of production of eLH/CG produced in Sf9 and MimicTM cells
MimicTM cells and Sf9 cells were coinfected with previously described NPV- (PH) and SLP10-ß (P10) recombinant baculoviruses (Legardinier et al. 2005).
Supernatants were collected daily from day 1 to day 7 post infection (pi) to determine by sandwich ELISA (four experiments) the production of secreted eLH/CG (Figure 3). The production of eLH/CG in Sf9 cells was almost maximal at day 2 postinfection (pi) with a production of 4.15 (0.49 µg.ml–1 reaching a steady-state which lasted until day 7. Time course of production of eLH/CG expressed in MimicTM cells was different with a progressive increase from day 1 to day 5 pi until a maximal production of 5.33 ± 1.32 µg.ml–1 at day 5 pi. As previously shown for glutathione s-transferase (GST) ManI (Aumiller et al., 2003), the expression of eLH/CG in MimicTM cells was found to be delayed compared to Sf9 cells.
|
Nickel affinity purification of recombinant alpha-His6 and beta-His6 subunits
About 100 ml of culture media from infected Sf9 or MimicTM cells containing 0.5 mg of recombinant alpha-His6 or beta-His6 subunits were loaded onto nickel-nitrilotriacetic acid (Ni-NTA) agarose and eluted with 500 mM imidazole. All eluted fractions were run on 14% SDS–PAGE and either silver-stained (Figure 4, left panels) or immunoblotted after transfer onto a nitrocellulose membrane (Figure 4, right panels).
|
The 22 kDa (Figure 4A) and 26 kDa (Figure 4C) proteins correspond to alpha-His6 proteins from Sf9 and MimicTM cells as shown by western blot (Figure 4B and D) using mAb 89A2 under nonreducing conditions.
The presence of a broad range of proteins between 20 and 50 kDa in E1 and E2 eluted fractions (Figure 4E) and in E1 to E4 (Figure 4G) corresponds to different glycoforms of beta-His6 proteins from Sf9 and MimicTM cells infected with SLP10-ßHis6 (P10) as first controlled by western blotting using heterodimeric eCG Ab under nonreducing conditions (Cahoreau and Combarnous, 1987) (data not shown). To confirm the presence of purified beta-His6 subunit in eluted fractions from Ni-NTA purification, we used specific rabbit ß1–9 Ab (Figure 4F and H) under reducing conditions.
Silver staining gels (left panels) show the presence of high MW (66–100 kDa) proteins not recognized by antibodies in all fractions of purified subunits (Figure 4, left panels, lanes E1–E4).
N-terminal sequences of purified alpha-His6 and beta-His6 subunits
The FPDGEFTTQD sequence was found for the 22 kDa positive band of purified alpha-His6 subunit expressed in Sf9 cells, and the SRGPLR was found for the 26 kDa positive band of purified LH/CG beta-His6 subunit expressed in Sf9 cells. These sequences correspond to the full cDNA sequences of equine alpha and LH or CG beta subunits respectively.
Lectin enzyme-linked immunosorbent assay
The carbohydrate binding specificities of the different lectins used are indicated in Table I. Recombinant heterodimeric hormones (eLH/CG±CTP) and free alpha subunits (His6-tagged or not) were produced in Sf9 and MimicTM cells as previously described (Legardinier et al., 2005) and were compared in the same assays to their natural counterparts eCG and alpha subunit for their ability to bind to different plant lectins (Table II, Figure 5). Microtitration plates were coated with the highly specific mAb directed against eCG alpha subunit ( mAb 89A2) to catch natural and recombinant dimeric hormones and alpha subunits secreted in the serum-containing supernatant from infected Sf9 or MimicTM cells. Tris buffer saline-Tween (TBS-T) and the supernatant from Sf9 cells infected with wild-type baculovirus Autographa californica multiple nuclear polyhedrosis virus(AcMNPV) were used as negative controls.
|
|
|
The location and the structures of carbohydrate side chains of natural eCG, its subunits, and recombinant counterparts produced in Sf9 and MimicTM cells are detailed in Figure 6.
|
Dose-dependent response.
A significant increase of the absorbance at 450 nm reveals that the tested lectin binds to specific saccharidic residues of carbohydrates of the glycoprotein caught by antibody or Ni-NTA. This recognition was dose-dependent as shown in Figure 7 with the lectin Galanthus nivalis agglutinin (GNA) that specifically bound to ending mannose residues of alpha-His6 subunit produced in Sf9 cells and not to alpha-His6 produced in MimicTM cells.
|
Binding to mannoses.
As shown in Table II and Figure 5, recombinant-free alpha subunits (alpha and alpha-His6) and dimeric hormones (eLH/CG±CTP) produced in Sf9 cells specifically interacted with GNA that is known to bind to terminal mannoses of N-glycans (Shibuya et al., 1988). On the contrary, the recombinant counterparts produced in MimicTM cells and natural eCG and alpha subunit did not bind to GNA, suggesting a complex-type N-glycosylation.
Binding to galactoses.
The highest absorbance variation was observed with the galactose-binding lectin ricinus communis agglutinin I (RCA I) (Green et al., 1987). Recombinant-free alpha (alpha and alpha-His6) and dimeric hormones (eLH/CG±CTP) produced in MimicTM cells are very strongly recognized by RCA I, contrary to the counterparts expressed in Sf9 cells, whose absorbance was more than five-fold less (close to the base line). Although natural complete eCG was not reactive with RCA I, its dissociated alpha subunit bound very weakly.
Binding to sialic acids.
The combination of the two lectins Maackia amurensis agglutinin (MAA; Wang and Cummings, 1988) and Sambucus nigra agglutinin (SNA; Shibuya et al., 1987) enables to distinguish the type of sialic acid linkage on sialylated N- and/or O-linked carbohydrate side chains. Whatever the lectin used, MAA or SNA, recombinant glycoproteins produced in Sf9 and MimicTM cells gave no significant increase of the absorbance, indicating an absence of interaction. On the contrary, natural eCG reacted with both MAA and SNA, whereas its dissociated alpha subunit only bound to MAA.
Binding to O-glycans.
Arachis Hypogaea agglutinin (PNA) binds the core Galß1–3GalNAc disaccharide of O-glycans (Goldstein and Hayes, 1978) only in the absence of terminal sialic acids as shown in Figure 6. Natural eCG and alpha subunit did not bind to PNA, contrary to all tested recombinant proteins, namely the dimeric hormones (eLH/CG±CTP) as well as free alpha subunits (alpha and alpha-His6). The binding values for eLH/CG without CTP were significantly lower.
Presence of N-carbohydrate chains.
Figure 8 (panel A) shows that natural and recombinant purified alpha-His6 subunits (lanes 1–3) exhibit lower MW after PNGase F treatment. Under reducing conditions, alpha-His6 subunit from Sf9 cells (lane 2) appeared at 18 kDa, whereas natural eCG alpha subunit (lane 1) and recombinant alpha-His6 subunit from MimicTM (lane 3) cells were higher, at 22 kDa.
|
After PNGase F treatment, recombinant alpha-His6 subunits from Sf9 and MimicTM cells appeared as doublets with bands at 15 and 16 kDa (lanes 2 and 3), whereas deglycosylated natural eCG alpha subunit was revealed at 14 kDa (lane 1). Thus, PNGase F induced a drop in the apparent MW comprised between 2 and 8 kDa in agreement with the elimination of N-carbohydrate chains. We can notice that a trace amount of protein was visualized at 17 kDa for deglycosylated natural eCG alpha subunit (lane 1), that could be attributed to a partially deglycosylated form. We used exactly the same samples in silver staining and in western blot analysis.
Lectin western blot
In lectin blot analyses, natural alpha subunit purified from dissociated eCG and recombinant Ni-NTA purified alpha-His6 subunits from serum-containing supernatant of infected Sf9 and MimicTM cells were tested for their ability to bind to seven different lectins (Figure 8, panels B–H), whose specificity was confirmed by using two purified commercial glycoproteins (negative and positive controls). After SDS–PAGE under reducing conditions and electrotransfer onto nitrocellulose membrane, complete glycoproteins or N-deglycosylated ones (pretreated with PNGase F) were revealed after incubation with one specific plant lectin (Figure 8). Lectin binding specificities are indicated in Table I and detailed in Figure 6.
Detection of mannose residues.
Con A, that binds to -linked mannoses even if the N-glycans are of complex type (Baenziger and Fiete, 1979), did not recognize natural eCG alpha subunit (lane 1). By contrast, alpha-His6 Sf9 (lane 2), alpha-His6 Mimic (lane 3) and carboxypeptidase Y (CPY) were detected and only before PNGase F treatment.
GNA binding was positive only for recombinant alpha-His6 subunit produced in Sf9 cells, an unique spot being visible at 18 kDa (lane 2). The recognition of CPY (63 kDa) confirms the high specificity of this lectin, which recognizes "terminal mannose linked to mannose" structures (Table I). In recombinant Sf9 alpha subunit and in CPY, this structure was clearly associated with N-glycans because there was no glycoprotein recognition after PNGase F treatment.
Detection of galactose residues.
RCA I recognized natural eCG alpha subunit as a large spot comprised between 20 and 30 kDa (lane 1). A spot near 43 kDa was also detected (lane 1) and could not be attributed to an alpha dimer because the electrophoresis was run under reducing conditions. Recombinant alpha-His6 subunit produced in Sf9 cells showed no signal near 18 kDa (lane 2). By contrast, the one produced in MimicTM cells (lane 3) strongly bound to RCA I. The positive control asialofetuin (aSF) (three glycoforms) responded to the lectin and still gave signals after enzymatic hydrolysis but at lower MW. After PNGase F treatment, a residual signal appeared at a MW of 17 and 20 kDa for natural and recombinant MimicTM alpha subunits respectively.
Datura stramonium agglutinin (DSA) that recognizes the ramified structure Galß1–4 (GlcNAc)2 (Crowley et al., 1984), strongly reacted with natural eCG alpha subunit but only before PNGase F treatment and did not bind to recombinant alpha-His6 subunits.
Glycoproteins (50–100 kDa), coeluted with alpha-His subunits during Ni-NTA purification, reacted with both lectins RCA I and DSA but were not recognized by the specific mAb 89A2 (Figure 4B and D). These signals were more intensive for purified alpha-His6 subunit from MimicTM cells, cultured with two-fold more serum than for Sf9 cells.
Detection of sialic acid residues.
SNA bound to natural eCG alpha subunit because the signal near 24 kDa was very strong although the dose of gonadotropin loaded on the membrane was the lowest (125 ng) tested. However, another signal was seen for natural eCG alpha at 43 kDa, as previously shown for RCA I (lane 1). For recombinant alpha-His6 subunit produced in Sf9 and MimicTM cells, no signal was detected at expected MW (18–22 kDa), but several spots always appeared at high MW (>45 kDa; lanes 2–3). Several bands of the positive control fetuin (three glycoforms) were detected. After PNGase F treatment, signals are largely attenuated, although not eliminated because one band was still visible for this control. Enzymatic hydrolysis of natural eCG alpha subunit showed one band near 17 kDa (see silver staining, panel A), a MW which is higher than that of completely deglycosylated natural alpha subunit (14 kDa).
MAA gave slight band near 22 kDa for natural eCG alpha (lane 1), which disappeared after PNGase F treatment, but no band appeared for recombinant ones (lanes 2 and 3). As observed with SNA, only signals at high MW were observed (lanes 2 and 3). Fetuin possesses Sia2,3Gal recognized by the lectin.
We oberved the same high MW glycoproteins as with galactose-binding lectins.
Detection of O-glycans.
PNA recognized the two recombinant alpha-His6 subunits before PNGase F treatment. After enzymatic hydrolysis, signals are seen for the two recombinant tagged subunits at 15–16 kDa, which is the MW of their N-deglycosylated forms (as previously shown, panel A).
|
Discussion |
|---|
|
TopAbstractIntroductionResultsDiscussionMaterials and methodsReferences |
|---|
Recombinant eLH/CG and its alpha and beta subunits have been successfully expressed in MimicTM cells (Legardinier et al. 2005
), and it was the first time that the use of this commercial source of cells was reported following manufacturer’s recommendations (Invitrogen). MimicTM cells are Sf9 cells stably transformed with five mammalian genes encoding GlcNAcT-I, GlcNAcT-II, ß4GalT, ST3 and ST6 and reported to produce mammalian-like N-linked carbohydrates up to terminal 2,6-linked sialic acid residues when cultured in the presence of fetal bovine serum (Hollister et al. 2002). Although major insect cells can not synthesize sialic acids and CMP-sialic acid (Jarvis, 2003; Lawrence et al., 2001; Tomiya et al. 2003), MimicTM cells are claimed to salvage sialic acids from extracellular sialoglycoproteins found in serum to produce their own CMP-sialic acid (Hollister et al., 2003). Our data demonstrate that mammalian complex-type N-glycans are indeed conjugated to the equine gonadotropin eLH/CG, and its subunits expressed in MimicTM cells, but terminal sialic acid residues were demonstrated to be missing. We reached this conclusion on the basis of lectin binding after separation either by capture (with a specific antibody or Ni-NTA coated plates) or by chromatography of alpha-His6 subunits from major contaminating materials (by Ni-NTA purification followed by lectin western blot). Lectin western blot analyses were particularly powerful because it was possible to detect lectin-positive bands that were not silver-stained, such as the spot near 43 kDa for natural eCG alpha subunit. Indeed, the lectins RCA I and SNA exhibited a very high affinity since trace amount of glycoproteins were detected.
In this study, glycoproteins with high MW (50–100 kDa), which were coeluted with alpha-His subunits during nickel affinity purification step, reacted with lectins specific of galactose (RCA I) and sialic acid (MAA and SNA) residues but were not recognized by the highly specific monoclonal anti-alpha antibody. Thus, these high MW signals could not be attributed to homodimers or aggregates of gonadotropins subunits. Moreover, these signals seem to be proportional to the initial quantity of serum in the supernatants of infected cells, because high MW glycoproteins gave more intensive signals for MimicTM cells cultured with two-fold more serum (Figure 8, lane 3) than that for Sf9 cells (Figure 8, Lane 2).
Reaction of natural eCG and its free alpha subunit with SNA, RCA I, MAA and DSA
In lectin ELISA and lectin western blot, natural eCG alpha subunit was found to react with SNA and RCA I, confirming the presence of terminal sialic acid (sia2–6Gal) and galactose (Galß1–4GlcNAc) residues on its N-glycans. The presence of sia2–6Gal did not totally abolish RCA I interaction with Galß1–4GlcNAc (Green et al., 1987); consequently, it is not possible to appreciate what proportion of the chains, if any, do not bear terminal sialic acid.
Natural alpha subunit from heat dissociated eCG did not react with MAA in lectin western blot and lectin ELISA, implicating the absence of sia2–3Gal termini. A spot at 43 kDa was observed with RCA I and SNA in western blot, but was not detectable by silver staining (Figure 8, panel A). It can be attributed to a very low residual contamination by eCG beta subunit not detected by silver staining. After PNGase F treatment, the two lectins RCA I and SNA revealed a spot at 17–18 kDa, a MW higher than that of the completely deglycosylated alpha subunit (14 kDa). This spot was also detected by silver staining and can be attributed to a small amount of alpha subunit still bearing one of its two N-saccharide chains. The important point is that no signal is detected at 14 kDa for completely deglycosylated natural eCG alpha subunit, thus confirming the presence of N-glycannic chains with sia2–6Gal termini and an undetermined proportion of non-sialylated Galß1–4GlcNAc termini. The strong positive reaction of natural eCG alpha subunit with DSA in lectin western blot suggests the presence of ramified Galß1–4(GlcNAc)2 structure (Table I) in addition to the lactosamine (Galß1–4GlcNAc) structure recognized by RCA I.
Natural eCG binds to both SNA and MAA, whereas natural alpha subunit binds only to SNA. Therefore, sia2–3 Gal endings must come from the beta subunit. Considering the mass proportion of O-glycans versus N-glycans, it can be supposed that sia2–3Gal are mainly borne by the twelve sialylated O-glycans of the beta subunit CTP (Hokke, et al., 1994). Using 2,3 specific neuraminidases, it was published that N-glycans of total eCG were predominantly of the 2,3 type (Smith et al., 1993), whereas previous data showed that N-glycans of eCG beta were predominantly of 2,6 type and O-glycans of 2,3 type (Damm et al. 1990).
Positive reaction of alpha-His6 and eLH/CG Sf9 with GNA and Con A
Recombinant alpha-His6 subunit and eLH/CG with or without CTP produced in Sf9 cells showed a strong affinity for GNA, confirming the presence of tri- or high-mannose structures on its N-glycans. These structures explain the high affinity of alpha-His6 Sf9 for Con A. It is known that the interaction with Con A (Baenziger and Fiete, 1979) is weaker when the chain is enriched by N-acetylglucosamine and galactose residues (Debray et al., 1981) explaining the very weak reaction of natural eCG alpha subunit. The absence of reaction with SNA, MAA, DSA, and RCA I is in keeping with the absence of terminal galactose or sialic acids in glycoproteins synthesized in Sf9 cells (Altmann et al., 1999).
Positive reaction of recombinant eLH/CG and alpha-His6 and beta-His6 subunits produced by MimicTM cells with RCA I but not with GNA, SNA, MAA and DSA
Recombinant eLH/CG with or without CTP and alpha-His6 subunit produced in MimicTM cells do not possess terminal sialic acids as shown by the lack of SNA and MAA binding in lectin ELISA and/or lectin western blot. Alpha-His6 subunit produced in MimicTM cells was poorly recognized with Con A demonstrating the absence of mannose-rich N-glycans, and it was well recognized by RCA-I showing the presence of terminal Galß1–4GlcNAc. Therefore, in serum-containing medium, adherent MimicTM cells elaborate complex-type unsialylated N-glycans up to galactose residues, and probably on both mannoses 1,3 and 1,6 of the common pentasaccharidic (Man)3(GlcNAc)2-Asn motif (negative reaction with GNA) suggesting that GlcNAcT-I, GlcNAcT-II and ß4galT are active. However, alpha-His6 subunit from MimicTM cells exhibited no DSA binding, proving that it does not bear ramified Galß1–4(GlcNAc)2 structure in contrast to natural eCG alpha subunit.
Mimic recombinant dimeric eLH/CG and its subunits were not sialylated, either because ST6 was not active (as it was the case of ST3) (Hollister et al. 2002) or because it was not expressed, or because MimicTM cells in the conditions used can not salvage sialic acids from extracellular sialoglycoconjugates, such as fetuin (Hollister et al. 2003, 2002, Jarvis, 2003). Indeed, parental Sf9 cells can not synthesize their own source of sialic acids: they are deficient in sialic acid synthase (SAS) and cytidine monophosphate sialic acid synthase activities (CMP-SAS) involved in the biosynthesis of sialic acid and CMP-sialic acid respectively. To overcome this limitation, genes encoding SAS and/or CMP-SAS were expressed in insect cells (Lawrence et al. 2001; Tomiya et al. 2004; Viswanathan et al. 2003). Moreover, there is not direct evidence that Sf9 cells can transport CMP-sialic acid from cytosol to the Golgi apparatus (Aumiller, 2003).
Negative reaction of eCG and its dissociated alpha subunit with PNA
Lectin ELISA and lectin western blot showed no binding of natural eCG and its dissociated alpha subunit with PNA. Indeed, natural dimeric eCG possess O-glycans only on the C-terminal beta subunit, and these O-glycans are sialylated poly-N-acetyllactosamine structures (Hokke et al., 1994), not recognized with PNA. Equine alpha subunit, as numerous glycoprotein hormones alpha subunits, possesses a potential O-glycosylation site at Threonine 43, but this site is not occupied when alpha is associated with beta subunit. Indeed, free alpha subunit, not combined with beta subunit in extracts of bovine pituitaries, was shown to be O-glycosylated (Parsons and Pierce, 1984), preventing from association with beta subunit. This is in agreement with the fact that the alpha subunit from dissociated natural eCG is not O-glycosylated.
Positive reaction of recombinant eLH/CG and alpha-His6 subunit from Sf9 and MimicTM cells with PNA
PNA binding to recombinant alpha-His6 subunits shows the presence of Galß1–3GalNAc O-glycans in the free secreted Sf9 and Mimic alpha and alpha-His6 subunits. This is evidenced in both lectin western blot and lectin ELISA. The presence of this additional O-glycan and of a polyhistidin tag could explain the comigration of Mimic alpha-His6 subunit with natural sialylated eCG alpha subunit. The presence of O-linked carbohydrate had been previously observed in rat glycoprotein hormone alpha subunit expressed in Sf9 cells (Delahaye et al., 1996). Previous studies showed that O-glycans with the unique GalNAc residue were generated in insect cells like Sf9 cells, on natural glycoproteins secreted from noninfected cells (Lopez et al., 1999) or recombinant glycoproteins (Thomsen et al., 1990). However, in these studies, the synthesis of Galß1–3GalNAc structures at very low amounts in Sf9 cells has also been described.
Our results with PNA suggest a significant activity of the Core1ß3GalT (EC 2.4.1.122
[EC]
) (Ju et al., 2002) adding galactose residues on proximal GalNAc-Ser/Thr in both Sf9 and MimicTM cell lines. Thus, the difference of MW between recombinant N-deglycosylated alpha-His6 subunits (15-16kDa) and natural N-deglycosylated eCG alpha subunit (14kDa) can be ascribed not only to the His6 tag plus the Factor Xa proteolytic site (HHHHHHIEGR: MW 1296) but also to the presence of an additional Galß1–3 GalNAc O-glycan chain (GalNAc, Gal: MW 365).
Recombinant eLH/CG strongly reacted with PNA in lectin ELISA supporting the presence of Galß1–3GalNAc O-glycans on the CTP of the beta subunit. Nevertheless, recombinant eLH/CG without its O-glycosylated CTP produced in the serum-containing supernatant of Sf9 and MimicTM cells still significantly reacted with PNA. This must be due to the recognition of the O-glycans of free alpha subunit produced during coinfection of insect cells with alpha and truncated beta baculoviruses, or it could be due to one remaining O-glycosylation site located at Ser118 upstream the (122–149) CTP region (Bousfield and Butnev, 2001; Bousfield et al., 1996). The absence of PNA reaction with natural eCG suggest either the absence or shielding of the Galß1–3GalNAc structures.
In conclusion, N-glycans in eLH/CG and its subunits produced by Sf9 cells were found to be terminated by mannosyl residues suggesting the presence of poly or pauci-mannose structures. By contrast, N-glycans were of complex type in eLH/CG produced in MimicTM cells; nevertheless, they were terminated by galactose residues and not sialic acids as expected. Interestingly, O-glycans exhibited the Galß1–3GalNAc structure in recombinant-free alpha and eLH/CG from both Sf9 and MimicTM cell lines. This comparative study led us to conclude that natural eCG possess carbohydrate side chains ended with sialic acids exhibiting the two types of linkages (2,3 or 2,6) and that N-glycans of alpha subunit from dissociated eCG are only composed of Sia2,6Gal.
Recently, a new transgenic insect cell line (SfSWT-3) was described (Aumiller et al., 2003), that corresponds to the transformation of MimicTM cells (SfSWT-1) with two additional mammalian genes encoding SAS and CMP-SAS. It has been claimed to correctly sialylate a recombinant glycoprotein on its two antennary N-carbohydrate chains in the absence of fetal bovine serum. It could thus be of utmost interest for the production of recombinant gonadotropins with in vivo activity.
|
Materials and methods |
|---|
|
TopAbstractIntroductionResultsDiscussionMaterials and methodsReferences |
|---|
Introduction of a cassette encoding a polyhistidine sequence in the baculovirus transfer vector p119
The p119 baculovirus transfer vector (Chaabihi et al., 1993
) contains the very late promoter and polyadenylation signal of the protein P10, that are flanked with specific sequences of P10 gene allowing homologous recombination with wild-type AcMNPV DNA. We introduced in HindIII restriction site a new fragment encoding the factor Xa proteolytic site IEGR followed with six histidine residues (His6) and TAA stop codon (Figure 1). Briefly, the p119 baculovirus transfer vector was digested using HindIII restriction enzyme (Roche, Meylan, France) and ligated with the purified prehybridized A,B,C,D oligonucleotides (Eurogentec Seraing, Belgium) (Figure 1). The A and B complementary oligonucleotides were 5'-AGC TTA TCG AAG GCA GGC AC-3' and 3'-ATA GCT TCC GTC CGT GGT AG-5', respectively and contain a sequence encoding the factor Xa proteolytic site IEGR. The C and D complementary oligonucleotides were 5'-CAT CAT CAC CAC CAT TAA C-3' and 3'-TAG TGG TGG TAA TTG TCG A-5', respectively and contain a sequence encoding six consecutive histidine residues (His6). About 1µg of each oligonucleotide was separately heated for 10 min at 80°C. A,B,C,D oligonucleotides were equally mixed and incubated for 10 min at 80°C just before hybridization that took place overnight during progressive decrease of temperature down to room temperature. About 50 ng of digested modified p119 transfer vector was ligated with 500 ng of prehybridized A,B,C,D oligonucleotides overnight at 4°C using 5 units of T4 DNA ligase (QBIOgene, Illkirch, France).
Cloning of equine alpha and LH/CG beta subunits cDNAs into the modified p119 transfer vector
A chimeric cDNA encoding equine gonadotropin alpha subunit with its natural equine amino acid sequence was previously cloned into p119 and pGmAc116T baculovirus transfer vectors (Legardinier et al., 2005). Chimeric cDNA encoding the equine alpha subunit was used as a template for PCR (polymerase chain reaction) using 5' (CTT AAC ACT AAG ATC TTC TAG ACT G) and 3' (GGT GTT CTA ATT CGA AAA AGT GGT TC) primers to generate modified cDNA ends with BglII site in 5' and HindIII in 3' instead of stop codon. The amplified fragment was subcloned into pGEM-T (Promega, Charbonnières, France) and directly cloned into the modified p119 transfer vector using BglII and HindIII to obtain the specific transfer vector p119-His6.
cDNAs encoding complete equine LH/CG beta subunit (Chopineau et al., 1995) or truncated LH/CG beta one missing the 122–149 CTP were previously cloned into p119 transfer vector downstream the p10 protein promoter (P10) using BglII and HindIII restriction sites (Legardinier et al. 2005) leading to the transfer vector p119-ß and p119-ß
CTP.
PCR procedures using 5' (CTA AGA TCT AGA ACC AAG GAT GGA G) and 3' (CTG AAG ATT CGA ATG TTT CGA AAA C) primers were used to amplify the complete LH/CG beta cDNA introducing HindIII restriction site instead of stop codon. The amplified product was subcloned into pGEM-T and directly cloned into the modified p119 transfer vector using BglII and HindIII restriction sites to obtain the specific transfer vectors p119-ßHis6.
Plasmid constructs were amplified and purified using Wizard Plus SV Minipreps DNA purification system kit (Promega, Charbonnières, France) and Nucleobonds AX kit (Macherey Nagel, Hoerdt, France). PCR were carried out using Taq polymerase (Promega, Charbonnières, France), and the sequences of all the amplified products were checked by the dideoxy chain termination method (Genome Express, Meylan, France).
Cells and viral infection
The Sf9 subclone of Spodoptera frugiperda Sf21 cells (Vaughn et al., 1977) was maintained at 28°C in supplemented TC-100 growth medium containing 5% heat-inactivated fetal bovine serum and 50 units ml–1 penicillin G and 50 µg ml–1 streptomycin. MimicTM cells (Invitrogen), that are stably transformed Sf9 cells previously denominated SfSWT-1 (Hollister et al., 2002), were cultured in complete Grace’s Insect Medium supplemented with 10% fetal bovine serum and antibiotics as recommended by the manufacturer (Invitrogen). Both insect cell lines were maintained as adherent cells at a density of 2.0–2.5 x 106 cells per ml in 25-cm2 flasks and passed over 80 percent confluency three times a week. To reach the same cell density as that obtained with Sf9 cells, more MimicTM cells had to be passed because of their slower growth and this never more than 30 times as recommended by the manufacturer.
The viruses were propagated in both cell lines and recovered as described previously (Summers and Smith, 1987). Cells were infected with recombinant baculoviruses at a multiplicity of infection (MOI) ranging from 5 to 10 plaque forming unit (Pfu)/cell in 25- or 75-cm2 flasks. After 45-min incubation with viral suspensions, 4 or 12 ml of fresh culture medium was added, and cells were incubated at 28°C until day 5 after infection. The viral titers were determined by plaque assay (Summers and Smith, 1987).
Time course of production of recombinant dimeric eLH/CG expressed in insect cells
Sf9 and MimicTM insect cell lines were coinfected with recombinant baculoviruses NPV-(PH) and SLP10-ß (P10) (Legardinier et al., 2005) at a MOI of 5–10 Pfu per cell as described above. Supernatants were recovered by centrifugation at 100 x g for 5 min and stored at 4°C as indicated times (Figure 3). Samples from four experiments were assayed in duplicates in sandwich ELISA, and quantitations were made using eCG NZY-01 (Lecompte et al., 1998) as standard preparation.
Cotransfection and purification of the recombinant baculoviruses
Sf9 cells were cotransfected with 5 µg of transfer vector DNA and 500 ng of purified AcSLP10 viral DNA (Chaabihi et al., 1993) using a lipofection method (DOTAP, Roche, Meylan, France). This viral DNA possessing only very late P10 promoter driving the expression of the polyhedrin gene was transfected separately with the specific transfer vectors such as p119-His6 and p119-ßHis6. In each case, recombinant baculoviruses were selected by plaque assay and distinguished from the wild-type progeny by their occlusion body-negative phenotype (Summers and Smith, 1987). The screening and purification of the recombinant baculoviruses SLP10-His6 (P10) and SLP10-ßHis6 (P10) were carried out as previously described for recombinant baculoviruses NPV- (PH), SLP10-ß (P10) and SLP10-ßCTP (P10) (Legardinier et al., 2005).
SDS–PAGE and western blot analysis
Sf9 or transgenic MimicTM insect cells were seeded in 25-cm2 flasks and infected separately with recombinant baculoviruses SLP10-His6 (P10) or SLP10-ßHis6 (P10) to obtain recombinant alpha-His6 or beta-His6 subunits. LH/CG heterodimers were obtained by coinfection with recombinant baculoviruses NPV- (PH) and SLP10-ß (P10) or SLP10-ßCTP (P10). Cells infected with wild-type baculoviruses were used as control. Supernatants (medium fraction) were recovered by centrifugation at 100 x g for 5 min, and aliquots were diluted in Laemmli’s 4x buffer (Laemmli, 1970) under nonreducing or reducing conditions. Samples were unheated or heated for 10 min at 100°C and electrophoresed in 14 or 15% SDS–PAGE. Gels were either silver-stained or transferred overnight at 4°C onto nitrocellulose membranes (Schleicher and Shuell, Ecquevilly, France) to probe the different glycoproteins with specific anti-alpha (1µg ml–1), anti-beta (1:5000), or anti-eCG (1:50000) antibodies, then with 1:5000 peroxidase conjugated goat antirabbit or antimouse IgG antibodies. Membranes were visualized using the ECLTM Western Blotting Detection Reagents (Amersham Biosciences Orsay, France).
Antibodies
The antibodies used in this study were as follows: 89A2 mouse monoclonal anti-eCG alpha antibody ( mAb 89A2) recognizing the nondenaturated alpha chain as single subunit or as heterodimer in eCG (Chopineau et al., 1993), a rabbit polyclonal antibody raised against a synthetic LH/CG ß1–9 peptide (ß1–9 Ab) representing the nine N-terminal residues of LH and CG ß-subunits from numerous species and detecting reduced subunits (Legardinier et al., 2005), a rabbit anti-eCG polyclonal antibody (eCG Ab) (Cahoreau and Combarnous, 1987), peroxidase-conjugated goat anti-rabbit or anti-mouse IgG (Jackson ImmunoResearch, Interchim, Montluçon, France) and anti-digoxigenin Fab fragments (Roche, Meylan, France).
Immunological quantitation of recombinant monomers and heterodimers
The concentrations of recombinant alpha-His6 and beta-His6 subunits produced in insect cell culture media were determined in specific competitive ELISA. Briefly, the microtiter plates were coated with 0.1 IU (60 ng) eCG NZY-01 in 100 µl of 0.1 M sodium carbonate/bicarbonate buffer, pH 9.6. After extensive washing in a 0.1% Tween 20-phosphate buffer saline (PBS-T) and saturation with 0.2 % bovine serum albumin (BSA, Sigma, Saint-Quentin Fallavier, France), 50 µl per well of standard eCG NZY-01 or media containing recombinant alpha-His6 or beta-His6 subunit were incubated at different concentrations in the saturation buffer for 4 h at 4°C either with 50 µl per well of mab 89A2 (1µg ml–1) or with ß1–9 Ab (1:5000). After washing, 100 µl of peroxidase-conjugated goat antimouse or antirabbit IgG were added at a 1:10,000 dilution in the saturation buffer for 1 h at 4°C. After rinsing, 100 µl per well of the SureBlueTM TMB peroxidase substrate (KPL, Eurobio, Les Ulis, France) were incubated for 20 min at room temperature in the dark. The reaction was stopped with 50 µl 2N H2SO4, and the absorbance was measured at 450 nm using a SpectraCountTM (Packard; Downers Grove, IL, USA) spectrophotometer, and data were analyzed by the I-SMART software (Packard; Downers Grove, IL, USA).
Natural eCG and recombinant eLH/CG with or without CTP were estimated in a sandwich ELISA, as previously described (Legardinier et al., 2005).
Determination of the N-terminal sequences of purified alpha-His6 and beta-His6 subunits
After separation by 15% SDS–PAGE gels, recombinant alpha-His6 and beta-His6 proteins purified from media of infected Sf9 cells on a Ni-NTA column were electroblotted onto PVDF membranes (ImmobilonP, Millipore, St Quentin en Yvelines, France). After staining with Ponceau red (alpha-His6) or Coomassie blue (beta-His6), the 22 and 26 kDa positive bands corresponding to alpha-His6 and beta-His6 proteins respectively were cut out from the polyvinylidene fluoride (PVDF) membrane. N-terminal sequences were determined with an Edman automated sequencer (Beckman/Porton LF 3000).
Nickel affinity purification of recombinant alpha-His6 and beta-His6 proteins under nondenaturing conditions
Sf9 and MimicTM cells were seeded in 75-cm2 flasks and infected with either recombinant baculoviruses SLP10--His6 or SLP10-ßHis6 at a MOI of 5–10 Pfu per cell in 12 ml final volume. Five days after infection, 80–100 ml supernatants were recovered by centrifugation at 100 x g and then at 1000 x g for 10 min, desalted using five PD10 gel filtration columns used for several runs (Amersham, Orsay, France), exchanged with a binding buffer (25 mM Tris–HCl, 500 mM NaCl, 5mM Imidazole, and pH 8.0), and loaded on 2.5 ml of Ni-NTA agarose (QIAGEN, Courtaboeuf, France) column preequilibrated in the same buffer. After extensive washing in binding buffer, nonspecific proteins were removed from column with washing buffer (25 mM Tris–HCl, 500 mM NaCl, 25 mM Imidazole, and pH 8.0). Bound proteins were eluted by increasing to 500 mM Imidazole in the same buffer at a rate of 50 ml h–1. One-milliliter fractions were collected and stored at 4°C until use. Those containing recombinant alpha-His6 or beta-His6 subunit were identified by immunoassays (ELISA and western blot) and their carbohydrates characterized by lectins.
The sensitivity of gonadotropins and their subunits to proteolysis is known to be very low. Thus, protease inhibitors are not required.
Glycosylation analyses
In vitro endoglycosidase treatment.
Elution fractions containing recombinant Sf9 and Mimic alpha-His6 subunits from Ni-NTA chromatography column were treated separately. About 5 µg of these eluted alpha-His6 subunits (as determined by ELISA) were denaturated in a final volume of 200 µl containing 0.1% w/v SDS and 10 mM ß-mercaptoethanol and boiled for 10 minutes. Enzymatic hydrolysis was carried out with 5 IU PNGase F (N-glycanase F, Roche, Meylan, France) after addition of 10 mM EDTA and 0.7% w/v NP-40 (Sigma, St Quentin Fallavier, France). Samples were incubated for 18 h at 35°C.
A sample of each glycoprotein was treated in the same way without enzyme. For lectin blotting, control positive or negative glycoproteins (see Lectin analysis section) were submitted to the same treatment. An aliquot of purified natural eCG alpha, after dissociation of a natural heterodimeric eCG (Herve et al., 2004) was treated in the same way and served as control in glycosylation analysis. We used exactly the same samples in silver staining and in western blot analysis.
Lectin analysis of glycosylation.
A panel of biotinylated (Vector Laboratories, AbCys, Paris, France) and digoxigenin labelled plant lectins (DIG glycan differentiation kit, Roche, Meylan, France) with various specificities were used in two types of experiments:
First to analyse nondenaturated glycoproteins in culture medium, we used an enzyme-linked lectin immunoassay. The microtitration plates were coated overnight at 4°C with 100 µl per well of mAb 89A2 (1 µg ml–1) in 0.1 M sodium carbonate/bicarbonate buffer, pH 9.6. After extensive washing, nonspecific sites were saturated for 1 h at 4°C in TBS-T (0.05% Tween20 TBS, 25 mM Tris, 140 mM NaCl, 3 mM KCl, and pH 7.4) containing 2% w/v polyvinylpyrrolidone K30 (Fluka, Sigma, St Quentin Fallavier, France). About 100–200 ng recombinant alpha or dimeric eLH/CG (100 µl supernatant per well) were incubated for 1 h at 4°C in the saturation buffer. Similar quantities of natural standard eCG NZY-01 and purified natural eCG alpha were used.
After washing, the microtitration plates were incubated with different concentrations of digoxigenin labelled lectins (0.2 µg ml–1 GNA, 0.2 µg ml–1 SNA, and 2 µg ml–1 MAA) or biotinylated lectins (0.2 µg ml–1 RCA I and 0.4 µg ml–1 PNA) for 1 h at 4°C in TBS-T with 1 mM MgCl2, 1 mM MnCl2, and 1 mM CaCl2. After rinsing, 1 ng (100 µl) peroxidase labelled antidigoxigenin or NeutrAvidinTM (Pierce Interchim, Montluçon, France) were added for 1 h at 4°C. After the addition of SureBlueTM TMB peroxidase substrate (KPL, Eurobio, Les Ulis, France), the reaction was stopped, and the absorbance was measured at 450 nm as described above.
In a second set of experiments, we analysed denaturated proteins treated or not with PNGase F by lectin western blotting.
Natural, recombinant and control (CPY, F, and asF) glycoproteins from DIG glycan differentiation kit, and