N-glycosylation influences the structure and self-association abilities of recombinant nucleolin Marie-Estelle Losfeld 1,2,3 , Arnaud Leroy 1,2,3,4 , Bernadette Coddeville 1,2,3 , Mathieu Carpentier 1,2,3 , Joe ¨ l Mazurier 1,2,3 and Dominique Legrand 1,2,3 1 Univ Lille Nord de France, Lille, France 2 USTL, UGSF, Villeneuve d’Ascq, France 3 CNRS, UMR, Villeneuve d’Ascq, France 4 EA 4529, Laboratoire de Biochimie applique ´ e, Faculte ´ de Pharmacie, Universite ´ Paris XI, Cha ˆ tenay-Malabry, France Keywords glycans; glycosylation; intermolecular interactions; nucleolin Correspondence D. Legrand, Unite ´ de Glycobiologie Structurale et Fonctionnelle, UMR CNRS 8576, IFR 147, Universite ´ des Sciences et Technologies de Lille, 59655 Villeneuve d’Ascq Cedex, France Fax: +33 3 20436555 Tel: +33 3 20434430 E-mail: dominique.legrand@univ-lille1.fr (Received 28 January 2011, revised 5 April 2011, accepted 12 May 2011) doi:10.1111/j.1742-4658.2011.08180.x Nucleolin is a major nucleolar protein involved in fundamental processes of ribosome biogenesis, regulation of cell proliferation and growth. Nucleolin is known to shuttle between nucleus, cytoplasm and cell surface. We have pre- viously found that nucleolin undergoes complex N- and O-glycosylations in extra-nuclear isoforms. We found that surface nucleolin is exclusively gly- cosylated and that N-glycosylation is required for its expression on the cells. Interestingly, the two N-glycans are located in the RNA-binding domains (RBDs) which participate in the self-association properties of nucleolin. We hypothesized that the occupancy of RBDs by N-glycans plays a role in these self-association properties. Here, owing to the inability to quantitatively pro- duce full-size nucleolin, we expressed four N-glycosylation nucleolin variants lacking the N-terminal acidic domain in a baculovirus ⁄ insect cell system. As assessed by heptafluorobutyrate derivatization and mass spectrometry, this strategy allowed the production of proteins bearing or not paucimannosidic- type glycans on either one or two of the potential N-glycosylation sites. Their structure was investigated by circular dichroism and fluorimetry, and their ability to self-interact was analyzed by electrophoresis and surface plasmon resonance. Our results demonstrate that all nucleolin-derived variants are able to self-interact and that N-glycosylation on both RBD1 and RBD3, or RBD3 alone, but not RBD1 alone, modifies the structure of the N-terminally truncated nucleolin and enhances its self-association properties. In contrast, N-glycosylation does not modify interaction with lactoferrin, a ligand of cell surface nucleolin. Our results suggest that the occupancy of the N-glycosyla- tion sites may contribute to expression and functions of surface nucleolin. Structured digital abstract l NCT binds to NCT by surface plasmon resonance (View interaction) l R3CT binds to R3CT by surface plasmon resonance (View interaction) l RCT binds to RCT by surface plasmon resonance (View interaction) l RCT binds to RCT by blue native page (View interaction) l NCT binds to lactoferrin by surface plasmon resonance (View interaction) l R3CT binds to R3CT by blue native page (View interaction) Abbreviations HRP, horseradish peroxidase; MOI, multiplicity of infection; NBT ⁄ BCIP, 5-bromo-4-chloro-3-indolyl phosphate ⁄ nitroblue tetrazolium; NCT, N-terminally deleted nucleolin; RBD, RNA-binding domain; RCT, N-terminally deleted nucleolin mutated on both Asn317 and Asn492 (N > A); R1CT, N-terminally deleted nucleolin mutated on Asn317 (N > A); R3CT, N-terminally deleted nucleolin mutated on Asn492 (N > A); SPR, surface plasmon resonance. 2552 FEBS Journal 278 (2011) 2552–2564 ª 2011 The Authors Journal compilation ª 2011 FEBS Introduction Nucleolin is a major and ubiquitous nucleolar protein of exponentially growing eukaryotic cells, which is involved in several fundamental processes of ribo- some biogenesis, regulation of cell proliferation and growth [1,2]. In spite of its major nuclear localiza- tion, nucleolin is also known to shuttle between the nucleus and the cytoplasm and, during this traffick- ing, it controls the organization of nucleolar chroma- tin, packaging of pre-RNA, rDNA transcription and ribosome assembly. In addition to its traffic between the nucleus and the cytosol, the presence of nucleolin at the surface of cells was formerly reported [3,4]. Cell surface expressed nucleolin has been reported as receptor or co-receptor of many proteins and molecules. Nucleo- lin interacts with apo-B and apo-E lipoproteins [4], matrix laminin-1 [5,6], attachment factor J [7,8] and endostatin [9], a domain of collagen XVIII, and it acts as a receptor of the anti-HIV cytokine midkine [10–12] and lactoferrin [13]. It has been proven that nucleolin interacts with lactoferrin through medium affinity and that nucleolin participates in endocytosis and nuclear targeting of lactoferrin [13], as was shown for midkine [10–12]. Furthermore, we recently demonstrated that interaction of nucleolin with its ligands may trigger signalization pathways [14]. HIV [15,16] and microbes [17–19] may also opportunisti- cally use surface nucleolin as an attachment molecule on cells. Knowledge of nucleolin structure is still patchy. Nu- cleolin is composed of three main domains [2]: the N-terminal domain is a highly acidic region with many TPXKK motifs which are involved in the interactions with histone H1 and chromatin; the central domain is composed of four RNA-binding domains (RBDs) able to interact with specific RNA targets like nucleolin rec- ognition element or evolutionary conserved motif [20]; and finally the C-terminal domain is rich in arginin and glycin residues and is able to interact with ribo- somal proteins [21] or ligands like lactoferrin [13]. Interestingly, the central domain has been shown to be involved in nucleolin self-interactions [22], a feature whose role in the protein structure, functions and trafficking is still unknown. Until now, the only three- dimensional structure of a recombinant protein con- sisting of domains RBD1 and RBD2 was resolved by NMR [23,24]. Recently, we reported that cytosolic and cell sur- face expressed nucleolins undergo complex N- and O-glycosylations [25]. Using nucleolin purified from Jurkat cells and techniques such as lectin recognition and MALDI-MS, it was demonstrated that nucleolin is glycosylated with two lactosaminic N-glycans local- ized in domains RBD1 (Asn317) and RBD3 (Asn492) and two sialylated O-glycans among five potential O-glycosylation sites in TPXKK motifs of the N-terminal domain [25]. A third potentially N-glycosylable site located between RBD1 and RBD3 (Asn478) was not found glycosylated [25]. This glycosylation is atypical because nucleolin has no signal sequence to direct it to the classical secretion pathway. It was not known until now how surface nucleolin, which is not an integral membrane protein, is expressed on the surface of cells and exerts its intracel- lular signaling properties on ligand interactions [14]. The evidence that surface nucleolin is exclusively gly- cosylated [14] strongly suggests that glycosylation is a crucial requisite for the expression and functions of nucleolin on cells. We thus hypothesized that glycosyl- ation could modify both structure and (self-)interaction properties of nucleolin. To assess this hypothesis, four secreted proteins cor- responding to human nucleolin deleted from its N-ter- minal acidic domain were expressed on the milligram scale by a baculovirus expression system. This strategy was used owing to the inability to produce full-length nucleolin. It also allows the convenient and quantita- tive preparation of recombinant glycoproteins from the culture medium. Protein mutations were designed to allow occupancy with glycans of one, two or none of the potentially N-glycosylable sites of nucleolin. These recombinant glycoproteins were produced and charac- terized by MALDI-TOF and GC ⁄ MS, CD and fluo- rimetry, and their self-interaction properties were investigated by PAGE and surface plasmon resonance (SPR) assays. Here, through monosaccharide composition and glycopeptide analysis, we show the production at the preparative level of nucleolin-derived proteins which carry or not paucimannosidic fucosylated N-glycans. This confirms the ability of the proteins to be gly- cosylated on only two N-glycosylation sites previ- ously evidenced on RBD1 and RBD3 in full-length surface nucleolin [25]. Fluorimetry and CD analysis demonstrate correct folding of the recombinant pro- teins, and the influence of N-glycosylation site occu- pancy in the structure of the proteins. SPR and PAGE studies confirm the ability of the nucleolin- derived proteins to self-interact, and show that this ability is strengthened depending on their N-glycosyl- ation status. M E. Losfeld et al. Glycosylation enhances nucleolin self-interactions FEBS Journal 278 (2011) 2552–2564 ª 2011 The Authors Journal compilation ª 2011 FEBS 2553 Results Production and purification of recombinant nucleolin-derived proteins, glycosylated or not, by a baculovirus ⁄ insect cell system Our previous attempts to produce recombinant full- size nucleolin in a baculovirus system resulted into low yield expression of intact proteins (not shown). This may be attributed to the propensity of the N-ter- minal highly acidic domain to cause nucleolin auto- proteolysis [26]. This, together with the previous evidence that the N-glycosylation sites are exclusively located in RBD1 and RBD3 of the central domain [25], led us to produce N-terminally truncated forms of nucleolin. This shorter form of the proteins, which encompasses residues 293–713 of nucleolin, also rids the molecule of the O-glycosylation sites located solely in the N-terminal domain of the protein [25]. To generate N-glycosylation variants of nucleolin, we have designed the production of four nucleolin- derived proteins in a baculovirus ⁄ insect cell system. In that aim, the truncated cDNA of native nucleolin was mutated for replacing Asn residues 317 and ⁄ or 492 with Ala residues. As illustrated in Fig. 1, three mutated proteins were obtained: R1CT mutated on Asn317 (hence N-glycosylable only on Asn492), R3CT mutated on Asn492 (hence N-glycosylable only on Asn317) and RCT mutated on both Asn 317 and 492 (not N-glycosylable). NCT corresponds to the non-mutated N-terminally truncated protein (hence fully N-glycosylable). All constructs are based on plasmid pAcSecG2T which contains a polyhedrin pro- moter and the secretion signal of glycoprotein p67, followed by the N-terminally truncated nucleolin sequence. Nucleolin-derived proteins were expressed using recombinant baculoviruses prepared as described in Materials and methods. Sf9 cells were used to propa- gate baculoviruses, which were then used to infect High Five cells. In these cells, a maximal level of recombinant protein secreted in cell culture superna- tant, estimated at 5–10 mgÆL )1 of medium, was observed after 5 days of infection (data not shown). All four recombinant proteins were purified by ion-exchange FPLC on an S-Sepharose Fastflow column (GE Healthcare Pharmacia, Uppsala, Sweden) as described in Materials and methods. Purified pro- teins were concentrated and analyzed by SDS ⁄ PAGE. As shown in Fig. 2 (Coomassie Blue staining), the purified proteins migrated as thick single bands with slight differences in their migrations, in the apparent molecular mass order NCT > R1CT ‡ R3CT ‡ RCT. This may be attributed to the presence of two glycans (NCT), one glycan (R1CT, R3CT) or no glycan (RCT) in the proteins. Immunostaining with anti-nucleolin IgG confirms that these bands are related to nucleolin, and also shows minor degradation products between 30 and 37 kDa (Fig. 2, Anti-nucleolin). Purity of pro- teins in preparations was estimated at over 95%. Characterization of the glycosylation of recombinant proteins In order to check the occupancy of glycosylation sites and the structure of glycans, GC ⁄ MS, MALDI-TOF Fig. 1. Schematic structural representation of (A) human nucleolin and (B) the four recombinant nucleolin-derived proteins (NCT, R1CT, R3CT and RCT) produced by a baculovirus ⁄ insect cell system. The different domains of full-size human nucleolin are represented: the acidic N-terminal domain (N-terminal), RBD1–RBD4 and the C-terminal RGG domain (C-terminal). The nuclear localization sequence is represented as NLS. The N-terminal domain was deleted in all NCT, R1CT, R3CT and RCT recombi- nant nucleolin-derived proteins. Positions of the mutated (N > A) and non-mutated (N) Asn residues of N-glycosylation sites are indicated. Glycosylation enhances nucleolin self-interactions M E. Losfeld et al. 2554 FEBS Journal 278 (2011) 2552–2564 ª 2011 The Authors Journal compilation ª 2011 FEBS analyses and immunostaining assays were performed. Monosaccharide composition of the glycans was deter- mined using GC ⁄ MS analysis after heptafluorobutyric anhydride treatment (Table 1). According to this method [27], the N-acetylglucosamine (GlcNAc) resi- due forming the N-glycosidic bond is converted to glu- cosamine (GlcNH 2 ). The GlcNH 2 peak (retention time of 12.39 min) was used as a reference, and its corrected area was considered as corresponding to one monosac- charide residue. The results summarized in Table 1 suggest the presence of mannose (Man), fucose (Fuc) and GlcNAc residues. Only three Man residues were detected, for one GlcNH 2 residue, which suggests the presence of paucimannosidic structures rather than high mannose structures. We detected 1.1 GlcNAc that may correspond to the GlcNAc residue linked to the N-linked GlcNAc. Furthermore, we detected 0.8 or 0.9 residue of Fuc, which would correspond to a Fuc linked in a1-3 on the N-glycan core, a feature specific to insect glycans [28]. In the RCT protein, as expected, no monosaccharide was found. This result also demon- strates that the third consensus site (Asn478), not mutated in any construction, is not occupied. This is in accordance with the absence of glycosylation on Asn478 observed in the glycosylated nucleolin isoforms from human cells [25]. In order to verify the presence of the fucosylated core, specific immunostaining of this structure was per- formed with polyclonal anti horseradish peroxidase (HRP) IgG as described by Kurosaka et al. [29]. As shown in Fig. 2 (Anti-HRP), the presence of a fucosy- lated core was revealed for the glycosylable NCT, R1CT and R3CT forms, but not for non-glycosylable protein RCT. Other faint bands probably correspond to both nucleolin-derived degradation products (evi- denced in Fig. 2, Anti-nucleolin) and contaminating insect cell fucosylated proteins. Taken together, these results strongly suggest the presence of fucosylated paucimannosidic glycans on the potentially N-glycosy- lable proteins. Lastly, occupancy of the glycosylation sites was analyzed by MS. MALDI-TOF analysis of glycopeptides from tryptic digest of recombinant protein was used to localize the N-glycosylated sites from the NCT protein, in order to verify, by compari- son with RCT, that the third consensus sequence con- taining Asn478 was not occupied. After enrichment of glycopeptides as described in Materials and methods, MS analysis (Fig. 3) of NCT tryptic peptides revea- led mass peaks at 3373.3, 3520.3, 3629.4, 3679.4 and 3775.5 Da which could correspond to peptides: Fig. 2. SDS ⁄ PAGE and immunoblotting of purified recombinant nucleolin-derived proteins NCT, R1CT, R3CT and RCT. Proteins were sepa- rated by SDS ⁄ PAGE (10%) and processed as described in Materials and methods. Proteins were either stained with Coomassie Blue or transferred onto nitrocellulose membranes for immunostaining with mouse monoclonal anti-nucleolin IgG (anti-C23, MS-3) by electrochemilu- minescence (Anti-nucleolin), and with rabbit polyclonal anti-HRP IgG followed by NBT ⁄ BCIP staining (Anti-HRP). Anti-HRP staining reveals a1–3 fucosylated glycans present in insect-expressed glycoproteins [29]. DF, dye front. Table 1. Monosaccharide composition of the four nucleolin-derived recombinant isoforms as determined by heptafluorobutyrate deriva- tion and GC ⁄ MS. Monosaccharide Recombinant protein NCT R1CT R3CT RCT GlcNH 2 11 1 0 Man 3 2.9 2.9 0 Fuc 0.8 0.9 0.9 0 GlcNAc 1.1 1.2 1.2 0 M E. Losfeld et al. Glycosylation enhances nucleolin self-interactions FEBS Journal 278 (2011) 2552–2564 ª 2011 The Authors Journal compilation ª 2011 FEBS 2555 298-VEGTEPTTAFNLFVGNLNFNK-318 ( 2312.16 Da) and 487-TLVLSNLSYATEETLQEVFEK-508 (2501. 27 Da), not phosphorylated or tri-phosphorylated, and carrying paucimannosidic mono-, di-fucosylated or non-fucosylated N-glycans (Fig. 3A). We hypothesize that phosphorylation of the recombinant proteins would result from the presence in the cell culture med- ium of kinases released from lysed baculovirus-infected cells. The slight difference between the theoretical and observed masses can be explained by the use of the linear positive-ion mode for mass determination of glycopeptides which could impair precision of this determination. In contrast, no peptides between 1887 and 2049 Da were found (Fig. 3B, brackets) which would correspond to a 995 Da peptide containing Asn478 linked to 892–1054 Da paucimannose fuco- sylated or non-fucosylated glycans. This result con- firms that Asn478 is not glycosylated. In conclusion, the structural analyses of the four recombinant nucleolin-derived proteins have demon- strated the presence of paucimannosidic fucosylated glycans on both Asn317 and Asn492 of the NCT pro- tein, on Asn492 of R1CT and on Asn317 of R3CT, and their absence on the RCT protein. Since directed mutagenesis can sometimes lead to protein misfolding, CD and fluorimetry were used to investigate the folding of the recombinant nucleolin- derived proteins. Interference of N-glycosylation with the structure of the recombinant nucleolin-derived proteins The two minima observed at 210 and 222 nm in the CD spectra (Fig. 4) show a dominant a-helical content in all four recombinant nucleolin isoforms with, how- ever, significant ellipticity differences between the pro- teins (Fig. 4). These differences can be attributed to conformation discrepancies that result from the state of glycosylation of the protein. Indeed, a minimal ellip- ticity at 222 nm for the unmutated fully glycosylated protein NCT and a maximal ellipticity at 222 nm for the single mutated protein R3CT (glycosylated on Asn317) can be observed (Fig. 4). Close and intermedi- ary ellipticities on the whole spectrum can be observed for R1CT (glycosylated on Asn492) and double mutated unglycosylated RCT (Fig. 4). These differ- ences in ellipticity of the proteins suggest that glycosyl- ation on both Asn317 and Asn492 probably interfere with conformation of the protein. Glycosylation of Asn317 increases the a-helical content of R3CT whereas glycosylation of Asn492 decreases the a-helical content of both R1CT and NCT. These differences in the measured ellipticity could also be interpreted as a consequence of the glycosylation on the propensity of NCT to oligomerization when glycosylated in position 492. This glycosylation may result, for example, in a loss of a-helical structure for b-sheet structure Fig. 3. MALDI-TOF mass spectra of the glycopeptides released from tryptic digest of recombinant nucleolin-derived proteins NCT and RCT. As described in Materials and methods, nucleolin iso- forms were separated by SDS ⁄ PAGE. The bands were excised and treated with trypsin. Glycopeptides were enriched by chromatogra- phy on Sepharose 4B and, after purification, mixed with a sinapinic acid matrix. Each sample was analyzed on a Voyager DE-STR MALDI-TOF instrument in the linear positive-ion mode using an accelerating voltage of 25 kV. (A) Spectra of glycopeptides from NCT and RCT between 2500 and 4000 Da containing peptides potentially glycosylated on Asn317 and Asn492: 3373.3 = 2312 Da + GlcNAc 2 Man 3 Fuc + Na + ; 3520.3 = 2312 Da + GlcNAc 2 Man 3 Fuc 2 +Na + ; 3629.4 = 2501 Da + GlcNAc 2 Man 3 triphosphate; 3679.4 = 2312 Da + GlcNAc 2 Man 4 Fuc 2 +Na + ; 3775.5 = 2501 Da + Glc- NAc 2 Man 3 Fuc triphosphate. (B) Spectra of glycopeptides from NCT and RCT between 1000 and 2500 Da. The bar marks the region potentially containing glycopeptide glycosylated on Asn478. Glycosylation enhances nucleolin self-interactions M E. Losfeld et al. 2556 FEBS Journal 278 (2011) 2552–2564 ª 2011 The Authors Journal compilation ª 2011 FEBS currently observed in self-associated proteins. We therefore checked this hypothesis by studying the self- association of the four recombinant isoforms (see the role of N-glycosylation on the self-association proper- ties of nucleolin later). The nucleolin sequence possesses two tryptophan res- idues at positions 481 and 644 [30]. Figure 5 shows the fluorescence emission spectra of the four nucleolin- derived proteins in the wavelength range 300–440 nm. We can observe maximal emission between 340 and 360 nm. A similar tendency between the fluorescence intensity and CD spectra is observed. A minimal inten- sity of fluorescence at about 350 nm is observed for NCT and a maximal intensity for R3CT. Close and intermediary fluorescence intensities are observed for R1CT and RCT (Fig. 5). Interestingly, the wavelength of maximal emission for both R1CT and NCT (both glycosylated on Asn492) undergoes a red shift com- pared with the wavelength of maximum emission for the double mutated unglycosylated RCT and for R3CT (both unglycosylated on Asn492) (Fig. 5). This is not surprising as one of the Trp residues (Trp481), which is close to the glycosylated site Asn492, becomes probably more hydrated and therefore more exposed to the sol- vent when Asn492 is glycosylated. This more exposed Trp observed when Asn492 is glycosylated can explain the significantly modified structures of NCT and R1CT, compared with R3CT and RCT (Fig. 5). It also indicates that the occupancy of Asn492 by an N-glycan may either modify the con- formation of proteins or facilitate the oligomerization of NCT and R1CT, or both (see the next section). Studies of the role of N-glycosylation on the self-association properties of the recombinant nucleolin-derived proteins To assess the influence of N-glycans on the self-associ- ation properties of nucleolin, we investigated the abil- ity of the four nucleolin isoforms to interact through PAGE in native conditions and SPR assays. In a first step, we investigated the PAGE behavior of the recombinant nucleolin-derived proteins in native conditions. Since the theoretical isoelectric point of the proteins is estimated at 8.8 (calculated according to http://expasy.org/tools/pi_tool.html), PAGE was per- formed at a lower pH (pH 7.0). As shown in Fig.6, unlike R3CT and RCT which markedly migrated as sin- gle major bands towards the cathode, NCT and R1CT exhibited no migration or very poor migration on the gel. Only faint bands can be observed in the upper part of the gel. This suggests that both proteins may be self- associated. The integrity of proteins separated in non- denaturing conditions was checked by submitting the proteins to a second electrophoresis on SDS ⁄ PAGE with b-mercaptoethanol as described in Materials and methods. All proteins migrated as single 45 kDa bands (data not shown). Taken together, these results suggest that the presence of a glycan on the RBD3 domain modifies the propensity of nucleolin to self-associate. Furthermore, our spectroscopic st udies (see a bove) s uggest that the presence of a glycan located on Asn492 changes the exposure to the solvent of the protein region close to Trp481 that seems responsible for its self-association. To confirm the results obtained by PAGE, SPR studies were undertaken on a BIAcore 3000 system using NCT, R1CT, R3CT and RCT immobilized on a Fig. 4. CD spectra of the recombinant nucleolin-derived proteins NCT, R1CT, R3CT and RCT. Spectra were obtained on 2 mg of purified proteins with a 1 mm cell at 25 °C and recorded between 190 and 250 nm with an increment of 0.5 nm and an integration time of 2 s. Fig. 5. Fluorescence spectra of the recombinant nucleolin-derived proteins NCT, R1CT, R3CT and RCT. Fluorescence of tryptophan was recorded on purified proteins NCT, R1CT, R3CT and RCT between 300 and 450 nm, as described in Materials and methods. M E. Losfeld et al. Glycosylation enhances nucleolin self-interactions FEBS Journal 278 (2011) 2552–2564 ª 2011 The Authors Journal compilation ª 2011 FEBS 2557 CM5 sensor chip (ligands) and injected as analytes. Various concentrations between 0.25 and 4 lm were injected. Figure 7 shows a representative experiment from three sets of experiments performed on different CM5 sensor chips. Interestingly, self-interactions between all nucleolin-derived isoforms can be observed. However, the affinity of NCT, R1CT or R3CT self-interactions was about 100-fold higher than that of RCT. Furthermore, although the K d of R3CT self-interactions was similar to that of NCT and R1CT, R3CT self-binding was about four-fold and three-fold lower than that of NCT and R1CT, respec- tively. These results are in agreement with the PAGE results (Fig. 6) that demonstrate different migration patterns for NCT and R1CT, compared with R3CT and RCT. Hence, our results confirm the observation that nucleolin is able to interact with itself and to oligomerize [22], a property that could play an impor- tant role for its receptor functions. We show here that N-glycosylation, and most particularly RBD3 Fig. 6. Migration patterns of the recombinant nucleolin-derived pro- teins NCT, R1CT, R3CT and RCT in PAGE in non-denaturing condi- tions. Purified recombinant proteins (10 lg of each protein per lane) were loaded onto a PAGE 7.5% gel and submitted to electrophore- sis in non-denaturing conditions as described in Materials and methods. Proteins were revealed by Coomassie Blue staining. +, cathode of the generator; ), anode of the generator. Fig. 7. Study by SPR (BIAcore) of the self-interactions of the recombinant nucleolin-derived proteins NCT, R1CT, R3CT and RCT. The SPR sensorgrams shown are from an experiment which is rep- resentative of a set of three separate experiments on different sen- sor chips with similar results. Recombinant nucleolin-derived proteins NCT, R1CT, R3CT and RCT were immobilized onto a CM5 sensor chip. The same proteins were used as analytes at different concentrations (250–4000 n M) at a flow rate of 5 lLÆmin )1 ,as described in Materials and methods. Association was studied during 3 min and dissociation during 10 min at a flow rate of 5 lLÆmin )1 of HBS. Constants were estimated using BIAEVALUATION 3.1 with the Langmuir 1 : 1 formula. The Scatchard plots derived from these data at equilibrium are presented as inserts. RU, response unit. Glycosylation enhances nucleolin self-interactions M E. Losfeld et al. 2558 FEBS Journal 278 (2011) 2552–2564 ª 2011 The Authors Journal compilation ª 2011 FEBS N-glycosylation, strongly enhances self-interactions of the nucleolin-derived proteins. N-glycosylation does not influence the interactions of the recombinant nucleolin-derived proteins with lactoferrin Cell-surface expressed nucleolin has been described as receptor or co-receptor of many proteins and molecules [1,2]. We previously demonstrated that nucleolin inter- acts with lactoferrin with medium affinity (K d = 0.24 lm) through the RGG C-terminal domain of nucleolin, and that nucleolin participates, like midki- ne [10–12], in endocytosis and nuclear targeting of its ligand [13]. Since N-glycosylation interferes with the structure and the self-interaction properties of the nucleolin-derived proteins, we investigated the ability of the different proteins to interact with lactoferrin using SPR (BIAcore 3000). With this aim, NCT, R1CT, R3CT and RCT were immobilized onto a CM5 sensor chip and concentrations ranging from 50 to 800 nm of human lactoferrin were injected. The results of a repre- sentative study among three separate studies are shown in Fig. 8. We observe that all nucleolin-derived proteins, regardless of their N-glycosylation status, interact with similar binding parameters, with an affinity of about several hundred nanomoles, similar to that of native full-size nucleolin reported by Legrand et al. [13]. These results suggest that N-glycosylation does not interfere with the ability of the nucleolin-derived pro- teins to interact with lactoferrin, one of its major sur- face ligands. Discussion Intracellular nucleolin is a ubiquitous protein that par- ticipates in important cellular events, like ribosome biogenesis, chromatin organization, apoptosis and reg- ulatory activity of transcription factors [1,2]. Cell-sur- face-expressed nucleolin has been described as interacting with extracellular components like laminin [5,6] or proteoglycans [31], but also as a receptor for apolipoproteins [4] or l-selectin [32] and as an internal- izing receptor for lactoferrin [13], midkine [10–12] or the gp120 protein of HIV [15,16]. However, many aspects of the structure–function relationships of sur- face nucleolin remain unknown. In fact, only the struc- tures of RBD1 and RBD2 have been determined [24]. Our group has previously demonstrated that nucleolin undergoes N- and O-glycosylations [25] and that N-glycosylation is an essential requirement for its cell surface expression [14]. Furthermore, we have recently described the ability of nucleolin to trigger calcium Fig. 8. Study by SPR (BIAcore) of the interactions of human lac- toferrin with the recombinant nucleolin-derived proteins NCT, R1CT, R3CT and RCT immobilized onto a CM5 sensor chip. The SPR sensorgrams shown are from an experiment which is repre- sentative of a set of three separate experiments on different sen- sor chips with similar results. Recombinant nucleolin-derived proteins NCT, R1CT, R3CT and RCT were immobilized onto a CM5 sensor chip. Human lactoferrin was used as analyte at dif- ferent concentrations (50–800 n M) at a flow rate of 5 lLÆmin )1 ,as described in Materials and methods. Association was studied dur- ing 3 min and dissociation during 10 min at a flow rate of 5 lLÆmin )1 of HBS. Constants were estimated using BIAEVALUATION 3.1 with the Langmuir 1 : 1 formula. The Scatchard plots derived from these data at equilibrium are presented as inserts. RU, response unit. M E. Losfeld et al. Glycosylation enhances nucleolin self-interactions FEBS Journal 278 (2011) 2552–2564 ª 2011 The Authors Journal compilation ª 2011 FEBS 2559 entry into cells through ligand binding, probably by interacting with other still-unknown cell surface signal- ing molecules [14]. The exact role(s) of N-glycans on surface nucleolin remain(s) unspecified but we hypoth- esized that they would interfere in the structure of nucleolin and hence in its functions and interaction abilities. Here, in agreement with previous observations made with recombinant nucleolin variants produced in Esc- herichia coli [22], we confirm that all the nucleolin- derived proteins are able to self-interact. Our results also indicate that the absence of residues 1–292 does not prevent the self-interactions, and hence support previous evidence that the central domain of nucleolin is involved in these interactions [22]. Most importantly, we demonstrate that N-glycosylation strongly enhances the self-interactions of nucleolin. This evidence was gained by producing recombinant nucleolin isoforms carrying or not paucimannosidic fucosylated N-glycans by a baculovirus⁄ insect cell system. Although such N-glycans are obviously different from those of natural nucleolin [25], we hypothesized that their mere pres- ence within the N-glycosylation sites would modify the ability of the nucleolin-derived proteins to oligomerize. The propensity of the nucleolin-derived proteins to self-interact [22], regardless of their glycan content, was ascertained through SPR which showed self-inter- actions of the non-glycosylated isoform (RCT) with a K d of 18.3 lm. Most interestingly, the fully N-glycosy- lated isoform (NCT) or N-glycosylated on the sole RBD3 (R1CT) exhibited 100-fold higher affinities than RCT. Similar affinity but with a lower binding capac- ity than NCT and R1CT was observed with the iso- form N-glycosylated on the sole RBD1 (R3CT). This suggests that the occupancy of the glycosylation sites, and most particularly Asn492 (RBD3), is a requisite for strong nucleolin self-interactions. These stronger interactions of the NCT and R1CT isoforms, com- pared with R3CT and RCT, were confirmed by native PAGE analysis. To explain this enhancement of NCT and R1CT self-interactions, and taking into account CD and fluorimetry results, we hypothesize that N-gly- cans, in particular the N-glycan on Asn492, modify the structure of the protein. Such modification could expose hydrophobic regions of the protein involved in high affinity nucleolin self-interactions. Whereas cell surface nucleolin appears to be located near the lipid raft or associated with it [10], it is not known how the molecule is presented at the cell sur- face. Although nucleolin is not an integral membrane protein, it is able to act as a receptor for a large num- ber of molecules and to internalize some of them [7,10–13] or, most unexpectedly, to induce cellular events [7,14] by triggering calcium entry in cells [14]. To explain this later observation, we hypothesized that nucleolin could act as a co-receptor able to interact with cell surface signaling receptors following ligand binding [14]. This implies a redistribution of nucleolin at the surface of cells following ligand binding, and a possible role of glycosylation in the topology of nucle- olin on cells. In strong support of this hypothesis, our results suggest that N-glycosylation probably influ- ences the self-interaction properties of nucleolin, while it does not affect the binding of ligands, such as lacto- ferrin. It may also be hypothesized that whereas the C-terminal domain is involved in the interactions of surface nucleolin with most of its ligands, the RBD- containing central domain is involved in the protein self-interactions. Interestingly, preliminary studies show that N-glyco- sylation may modulate binding of the nucleolin-derived protein to heparin (data not shown). This suggests that glycosylation could modulate the interaction of nucleo- lin with sulfated glycosaminoglycans, its distribution on the surface of cells, and its ability to interact with signaling molecules. Of course, we should bear in mind that both composition and length of the N-glycans of insect cells are different from human ones. It would thus be interesting to transpose our system in human- ized insect cells [33]. In conclusion, many questions remain to be resolved concerning the expression of nucleolin at the surface of cells and its trafficking, but also concerning the involve- ment of glycosylation in these processes. As demon- strated previously [14], glycosylation is an essential requirement for cell surface expression. Here, our results with N-terminally-truncated nucleolin suggest that N- glycosylation may influence the structure of natural nu- cleolin in a way that enhances its ability to self-interact. The exact roles of high affinity nucleolin self-interac- tions of surface nucleolin, in particular in the activation of signaling partners, are still to be elucidated. Materials and methods Cells Spodoptera frugiperda (Sf9) and Trichoplusia ni (High Five) insect cells were purchased from Invitrogen (Carlsbad, CA, USA). Sf9 and High Five cells were respectively grown in Max-XP medium (BD Biosciences, Le Pont de Claix, France) containing 1% (v ⁄ v) ultraglutamine (Lonza, Basel, Switzerland), 1% fetal bovine serum (Cambrex, Emerain- ville, France) and 50 lgÆmL )1 gentamycin, and in Express Five medium (Cambrex) containing 10% (v ⁄ v) ultragluta- mine (Lonza) at 27 °C. Glycosylation enhances nucleolin self-interactions M E. Losfeld et al. 2560 FEBS Journal 278 (2011) 2552–2564 ª 2011 The Authors Journal compilation ª 2011 FEBS DNA, plasmids, site-directed mutagenesis and baculovirus preparation Human nucleolin cDNA was previously obtained from MDA-MB231 cells (ATCC) in our laboratory and cloned into pTRE-HA plasmid from Clontech (Takara Bio Europe, Saint-Germain-en-Laye, France) [13]. Oligonucleotide prim- ers were synthesized by Eurogentec (Lie ` ge, Belgium) and restriction enzymes were from New England Biolabs (Evry, France). pAcSecDGST was obtained from plasmid pAc- SecG2T (Pharmingen-BD Biosciences, Le Pont de Claix, France) by inserting a BamH1 site before the GST sequence in order to remove it (Directed mutagenesis Quickchange XL II kit, Stratagene-Agilent, Santa Clara, CA, USA). The primers used were 5¢-CCTTTGCGGATCTGATGTCCCCT GGATCCGC-3¢ and 5¢-GCACAAGGCCCTTAATTTTC CAATAACCGGA-3¢. pTRE-HA containing the coding region of nucleolin cDNA was mutated, or not, on the glycosylation sites (Direc- ted mutagenesis Quickchange XL II kit). Mutation of Asn317 (N > A) was achieved with primers 5¢-CAATCTCT TTGTTGGAAACCTAAACTTTCAGAAATCTGCTCCT GAATTAAAAACTGG-3¢ and 5¢-CCAGTTTTTAATTCA GGAGCAGATTTCTGAAAGTTTAGGTTTCCAACAA AGAGATTG-3¢. Mutation of Asn492 (N > A) was achie- ved with primers 5¢-GGTGAATCAAAAACTCTGGTTTT AAGCCAGCTCTCCTACAGTGCAACAGAAGAAACTC-3¢ and 5¢-GAGTTTCTTCTGTTGCACTGTAGGAGAGCTG GCTTAAAACCAGAGTTTTTGATTCACC-3¢. The coding region of nucleolin cDNAs corresponding to the N-terminal truncated nucleolin sequence mutated or not on glycosylation sites was amplified by PCR with con- current introduction of BamH1 and EcoR1 sites at the 5¢ and 3¢ terminus, respectively. The primers used were 5¢-G CCAAACAGAAAGCAGCTCCTGGATCCAAGAAACA G-3¢ and 5¢-GTGCCTTCCACTTTCTGTTTCTTGGATCC AGGAGCTGCTTTC-3¢. After sequence verification, the BamH1- and EcoR1-digested inserts were cloned into sim- ilarly digested pAcSecDGST. The resulting constructs were used to co-transfect Sf9 cells with linearized baculovirus BD BaculoGoldÔ Bright (Pharmingen-BD Bio-sciences) by addition of 5 lL of FlyfectinÔ (OZ Biosciences, Marseille, France) for 500 ng of recombinant plasmid and 100 ng of linearized baculovirus. After plasmid recombina- tion and circularization of baculovirus, the infected cells are able to express the green fluorescent protein. The level of infection was thus verified by flow cytofluorimetry on a FACScalibur cytometer (BD Bioscience). The super- natant of transfected cells containing recombinant bacul- oviruses was collected after 5 days. Viruses were amplified by successive infections for obtaining a high multiplic- ity of infection (MOI) (‡ 2 · 10 5 virusÆlL )1 ). Expression of recombinant nucleolin-derived proteins was achieved by infecting High Five cells with the baculovirus suspensions. Expression and purification of recombinant nucleolin-derived proteins Production of recombinant proteins was done in High Five cells grown in 175 cm 2 flasks. Cells were infected by addi- tion of 4% (v ⁄ v) of supernatant at high MOI in the culture medium of 70% confluent cells. After 5 days, the superna- tant was collected and stored at ) 20 °C. A chromatography procedure was used to purify the recombinant forms of nucleolin from insect cell superna- tants. The purification was made by FPLC (GE Healthcare Pharmacia, Uppsala, Sweden) using an ion exchange S-Sepharose Fast Flow 5 · 2 cm column (GE Healthcare Pharmacia). The supernatant were diluted in buffer A [Tris ⁄ HCl 50 mm pH 7.5; MgCl 2 5mm; EDTA 0.1 mm; Pefabloc 1 mm (Roche Diagnostics, Meylan, France)] with addition of 5% (v ⁄ v) of EDTA 100 mm and 2% (v ⁄ v) of NaOH 1 m. The elution was performed with a gradient from 0% to 60% of buffer B (Tris ⁄ HCl 50 mm pH 7.5; MgCl 2 5mm; EDTA 0.1 mm; NaCl 1 m; Pefabloc 1 mm) for 60 min at a rate of 1 mLÆmin )1 , followed by a gradient from 60% to 100% of buffer B for 15 min at a rate of 1mLÆmin )1 . The protein content of fractions was concen- trated by centrifugation (3000 g) on a Vivaspin (Sartorius AG, Goettingen, Germany) concentrator and dialyzed in ammonium bicarbonate 200 mm. GC ⁄ MS analysis of nucleolin monosaccharides The four nucleolin isoforms were separated by SDS ⁄ PAGE (4 ⁄ 7.5%, 2 lgÆlane )1 ) and electro-transferred on poly(vinyli- dene difluoride) membrane. After Ponceau S staining, the bands of nucleolin were cut out, washed and dried. The preparation of samples was as described previously [25]. Samples were submitted to methanolysis (20 h at 80 °Cin 500 lL of anhydrous methanol containing 0.5 m gaseous HCl). After the samples were dried under a stream of nitro- gen, they were added to 200 lL of acetonitrile and 25 lL of heptafluorobutyric anhydride and heated for 30 min at 150 °C. After evaporation of the solvents, the samples were dis- solved in 200 lL of dried acetonitrile and 1 lL was injected in the Ross injector (260 °C) of a Carlo Erba GC 8000 gas chromatograph (Carlo Erba, Sabadell, Spain) equipped with a 25 m · 0.32 mm CP-Sil5 CB low bleed ⁄ Ms capillary column, 0.25 lm film phase (Chrompack, Les Ulis, France). The sample was analyzed using a program starting at 90 °C for 3 min, followed by an increase (50 °CÆmin )1 ) until 260 °C was reached. The column was coupled to a Finni- gan Automass II mass spectrophotometer (Thermo Finni- gan, Courtaboeuf, France). Analysis was performed in the electron impact mode (ionization energy 70 eV; source tem- perature 150 °C). Quantitation of various constituents was performed using the total ion count of the MS detector and the xcalibur software (Thermo Finnigan). M E. Losfeld et al. Glycosylation enhances nucleolin self-interactions FEBS Journal 278 (2011) 2552–2564 ª 2011 The Authors Journal compilation ª 2011 FEBS 2561 [...]... lLÆmin)1 and for the dilution of ligands and analytes For the binding assays, N-terminal truncated recombinant forms of nucleolin were immobilized onto the BIAcore sensor chip CM5 using an amine-coupling kit (BIAcore) according to the manufacturer’s instructions Recombinant nucleolin was immobilized at a concentration of 1– 3 lgÆmL)1 in 10 mm sodium acetate, pH 3.8, at a 5 lLÆmin)1 flow rate of HBS Covalent... of 1000 ± 100 resonance units An empty flow cell was used as a control for non-specific binding and bulk effects For the studies of self-interactions of nucleolin isoforms, the ligands were injected at several concentrations (ranging from 0.5 to 4 lm in HBS), at a 5 lLÆmin)1 flow rate during 3 min Dissociation was studied during 10 min at a flow rate of 5 lLÆmin)1 of HBS After the dissociation phase, the. .. was regenerated by injection of 10 lL of NaOH 20 mm at a 5 lLÆmin)1 flow rate For the study of lactoferrin binding to the nucleolin isoforms, human lactoferrin was injected at concentrations ranging from 50 to 800 nm in HBS at a 5 lLÆmin)1 flow rate during 3 min Regeneration was done by injection of 10 lL of NaOH 20 mm at a 5 lLÆmin)1 flow rate The dissociation constant (Kd) and Rmax ± SEM were calculated... integration time of 2 s, FEBS Journal 278 (2011) 2552–2564 ª 2011 The Authors Journal compilation ª 2011 FEBS M.-E Losfeld et al and a constant band-pass of 2 nm The signal from the blank scan was subtracted from the corresponding sample scan Steady-state fluorescence of tryptophan was monitored on a Fluoromax-2 (Jobin Yvon SPEX) spectrometer at 25 °C An excitation wavelength of 295 nm was used and the emission... (2000) The C-terminal domain of nucleolin accelerates nucleic acid annealing Biochemistry 39, 15493–15499 2564 23 Allain FH, Bouvet P, Dieckmann T & Feigon J (2000) Molecular basis of sequence-specific recognition of preribosomal RNA by nucleolin EMBO J 19, 6870–6881 24 Allain FH, Gilbert DE, Bouvet P & Feigon J (2000) Solution structure of the two N-terminal RNA-binding domains of nucleolin and NMR... Community (FEDER), the ´ Region Nord-Pas de Calais (France), the CNRS and Glycosylation enhances nucleolin self-interactions ´ the Universite des Sciences et Technologies de Lille We thank Frank Wien of the synchrotron facilities SOLEIL at Saint Aubin, France, for his advice and fruitful scientific discussions for the circular dichroism experiments References 1 Mongelard F & Bouvet P (2007) Nucleolin: a multifaceted... Elass E, Masson M, Slomianny MC, Carpentier M, Briand JP, Mazurier J & Hovanessian AG (2004) Surface nucleolin participates in both the binding and endocytosis of lactoferrin in target cells Eur J Biochem 271, 303–317 Losfeld ME, Khoury DE, Mariot P, Carpentier M, Krust B, Briand JP, Mazurier J, Hovanessian AG & Legrand D (2009) The cell surface expressed nucleolin is a glycoprotein that triggers calcium... Briand JP, Dupuis M & Charbit A (2008) A novel receptor–ligand pathway for entry of Francisella tularensis in monocytelike THP-1 cells: interaction between surface nucleolin and bacterial elongation factor Tu BMC Microbiol 8, 145 de Verdugo UR, Selinka HC, Huber M, Kramer B, Kellermann J, Hofschneider PH & Kandolf R (1995) Characterization of a 100-kilodalton binding protein for the six serotypes of. .. For bidimensional electrophoresis, 15 lg of proteins were loaded onto two lanes of PAGE in non-denaturing conditions at pH 7.0 (as described above) After migration, one of the two lanes was cut and stained with Coomassie Blue The second lane was cut and boiled in 3 · sample buffer with b-mercaptoethanol for 30 min After treatment, the lane was set at the top of a 7.5% SDS ⁄ PAGE Migration was performed... sulfatebinding protein on the surface of cancer cells Glycobiology 15, 1–9 32 Harms G, Kraft R, Grelle G, Volz B, Dernedde J & Tauber R (2001) Identification of nucleolin as a new L-selectin ligand Biochem J 360, 531–538 33 Harrison RL & Jarvis DL (2006) Protein N-glycosylation in the baculovirus-insect cell expression system and engineering of insect cells to produce ‘mammalianized’ recombinant glycoproteins . proteins. We therefore checked this hypothesis by studying the self- association of the four recombinant isoforms (see the role of N-glycosylation on the self-association. Fluorimetry and CD analysis demonstrate correct folding of the recombinant pro- teins, and the influence of N-glycosylation site occu- pancy in the structure of the