N-Glycosylation Is Required for Human CD2 Immunoadhesion Functions*

The T-lymphocyte glycoprotein receptor, CD2, mediates cell-cell adhesion by binding to the surface molecule CD58 (LFA-3) on many cell types including antigen presenting cells. Two domains comprise the CD2 extracellular segment, with all adhesion functions lo-calized to the amino-terminal domain that contains a single N-glycosylation site at Asn”. We have defined an important role for the N-linked glycans attached to Asns5 of this domain in mediating CD2-CD58 interactions and also characterize its N-glycotype structure. Analysis of deglycosylated soluble recombinant CD2 as well as a mutant transmembrane CD2 molecule containing a single A ~ n ~ ~ - G l n ~ ~ substitution demonstrates that neither deglycosylated CD2 nor the mutant CD2 transmembrane receptor binds CD58 or monoclonal antibodies directed at native CD2 adhesion domain epitopes.

The T-lymphocyte glycoprotein receptor, CD2, mediates cell-cell adhesion by binding to the surface molecule CD58 (LFA-3) on many cell types including antigen presenting cells. Two domains comprise the CD2 extracellular segment, with all adhesion functions localized to the amino-terminal domain that contains a single N-glycosylation site at Asn". We have defined an important role for the N-linked glycans attached to Asns5 of this domain in mediating CD2-CD58 interactions and also characterize its N-glycotype structure. Analysis of deglycosylated soluble recombinant CD2 as well as a mutant transmembrane CD2 molecule containing a single A~n~~-G l n~~ substitution demonstrates that neither deglycosylated CD2 nor the mutant CD2 transmembrane receptor binds CD58 or monoclonal antibodies directed at native CD2 adhesion domain epitopes. Electrospray ionization-mass spectrometry demonstrates that high mannose oligosaccharides ((Man).GlcNAc2, n = 5-9) are the only N-glycotypes occupying AS^'^ when soluble CD2 is expressed in Chinese hamster ovary cells. Based on a model of human CD2 secondary structure, we propose that Nglycosylation is required for stabilizing domain 1 in the human receptor. Thus, N-glycosylation is essential for human CD2 adhesion functions. Refs. 1 and 2). The T-cell-restricted CD2 molecule binds to the widely distributed CD58 (LFA3) structure found on many nucleated and non-nucleated cells (3)(4)(5). By promoting cell-cell contacts and hence, receptor interactions between apposing membranes of T-lymphocytes and their cognate partners, the CD2-CD58 adhesion pair plays a n important role in facilitating CD3-Ti antigen receptor Grants ROlGM45781 (to V. N. R.) and A121226 (to E. L. R.) The * This work was supported in part by National Institutes of Health costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

CD2 (T11) is a transmembrane glycoprotein receptor that mediates adhesion between T-lymphocytes and antigen-presenting cells (reviewed in
$ To whom correspondence should be addressed. (TCR)' recognition of antigens presented via the major histocompatibility complex (6)(7)(8). Recent evidence suggests that CD2-CD58 interaction initiates primary cell-cell adhesion prior to TCR engagement (9). Conjugate formation stabilized primarily by CD2-CD58 interactions permits efficient antigen/major histocompatibility complex recognition by the TCR and subsequent up-regulation of the binding affinity between the LFA1-ICAM1 adhesion pair (9)(10)(11). Thus, by boosting the efficiency of cell-cell interactions, T-cell activation can effectively be accomplished via specific TCR-major histocompatibility complex binding with much lower antigen concentrations (8). The importance of the CD2-CD58 adhesion pair in immune responses is underscored by studies demonstrating that anti-CD2 monoclonal antibodies (mAbs) which block CD58 binding are directed against a specific CD2 epitope (Tll'), and these mAbs also inhibit antigen-dependent T-cell proliferation and cytolytic effector functions (12)(13)(14)(15).
CD2 also plays a role in T-cell lymphocyte activation. Perturbation of the CD2 extracellular domain with a combination of antibodies against two other CD2 extracellular epitopes (Tlln, Tl13), which are distinct from the T l l l site, or interaction of CD2 with CD58 bearing cells together with mAbs specific for the Tlla epitope can stimulate T-cell proliferation, IL-2 secretion, and clonal expansion (12,(16)(17)(18)(19). Stimulation of T-cells through CD2 also transduces activation signals that synergize with those mediated by the TCR (20)(21).
We previously established that the structural basis for CD58 recognition lies within the first extracellular domain of the CD2 polypeptide using recombinant soluble CD2 proteins (22). Mutagenesis studies also defined two distinct regions within the CD2 adhesion domain, centered about L Y S~~ and GlygO of the mature polypeptide that are implicated in CD2-CD58 binding (23, 24). However, synthetic peptides spanning 21-30-residue sequences within the adhesion domain between IleZi3 and Leu'"" failed to bind either CD58 or anti-CD2 mAbs (22), suggesting that the CD58 binding site on CD2 may be comprised of a combination of nonlinear epitopes on the receptor surface or perhaps requires a conformationally constrained linear polypeptide. The amino acid sequence of hu-man CD2 (25-27) revealed the presence of a single consensus N-linked glycosylation site within the adhesion domain at Am6'. This N-carbohydrate binding site is conserved in murine CD2 (28,29) but not in rat CD2 (30), based on sequence alignment of homologous regions within the first domain of each receptor. Since this single N-glycosylation site is flanked by the two regions within the human CD2 polypeptide sequence implicated in both T-cell activation and adhesion functions, we investigated the functional role that N-linked glycans might play in mediating CD2-CD58 interactions.

EXPERIMENTAL PROCEDURES
Construction of sCDglR2 and Expression in Chinese Hamster Ovary Cells-Plasmid pMl/~CD2~8p was constructed by isolating a BamHI generated DNA fragment coding for s C D~~~~ from the plasmid pACY373/Tll,,, (31). The fragment was blunt-ended with T4 DNA polymerase, then all but four nucleotides from the 5"untranslated sequence were removed by digesting with DdeIIBglII and BglII alone. DNA fragments containing the 5' and 3' portions of the gene were gel purified and the TlleXa gene was reconstituted by ligation of these fragments into the expression plasmid pMI-gpt at its unique XhoI site. Chinese hamster ovary (CHO) DXB 11 (dhfr-minus) cells were cotransfected with 25 pg of pMI/sCD21R2 and 1 pg of pSV2-dhfr. Transfectants were selected for their conversion to a dhfr-positive phenotype and screened for co-expression of secreted sCD2,~z using a n anti-CD2 ELISA (22). High ~c D 2~~~-p r o d u c i n g CHO cell lines were selected from those exhibiting resistant to increasing concentrations of methotrexate. Clone D51 was cultured in CellPharm hollow fiber bioreactors (C. D. Medical, Miami Lakes, FL) or grown in suspension culture attached to microcarriers.
Purification of s C D~~R~-S C D~~R~ was purified using an anti-CD2 immunoaffinity column prepared as follows, 1.6 g of anti-T112 monoclonal antibody (10LD2-4C1) in RPMI, 5% fetal calf serum was bound to 200 ml of protein A-Sepharose (Repligen, Cambridge, MA) overnight a t 4 "C, cross-linked with 20 mM dimethypymilimidate in 0.2 M borate buffer, pH 9, for 30 min at 25 "C, blocked with 0.2 M ethanolamine, pH 8, for 2 h a t 25 "C, washed with 50 mM sodium citrate buffer, pH 3.0, and re-equilibrated with phosphate-buffered saline (PBS). CHO cell-conditioned medium was passed over the immunoaffinity column, nonspecifically bound protein was washed off the column with PBS, and S C D~~R~ was eluted with 50 mM sodium citrate buffer, pH 3.0. Fractions containing s C D~~R~ were pooled, diluted with an equal volume of 0.1% trifluoroacetic acid/water (solvent A), and loaded onto a preparative C4 reverse phase HPLC column (21 X 250 mm; Vydac) equilibrated in 75% solvent A, 25% solvent B (0.1% trifluoroacetic acid, 90% acetonitrile, 10% water). The column was eluted with a linear gradient from 25 to 50% solvent B over 50 min a t a flow rate of 10 ml/min. Fractions containing S C D~~~? were pooled, diluted with an equal volume of 10 mM sodium citrate buffer, pH 5.5, plus 3 volumes of distilled water and loaded onto CM-Sepharose FastFlow medium (Pharmacia LKB Biotechnology Inc.) equilibrated in the same buffer. s C D~~~~ was eluted by stepwise adjustment of the buffer to 500 mM NaCl, dialyzed into 10 mM sodium citrate buffer, pH 5.5, and stored a t -70 "C for subsequent studies. Purity was monitored by SDS-PAGE under reducing conditions (32) and s C D~~R~ antigen was quantitated using a sandwich ELISA employing two different anti-CD2 mAbs as previously described (22).
Antibody Capture ELISAS-96-well plates were coated with varying concentrations of antigen (0.2-1.5 pg/ml in PBS) overnight a t 4 "C, washed with PBS, 0.05% Tween 20, blocked with 2% bovine serum albumin, PBS, 0.05% Tween 20 for 2 h a t 25 "C, and washed. Plates were then probed with either anti-Tlll mAb horseradish peroxidaseconjugate or anti-Tllp mAb horseradish peroxidase conjugate prepared and developed with OPD substrate as described previously (22). Preparation of and Deglycosylated sCD210s-sCD210.S was prepared by dialyzing a 50-mg aliquot of sCD2,n2 (200 pg/ml) into PBS containing 1 mM CaCI2. Freshly prepared dithiothreitol was added to a final concentration of 2 mM and the sample was digested with 1.0 mg of clostripain (Promega: 280 units/mg) for 40 min a t 37 "C. The enzyme was inactivated by adding freshly prepared 20 mM iodoacetamide in 10% methanol, incubated for 30 min in the dark at 4 "C, dialyzed against 50 mM sodium acetate buffer, pH 5.5, and concentrated to 2 mg/ml in Centriprep concentrators (Amicon; 10,000 MWCO). The digestion mixture was applied to a 500-mI Sepharose S-100HR (Pharmacia) column (2.5 X 100 cm) equilibrated in the same buffer and eluted a t 2.5 ml/min. Aliquots containing the s C D~~R~ adhesion domain were pooled based on SDS-PAGE and TSK 3000SW HPLC analysis. Deglycosylated sCDZlo5 was prepared by dialyzing 1.0 mg of protein (200 pg/ml) into 0.5 M Tris-HCI buffer, pH 8, and digesting with 10 units of peptide N-glycosidase F (N-glycanase, Genzyme) overnight a t 37 "C followed by purification using a TSK 2000SW HPLC column equilibrated in 50 mM sodium acetate buffer, pH 5.5, containing 200 mM Na2S04. Protein concentrations were determined by Coomassie Blue dye binding assay (Bio-Rad).
Construction of CD2 Mutant and Expression in COS-I Cells-A 1.2-kilobase PstI-Bam-H1 fragment containing the full length s C D~~R~ cDNA was excised from plasmid pAcy373/T11e,2, blunt ended with T4 DNA polymerase and Klenow, and ligated to the large Klenow blunted XbaI fragment from the CDM8 expression vector. Recombinant phagemids were transformed into Escherichia coli CJ236/P3 (dut-ung-) and single stranded phagemid DNA was rescued from these transformants by infection with helper phage M13K07 (33). Oligonucleotide-directed mutagenesis using T7 DNA polymerase was performed as described (34,35). The mutagenic nucleotide (5'-CTATTTAAAsAGGGAACTCTG-3') generates two changes in the WT-CD2 nucleotide sequence at the positions underlined (A-C and T-G), resulting in the substitution of Gln for Asn at amino acid 65 of the mature CD2 polypeptide sequence. COS-1 cells were transfected with wild type CD2, mutant NGs:Q"-CD2, or CDM8 control plasmids as described (36) and assayed for surface expression of CD2 by staining with 20 pg/ml Protein-A purified anti-CD2 polyclonal rabbit antisera (M32B), anti-T1ll mAb (3T48B5), or anti-Tlln mAb (10LD2-4Cl) followed by fluorescein-conjugated goat anti-mouse (Caltag, 1:lOO dilution) or goat anti-rabbit IgG (Tago, 1:40 dilution) as appropriate. For each experiment, 10,000 cells were analyzed by indirect immunofluorescence flow cytometry on and FACScan (Becton Dickinson). SRBC rosetting of COS-1 cells transfected with either human WT-CD2 or N":QG5-CD2 plasmids was performed as described (27).
Electrospray Ionization-Mass Spectrometry-The instrument used in this study was a TSQ-700 instrument (Finnigan-MAT Corp., San Jose, CA) equipped with an electrospray ion source (ESI). Samples were dissolved in water or acetonitrile solutions and analyzed by syringe pump flow injection a t a rate between 5 pllmin and 20 pl/ min directly into the electrospray chamber through a stainless steel hypodermic needle. The voltage difference between the needle tip and the source electrode was -3.5 KV.

RESULTS
Purification and Characterization of sCD2,Az and sCD2105-A cDNA construct previously designed to produce a soluble two-domain human CD2 receptor ( s C D~~~~) encoding the first 182 amino acid residues of the molecule (31) was re-engineered into a new vector for high level expression of ~CD2~82 in Chinese hamster ovary (CHO) cells. Two-domain S C D~~~~ secreted from CHO cells is a mixture of monomeric glycoprotein species with apparent molecular mass between 32 and 36 kDa (Fig. 1, lune 2). The distribution of minor sCD2182 protein bands between 31 and 35 kDa varies depending on the preparation and represent multiple isoelectric glycoforms, since 21 - http://www.jbc.org/ Downloaded from treatment of sCD2,%, with neuraminidase collapses these bands to a single species on isoelectric focusing gels (data not shown). Therefore, one or more of the three predicted Nlinked glycosylation sites at A d s , Asn1I7, and Asn'" contain either hybrid or complex-type glycans endcapped with variable amounts of sialic acid.

14-
Digestion of s C D~~~~ with clostripain produces a homogeneous 105-amino acid residue CD2 adhesion domain ( s C D~~~~) resulting from specific cleavage of s C D~~~~ at the carboxyl terminus of ArgIos and Arg146 (two of the three predicted sites for clostripain specificity) (Fig. 1, lane 3 ) . An additional clostripain cleavage site within the adhesion domain a t Arg4' does not appear readily accessible, as evidenced by the stability of S C D~,~~ toward further degradation. This domain 1 fragment is nearly identical in size to the adhesion domain fragment (Tllpap) previously generated by papain digestion of (22), but does not have a heterogeneous carboxyl terminus found previously with papain-digested Tllpap. Incubation of purified S C D~,~~ with peptide N-glycosidase F reduces its molecular mass from -16 to -12 kDa (Fig. 1, lane  4), consistent with the removal of N-linked glycoconjugates from this site (predicted S C D~,~~ polypeptide Mr = 12,430).
The sCD210a polypeptide sequence contains a single consensus N-linked glycosylation site at and based on SDS-PAGE mobility of multiple preparations of S C D~,~~ derived from the two-domain s C D~, %~ expressed in CHO cells, this site always appears occupied with N-linked glycans. However, deglycosylated S C D~,~~ (~N ' " -S C D~,~~) is completely unreactive in a sandwich ELISA which employs two anti-CD2 mAbs (anti-T1ll, anti-TllJ that recognize distinct functional CD2 adhesion domain epitopes (12)(13)(14)(15)(16)(17)(18)(19) (Fig. 2 A ) . Furthermore, when dNG5-sCD210a is bound directly to ELISA plates and probed separately with either anti-Tl1, or anti-Tllz mAbs, no reactivity is observed with either mAb compared to native ~CD2~05, indicating that both T l l l and T11, epitopes are equally disrupted (Fig. 2, B and c).
Deglycosylated sCD2105 Does Not Bind CD58"The spontaneous aggregation or "rosette" formation between T-lymphocytes and sheep red blood cells (SRBC) originally identified T-lineage cells in the human before the advent of mAbs (37-40). This cell-cell interaction is now known to be the consequence of binding of human CD2 to sheep CD58 (41). In contrast to results obtained with native s C D~~~~ which, as expected, inhibit SRBC rosetting with human T-cells in a dose-dependent manner, purified dN65-sCD2105 fails to bind CD58 on SRBC (Fig. 3). Even with doses of dN65-sCD2105 as high as 100 PM, no inhibition of SRBC rosetting is observed. Therefore, enzymatic removal of the N-linked carbohydrates attached to the single glycosylation site within the human CD2 adhesion domain disrupts binding of dN65-sCD2105 to CD58 and to two independent anti-CD2 mAbs that recognize native surface structures implicated in CD2 activation and adhesion functions (12)(13)(14)(15)(16)(17)(18)(19).
Functional Analysis of a Transmembrane CD2 Glycosylation Mutant-To test whether a transmembrane CD2 receptor lacking N-linked carbohydrate within the adhesion domain could display functional cell-surface epitopes recognized by anti-CD2 mAbs and bind to CD58, a cDNA coding for a fulllength mutant CD2 having a single AsnG5:GlnGs substitution (N"':Q"-CD2) was constructed. This mutant N'j5:Q6'-CD2 cDNA was transfected into COS-1 cells and plates were screened for surface expression of N65:Q65-CD2 by flow cytometry analysis using various anti-CD2 antibodies (Fig. 4A). AS shown by indirect immunofluorescence analysis, approximately 50% of COS-1 cells transiently expressing either wildtype transmembrane CD2 (WT-CD2) or mutant NGs:Q6s-CD2 specifically react with a rabbit polyclonal anti-CD2 antisera (M32B) raised against two-domain S C D~,~~ which recognizes both native and denatured epitopes in SDS-PAGE Western blots. A similar percentage of WT-CD2 COS-1 cells also stain with anti-Tl1, and anti-Tl12 mAbs, with a small population staining brightly with both mAbs. In contrast, no staining of COS-1 cells expressing the mutant N6':QG5-CD2 was observed with either anti-Tl1, or anti-Tll, mAbs. The minor reactivity of the anti-T11, mAb with the mutant N65:Q65-CD2 population represents nonspecific binding as evidence by a similar shift in the anti-T1ll staining of COS-1 cells transfected with CDM8 vector alone (Fig. 4A). Therefore, mutation of the consensus N-glycosylation sequence at AmG5 which then precludes attachment of N-linked carbohydrate at this site does not prevent translocation of Nfi5:Q"-CD2 to the cell surface, but the mutant N":Qfi5-CD2 transmembrane molecules lack reactivity with anti-Tlll and anti-T112 mAbs. Moreover, when COS-1 cells transfected with NG5:Qfi5-CD2 are incubated with SRBC, no rosettes are observed (Fig. 4B). The rosette pattern is indistinguishable from mock transfections with CDM8 vector alone, while parallel analysis of WT-CD2 COS-1 transfectants showed multiple rosettes (Fig. 4B). The latter could be inhibited with micromolar concentrations of soluble S C D~~~. or S C D~, ,~, indicating the adhesion dependence of SRBC rosetting on the CD2-CD58 co-receptor pair (data not shown).

Characterization of the Clycoform Profile of SCD~~OS by ESI-
MS-Since the CD2 domain 1 glycans appear to be necessary for maintaining both native CD2 surface epitopes and CD58 adhesion functions, we characterized the molecular mass and glycoform profile of CHO-derived sCD21os by ESI-MS. Least squares deconvolution fitting of the ESI-MS data (42) in combination with the predicted polypeptide mass of sCD2lon identified a series of five glycoprotein isoforms (glycoforms) which indicate high mannose glycotypes, i.e. ((Man), GlcNAc') where n = 5-9 (Fig. 5). The most abundant glycoform species in this sample correspond to intermediate size high mannose glycotypes, ManBGlcNAc2 (-18%), ManfiGlcNAcn (-31%), and ManTGlcNAcn (-41%). Based on ESI-MS analysis of multiple sCD2105 samples the high mannose glycotype profile varied in their relative percentages; however, these three glycotypes remained the most abundant (>85%). Processed spectra indicate only high mannose glycotype, apparent from peak intervals of m/z 162 Da, with a notable absence of ions indicating hybrid or complex glycotypes (e.g. m/z 365,203, or 291 Da; lactosylamine, hexosylamine, and neuraminyl, respectively). These data suggest that high mannose glycans are the only glycotypes present within the CD2 adhesion domain when two-domain SCD~~R' is expressed in CHO cells. Preliminary ESI-MS evidence has indicated that the second domain of s C D~~R~, which possesses two consensus N-glycosylation sites, contains only complextype, polylactosamine neuraminyl capped glycans.'

DISCUSSION
We report here the purification of a recombinant sCD21os adhesion domain protein expressed in CHO cells, characterize ' €3. Reinhold   . Panel R, COS-1 cells transfected with CDMB control, human WT-CD2, or NG":Q"'-CD2 plasmids were incubated with SRBC and rosetting was performed as described (27). Magnification equals 160-fold. the glycoform profile of the glycan attached to the single Nlinked carbohydrate binding site in SCD~,~,, and examine the functional properties of both transmembrane and soluble CD2 proteins lacking carbohydrate at this site. Analysis of deglycosylated sCD2105 as well as a mutant transmembrane CD2 molecule containing a single Asnfis-Glnfi5 substitution demonstrates that neither binds CD58 or monoclonal antibodies directed at native CD2 adhesion domain epitopes. Electrospray ionization-mass spectrometry demonstrates that high mannose oligosaccharides ((Man),GlcNAcz, n = 5-9) are the only N-glycotypes occupying Amfi5 when recombinant twodomain S C D~~~' is expressed in CHO cells. These data allow us to extend our previous observations concerning the structural and functional properties of CD2 (22) by demonstrating the critical role that N-glycosylation plays in mediating the CD58 binding properties of the CD2 adhesion domain. The absence of high mannose N-glycans at the single N-glycosylation site (Amfi5) within the adhesion domain clearly leads to loss of both structural and functional properties in both recombinant sCD210s and transmembrane CD2 proteins.
Our results are particularly intriguing in light of the recently reported NMR solution structure of rat CD2 domain 1 showing an envelope of multiply charged ion clusters generated from the single glycopeptide. Molecular weight and satellite peak intervals within each ion cluster are a function of charge state and characteristic of the glycan glycotype and glycoform. Since these interval measurements are identical between all ion clusters, pattern recognition algorithms may be applied t o assign glycopeptide mass and glycotype (panel B ) . Panel B, cluster deconvolution of glycopeptide electrospray raw data (panel A ) accomplished by pattern recognition using a relative or level-2 entropy algorithm (37). Processing of the raw data in this manner collapses the spectra shown in panel A to a "root" glycopeptide glycotype and improves signal to noise. The prominent base ion, 1,530 Da larger (43). This receptor fragment was prepared by fusing amino acid residues 1-99 of the rat CD2 sequence to g1utathione-Stransferase and expressing the fusion protein in E. coli, where it fractionated in the soluble lysate of E. coli cells. Rat CD2 domain 1 was judged to be folded correctly after cleavage from its fusion partner by its reactivity with distinct mAbs raised against native rat CD2. Multidimensional NMR spectra obtained on this protein fragment yielded data permitting calculation of polypeptide structures that resemble an immunoglobulin (Ig) p-fold, similar to human CD4 domain 1 (44,45). Rat CD2 domain 1 contains three consensus N-linked carbohydrate binding sites and two are known to be glycosylated in vivo (30), but occupancy of these sites with N-glycans does not appear to be required for formation of native-like structure in the rat molecule.
However, the structural and functional intregrity of the human CD2 adhesion domain is clearly dependent on the presence of N-linked carbohydrate at A d 5 as evidenced by our data presented here. It is therefore not surprising that we were unable to prepare a functional human sCD2 protein when cDNAs coding for the CD2 adhesion domain were originally expressed in bacterial cells. When CD2 domain 1 was produced a t high levels as an E. coli intracellular protein, it aggregated into inclusion bodies which were readily isolated and the protein subsequently purified to homogeniety from urea-solubilized pellets via sequential chr~matography.~ Nevertheless, the E. coli produced CD2 domain 1 protein remained completely unreactive toward anti-CD2 mAbs in both sandwich and antibody capture ELISAs and did not bind CD58 in SRBC rosetting assays. Both s C D~~~~ and sCD2105 purified from CHO cells can be reversibly renatured after exposure to strong chaotropic agent^.^ However, attempts to renature functional binding activity of the E. coli produced human CD2 domain 1 using identical renaturation protocols were unsuccessful. We take these data as additional confirming evidence that N-glycosylation in the human CD2 adhesion domain plays a critical role both in forming and maintaining a functional CD58 binding site.
There are numerous examples whereby oligosaccharides are required for proper folding, transport, and biological function of either secreted glycoproteins or transmembrane glycoprotein receptors (reviewed in Refs. 46 and 47). Early studies with the transmembrane G protein of vesicular stomatitis virus demonstrated that either blockage of N-glycosylation with tunicamycin, elimination of N-glycosylation sites by sitedirected mutagenesis, or generation of variants containing novel glycosylation sites resulted in aggregation of nascent vesicular stomatitis virus-G chains in the endoplasmic reticulum and severe impairment of their intracellular transport (48-51). It has also been recently shown that expression of functional human CD4 on the cell surface requires glycosylation at either one of the two N-linked sites within the third Ig-like domain (52, 53). Mutations which eliminate both consensus N-glycosylation sites renders CD4 transport incompetent and improperly folded CD4 (as judged by mAb reactivity) is retained in the endoplasmic reticulum. However, our data indicates that elimination of N-glycosylation within the human CD2 adhesion domain alone does not inhibit surface expression of mutant N65:Q65-CD2, but the CD2 adhesion domain lacking N-carbohydrate is clearly non-native in conformation in either its transmembrane or soluble form.
Carbohydrates also serve directly as molecular determinants responsible for mediating cell-cell adhesion and lymphocyte trafficking via binding to the selectin family of adhesion receptors (54). ELAM-1 regulates adhesion of leukocytes to vascular endothelium by recognition of Sialyl-LeX (sialyl-Lewis X) (55,56), a carbohydrate ligand found on cell-surface glycoprotein and glycolipid groups of neutrophils. In contrast, N-linked glycosylation within domain 3 of ICAM-1 appears to specifically shield the ligand binding site for the leukocyte integrin MAC-1 (CDlS/CDllb) in that reagents which interfere with N-carbohydrate biosynthesis (or mutations that eliminate N-glycosylation sites in ICAM-1) enhance the binding of ICAM-1 to MAC-1 (57).
Based on the assumption (given -50% amino acid identity between rat and human CD2) that the three-dimensional structure of human CD2 domain 1 resembles its rat CD2 counterpart, Ig P-strand folding patterns for human CD2 were proposed (Fig. 6) (43). Preliminary data obtained on human CD2 domain 1 ( S C D~,~, ) by multidimensional NMR spectroscopy indeed suggests that the CD2 adhesion domain adopts an overall conformational structure characteristic of an Igfold? Taken together, these observations suggest that the Nglycans attached to A S I~~ project outward from a tight loop connecting p-strands D and E of a classical Ig P-sandwich Circles indicate Ig @-sandwich domain 1 ( D l , residues 1-105) and domain 2 ( 0 2 , residues 106-182), and the bull-stick symbols depict N-linked carbohydrate structures. The 4 cysteine residues are linked by two intradomain disulfide bridges, but the S-S pairings for these disulfide bonds have not been definitively mapped. Panel B , primary sequence of one-domain sCD2105 and two-domain sCDZIR2. The bars represent putative @-strand assignments predicted for human CD2 based on homology with the rat CD2 structure (43). The shaded residues indicate the position of the three N-linked glycosylation sites a t Asn'", Asn1l7, and Am"'. Thus, in this regard the high mannose N-glycans would not serve directly as a ligand in mediating CD58 binding but could influence the integrity of the CD58 binding site by some, as yet undefined, long-range stabilization effects. Alternatively, these high mannose N-glycans might project upward from the top of the D/E loop and stabilize flexible turn regions between either P-strands B and C or possibly C' and C" via hydrogenbonding with specific amino acid side chain residues within these loops. Based on our data, the presence of this N-glycan is clearly required for both the formation and maintenance of native structure. The precise role that N-linked carbohydrate plays in mediating the conformational stability of the human CD2 adhesion domain, and thereby influencing molecular recognition of CD58, may be clarified once the three-dimensional structure of human CD2 is solved.