1076 Dispatch

Protein secretion: Sorting sweet sortingSreenivasan Ponnambalam* and George Banting†

Membrane-spanning, lectin-like proteins in theeukaryotic secretory pathway seem to operate quality-control checkpoints by fine tuning protein exit orretention within each subcompartment.

Addresses: *The Brunskill Laboratory, Department of Biochemistry,Medical Sciences Institute, University of Dundee, Dundee DD1 4HN,UK. †Department of Biochemistry, Bristol University Medical School,University Walk, Bristol BS8 1TD, UK.

Current Biology 1996, Vol 6 No 9:1076–1078

© Current Biology Ltd ISSN 0960-9822

How are proteins sorted to different destinations in eukary-otic cells? Until recently, it was generally thought that theconstitutive flow of proteins to the cell surface occurs bydefault, and that proteins simply accumulate carbohydratemodifications as they pass sequentially through each com-partment along the secretory pathway [1]. This hypothesiswas supported by studies showing that inhibition of N-linked glycosylation by drugs or genetic defects in certaincarbohydrate modification enzymes did not stop proteinsecretion [1–3]. It is becoming evident, however, thatcytosolic and lumenal machinery play a crucial role indetermining protein flow through the secretory pathway[4]. The finding that membrane-spanning lectins residentin compartments along this route appear to regulateprotein transport between successive compartments is animportant new development [5–9].

The Golgi apparatus lies at the heart of the secretorypathway, and is a crucial sorting station for proteins des-tined for the cell surface and other intracellular compart-ments [10]. It comprises a central stack of fenestratedmembranes abutted by tubulo-vesicular subcompartmentsthat contain unique mixtures of resident enzymes andproteins. The intramolecular signals involved in the local-ization of proteins to the Golgi apparatus and other com-partments in the secretory pathway are gradually beingdefined [11]. With the exception of those enzymes thathave been shown to be involved in post-translational mod-ification, however, the precise functions and mechanismsof action of many of these proteins remain obscure. Oneinteresting clue to their function is the finding that thelumenal domains of many such proteins bind to specificcarbohydrate structures.

Elegant studies by Helenius, Parodi and co-workers [5–7]showed that the endoplasmic reticulum (ER) resident pro-teins calreticulin and calnexin bind nascent, monoglucosy-lated glycoproteins as part of the process which ensuresthat only correctly folded proteins are allowed to exit the

ER. Thus, calreticulin and calnexin behave as ‘retention’lectins by binding newly synthesized, transiently glucosy-lated proteins and preventing entry into ER-derived trans-port vesicles. By contrast, other integral membraneproteins, such as the well-characterized mannose-6-phos-phate receptor, behave as lectin-like ‘cargo receptors’. Themannose-6-phosphate receptor binds glycoproteins bearingappropriately modified mannose residues on their N-linkedglycans and delivers them to pre-lysosomal compartmentsfrom the trans-Golgi network [2].

New lectinsMore recently, evidence has emerged that other proteinslocalized to particular membranes along the secretorypathway may also act as lectins. One important biochemi-cal marker of the early stages of the secretory pathway isERGIC-53, a 53 kDa membrane protein predominantlylocated in the ER–Golgi intermediate compartment(ERGIC) [12]. The isolation and characterization of cDNAclones encoding the human, murine and Xenopus laevisorthologues of ERGIC-53 showed it to be a type I integralmembrane protein that contains intra-molecular signalsresponsible for its appropriate location [12,13]. Simultane-ous characterization and cloning of a human mannose-binding protein led to the surprising discovery that it isidentical in sequence to ERGIC-53 and has some similar-ity to certain animal and plant lectins [14]. Subsequentbiochemical characterization of ERGIC-53 confirmed itslectin properties, showing an affinity for mannose sugarsand a two-fold lower affinity for glucose and glucosamineresidues [15].

Intriguingly, this ability to bind mannose residues isdependent on Ca2+ concentration, suggesting that differ-ences in Ca2+ concentrations between ER and post-ERcompartments may regulate the binding and release ofligands by ERGIC-53 [15]. Other modulators of ERGIC-53 interactions with terminally mannosylated glycopro-teins might be mannosidases I and II. These enzymesremove terminal mannose residues from the core N-linkedoligosaccharide units attached to glycoproteins early in thesecretory pathway. Thus, an integral membrane lectin thatrecycles between early compartments along the secretorypathway [13] might function as a cargo receptor by selec-tively binding glycoproteins containing terminal mannosesugars; removal of these residues by mannosidases I and/orII would help to release proteins from ERGIC-53 intransit. Importantly, mannosidases I and II remove a1→2linked mannose residues. This suggests that ERGIC-53could discriminate between the a1→2 and a1→6 linkedmannose sugars on N-linked carbohydrate structures.

VIP36 is another type I integral membrane protein withlimited homology to plant and animal lectins (includingERGIC-53). VIP36 is a constituent of post-trans-Golginetwork vesicle membranes [16] and has a lumenaldomain that binds glycopeptides bearing galactose units[9]. This binding is dependent on divalent cations [9], sug-gesting that VIP36 binding and release of ligands in vivomay be dependent on terminal modification of carbohy-drate structures and Ca2+ levels. Interestingly, vesicle bio-genesis in the trans-Golgi network and fusion with targetmembranes depend critically on Ca2+ concentrations [17].VIP36 binding to glycopeptides is inhibited by N-acetyl-galactosamine, and Fiedler and Simons [9,18] suggest thatVIP36 may sort O-glycosylated proteins, which contain N-acetylgalactosamine residues, at the point of exit from thetrans-Golgi network.

Most O-glycosylation, however, results in blockage or sub-stitution of N-acetylgalactosamine with glucose, N-acetyl-glucosamine or other sugars, leaving few, if any, terminalN-acetylgalactosamine residues. One possible function ofVIP36, therefore, is to recycle incorrectly or incompletelyglycosylated proteins containing free, terminal N-acetyl-galactosamine residues to earlier compartments for thecorrect carbohydrate modifications to occur before secre-tion. ERGIC-53 may also function to retrieve incorrectlyglycosylated proteins along the secretory pathway. Alter-natively, both proteins might function as ‘cargo receptors’for the transport of correctly glycosylated proteins inanterograde transport vesicles. To test these hypotheses, itwill be necessary to investigate further the ligand-bindingproperties of these lectins.

Quality controlThe finding that lectins with apparently different carbo-hydrate specificities are located in specific membranesalong the secretory pathway suggests an additional level ofquality control on protein secretion. Cytosolic and lumenalmachinery may effect protein secretion by a process akinto that of distillation [4]; however, resident lectins withineach compartment may further regulate this process bytransient interactions with incorrectly glycosylated and/orfolded proteins. The net effect would be to ensure effec-tive secretion or exit only of correctly glycosylated pro-teins; partially or incorrectly glycosylated proteins thatreach or escape from a specific compartment would bebound by a specific ‘retention’ or ‘retrieval’ lectin untilthey receive the necessary modification by enzymes inthat compartment.

This hypothesis seems to be at odds with studies on well-known mutant Chinese hamster ovary (CHO) cell linesdefective for one or more N-linked glycosylation enzymes,which show that aberrantly N-glycosylated proteins appearto be secreted normally [3]. One explanation might bethat anterograde transport of glycosylated proteins occurs

even when transient interactions with ‘retention’ or‘retrieval’ lectins slows the net rate of transport through thesecretory pathway. Alternatively, lectin-mediated retentionor retrieval mechanisms may be more important for O-gly-cosylated proteins. Some CHO mutant cell lines generatedby Krieger and co-workers [19,20] are defective in both N-and O-linked glycosylation. Careful examination of the rateof secretion and transport of a variety of N- and O-glycosy-lated soluble and membrane proteins in these cell linesmay yield valuable information on lectin-mediated regula-tion of protein secretion.

Correctly glycosylated proteins may be bound by ‘cargoreceptor’ lectins, which act to concentrate newly synthe-sized proteins in anterograde transport vesicles. Studies inthe mid-1980s on virally infected cells suggested thatnewly synthesized viral spike glycoproteins were activelyconcentrated in ER transport vesicles [21]. This wasfurther supported by the more recent work of Balch andco-workers [22]. If so, the original default hypothesis mightbe only partially correct. Newly synthesized proteins mightacquire carbohydrate modifications and progress to subse-quent compartments by vesicular transport in the absenceof specific signals; however, the packaging of proteins intotransport vesicles is also regulated by the nature of thefolding and glycosylation state of each protein.

Choices in the secretory pathwayHow proteins are sorted to different destinations at theexit site of the Golgi apparatus — the trans-Golgi network— is key to understanding protein secretion. Before thispoint in the secretory pathway, proteins can go only for-wards, backwards or stay in the same membrane. Once atthe trans-Golgi network, the number of choices increasesas proteins may be delivered to either the apical or baso-lateral surface membrane of a polarized cell, they may bedelivered to destinations along the endocytic pathway, orpackaged into regulated secretory vesicles. One hypothe-sis is that during regulated secretion, proteins that accu-mulate in dense-core secretory granules do so by a processof aggregation and protein condensation [17]. This processseems to be sensitive to pH and Ca2+ levels, suggestingthat the trans-Golgi network milieu may be crucial inprotein sorting and dense-core secretory granule forma-tion. But this does not explain why proteins and hormonesare selectively sorted into different types of vesicle.

Simons and co-workers [8] have shown that the addition ofN-linked glycosylation sites to rat growth hormone dramat-ically increases the efficiency of its secretion by the apicalroute in Madin-Darby Canine Kidney cells, suggesting thatthese N-linked carbohydrate moieties are recognized bycellular sorting machinery in the trans-Golgi network inthese polarized cells. Other recent work by Rodriguez-Boulan and co-workers [23] showed sorting of viral glyco-proteins into different vesicles at the trans-Golgi network

Dispatch 1077

1078 Current Biology 1996, Vol 6 No 9

en route to the cell surface in both polarized and non-polarized cells. Intriguingly, this work also suggests thatthe cellular machinery in a non-polarized cell line dis-criminates between two constitutively secreted plasmamembrane proteins before packaging and export from thetrans-Golgi network [23]. These findings suggest thatspecific ‘cargo receptors’ or lectins may sort N-glycosy-lated proteins into the different transport vesicles leavingthe trans-Golgi network.

Membrane-spanning mammalian lectins thus appear tofunction as sentinels stationed at specific exit sites alongthe secretory pathway, acting either to accelerate thetransport of correctly processed glycoproteins or to delaythe transport of their incorrectly processed counterparts.Proteins that receive the correct glycosylation in one com-partment escape the embrace of the relevant ‘retention’ or‘retrieval’ lectin, such as calnexin or calreticulin, and maybe recognized by a specific lectin ‘cargo receptor’, such asVIP36. Incorrectly or incompletely glycosylated proteins,however, are detained by the relevant lectin until they arecorrectly processed, retrieved by retrograde transport to anearlier compartment, and/or targeted for degradation.

Such mechanisms would provide a framework for the finecontrol of temporal and spatial development in multicellu-lar organisms by allowing them to regulate exquisitely theexpression of cell-surface receptors and secreted proteins.The coordinated, differential expression of specific glyco-syltransferases and lectins along the secretory pathwaywould allow a cell to influence the nature and rate of flowof secreted molecules in different tissues, thus influencingthe way in which a cell interacts with its local environmentand thereby affecting tissue morphology and function.The complexity of a multicellular organism does not ariseby default; it may not be the case that “little girls are madeof sugar and spice and all things nice”, but the sugar maybe more important than previously realised!

AcknowledgementsWe thank members of the Banting group, the Brunskill Laboratory, and MikeFerguson for comments and critical review of the manuscript.

References1. Rothman JE, Orci L: Molecular dissection of the secretory pathway.

Nature 1992, 355:409–415.2. Kornfeld S, Mellman I: The biogenesis of lysosomes. Annu Rev Cell

Biol 1989, 5:483–525.3. Stanley P: Glycosylation mutants of animal cells. Annu Rev Genet

1984, 18:525–552.4. Rothman JE, Wieland FT: Protein sorting by transport vesicles.

Science 1996, 272:227–234.5. Herbert DN, Foellmer B, Helenius A: Glucose trimming and re-

glucosylation determine glycoprotein association with calnexin inthe endoplasmic reticulum. Cell 1995, 81:425–433.

6. Peterson JR, Ora A, Van PN, Helenius A: Transient, lectin-likeassociation of calreticulin with folding intermediates of cellularand viral glycoproteins. Mol Biol Cell 1995, 6:1173–1184.

7. Labriola C, Cazzulo JJ, Parodi AJ: Retention of glucose units addedby the UDP-GLC: glycoprotein glucosyltransferase delays exit ofglycoproteins from the endoplasmic reticulum. J Cell Biol 1995,130:771–779.

8. Scheiffele P, Peranen J, Simons K: N-glycans as apical sortingsignals in epithelial cells. Nature 1995, 378:96–98.

9. Fiedler K, Simons K: Characterization of VIP36, an animal lectinhomologous to leguminous lectins. J Cell Sci 1996, 109:271–276.

10. Griffiths G, Simons K: The trans-Golgi network: sorting at the exitsite of the Golgi complex. Science 1986, 234:438–443.

11. Nilsson T, Warren G: Retention and retrieval in the endoplasmicreticulum and the Golgi apparatus. Curr Opin Cell Biol 1994,6:517–520.

12. Lahtinen U, Hellman U, Wernstedt C, Saraste J, Pettersson RF:Molecular cloning and expression of a 58-kDa cis-Golgi andintermediate compartment protein. J Biol Chem 1996,271:4031–4037.

13. Itin C, Schindler R, Hauri HP: Targeting of protein ERGIC-53 to theER/ERGIC/cis-Golgi recycling pathway. J Cell Biol 1995,131:57–67.

14. Arar C, Carpentier V, Lecaer JP, Monsigny M, Legrand A, Roche AC:ERGIC-53, a membrane protein of the endoplasmicreticulum–Golgi intermediate compartment, is identical to MR60,an intracellular mannose-specific lectin of myelomonocytic cells. JBiol Chem 1995, 270:3551–3553.

15. Itin C, Roche AC, Monsigny M, Hauri HP: ERGIC-53 is a functionalmannose-selective and calcium-dependent human homolog ofleguminous lectins. Mol Biol Cell 1996, 7:483–493.

16. Fiedler K, Parton RG, Kellner R, Etzold T, Simons K: VIP36, a novelcomponent of glycolipid rafts and exocytic carrier vesicles inepithelial cells. EMBO J 1994, 13:1729–1740.

17. Bauerfeind R, Huttner WB: Biogenesis of constitutive secretoryvesicles, secretory granules and synaptic vesicles. Curr Opin CellBiol 1993, 5:628–635.

18. Fiedler K, Simons K: The role of N-glycans in the secretorypathway. Cell 1995, 81:309–312.

19. Kingsley DM, Kozarsky KF, Segal M, Krieger M: Three types of lowdensity lipoprotein receptor-deficient mutant have pleiotropicdefects in the synthesis of N-linked, O-linked, and lipid-linkedcarbohydrate chains. J Cell Biol 1986, 102:1576–1585.

20. Hobbie L, Fisher AS, Lee S, Krieger M: Isolation of three classes ofconditional Chinese hamster ovary cell mutants withtemperature-dependent defects in low density lipoproteinreceptor stability and intracellular membrane transport. J BiolChem 1994, 269:20958–20970.

21. Quinn P, Griffiths G, Warren G: Density of newly synthesisedplasma membrane proteins in intracellular membranes. II.Biochemical studies. J Cell Biol 1984, 98:2142–2147.

22. Balch WE, McCaffery JM, Plutner H, Farquhar MG: Vesicularstomatitis virus glycoprotein is sorted and concentrated duringexport from the endoplasmic reticulum. Cell 1994, 76:841–852.

23. Musch A, Xu H, Shields D, Rodriguez-Boulan E: Transport ofvesicular stomatitis virus G protein to the cell surface is signalmediated in polarized and non-polarized cells. J Cell Biol 1996,133:543–558.