REVIEWS (1995) Biechem. Biophys. Res. Commun. 214, 576-581

37 Muscatelli, F. et al. (1994) Nature 372, 672-676

38 Bardoni, B. et al. (1994) Nat. Genet. 7 ,497-501 39 King, V. et al. (1995) Curr. Biol. 5, 37-39 40 Ryner, L. C. and Swain, A. (1995) Cell 81,

483-493

41 Sockanathan, S., Cohen-Tannoudji, M., Colignon, J. and Lovell-Badge, R. (1993) Genet. Res. 61, 149

42 Kamachi, Y. et al. (1995) EMBO J. 14, 3510-3519

43 Denny, P. et al. (1992) EMBO J. 11, 3705-3712 44 van de Wetering, M., Oosterwegel, M.,

van Norren, K. and Clevers, H. (1993) EMBO J.

TIBS 2 1 - AUGUST1996

12, 3847-3854 45 Affara, N. A. et al. (1993) Nucleic Acids Res. 2,

785-789 46 Harley, V. R. et al. (1992) Science 255,453-456 47 Braun, A. et al. (1993) Am. J. Hum. Genet. 52,

578-585 48 Zeng, Y. et al. (1993) J. Med. Genet. 30, 655--657 49 Pou~at, F. et al. (1994) Hum. Mutat. 3, 200-204

Why mammalian cell surface proteins are glycoproteins

Carl G. Gahmberg and Martti Tolvanen Most proteins presented at the external surface of mammalian cells con- tain carbohydrate. The reason for this is not fully understood, but recent work has shown that such carbohydrate has two major functions. Inside the cell, it helps proteins fold and assemble correctly in the endoplasmic reticulum, and it might also act as a signal for the correct migration of glycoproteins. Outside the cell, it provides specific recognition structures for interaction with a variety of external ligands.

EARLY WORK IN the 1960s on the mor- phology of mammalian cells showed that the external surface of the plasma membrane is rich in carbohydrate, whereas the inner side is devoid of con- ventional-type oligosaccharides 1. Until recently, it was unclear whether the carbohydrate was confined to few or many different cell surface glycoconju- gates 2, and the development of radio- active techniques in particular has al- lowed the carbohydrate portions of exposed cell surface glycoproteins and glycolipids to be labeled 3,4. It was then possible to study the larger number of glycoconjugates specifically presented at the surface of various cells. Also, it became apparent that cell membranes contain a multitude of glycoproteins, many more than previously thought 5,6.

In 1976, after studying human eryth- rocytes and other cells, it was proposed that cell surface proteins are always glycoproteins s. A similar proposal was made independently by Bretscher and Raff 2. As more and more mammalian cell membrane proteins and the genes

c. G. Gahmberg and M, Tolvanen are at the Department of Biosciences, Division of Biochemistry, P.O. Box 56, Viikinkaari 5, FIN-O0014, University of Helsinki, Finland.

308

encoding them have been characterized, cloned and sequenced, this proposal has turned out to be largely correct.

Most glycoproteins are N-glycosylated, i.e. they contain asparagine-linked oligo- saccharides located at the peptide se- quence(s) NxS/T (where x stands for any amino acid except for proline) at the external aspect of the membrane. Some membrane proteins are O-glycosylated, with the carbohydrate chains attached to serine or threonine residues, which are often clustered in distinct regions of the polypeptides. Most O-glycosylated pro- teins also contain one or more N-glyco- sidic oligosaccharides 7. The importance and requirements for O-glycosylation have yet to be elucidated.

Exceptions to the rule Early searches of the literature for un-

glycosylated surface proteins in mam- malian cells met with little success. How- ever, in 1982, the human red cell Rh(D) (Rhesus) protein, with an apparent mol- ecular weight of 30-32 kDa, was identi- fied s,9. Importantly, no evidence for the presence of carbohydrate was found 1~ and subsequent cloning and sequencing of its cDNA and that of other polypep- tides belonging to the Rh-blood group system, showed that they do indeed lack N-glycosylation sequences u32.

�9 1996, Elsevier Science Ltd

Thus, this protein seemed to be an exception to the glycosylation rule.

However, more recent work has shown that this is not the case. There is now evidence that the Rh-polypeptides form part of a large glycopolypeptide complex, including among others the Rh50 glycoproteins and the Landsteiner- Wiener (LW) blood group glycoprotein (intercellular adhesion molecule 4) (Refs 13, 14). This situation is similar to that of ~2-microglobulin in class I trans- plantation antigens, where the unglyco- sylated protein associates with the heavy chains of the transplantation antigens. The importance of the associ- ation of the Rh-polypeptide with other glycosylated proteins is underscored by the fact that it has not yet been poss- ible to express the Rh cDNA in any mammalian cell expression system.

In a recent survey of the SWlSS-PROT database (release 33.0, April 1996) we found 1823 complete animal protein entries with reported extracellular fea- tures, of which 1671 (91.7%) were described as 'glycoproteins' in the key- word field; 1630 of these 1671 contained the N-glycosylation peptide sequence NxS/T. The remaining 8.3%, representing 152 potentially non-glycosylated plasma- membrane proteins, contained 116 pro- teins with multiple transmembrane re- gions, 15 proteins that are known to associate with glycosylated subunits in a complex such as CD3 chains and the Rhesus D-polypeptide, and seven that contained 5-38 potential N-glycosylation sites, i.e. polypeptides highly likely to be glycosylated, yet not marked as glyco- proteins. This leaves only 14 sequences (0.7%) that are candidates for non- glycosylated, non-complexed plasma membrane proteins with a single trans- membrane domain.

In another survey, we assessed whether this high representation of the N-glycosylation tripeptide sequence is more than should be found by chance alone*. To do this, we extracted all se- quence features marked as extracellu- lar domains from animal proteins in SWlSS-PROT 33.0, which resulted in 4259 stretches of sequence from 1933

PII: S0968-0004(96) 10034-7

REVIEWS TIBS 2 1 - AUGUST1996

proteins. By chance alone, this material is expected to contain 3343 potential N-glycosylation sites, but it actually con- tains 6725 sites, an over-representation of more than twofold. This high frequency of potential N-glycosylation sites in extra- cellular domains might reflect recent gene duplication and shuffling events as well as a possible evolutionary pressure to enrich for glycosylation sites.

For comparison we analysed all re- ported c~oplasmic sequences in ani- mal proteins in the same way. This ma- terial contained 1656 occurrences of the N-glycosylation sequence versus 1727 expected occurrences.

Glycosylation is essential in the endoplasmic reticulum

It has been difficult to understand why the glycosylation machinery, especially that of N-glycosylation, is so remarkably complex is. Is Nature wasteful? Briefly, dolichol-containing glycolipids are used to donate the initial glucose3-mannose 9- N-acetylglucosamine 2 oligosaccharide to asparagine residues in the lumen of the endoplasmic reticulum ~R) ~ig. 1). The peripheral glucose residues are subse- quently removed by cr I and II, followed by c~-mannosidase re- moving e-mannosyl residues; the final oligosaccharides ~ig. 2) are formed by the action of various glycosyltransfer- ases. If the protein is still unfolded after removal of the glucose residues, re- glucosylation takes place by the ER en- zyme UDPglucose: glycoprotein glucosyl transferase 16 ~ig. 1). Importantly, this enzyme acts only on denatured or un- folded substrates. The subsequent glycosylation reactions take place dur- ing the transport of the glycoproteins through the ER and Golgi apparatus en

route to the cell surface. Using glycosylation mutants, Stanley

and co-workers have convincingly shown that, whereas N-glycosylation is essen- tial for the viability of cells, hybrid and complex oligosaccharides are not an absolute necessity ~7. They are, how- ever, required for the whole organism to develop normally ]7. Thus, mice with inactivated N-acetylglucosamine trans- ferase ] die at mid-gestational age. This

*If the probability of a given tripeptide being an N-glycosylation sequence (N-gs) is Prigs, a sequence of n residues is considered as n - 2 tripeptides and a binomial distribution is assumed for the number of glycosylation sites in this population of tripep- tides, the expectation value of the number of glyco- sylation sites, p. = P~s• for N sequences of total length L. The assumption of a binomial distri- bution is fair when P~gs is small, but it gives a slight overestimation for the number of sites.

transferase is the key enzyme in the initiation of complex- and hybrid-type N-linked oligosaccharide biosynthesis.

It is still poorly under- stood how the biosynthesis of membrane oligosaccha- rides is regulated. Evidently, the activity of the glycosyl transferases and also the availability of nucleotide sug- ars are of key importance. The following example might illustrate this: human En(a-) red cells lack the gene that encodes for glycophorin A, the major er~hroc~e sialo- glycoprotein 18, but the cell compensates for this loss by making a larger-than-normal Band 3 (anion transport pro- tein) oligosaccharide of the polylactosamine type. Indi- viduals heterozygous for the glycophorin A defect (containing 50% the normal amount of glycophorin A) synthesize a Band 3 oligo- saccharide with a size be- tween that of normal and En(a-) cells 18. Although the carbohydrate chains in this example are profoundly different, the findings in- dicate that there is competition for acti- vated sugars, and the lack of one major polypeptide acceptor gives others the opportunity to get more carbohydrate.

Calnexin and calreticulin The realization that calnexin and cal-

reticu]in, both ER proteins and molecular chaperones, recognize and bind to the

Glucosidase II

GIc Glucosidase I ~ I c~1-2

GIc UDP-GIc: glycoprotein I c~1-3 / glucosyltransferase

G]c ~-/ i o~1-3

Man { z l -2 I

Man ~1-2l

Man ~1 - 3 \

Man Man ~1-2l I {x1-2

Man Man ~ 1 - 3 \ / (zl-6

Man / c~1-6

Man I 131-4

GIcNAc I p l - 4

GIcNAc IP

Figure 1 Structure of untrimmed N-glycosidic oligosaccharide. This oligosaccharide is later modified in the endoplas- mic reticulum (ER) by removal of the peripheral glu- cose by glucosidase I and of the other two glucose residues by glucosidase I1. The innermost glucose residue is essential for the interaction with calnexin and calreticulin. UDPglucose: glycoprotein glucosyl- transferase, which acts on unfolded proteins, can restore it. Abbreviations used: GIc, D-glucose; GIcNAc, N-acetyl-o-glucosamine; Man, D-mannose.

carbohydrate portions (specifically to the innermost glucose residue) of newly synthesized glycoproteins is of funda- mental importance ~9-23. Whereas the biosynthesis of the polypeptides is rela- tively quick, their subsequent folding and the association of subunits are much slower processes. By binding to newly synthesized glycopolypeptides, calnexin anchors the polypeptides in the ER until they have achieved their

SA (z2-3 I

Gal Gal #1-41 pl-41

GIcNAc GIcNAc p1-2 I pl-21

Man Man o~1-3 ~ / o ~ 1 - 6

Man 1#1-4

GIcNAc 1~1-4

Fuc c~1-6 GIcNAc

(a)

(Man) o~1-2 I

(Man) (z.1-2 I

Man c~1-3

(b)

(Man) (Man) c~1-21 1~1-2

Man Man c ~ 1 - 3 ~ / ' o d - S

Man / c r

Man I P1-4

GIcNAc IP l -4

GIcNAc

Figure 2 (a) Complex-type and (b) high-mannose-type N-glycosidic oligosaccharides. Many complex-type oligosaccharides are important for cell-ligand interactions, but also high-mannose-type structures can function as lectin and microbial ligands. Abbreviations used: Fuc, L-fucose; Gal, D-galactose; GIc, D-glucose; GIcNAc, NacetyI-D-glucosamine; Man, D-mannose; SA, sialic acid.

309

REVIEWS TIBS 2 1 - AUGUST1996

SA 1.2-6

SA ~_~3 Gal ~1-3 GIcNAc ~

SA.2-3 Gal ~1-4 GIcNAc I pl-6

SA ~ 3 Gel pl-3 GIcNAc ~

Figure 3 Structures of Oglycosidic oligosaccharides. O-glycosidic oligosaccharides are found in mucin-type glycoproteins, often in large numbers. The multi-valency is probably important in increasing their avidity for various interactions. A large number of sugar chains might also have protective functions against proteolysis. Abbrevi- ations used: Gal, D-galactose; GalNAc, N-acetyl-D-galactosamine; SA, sialic acid.

correct folding conformation and, where appropriate, associated with other poly- peptides into supramolecular complexes. However, if deglycosylation is pre- vented by glucosidase inhibitors like castanospermine and deox3mojirimycin, the proteins remain bound to calnexin and calreticulin, their transport is de- layed in the ER and the proteins are subsequently degraded. Treatment with tunicamycin, which completely inhibits the N-glycosylation machinery, blocks calnexin and calreticulin interactions with their substrates, resulting in in- correctly folded polypeptides, protein degradation and mislocalization of newly synthesized proteins. These find- ings do not exclude the fact that other molecular chaperones such as BiP Oinding protein) are also important.

In addition to these glycoprotein chaperones, a few proteins have been identified that can act as intracellular

lectins (chaperones?) later in the bio- synthetic pathway. These include the ERGIC-53 protein and VIP36, which are homologous to plant lectins and can bind high-mannose type oligosaccharides 24.

Little is known about the importance of O-glycosylation, but because those cell surface proteins that are O-glycosylated Gig. 3) are often heavily so, they in- fluence strongly the three-dimensional structure of these mucin-type membrane glycoproteins. Furthermore, O-glycosyl- ation efficiently protects such proteins from proteolysis. Olinked oligosaccha- rides might extend surface proteins into rod-like structures, which could be important in mediating or preventing cell-cell interactions. Moreover, carbo- hydrate--carbohydrate interactions 25 can be important during various stages of glycoconjugate biosynthesis, but little is lmown about their significance.

Exit from the cell It has been argued that plasma mem-

brane glycoproteins do not need any 'exit signal', but migrate by bulk flow without specific retention. This would mean that glycoproteins remain at the same concentration in the transport vesicles during intracellular migration. Wieland et al. 26 used an 1251-labeled, formylated N-glycosylation consensus se- quence tripeptide (NYS), which was taken up by intact cells, glycosylated and then secreted in 5-10 min. Whether this sequence employed the physiological migration route through the Golgi appa- ratus is not known. The fast, intracellu- lar migration by this simple molecule was interpreted to mean that secretion involves neither any specific retention signals, nor any signals that would target the tripeptide to the cell surface.

TA SI"ING byDr nWnsfHra;aBuSx

However, one could also interpret the results to suggest that the oligosac- charide of the N-glycosylated tripeptide in fact gives the molecule a positive migration signal. Support for such a function of oligosaccharides has been obtained using a recombinant chimeric membrane protein based on rat growth hormone. The soluble protein is nor- mally not glycosylated, and when a membrane-anchored form was engi- neered it remained intracellular 27. How- ever, the introduction of N-glycosylation sites resulted in its glycosylation, and presentation at the cell surface. Inter- estingly, the same glycosylated protein showed apical sorting in Madin-Darby canine kidney cells, whereas the non-glycosylated growth hormone was secreted from both the apical and basolateral membranes 28.

Oligosaccharide function at the cell surface A number of carbohydrate-specific

functions occur at the cell surface, many of which involve recognition events. These include cell adhesion, interactions between cells and soluble ligands, and between cells and various microbes. Such interactions often involve carbo- hydrate-binding lectins.

Although the presence of mammalian cell lectins has been known for a num- ber of years, their importance was ini- tially largely neglected. This was owing to the fact that rather few specificities were found, and the lectins were con- sidered important only in a few special cases. Only a few examples of specific carbohydrate-plasma membrane glyco- protein interactions can be mentioned here, and the reader is referred to more extensive recent reviews for additional reading29, 30.

The finding of a hepatoc~e lectin for de-sialylated serum glycoproteins opened the field 31. This lectin, presented at the surface of liver epithelial cells, binds to serum glycoproteins that have lost ter- minal sialic acids, resulting in exposure of galactosyl/N-acetylgalactosaminyl resi- dues. A corresponding macrophage ac- tivity involving binding of mannose- containing proteins was subsequently described. But more widespread inter- est in the importance of glycoprotein oligosaccharides arose when leukoc~e adhesion was found to involve initial carbohydrate-ligand interactions. Sev- eral studies have shown that 'rolling' of neutrophils and monocytes along capil- lary endothelia results from reversible interactions between selectins presented on endothelial cells and leukocytes, and

310

REVIEWS TIBS 2 1 - AUGUST1996

specific carbohydrate structures on the counter-ligand cells. These selectins 32,33 bind to sialyl Le x, sialyl Le a (Fig. 4), sul- fatides and related carbohydrate struc- tures. L-selectin is found on nucleated blood cells, whereas E- and P-selectins are mainly found on endothelial cells (P-selectins are also found on platelets).

Interestingly, many selectin receptors are cell surface proteins rich in O-glyco- sidic oligosaccharides. These include glycosylation-dependent cell adhesion molecule 1 (GlyCAM-1), CD34, mucosal vascular addressin cell adhesion mol- ecule 1 (MAdCAM-1) and P-selectin glycoprotein ligand 1 (PSGL-1) (Ref. 34). The selectin-mediated 'rolling' phen- omenon is essential for subsequent stronger binding, and for tissue migra- tion of leukocytes involving integrins and ligands of the immunoglobulin superfamily [intercellular adhesion mol- ecule (ICAM) and vascular cell adhe- sion molecule (VCAM)].

Additional leukocyte carbohydrate- binding proteins have recently been de- scribed; among them is the B lympho- cyte surface protein CD22, which binds ~2-6 sialic acid-galactosyl structures, and the macrophage sialo-adhesin with specificity for u2-3 sialic acid-containing glycoconjugates 35. Of particular interest is the binding of sperm to zona pellucida glycoproteins through cr resi- dues 33, although this field is still contro- versial and is probably more complex than currently thought.

However, it is anticipated that this list of important surface glycoproteins will grow over the next few years, along with the elucidation of their binding speci- ficities. In fact, the carbohydrate struc- tures of only a few cell surface glyco- proteins are currently known. This is, of course, largely owing to the fact that it is difficult to purify native glycoproteins from mammalian cell membranes in suf- ficient quantities. However, it is already evident that cell surface proteins origi- nating from the same cell can have very different oligosaccharide compositions. On one hand, CD45, a major cell surface glycoprotein of leukocytes, which con- tains tyrosine phosphatase activity on the inner aspect of the membrane, is enriched in ~2-6 sialic acid-galactosyl residues. On the other hand, the leuko- cyte CD11/CD18 integrins contain ct2-3 sialic acid-galactose structures, but no ~2-6 sialic acid. These facts are reflected in their binding specificities. Thus, CD22 binds to CD45 (and to some other pro- teins), but not to the integrins, whereas the opposite is true for E-selectin 36.

Currently, a large number of binding specificities of cell surface carbohydrate are known for unphysiologi- cal ligands such as bacteria, viruses, toxins etc. Although clinically important, these activities are obviously not physiological, but reflect evolutionary adaptations of various microbes.

Concluding remarks The main purpose of this

short review is to point out the requirement of cell sur- face proteins for carbohy- drate. For successful biosynthesis, fold- ing and intracellular migration, cell surface proteins (and most secreted proteins) need to be glycosylated or, we postulate, linked to a glycosylated protein. The few exceptions to this rule include cell surface proteins that span the membrane several times. Intra- cellular lectins such as calnexin and cal- reticulin are responsible for retaining glycoproteins on the ER until the time is right for their migration to the mem- brane, but carbohydrates might also be important as positive plasma membrane signals. However, many of the specific functions of mature cell surface glyco- protein oligosaccharides are physio- logically important, but not always essential for the protein function.

Although this research area is still largely in its infancy, it is rapidly devel- oping. Whether most (or all) carbo- hydrate structures present at the cell surfaces eventually will turn out to be important in interactions with sur- rounding cells, soluble ligands and infecting microbes remains to be seen.

Acknowledgements The original research from the au-

thors' laboratory was supported by the Academy of Finland, the Sigrid Jus61ius Foundation and the Finnish Cancer Society. We thank K. Simons (European Molecular Biology Laboratory) for use- ful comments on the manuscript, and Yvonne Heinil~ for secretarial assistance.

References 1 Rambourg, A., Neutra, M. and LeBIond, C. P.

(1966) Anat. Rec. 154, 41-71 2 Bretscher, M. S. and Raft, M. C. (1976) Nature

258, 43-49 3 Gahmberg, C. G. and Hakomori, S. (1973)

J. Biol. Chem. 248, 4311-4317 4 Gahmberg, C. G. and Andersson, L. C. (1977)

J. Biol. Chem. 252, 5888-5894 5 Gahmberg, C. G. (1976) J. Biol. Chem. 251,

510-515 6 Gahmberg, C. G., H~yry, P. and Andersson, L. C.

(a) SA (b) ] q.2.--3

Gal Fuc I~1--4 I c~1--4

Fuc ~-3 GIcNAc SA ~2-3 Gal ~-3 GIcNAc I I

Sialyl Le x Sialyl Le a

Rgure 4 Structures of the selectin ligands (a) sialyl Le x and (b) sialyl Le a. These oligosaccharides are important ligands for selectins and are found as terminal structures in several different oligosaccharides, both in glycoproteins and glycolipids. Abbreviations used: Fuc, L-fucose; Gal, D-galactose; GIcNAc, N-acetyl-D- glucosamine; SA, sialic acid.

(1976) J. Cell Biol. 68, 642-653 7 Kornfeld, R. and Kornfeld, S. (1976) Annu. Rev.

Biochem. 45, 217-237 8 Gahmberg, C. G. (1982) FEBS Lett. 140, 93-97 9 Moore, S., Woodrow, C. F. and McClelland, D. B. L.

(1982) Nature 295, 529-531 10 Gahmberg, C. G. (1983) EMBO J. 2, 223-227 11 Cherif-Zahar, B. et al. (1990) Proc. Natl. Acad.

Sci. U. S. A. 87, 6243-6247 12 Avent, N. D., Ridg~vell, K., Tanner, M. J. A. and

Anstee, D. J. (1990) Biochem. J. 271, 821-825 13 Moore, S. and Green, C. (1987) Biochem. J.

244, 735-741 14 Agre, P. and Cartron, J-P. (1992) In Protein

Blood Group Antigens of the Human Red Cell (Agre, P. and Cartron, J-P., eds), pp. 20-52, Johns Hopkins University Press

15 Kornfeld, R. and Kornfeld, S. (1985) Annu. Rev. Biochem. 54, 631-664

16 Parodi, A. J., Mendelzon, D. H., Lederkremer, G. Z. and Martin-Barrientos, J. (1984) J. Biol. Chem. 259, 6351-6357

17 Stanley, P. and Ioffe, E. (1995) FASEB J. 9, 1436-1444

18 Gahmberg, C. G. et al. (1976) J. Biol. Chem. 251, 6108-6116

19 Degen, E. and Williams, D. B. (1991) J. Cell Biol. 112, 1099-1115

20 Ou, W-J., Cameron, P. H., Thomas, D. Y. and Bergeron, J. J. M. (1993) Nature 364, 771-776

21 Hammond, C., Braakman, I. and Helenius, A. (1994) Proc. Natl. Acad. Sci. U. S. A. 91, 913-917

22 Hebert, D. N., Foellmer, B. and Helenius, A. (1995) Cell 81, 425-433

23 Fiedler, K. and Simons, K. (1995) Cell 81, 309-312

24 Fiedler, K. and Simons, K. (1994) Cell 77, 625-626

25 Eggens, I. et al. (1989) J. Biol. Chem. 264, 9476-9484

26 Wieland, F. T., Gleason, M. L., Serafini, T. A. and Rothman, J. E. (1987) Cell 50, 289-300

27 Guan, J-L., Machamer, C. E. and Rose, J. K. (1985) Cell 42, 489-496

28 Scheiffele, P., Per~nen, J. and Simons, K. (1995) Nature 378, 96-98

29 Paulson, J. C. (1989) Trends Biochem. Sci. 14, 272-276

30 Varki, A. (1993) Glycobiology 3, 97-130 31 Ashwell, G. and Harford, J. (1982) Annu. Rev.

Biochem. 51, 531-554 32 Bevilacqua, M. P. and Nelson, R. M. (1993)

J. Clin. Invest. 91, 379-387 33 Gahmberg, C. G., Kotovuori, P. and Tontti, E.

(1992) Acta Pathol. Microbiol. Immunol. Scand. 100, 39-52

34 McEver, R. P., Moore, K. L. and Cummings, R. D. (1995) J. Biol. Chem. 270, 11025-11028

35 Kelm, S. et al. (1994) Curr. Biol. 4, 965-972 36 Kotovuori, P. et al. (1993) Glycobiology 3,

131-136

311