PHYSIOLOGICAL REVIEWS   Vol. 79 No. 1 January 1999, pp. 73-98
Copyright ©1999 by the American Physiological Society

New Perspectives on Mechanisms Involved in Generating Epithelial Cell Polarity

CHARLES YEAMAN, KENT K. GRINDSTAFF, AND W. JAMES NELSON

Department of Molecular and Cellular Physiology, Beckman Center for Molecular and Genetic Medicine, Stanford University School of Medicine, Stanford, California

I. INTRODUCTION
II. SPATIAL CUES FOR EPITHELIAL CELL POLARITY: CELL ADHESION GENERATES ASYMMETRY AT THE CELL SURFACE AND THE AXES OF POLARITY
III. ASSEMBLY OF CYTOSKELETAL AND SIGNALING PROTEIN COMPLEXES AT SITES OF INTERACTION BETWEEN CELL ADHESION RECEPTORS AND EXTRACELLULAR CONTACTS
IV. CELL ADHESION INITIATES FORMATION OF STRUCTURES THAT SPECIFY DELIVERY AND RETENTION OF NEWLY SYNTHESIZED MEMBRANE PROTEINS
    A. Targeting Patches
    B. Membrane Skeletons
    C. Tight Junctions
V. TRANSLATION OF LOCALIZED SPATIAL INFORMATION INTO GLOBAL CHANGES IN CELL MORPHOLOGY
    A. Actin Reorganization
    B. Microtubule Reorganization
VI. SORTING AND TRANSPORT OF PROTEINS TO APICAL AND BASOLATERAL MEMBRANE DOMAINS REINFORCES AND MAINTAINS CELL SURFACE ASYMMETRY ESTABLISHED BY EXTRACELLULAR CONTACTS
VII. ROLE OF THE CYTOSKELETON IN VESICLE TRANSPORT
VIII. A HIERARCHY OF STAGES FOR ESTABLISHING CELL POLARITY
REFERENCES

	ABSTRACT

Top References

Yeaman, Charles, Kent K. Grindstaff, and W. James Nelson. New Perspectives on Mechanisms Involved in Generating Epithelial Cell Polarity. Physiol. Rev. 79: 73-98, 1999.  Polarized epithelial cells form barriers that separate biological compartments and regulate homeostasis by controlling ion and solute transport between those compartments. Receptors, ion transporters and channels, signal transduction proteins, and cytoskeletal proteins are organized into functionally and structurally distinct domains of the cell surface, termed apical and basolateral, that face these different compartments. This review is about mechanisms involved in the establishment and maintenance of cell polarity. Previous reports and reviews have adopted a Golgi-centric view of how epithelial cell polarity is established, in which the sorting of apical and basolateral membrane proteins in the Golgi complex is a specialized process in polarized cells, and the generation of cell surface polarity is a direct consequence of this process. Here, we argue that events at the cell surface are fundamental to the generation of cell polarity. We propose that the establishment of structural asymmetry in the plasma membrane is the first, critical event, and subsequently, this asymmetry is reinforced and maintained by delivery of proteins that were constitutively sorted in the Golgi. We propose a hierarchy of stages for establishing cell polarity.

	I. INTRODUCTION

Top Next References

Transporting epithelia form permeability barriers between two compartments in the body (e.g., ultrafiltrate and blood supply in the kidney) and vectorially transport ions and solutes between these compartments. These functions require a unique structural and functional organization of the cell. First, specialized cell adhesion complexes and cytoskeleton organizations are required to maintain cell-cell adhesion and cell attachment to the extracellular matrix (ECM). Second, the plasma membrane is divided into two structurally and functionally different domains, termed apical and basolateral. Each domain is comprised of a distinct subset of proteins, which regulate ion and solute transport across the epithelium. Differences in the distributions of membrane proteins between apical and basolateral plasma membrane domains are maintained by protein sorting from intracellular compartments and selective retention in the membrane (298).

Generally, epithelial cell polarity has been viewed from the standpoint of protein sorting in the trans-Golgi network (TGN) and endosomes (218, 301). For over 20 years (300, 302), it has been known that apical and basolateral proteins are sorted into distinct TGN-derived transport vesicles that traffic to different membrane domains. The accepted view is that this sorting event is a specialized process in polarized cells and is the basis for generating cell surface polarity (327). However, recent results indicate that sorting of apical and basolateral proteins in the TGN is a constitutive process, occurring in nonpolarized cells such as fibroblasts as well as polarized epithelial cells (244, 390). Moreover, membrane proteins in resting fibroblasts are not segregated into distinct membrane domains, despite sorting in the TGN, in contrast to their highly organized distributions in polarized epithelial cells. Therefore, additional processes must be involved in establishing and maintaining membrane domains in polarized epithelial cells.

In this review, we highlight the contributions of two processes that occur at the plasma membrane and that are critical for generating membrane asymmetry, namely, vesicle docking/fusion and selective protein retention. We also consider each of these processes in the broader context of how membrane polarity is generated de novo. We focus on the role of spatial cues at the cell surface in triggering a molecular cascade that establishes cell polarity. We propose that this cascade is initiated by a combination of extrinsic spatial cues mediated by cell-cell and cell-substratum adhesion. These cues physically and molecularly define contacting and noncontacting cell surfaces, which are the precursors of basolateral and apical membrane domains, respectively. These spatial cues initiate localized assembly of specialized cytoskeletal and signaling protein complexes at the contacting membrane, which position other cytoskeletal complexes and protein sorting compartments (TGN, endosomes) in the cytoplasm relative to the spatial cues. Subsequently, protein sorting from these compartments to the cell surface reinforces and maintains the structural and functional specialization of these membrane domains.

	II. SPATIAL CUES FOR EPITHELIAL CELL POLARITY: CELL ADHESION GENERATES ASYMMETRY AT THE CELL SURFACE AND THE AXES OF POLARITY

Top Previous Next References

We propose that the generation of epithelial cell polarity begins with the establishment of physical and molecular asymmetry at the cell surface (see Fig. 1). Single epithelial cells in suspension culture, in which there is neither cell-cell nor cell-substratum contact, exhibit nonpolarized distributions of marker proteins of the apical and basolateral membrane domains (299, 371). Extracellular contacts, either between a single cell and the ECM or between two cells in the absence of matrix, initiate the segregation of proteins into contacting and noncontacting (free) surface domains. Cell-cell and cell-substratum contacts are specified by adhesion receptor proteins. Adhesion between epithelial cells is mediated principally by E-cadherin, a member of the Ca²⁺-dependent cadherin superfamily of adhesion receptors (reviewed in Ref. 179). Cell adhesion to ECM is mediated by the integrin superfamily of adhesion receptors (reviewed in Ref. 162). We propose that interactions between these receptors and their corresponding extracellular ligands constitute the primary spatial cues that mark the contact sites on the plasma membrane and serve to distinguish these sites from noncontacting sites. Data supporting this proposal are summarized in section II.

View larger version (32K): [in this window] [in a new window]   FIG. 1.   Cell adhesion initiates cellular asymmetry and establishes axes of polarity. Schematic representation of cellular reorganization that occurs after establishment of cell-cell and/or cell-substrate adhesion (see text for details). Establishment of cell adhesion contacts either between cells or with extracellular matrix (ECM) leads to establishment of distinct cellular morphologies. As discussed in text, cell-cell or cell-substrate adhesion is sufficient to generate an initial level of cellular asymmetry that is oriented with respect to noncontacting and contacting membranes. However, both adhesion events are required to establish and maintain overall molecular and structural asymmetry characteristic of polarized epithelial cells (see text for details). TV, transport vesicles; N, nucleus; MTOC, microtubule organizing center; MT, microtubule; MF, microfilament; Ap-endo, apical endosome; Bl-endo, basal endosome.

Cell-cell adhesion is required to restrict the localization of basolateral membrane proteins in the plasma membrane. This was first demonstrated by examining the distributions of endogenous marker proteins of the basolateral membrane domain in Madin-Darby canine kidney (MDCK) monolayers cultured at low cell density (14, 136, 254) or after extracellular calcium depletion (266, 363) to reduce cadherin-dependent cell-cell adhesion. Under these culture conditions, cells fail to form extensive cell-cell contacts, and despite cell attachment to the ECM, basolateral membrane proteins are randomly localized on the cell surface. Even in the absence of cellular contact with the ECM, cell-cell adhesion is sufficient to initiate the segregation of both apical and basolateral membrane proteins into different membrane domains. Small clusters of MDCK cells grown in suspension restrict the budding of influenza virus to the free surface, whereas vesicular stomatitis virus (VSV) budding occurs only along the contacting surface (299). Cell-cell adhesion also initiates asymmetry in endogenous plasma membrane proteins. Adhesion between two cells in suspension is sufficient to cause an endogenous apical membrane protein (gp135) to localize to the free surface domain facing the culture medium and two basolateral proteins (Na⁺-K⁺-ATPase and E-cadherin) to localize to the cell-cell contacting membranes (371). The central role that E-cadherin plays in specifying membrane asymmetry was demonstrated by expressing E-cadherin ectopically in L-cell fibroblasts (229). Normally, these cells lack surface domains that are characteristic of polarized epithelial cells. However, expression of E-cadherin caused the redistribution of certain endogenous proteins (e.g., Na⁺-K⁺-ATPase, fodrin) to cell-cell contacts similar to their distributions after cell-cell adhesion in polarized epithelial cells.

Cell adhesion to ECM, which is mediated by the integrin superfamily of adhesion receptors, is also important for generating membrane asymmetry. Upon attachment of single MDCK cells to the ECM, several marker proteins of the apical membrane become rapidly localized to the free cell surface and excluded from the membrane in contact with the substratum, and apical microvilli form on the contact-free surface (266, 363). This is in contrast to basolateral membrane proteins, which remain randomly distributed on both membrane domains (254).

Cell adhesion to the ECM is particularly important for organizing the apicobasal axis of polarity. This axis defines the orientation of apical and basolateral membrane domains relative to the biological compartments separated by the epithelium, and hence the direction of ion and solute transport. An important indicator of the apicobasal polarity axis is the position of the tight junction at the apicolateral membrane boundary. As noted above, epithelial cell aggregates grown in suspension culture (i.e., in the presence of cell-cell contacts) form different membrane domains relative to the contact site between cells. Although the tight junction location would be expected at the boundary of the apical (outside surface) and lateral (cell-cell contact) membrane domains, the tight junction-associated protein ZO-1 is localized all along the cell-cell contacts similar to E-cadherin (371). After cells have accumulated an endogenously secreted ECM (type IV collagen, laminin) in the center of the cyst, the location of the tight junction becomes restricted to the apicolateral membrane boundary. Thus the apicobasal axis of polarity is established in these cell aggregates only after an ECM accumulates on one side of the cells. This conclusion is further supported by the observation that the establishment of an apico-basal axis is blocked in these cultures by addition of cis-OH-proline, an inhibitor of collagen secretion (371). Another example of the importance of ECM location in organizing the apicobasal axis of polarity is during reversal of polarity after cysts formed in suspension culture are transferred into a collagen gel matrix (372). Apical membrane proteins are internalized from the outside surface (now in contact with the collagen gel) and degraded. A new apical surface forms on the luminal surface and the tight junction relocates to the new apicolateral membrane boundary (372).

These changes in the axis of polarity, induced by the orientation of ECM relative to different epithelial cell surfaces, are mediated by integrins. For example, the reversal of polarity induced by placing suspension-grown cysts in collagen matrix is inhibited by antibodies against the ₁-integrin subunit, suggesting that the collagen receptor plays a central role in establishing the apicobasal axis (268). Moreover, when expression of ₂₁-integrin is inhibited by expression by ₂-integrin subunit antisense RNA, MDCK cells fail to establish an apicobasal axis and form only small rudimentary cysts lacking a central lumen (309). Further evidence for the central role that integrins play in establishing the apicobasal axis of polarity has been provided by studies on tubulocyst formation by MDCK cell monolayers when overlaid with type I collagen gel. Upon contact of the apical cell surface with collagen, the apical plasma membrane protein gp135 is rapidly internalized and subsequently reappears at sites of lateral membrane cell-cell contact where a microvilli-lined apical luminal compartment eventually forms (265). The ₂₁-integrin is primarily localized on the basolateral membrane domain (320), but remodeling of the cell surface in response to collagen is dependent on expression of a relatively minor amount of ₂₁-integrin on the apical plasma membrane. Antibodies against either the ₂- (322) or ₁-integrin subunits (322, 392) inhibit endocytosis of apical plasma membrane and formation of the apical lumen. Furthermore, a clone of MDCK cells that does not express ₂₁-integrin on the apical plasma membrane does not bind the type I collagen gel and fails to form tubulocyts (392).

Cell contact with specific ECM proteins can also direct the polarized distribution of specific plasma membrane proteins in the absence of a complete remodeling of cellular architecture. One example is hensin, a 230-kDa ECM protein secreted by intercalated epithelial cells (6, 340, 357). Cells plated on a substrate containing hensin express an "alpha" phenotype characterized by active endocytosis at the apical plasma membrane and express the anion exchanger AE1 on the basolateral membrane and H⁺-ATPase on the apical domain (357). In contrast, cells plated on substrates lacking hensin reverse the polarity of both proteins and do not undergo apical endocytosis (357). Together, these results strengthen the conclusion that although cell-cell adhesion can initiate the segregation of proteins within the plane of the membrane, cell adhesion to ECM is required to provide a spatial cue for the establishment of the axis of polarity in polarized epithelial cells.

	III. ASSEMBLY OF CYTOSKELETAL AND SIGNALING PROTEIN COMPLEXES AT SITES OF INTERACTION BETWEEN CELL ADHESION RECEPTORS AND EXTRACELLULAR CONTACTS

Top Previous Next References

How are spatial cues from extracellular contacts propagated through cell adhesion receptors to the rest of the cell surface and cell interior? Integrin- and cadherin-mediated adhesion induce localized assembly of networks of specialized cytoskeletal and signaling proteins at contacting membranes, which direct both local and global changes in the organization of the cell surface and the secretory apparatus (reviewed in Refs. 16, 54, 70, 276, 387).

The actin cytoskeleton is associated with both cadherin and integrin adhesion receptors. We propose that these interactions serve to reinforce the spatial cues provided by extracellular contacts by several mechanisms. First, associations between adhesion receptors and an actin-based cytoskeleton strengthen cell-cell and cell-ECM contacts. Second, localized cytoskeleton assemblies serve as a scaffold for recruitment and binding of signaling proteins that further define different membrane domains. Third, these interactions promote the assembly of structures that physically restrict intermixing of newly synthesized apical and basolateral membrane proteins. Each of these mechanisms is described in separate sections below.

Integrins bind to three cytoskeletal proteins in vitro: -actinin (271), talin (150), and paxillin (195). Conserved sequences in the cytoplasmic domains of integrin -subunits mediate each of these associations (195, 272) and contribute to integrin localization at cell-ECM contact sites (195, 292). -Actinin is a homodimer that binds and cross-links actin. Furthermore, -actinin binds vinculin (369), which is another actin-binding protein that also binds talin (44), paxillin (352), and tensin (171). Thus actin filaments may be linked to integrins directly through -actinin and talin or indirectly via vinculin and tensin. These assemblies of integrins with associated structural proteins are thought to be essential for stabilizing cell-ECM adhesion and regulating cell shape and polarity. Support for this notion was provided by studies in which fibroblasts were microinjected with proteolytic fragments corresponding to either the integrin- or actin-binding domains of -actinin (279). Both fragments caused stress fibers and focal contacts to reversibly disassemble, presumably by interfering with the activity of endogenous -actinin or associated proteins. Interestingly, mice lacking tensin develop large cysts in proximal kidney tubules resulting from weakened cell-ECM contacts and loss of cell polarity (200).

Engagement and clustering of integrins induces the assembly of a network of signaling proteins that transmit spatial information from the cell-ECM contact site to the interior of the cell (54, 70, 233, 276, 387). Integrin-ECM binding promotes rapid tyrosine phosphorylation of several proteins (45, 55, 259, 283). One of these proteins is focal adhesion kinase (FAK) (183, 198, 315). Focal adhesion kinase is a tyrosine kinase that localizes to focal adhesions through a carboxy-terminal focal adhesion targeting (FAT) domain (137) that partially overlaps with a paxillin-binding motif (138). Focal adhesion kinase serves to recruit several other signaling proteins to contact sites, including nonreceptor tyrosine kinases (e.g., Src, Fyn, and Csk) (58, 123, 234, 386), components of the RAS-MAP kinase pathway (e.g., SOS and Grb2) (234), as well as GTP-binding proteins (e.g., rho) (234) and associated proteins (139). Although the molecular interactions between integrins, FAK, and several other signaling proteins are well documented, the precise function of each of these associations in orchestrating changes in cell morphology is less clear. Overexpression of a dominant-negative construct containing the FAT domain results in displacement of FAK from focal adhesions and a concomitant decrease in tyrosine phosphorylation of FAK, paxillin, and tensin (111). However, these cells still assemble focal adhesions, although the rate at which they do so is reduced. Furthermore, cells lacking FAK are able to form normal focal adhesions (166). It is possible that FAK-related kinases could still be functioning in these cells, however. In contrast, a signaling pathway that is independent of FAK involves activation of rho and myosin light-chain phosphorylation (53). These events increase contractility and, together with integrin binding to ECM proteins, appear to be required for formation of focal adhesions and changes in cell morphology (43, 151).

Cadherin-mediated cell adhesion also results in localized assembly of cytoskeletal and signaling networks. A family of related cytoplasmic proteins, termed -catenin, plakoglobin, and p120^CAS, bind tightly to the cytoplasmic domain of cadherins (227, 273, 334). -Catenin and plakoglobin link cadherins to -catenin (1, 141), which can bind to and bundle F-actin filaments (294). However, -catenin and plakoglobin form distinct complexes with E-cadherin (140) and likely serve distinct functions, since plakoglobin does not appear to be involved in E-cadherin-mediated early cell-cell contacts (4). -Actinin is also a component of the cadherin/catenin complex and may provide an additional link between the complex and actin filaments (182, 258). The requirement for these associations in stabilizing cell-cell contacts was demonstrated by expressing E-cadherin constructs that either contained or lacked the cytoplasmic domain in L-cell fibroblasts (245, 246, 273). Cells expressing full-length E-cadherin formed stable cell-cell contacts, whereas cells expressing the truncated protein did not. In addition to strengthening cell-cell contacts, these associations serve to recruit signaling molecules to sites of cell-cell adhesion (reviewed in Refs. 1, 16). Tyrosine kinases (e.g., c-fyn and c-src) (351), kinase substrates (e.g., p100/120) (5, 323, 334), and tyrosine phosphatases (e.g., PTPµ, PTP-LAR) (30, 187) have been localized to cell-cell contacts, but an understanding of their functions in signaling at these sites is currently less well developed than at integrin-based adhesion sites.

	IV. CELL ADHESION INITIATES FORMATION OF STRUCTURES THAT SPECIFY DELIVERY AND RETENTION OF NEWLY SYNTHESIZED MEMBRANE PROTEINS

Top Previous Next References

Thus far, we have discussed how cell surface asymmetry is initiated by extrinsic spatial information provided by cell-cell and cell-ECM adhesion, and how this information is interpreted by adhesion receptors and transmitted to the cell interior by complex assemblies of cytoskeletal and signaling proteins. We propose that an early consequence of this spatial signaling is the generation of three types of membrane-associated structures that serve to reinforce membrane asymmetry by ensuring that newly synthesized membrane proteins are targeted to and retained in the appropriate membrane domain. These structures are 1) targeting patches, which specify docking and fusion of transport vesicles carrying either apical or basolateral membrane proteins; 2) membrane skeletons, which direct the retention and accumulation of specific proteins in different membrane domains; and 3) tight junctions, which provide a physical barrier at the boundary between the apical and lateral membrane domains, thereby preventing the intermixing of membrane proteins as well as lipids in the outer leaflet of the bilayer. Each of these structures is discussed individually in sections IVA-C.

Little is known about molecular interactions between TGN-derived transport vesicles and the plasma membrane of polarized epithelial cells. The SNARE hypothesis has provided a conceptual framework to begin characterizing the mechanisms involved (308). This model postulates that correct pairing of addressing proteins on the transport vesicle (termed v-SNAREs) with cognate receptors on the target membrane (termed t-SNAREs) determines the specificity of vesicle docking and fusion (46, 332). Vesicle-target membrane fusion also requires the activity of an ATPase, N-ethylmaleimide-sensitive fusion protein (NSF) (23, 206), and soluble NSF attachment proteins (SNAP) (56, 375). The SNAREs were identified as membrane-anchored receptors for SNAP (332, 333, 379, 381).

Vesicle-associated membrane proteins, synonymously referred to as VAMP-1 (349) and VAMP-2 (79), or synaptobrevin 1 (18) and synaptobrevin 2 (338), are the founding members of a growing family of related v-SNAREs (129, 130, 207, 230, 247, 274). These are small (~18 kDa), type II membrane proteins that have single carboxy-terminal transmembrane domains and most of their mass extending into the cytoplasm. Syntaxins, SNAP-25, and related proteins comprise an extended family of t-SNAREs defined by a conserved 60-amino acid "t-SNARE" homology domain (377). Syntaxins are type II membrane proteins, ~35 kDa in size, and occupy distinct cellular compartments (20, 21). As with the v-SNAREs, the list of syntaxin-related proteins is constantly growing (25, 26, 65, 129, 130, 157, 168). The SNAP-25 and related proteins (23-25 kDa) (33, 47, 51) are cytoplasmically oriented proteins that are anchored to the membrane by palmitoyl groups attached to central cysteine residues (366).

Endogenous t-SNAREs have polarized distributions in several epithelial cell types, including hepatocytes (94), enterocytes (71), gastric parietal cells (48, 282), renal collecting duct cells (90, 91, 170, 208, 209, 257), and pancreatic acinar cells (31, 99-102). In addition, exogenously expressed syntaxin isoforms have distinct distributions in polarized MDCK cells (203). Syntaxin 3 has been found primarily on the apical plasma membrane in MDCK cells (203), enterocytes (71), and hepatocytes (94) as well as on tubulovesicles (282) and zymogen granules (99) that will fuse with apical plasma membrane during regulated secretion in gastric parietal cells and pancreatic acinar cells, respectively. In contrast, syntaxin 4 is predominantly expressed on the basolateral membrane domain of MDCK cells (203), hepatocytes (94), and pancreatic acinar cells (99). Interestingly, a reversed polarity of these two syntaxin isoforms has been reported in renal collecting duct epithelia (208, 209). A ubiquitously expressed homolog of SNAP-25, SNAP-23, is expressed primarily on the basolateral membrane domain in rat pancreatic acinar cells (102) but is nonpolarized in MDCK and Caco-2 cells (104, 204).

Because t- and v-SNARE components have distinct polarized distributions in several epithelial cell types, it has been suggested that these proteins serve to specify the correct delivery of transport vesicles to apical and basolateral membrane domains (203, 376). However, this conclusion is almost certainly an oversimplification. First, although syntaxin isoforms are differentially distributed between distinct plasma membrane domains, their localizations are not restricted to a single membrane domain, and their expression patterns clearly overlap (94). Second, it has been suggested that exocytosis at the apical membrane domain is independent of SNAREs (165). Treatment of permeabilized MDCK cells with either tetanus toxin or botulinum neurotoxin serotype F, which cleave VAMP-1 and -2 and cellubrevin (196, 228), significantly reduced the efficiency of VSV G protein transport to the basolateral membrane but had no inhibitory effect on delivery of influenza virus hemagglutinin (HA) to the apical membrane (165). This result suggested that v-SNAREs are not required for apical secretion. It was also proposed that a novel mechanism, employing annexin XIIIB, is required for transport vesicle fusion with the apical plasma membrane (83). However, the possibility remains that a VAMP homolog that is insensitive to toxin cleavage is required for transport to the apical plasma membrane. This possibility is bolstered by the recent identification of a toxin-insensitive VAMP (TI-VAMP) in Caco-2 cells, which forms a complex with the apical t-SNARE syntaxin 3 and SNAP-23 (104). However, a function for TI-VAMP in apical transport has not yet been demonstrated.

The conclusion that apical exocytosis is independent of SNAREs is also based on the finding that neither NSF nor -SNAP appears to be required for apical transport of influenza HA (165). Several anti-NSF antibodies, when added to permeabilized MDCK cells, reduced the efficiency of VSV G protein delivery to the basolateral plasma membrane but had no effect on apical delivery of HA. A trivial explanation for this result is that an NSF homolog not recognized by these antibodies mediates fusion at the apical membrane domain (376). We suggest an alternative explanation, in which both NSF and -SNAP are required before vesicle docking and fusion with the apical plasma membrane. It is possible that NSF and -SNAP act to prime v- and t-SNAREs on the plasma membrane and on nascent apical transport vesicles during incubation at 19.5°C, before cell permeabilization. If true, then fusion of post-Golgi transport vesicles with the apical plasma membrane would occur by a mechanism analogous to yeast vacuolar fusion (121, 221, 256, 355). Thus, although the permeabilized cell transport assay does reveal differences in the mechanisms of vesicle docking and fusion at the apical and basolateral membrane domains, the protein components (i.e., NSF and -SNAP) involved in docking and fusion need not be distinct.

Although no evidence for the involvement of v- or t-SNAREs in constitutive transport from the TGN to the apical plasma membrane has been reported, both regulated apical secretion and apical secretion from endosomes (transcytosis) are dependent on SNAREs. Vesicle-associated membrane protein-2 is required for regulated apical secretion in pancreatic acinar cells (100), as well as fusion events responsible for vasopressin-stimulated aquaporin-2 trafficking to the apical membrane domain in the kidney papilla (170). Furthermore, transcytotic delivery of polymeric immunoglobulin receptors (pIgR) in MDCK cells is reduced by treatments that inhibit the functions of SNAP-23 and NSF (8, 204). This result is consistent with the localization of syntaxin 3 to the apical plasma membrane and endosomes in hepatocytes (94) and enterocytes (71), two cell types that rely on the transcytotic pathway for delivery of newly synthesized proteins to the plasma membrane (154, 190, 215).

Although evidence that SNARE complexes determine vesicle/target membrane fusion is strong, it is unclear whether interactions between cognate v- and t-SNAREs are sufficient to specify the site of vesicle docking with the target membrane. Much of the work on SNARE interactions has focused on events at the synapse, where synaptic vesicles are generated from a local endosome and are predocked at the synaptic membrane (reviewed in Refs. 47, 68). In most other cell types, post-TGN transport vesicles must traffic relatively long distances to their target membrane. Such is the case for vesicle delivery during bud formation in S. cerevisiae. During the cell cycle, the mother cell forms a new daughter cell bud through asymmetric growth of the plasma membrane and cell wall (115). Bud formation requires directed delivery of post-TGN transport vesicles from the mother to the daughter cell. Significantly, docking/fusion of these transport vesicles is restricted to a small region of the plasma membrane at the tip of the growing bud. If t-SNAREs alone specify transport vesicle docking with the bud tip, they should have a distribution that is restricted to that site. However, the plasma membrane t-SNARE, Sso1/2, and Sec9 are distributed uniformly over the mother cell and bud plasma membranes (33). Similarly, in neurons, syntaxin 1 and SNAP-25 have uniform distributions along the length of the axon, although the sites of synaptic transmission are restricted (103, 106). These results suggest that v- and t-SNAREs define a set of membranes that have the capacity to fuse, but that other proteins are required to define the site at which fusion occurs.

How are the sites of vesicle docking and fusion with plasma membrane selected? This is a matter of some speculation. It could be that t-SNAREs must be activated before fusion, perhaps by associating with certain proteins or dissociating from other proteins, to facilitate interactions with v-SNAREs (307). These accessory proteins could be restricted to sites of fusion. Alternatively, t-SNAREs may be competent for fusion everywhere on the plasma membrane, but other structures such as microtubules, actin, or septin filaments direct vesicles only to sites where fusion occurs (3). Apart from the question of how vesicle docking and fusion is controlled is the question of where targeting patches are located. Insight into each of these questions is gained by considering other proteins that play a role in vesicle docking and fusion.

One set of proteins that may control SNARE function includes homologs of S. cerivisiae Sec1, Drosophila melanogaster Rop, and Caenorhabditis elegans Unc-18 proteins. Initially identified as brain-specific syntaxin-binding proteins (105, 127, 286), isoforms of these 68-kDa proteins are ubiquitously expressed (128, 342). One isoform appears to be enriched at the apical plasma membrane of polarized epithelial cells (293). Mammalian Unc-18 homologs (Munc-18) lack membrane-spanning domains but may be recruited from the cytosol to the plasma membrane by association with syntaxins (286, 293). Munc-18a (also known as n-sec1/rb-sec1) and Munc-18b bind syntaxins 1a, 2, and 3 (128, 285), whereas Munc-18c binds syntaxins 2 and 4 (341). These in vitro binding studies reveal affinities in the nanomolar range (285, 341), although the complexes appear to be unstable in vivo (106). Interest in Munc-18 isoforms stems from the finding that Munc-18 binding to syntaxin inhibits syntaxin binding to SNAP-25 and VAMP (285). Therefore, Munc-18 proteins could regulate vesicle fusion by controlling the ability of syntaxins to interact with other components of the SNARE complex. A further level of control likely involves regulation of Munc-18/syntaxin binding by phosphorylation (95, 142), or Munc-18 binding proteins such as Mints (269) and DOC2 proteins (367). Regulatory proteins, such as Mints, may be restricted to sites of vesicle docking and fusion through interactions with specific plasma membrane proteins via PDZ domains (269).

A second set of proteins required for transport vesicle docking and fusion with the plasma membrane consists of a family of small GTP-binding proteins related to the yeast Ypt1/Sec4 proteins (260). Sec4p is present on post-Golgi transport vesicles and plasma membrane (114), and these vesicles accumulate in the daughter cell bud adjacent to the bud tip in sec4 mutants (312). A closely related mammalian homolog, rab8, is present on TGN-derived basolateral transport vesicles and on the basolateral membrane domain of polarized MDCK cells (155). Perturbation of rab8 function in permeabilized MDCK cells caused a significant reduction in the efficiency of vesicle transport to the basolateral but not the apical plasma membrane domain (155). Therefore, Sec4, rab8, and closely related proteins such as rab10 (52) and rab13 (391) are likely to be involved in vesicle docking and fusion with the plasma membrane. Although the precise role of these proteins is unclear, they have been proposed to potentiate the binding of v- and t-SNAREs (307, 331). Therefore, these proteins could be key regulators of either the timing or localization of vesicle fusion. Interestingly, both rab8 and rab13 are enriched on the plasma membrane at the apical junctional complex, which includes the tight junction and adherens junctions (155, 391). Furthermore, rab3B, an isoform of a brain-specific protein involved in synaptic vesicle fusion, also has a very restricted distribution at the tight junction (374). Disruption of cell-cell contacts results in redistribution of both rab3B and rab13 from the plasma membrane to the cytoplasm, indicating that the restricted distribution of these proteins is dependent on cell-cell contact (374, 391).

Although the rab proteins are likely to be involved in regulating vesicle docking and fusion, they are vesicle-associated proteins and therefore cannot, by themselves, mark a site on the plasma membrane for vesicle docking. It is true that certain rab proteins, such as rab3B, rab8, and rab13, are enriched on the plasma membrane at the tight junction (see above). However, it is not known whether these proteins are recruited to targeting patches before transport vesicles or if they are carried to exocytic sites on the vesicles themselves. We propose that rab3B, rab8, and rab13 are deposited at the apical junctional complex after vesicle docking and fusion at this site (see below). Steady-state localization of rab proteins at the apical junctional complex may result from slow removal from the plasma membrane relative to the rate of vesicle fusion.

A spatial landmark for vesicle docking and fusion must have the property that it is restricted to sites of exocytosis on the plasma membrane before arrival of transport vesicles. A further requirement for a spatial landmark involved in establishing plasma membrane domains during development of epithelial cell polarity is that the proteins that constitute the landmark should become spatially restricted after initiation of cell-cell or cell-ECM contact. Recent work from our laboratory indicates that a multiprotein complex consisting of mammalian homologs of the yeast Sec3, Sec5, Sec6, Sec8, Sec10, Sec15, and Exo70 gene products meets these criteria (118). In yeast, this protein complex (termed the "exocyst"; Ref. 343) is present on the plasma membrane and is restricted to sites of active exocytosis, including the tips of growing buds and the motherbud neck of large-budded cells (87, 344). Recently, mammalian homologs of Sec5, Sec6, Sec8, Sec10, Sec15, and Exo70 have been identified in neurons (119, 133, 175, 348). Similar to yeast, these proteins are present in a large complex in neurons and MDCK cells (118, 152). In single MDCK cells in contact with the substratum, the Sec6/8 complex is diffusely distributed in the cytosol but is rapidly recruited (half-life, ~3-6 h) to the plasma membrane to sites of cell-cell contact after the induction of calcium-dependent cell adhesion (118). In early contacting cells, the Sec6/8 complex is codistributed with E-cadherin and the tight junction-associated protein ZO-1 along the length of each cell-cell contact, but does not extend beyond the boundary of these contacts. As the monolayer becomes polarized, the distribution of Sec6/8 becomes restricted to the apex of the lateral membrane and no longer extends along the length of each cell-cell contact. Although this distribution is distinct from that of E-cadherin in polarized cells, localization of the Sec6/8 complex at the apical junctional complex is still dependent on calcium-dependent cell-cell adhesion. Treatment of cells with EGTA disrupts E-cadherin-mediated cell-cell contacts and results in dissociation of Sec6/8 from the plasma membrane. After readdition of calcium to the culture medium, the Sec6/8 complex is rerecruited to the plasma membrane and relocalizes to the apical junctional complex with kinetics similar to those observed for ZO-1.

In steptolysin O (SLO)-permeabilized MDCK cells, antibodies to Sec8 significantly reduced delivery of low-density lipoprotein (LDL) receptor-containing vesicles to the basolateral membrane, whereas targeted delivery of p75^NTR to the apical plasma membrane was unaffected (118). We speculate that antibodies bound to the Sec6/8 complex inhibit delivery of LDL receptor-containing vesicles by sterically hindering vesicle docking. Because the Sec6/8 complex is restricted to the apex of the lateral membrane, this result implies that vesicle docking and fusion occur near the apical junctional complex. This observation is consistent with previously reported localization of multiple rab proteins at this site (see above; Refs. 155, 374, 391). Furthermore, the observed redistribution of rab3B and rab13 into cytosolic vesicular structures after extracellular calcium depletion (374, 391) could reflect the accumulation of transport vesicles that are unable to dock efficiently with the plasma membrane in the absence of the Sec6/8 complex. If the Sec6/8 complex specifies a subdomain of the plasma membrane for efficient vesicle docking, we anticipate that basolateral proteins would be delivered less efficiently to the plasma membrane in contact-naive MDCK cells because the Sec6/8 complex is not membrane bound. Similarly, when the Sec6/8 complex is recruited to the plasma membrane to sites of cell-cell contact, we would anticipate an increased efficiency of delivery of basolateral proteins. These predictions are supported by previously published results. Detailed studies of the trafficking of newly synthesized desmoglein I, a basolateral membrane protein, showed than <10% was transported to the plasma membrane of contact-naive MDCK cells during a 1-h chase (278). However, upon initiation of cell-cell contacts, desmoglein I was rapidly (half-life ~30 min) and efficiently (>90%) delivered to the plasma membrane (278).

Mechanisms regulating assembly of targeting patches for transport vesicles on different membrane domains are unknown. Detailed studies of the kinetics and fidelity of membrane protein trafficking between the TGN and the basolateral membrane indicate that direct delivery occurs within 6 h of induction of cell-cell adhesion (117, 224). That this process is so rapid indicates that cellular components involved in sorting and targeting of membrane proteins to the basolateral membrane domain are present in nonpolarized cells before cell-cell adhesion. Furthermore, because direct targeting requires cadherin-mediated cell-cell adhesion, it is likely that components of the targeting patch, at least for vesicles containing basolateral proteins, are somehow associated with the cadherin adhesion complex. Specialized cytoskeletal networks assembled at different membrane domains in response to spatial cues may organize vesicle targeting patches. In turn, these targeting patches specify protein delivery from sorting compartments, thereby reinforcing and maintaining differences in cell surface protein distributions that were initiated by the spatial cue(s).

If cell-cell adhesion promotes the localized assembly of a basolateral-specific vesicle targeting patch on the contacting membrane, how does the apical membrane form? Developing epithelial tissues in vivo arise from cell aggregates, which lack a free membrane domain (298, 373). Recruitment of the Sec6/8 complex to sites of cell-cell adhesion likely establishes a targeting patch for basolateral transport vesicles, but several lines of evidence indicate that apical transport vesicles fuse very poorly with these contacting plasma membranes. First, contact between a cell and the ECM is sufficient to restrict apical membrane proteins to the free, noncontacting surface (299, 371). Second, epithelial cell aggregates grown in suspension culture restrict apical membrane proteins to the free surface (371). Transfer of these aggregates into a collagen gel deprives the cells of their free surface, and apical membrane proteins are rapidly removed from the surface and degraded (372). Finally, treatment of epithelial cells with nocodazole significantly reduces the efficiency of vesicle transport to the apical plasma membrane (362). However, apical vesicles do not fuse with the basolateral membrane, but rather accumulate intracellularly and may in time fuse with each other to form large vacuole-like structures (110). Therefore, for apical transport vesicles to fuse with the plasma membrane, a cell contacted on all sides by other cells and the ECM must first establish a noncontacting membrane. The SNARE hypothesis may also account for the de novo formation of such a membrane domain.

We propose the following scenario for how an apical surface forms de novo. Trans-Golgi network-derived transport vesicles containing previously sorted apical protein cargo would fuse very inefficiently with the plasma membrane along cell-cell contacts. If all surfaces of the cell were in contact with other cells or with the ECM, then TGN-derived apical transport vesicles would accumulate in the cytoplasm. Under such conditions, these vesicles may tend to fuse with one another by homotypic vesicle fusion. Homotypic vesicle fusion has been most thoroughly described for yeast vacuoles, where it is clear that both v- and t-SNARE proteins are associated in a complex on both "donor" and "target" membranes (121, 221, 256, 355). This protein complex is dissociated by the action of NSF and -SNAP before docking, which serves to activate the SNARE proteins for subsequent membrane fusion. Because v- and t-SNAREs have been described together on a number of types of vesicles, it is conceivable that post-Golgi apical transport vesicles contain both v- and t-SNAREs (25, 48, 99, 100, 282, 370). Under appropriate conditions, these could mediate homotypic fusion between apical transport vesicles, leading to the generation of a large intracellular luminal compartment. Such compartments have been described (110, 364, 373). Delivery of this luminal compartment to the plasma membrane may be mediated by the low levels of apical t-SNAREs present on the basolateral plasma membrane. Insertion of apical luminal compartments into the lateral membrane has been visualized (110, 265, 365). Once inserted in the plasma membrane, the preformed apical membrane may be prevented from mixing with the basolateral membrane domain by the rapid movement and assembly of tight junction components to a site between the two membrane domains (see sect. IVC).

Localized assembly of cytoskeletal and signaling complexes at sites of cell adhesion has implications for local and global changes in cellular organization. Locally, formation of an actin cytoskeleton precedes the assembly of a fodrin-based membrane skeleton at sites of cell adhesion (254). Fodrin is a long, rod-shaped protein that assembles into a protein skeleton with a complex of actin, protein 4.1, adducin, and other proteins (reviewed in Ref. 22). The fodrin membrane skeleton is associated with the cadherin/catenin complex (253), and although the associations are not well defined, the linkage could be mediated through binding of either actin or fodrin to -catenin (202). Fodrin also binds to ankyrin, which in turn binds with high affinity to integral membrane proteins, including Na⁺-K⁺-ATPase (255) and Cl/HCO₃ exchanger (reviewed in Ref. 22). These interactions result in localized assembly of the fodrin-based membrane skeleton and restriction of specific membrane proteins to sites of cell adhesion.

Assembly of the fodrin-based membrane skeleton may play a direct role in the early formation of a basolateral membrane domain at sites of cell-cell contact by directing the retention and accumulation of specific proteins that have affinity for the fodrin lattice. Several lines of evidence highlight the importance of interactions between cadherin/catenin complexes and the membrane skeleton in localization of certain proteins, such as Na⁺-K⁺-ATPase, to sites of cell-cell adhesion. Ectopic expression of E-cadherin in nonpolarized fibroblasts (229) or polarized retinal pigmented epithelial cells (213) induces the assembly of a membrane skeleton at cell-cell contact sites and leads to a restricted localization of Na⁺-K⁺-ATPase to these sites. However, when a truncated E-cadherin form lacking the catenin binding domain was expressed, neither fodrin nor Na⁺-K⁺-ATPase was localized to cell-cell contacts (229). Significantly, overexpression of the actin-binding domain of fodrin appears to competitively inhibit the association of actin with the membrane and results in both the disruption of membrane skeleton organization and the relocalization of Na⁺-K⁺-ATPase into cytoplasmic vesicles (153). The concept of the membrane skeleton as a protein sorting machine is conceptually attractive, since it predicts that certain proteins, such as Na⁺-K⁺-ATPase, would be preferentially excluded from endocytic pathways and accumulate at cell-cell contact sites (19). In contrast, mistargeted membrane proteins would not be incorporated into the membrane skeleton and consequently be internalized more rapidly. Data to support these predictions have been presented (122, 252). Thus assembly of a membrane skeleton may be one of the first steps in propagating signals from extrinsic spatial cues to initiate both localized assembly of a specialized membrane domain at sites of cell adhesion and global changes in the organization of other cytoskeletal complexes and membrane compartments (see sect. V).

The organization of membrane-associated actin cytoskeleton on the apical membrane has been well characterized (236). A central core of bundled actin filaments inserts into each microvillus that protrudes from the apical surface. This actin filament organization is regulated by the actin-associated proteins fimbrin and villin (34, 36, 37). Each actin bundle inserts into a fodrin-actin meshwork underlying the apical membrane (terminal web) (143-145, 281). Although the specialization of the actin cytoskeleton at the apical membrane is important for increasing cell surface area, less is known about linkage of membrane proteins to this membrane skeleton, or whether it functions to maintain polarized protein distributions.

The tight junction (zonula occludens), the most apical structure of the junctional complex, acts as a selective permeability barrier to the paracellular space (gate function; Ref. 75) and an intramembranous fence to prevent free mixing of membrane domain-specific proteins and lipids (fence function; Refs. 76, 360, 361).

The tight junction has a well described but poorly understood structure. Freeze-fracture electron microscopy has depicted the tight junction as a continuous network of parallel and interconnected strands that circumscribe the apex of lateral membranes of adjacent cells (57). Early studies sought to establish a relationship between tight junction strand number and function (57), but it has been difficult to extrapolate generalizations to all cell types (335). Little is known about how tight junction functions are regulated in different physiological or pathological conditions, nor the molecular basis for differences in the paracellular permeability of "leaky" (e.g., proximal renal tubule) and "tight" epithelia (e.g., distal renal tubule).

The protein composition and organization of tight junction are being uncovered. Occludin is the only tight junction transmembrane protein identified so far (96). However, its function in regulating tight junction structure is uncertain (310). Immunoelectron microscopy combined with freeze fracture has shown that occludin is localized to tight junction strands (93). Overexpression of full-length occludin or mutant occludin lacking the cytoplasmic domain (15, 226) or addition of a synthetic peptide corresponding to the second extracellular loop of occludin (384) perturbs tight junction gate function. How occludin might seal the extracellular space remains unknown, although hydrophobic interactions through the glycine-rich extracellular domains of occludin may be important (384). Significantly, ectopic expression of occludin in fibroblasts confers cell-cell adhesion, indicating that occludin also has an adhesive function (359).

Several cytoplasmic proteins interact either directly or indirectly with occludin, including ZO-1, ZO-2, and ZO-3 (97, 126, 167, 169, 380). These proteins may act as a protein scaffold to organize occludin at the tight junction. For example, the spatial organization of occludin on the membrane of fibroblasts requires coexpression of ZO-1 (359). ZO-1, ZO-2, and ZO-3 also contain PDZ domains and SH₃ domains (81, 126, 167, 169, 380) and may recruit signaling molecules and the actin cytoskeleton to the tight junction.

A number of signaling molecules have been implicated in the regulation of tight junction function, including tyrosine kinases, cAMP, Ca²⁺, protein kinase C, heterotrimeric G proteins, and phospholipase C (7, 50). One target common to these molecules is actin (7) and the perijunctional actin-myosin ring located subjacent to the apical junction complex (205). It has been proposed that contraction of perijunctional actin filaments and the resulting centrifugal traction on tight junction membrane regulate tight junction permeability (205).

	V. TRANSLATION OF LOCALIZED SPATIAL INFORMATION INTO GLOBAL CHANGES IN CELL MORPHOLOGY

Top Previous Next References

Globally, the generation of physical and molecular asymmetry at the cell surface by cell adhesion leads to formation of the apical membrane in the noncontacting (free) cell surface and the basolateral membrane along the contacting cell surface. Although our understanding of how the basolateral membrane forms is increasing, much less is known about how formation of the apical membrane is induced. Because the apical membrane is not in contact with other cells or the ECM, it is likely that the mechanisms involved in apical membrane formation are different from those that establish the basolateral membrane domain. We propose that the apical membrane may arise from exocytosis of a preformed intracellular luminal compartment (see sect. IVA). It is possible that spatial information from adhesive contacts on the basal (lateral) membrane is somehow transduced through the cytoskeleton to the free cell surface. Underlying these cell surface changes is a reorganization of cytoskeletal structures (Fig. 2). The importance of the cytoskeleton in eukaryotic cells is demonstrated by its involvement in cellular events ranging from, but not limited to, mitosis, cytokinesis, cell motility, maintenance of cell shape, organelle and protein distribution, secretion, and endocytosis (35, 39, 42, 59, 178, 223, 337).

View larger version (34K): [in this window] [in a new window]   FIG. 2.   Global changes in cytoskeletal organization during establishment of a polarized monolayer. In absence of cell-cell contacts, cytoskeleton and secretory compartments in an epithelial cell are organized similar to that in a fibroblast. Microtubule network originates from a broad perinuclear region, concomitant with localization of Golgi complex, and extends outward to cell periphery. Actin cytoskeleton (microfilaments) is primarily organized into focal contacts that arise from sites of cell-substrate interactions. After initiation of cell-cell contacts, microtubules reorganize to form a dense mat of short, randomly oriented microtubules in subapical membrane cytoplasm and along basal membrane. Long cortical bundles of microtubules form parallel to lateral membrane along that apical-basal axis of cell. Coincident with microtubule reorganization, Golgi complex becomes localized to apical cytoplasm. At apical membrane, actin microfilaments form core of long membrane protrusions, termed microvilli, that are characteristic of apical membrane of absorptive epithelia. Actin cytoskeleton is also distributed in a ring around apex of lateral membrane at zonula adherens just beneath zonula occludens. Along basolateral membrane, microfilaments are associated with membrane cytoskeleton (M-CYTO) and focal adhesion sites. This distribution is linked to localized assembly of cytoskeletal structures and signaling complexes at sites of cell-cell and cell-substrate contacts as discussed in text.

Morphogenesis of the apical membrane involves assembly of a unique, actin-based cytoskeletal structure that forms the core of long membrane protrusions, termed microvilli, that are characteristic of the apical membrane of absorptive epithelia (236). As described above, the microvillus core comprises a bundle of actin filaments that are cross-linked by villin (37) and fimbrin (34, 36) and linked laterally to the membrane by myosin I (49, 61). These actin bundles extend to the base of each microvillus and are linked together by myosin I and a fodrin lattice called the terminal web (143-145, 281). Ectopic expression of villin in fibroblasts by transient transfection (92) or microinjection (89) causes a redistribution of actin from stress fibers into new structures resembling apical microvilli that also contain fimbrin and ezrin. This effect was shown to require the actin bundling activity of villin (89, 92). Furthermore, expression of villin anti-sense RNA in polarized Caco-2 intestinal epithelial cells causes a reversible loss of apical microvilli (62). The precise role of villin in morphogenesis of apical microvilli is uncertain, however, because mice lacking villin have intact microvilli (287). Therefore, it is likely that in the absence of one key component of the microvillus core, other proteins with overlapping or redundant functions can fill in.

It is unknown how proteins that comprise the microvillus core become localized to the forming apical membrane. Before the appearance of microvilli in developing epithelia, villin is found in the apical cortex and near the adherens junction at sites of cell-cell contact (80, 135). Because the adherens junction contains a high concentration of cadherin/catenin complexes and actin filaments, villin may become localized to the apical cortex through its association with these cytoskeletal complexes. In addition, assembly of the terminal web may be linked to formation of the membrane skeleton on the lateral membrane, since both structures share fodrin as a common component (143, 254, 297). Similarly, the cortical actin that surrounds single cells attached to ECM could act to nucleate the assembly of actin structures on the free cell surface. As well as forming the core of individual microvilli in polarized epithelial cells, the actin cytoskeleton is distributed in a ring around the apex of the lateral membrane at the site of the adherens cell-cell adhesion junction, and along the lateral and basal membranes (223, 251). This distribution is linked to localized assembly of cytoskeletal structures and signaling complexes at sites on cell-cell contacts as discussed above.

After the establishment of cell-cell and cell-substratum contacts, microtubules become reorganized in the apical and basolateral cytoplasm. In nonpolarized MDCK cells, microtubules originate from a broad perinuclear region of the cytoplasm and extend outward to the plasma membrane in a uniform polarized array (plus ends of microtubules at the cell periphery). After the initiation of cell-cell and cell-substratum contacts, the microtubules reorganize to form a dense mat of short, randomly oriented microtubules in the subapical membrane cytoplasm and a randomly organized mat of short microtubules along the basal membrane. Long cortical bundles of microtubules with the same polarity (minus ends of microtubules oriented toward the apical surface of the cell) form parallel to the lateral membrane along that apicobasal axis of the cell (13, 117). Thus microtubules apparently reorganize around subapical sites during the establishment of polarity. Microtubule reorganization may be linked to the transition from a flat fibroblast-like morphology to the columnar cell shape characteristic of a polarized epithelial cell. Mechanisms that regulate microtubule reorganization after cell adhesion are not known. It is possible that the orientation of microtubule assembly in the apicobasal axis is determined by linkage of microtubules to cytoskeletal complexes on lateral and basal membranes, or to the actin cytoskeleton forming under the apical membrane domain.

Coincident with microtubule reorganization, compartments of the secretory apparatus that are involved in protein sorting become localized to the apical (Golgi, TGN, and apical endosome) and basal (basolateral endosome) cytoplasm. Microtubule organization is required to maintain these distributions, but again, the mechanisms involved are not well understood. Microtubule motor proteins, such as dynein or kinesin, may translocate these compartments toward the minus or plus ends of polarized microtubules, resulting in their distribution in the apical and basal cytoplasm, respectively (reviewed in Ref. 24). In addition to its spatial distribution, the three-dimensional structure of the Golgi apparatus appears to be dependent on microtubules (184). Treatment of cells with microtubule-depolymerizing drugs results in fragmentation and redistribution of the Golgi apparatus throughout the cytoplasm (280, 347, 353). These Golgi membrane fragments do not appear to be randomly distributed but rather are organized around peripheral sites that correspond to endoplasmic reticulum exit sites (59). Trafficking of proteins through the Golgi apparatus is inhibited immediately after microtubule disruption but is subsequently restored after the reorganization of Golgi fragments at peripheral sites adjacent to the endoplasmic reticulum (59).

	VI. SORTING AND TRANSPORT OF PROTEINS TO APICAL AND BASOLATERAL MEMBRANE DOMAINS REINFORCES AND MAINTAINS CELL SURFACE ASYMMETRY ESTABLISHED BY EXTRACELLULAR CONTACTS

Top Previous Next References

To compensate for protein turnover at the cell surface, and to maintain polarized distributions of cell surface constituents, newly synthesized proteins are sorted in the Golgi complex and endosomes and delivered to either the apical or basolateral membrane. Targeted delivery of proteins to different membrane domains involves three distinct processes: protein sorting into separate transport vesicles in the sorting compartment, targeting of vesicles to specific membrane domains, and docking and fusion of vesicles with the correct membrane domain. The fidelity and efficiency of vesicle targeting is facilitated by the organization of the secretory apparatus and microtubules (see sect. VII). Docking and fusion of transport vesicles with the correct plasma membrane domain is ensured by domain-specific vesicle targeting patches that we propose are established in response to cell adhesion (see sect. IVA).

Sorting of apical and basolateral membrane proteins occurs in the TGN and endosomes (161, 218, 301). Basolateral sorting in each of these compartments is dependent on identical cytoplasmic amino acid sequences, indicating that the sorting machinery is likely similar in the two organelles (11, 219). However, recent studies indicate that this may be an oversimplification (261). Basolateral sorting signals frequently contain a critical tyrosine residue within the amino acid sequence NPXY or YXXO/ (where X is any amino acid and O/ is an amino acid with a bulky hydrophobic group) (350). In several cases, these motifs are predicted to form a structure known as a "tight -turn" (60, 77). Tyrosine-based motifs mediate basolateral sorting of several proteins [e.g., receptors for LDL (159, 217) and asialoglycoproteins (108), lysosomal acid phosphatase (289), lysosomal membrane glycoprotein LAMP-1 (147), envelope glycoproteins of VSV G (345, 346) and human immunodeficiency virus (201), and mutated forms of p75^NTR (191, 235) and HA (38)]. In some cases, a dihydrophobic motif serves to target the protein basolaterally [e.g., Fc receptor (158, 220) and myosin heavy chain class II invariant chain (262, 328)]. Finally, a growing number of basolateral sorting signals bear little or no sequence similarity to these motifs [e.g., pIgR (10), neural cell adhesion molecule (193), transferrin receptor (261)]. Some of these latter signals have been suggested to form a peptide structure that comprises a tight -turn and therefore may resemble the tyrosine-based signals (10, 193). However, it is unknown whether these structures are present in the context of the native protein.

Basolateral sorting signals superficially resemble signals responsible for endocytosis and protein sorting to diverse cellular locations, including early and late endosomes, lysosomes, TGN, and specialized organelles such as phagosomes and the antigen-processing compartment in B lymphocytes (reviewed in Ref. 211). Although YXXO/-based signals mediate sorting to many organelles, the specificity of these signals is restricted by other factors, including surrounding amino acid sequences (125, 147, 220, 284), proximity of the targeting signal to the transmembrane domain (303), and phosphorylation (27, 192, 212, 263). Furthermore, the oligomeric state of the transported protein can alter the sorting activity of dihydrophobic signals (9).

Both YXXO/- and dihydrophobic-based sorting signals selectively bind clathrin adaptor protein complexes AP-1 and AP-2 (27, 146, 263, 264, 296). Very solid evidence exists for the interaction between tyrosine-based signals and the medium (µ) chains of these complexes (27, 263, 264). Although both tyrosine- and dihydrophobic-based signals can bind AP-1 and AP-2 complexes in vitro, proteins with one type of signal do not cross-compete with proteins carrying the other type of signal, implying that distinct mechanisms mediate sorting of each class of signals (212). In support of this conclusion, recent evidence indicates that dihydrophobic signals bind to the -chains rather than the µ-chains of adaptor complexes (291). Although interactions between tyrosine- and dihydrophobic-based signals and AP-1/AP-2 complexes are responsible for cargo-selective sorting into TGN- and plasma membrane-derived clathrin-coated vesicles, the mechanisms responsible for basolateral sorting remain unknown. Certain basolateral proteins are delivered to early endosomes before arrival at the plasma membrane (98, 194). However, one protein that follows this route to the plasma membrane, transferrin receptor, does not bind the µ₁-chain of TGN AP-1 complexes (263), indicating that AP-1/clathrin-coated vesicles do not mediate basolateral transport of transferrin receptors. A recently described adaptor complex, AP-3 (73, 74, 329, 330), also interacts with tyrosine- and dihydrophobic-based signals (73, 148, 329). Initial studies suggested that AP-3 is a component of nonclathrin coats (329, 330), but recent work has demonstrated an association between the ₃-appendage domain of AP-3 and clathrin (72). Thus AP-3 appears to be involved in the formation of a novel class of clathrin-coated vesicles in the TGN and endosomes (72, 336). The AP-3 complex is required for protein sorting to the yeast vacuole (63) and pigment granule formation in Drosophila (270), but a function in basolateral protein sorting has not yet been shown.

Protein sorting to the apical plasma membrane is dependent on motifs present in the luminal domain or membrane anchor, rather than the cytoplasmic domain (reviewed in Ref. 176). Certain membrane proteins are anchored to the lipid bilayer by a glycosylphosphoinositol (GPI) group. Strong evidence supports a role for this structure in apical sorting in MDCK cells (40, 199, 382), although this signal does not function as an apical sorting determinant in all epithelial cells (393). The identity of signals responsible for apical sorting of membrane-spanning proteins is much less clear. Whereas all basolateral sorting signals characterized to date are short cytoplasmic peptides, apical sorting signals appear to be located in the large, highly folded extracellular domain. This conclusion is based on findings that viral protein chimeras (231, 305), "tail-minus" mutants of basolateral membrane proteins (159, 239, 289), and soluble extracellular domains of apical and basolateral membrane proteins (40, 193, 199, 238, 288, 368, 378, 389) are all targeted to the apical membrane of polarized MDCK epithelial cells. Peptide-based apical sorting signals have not been identified in any of these proteins.

Early studies examined a potential role of N-glycans in apical sorting (116, 304). Correlative evidence for an involvement of these structures has been reviewed elsewhere (85). Experimental evidence for the involvement of N-glycans in apical sorting of certain proteins has been presented (120, 181, 316, 356), but apical sorting of many proteins occurs independently of N-glycosylation (112, 156, 214, 290, 354, 389). For at least one protein, the neurotrophin receptor p75^NTR, apical sorting information has been localized to an O-glycosylated "stalk" that is adjacent to the transmembrane domain (389). When this signal is removed, p75^NTR is very efficiently (>95%) targeted to the basolateral membrane domain. The mechanisms of carbohydrate-mediated apical sorting are not known. One potential mechanism involves recognition of carbohydrate moieties by sorting lectins, such as VIP36 (84, 86), which is hypothesized to partition into microdomains of the TGN along with other apical membrane proteins and lipids (see below).

Regardless of the nature of the sorting signal, the mechanism of protein sorting in the TGN and endosomes likely employs a simple and common theme. Sorting mechanisms are poorly understood, but the basic principle may involve segregation of different proteins by localized clustering in the plane of the lipid bilayer. Apical sorting has been proposed to involve formation of glycolipid- and cholesterol-containing membrane domains, or "rafts," in the exoplasmic leaflet of the Golgi (301, 326). The biochemical properties of these rafts have been reviewed recently (325). Some apical membrane proteins are clustered into glycosphingolipid rafts during transport through the Golgi. Clustering may be mediated by transmembrane domains [e.g., influenza virus HA (317) and neuraminidase (185)] or GPI anchors (41, 107, 394). After insertion into the apical plasma membrane, GPI-anchored proteins are initially clustered but gradually disperse in the plane of the bilayer (124), whereas HA apparently remains clustered (317). The formation of rafts can be disrupted by inhibiting sphingolipid synthesis with fumonisin B₁ (225) or by depleting cells of cholesterol after treatment with methyl--cyclodextrin (317). These treatments inhibit apical sorting of GPI-anchored proteins but have different effects on the secretory protein gp80/clusterin (177, 225). Furthermore, disruption of raft formation by cholesterol depletion causes missorting of influenza virus HA (177), but apical transport of many other proteins occurs independently of rafts (12, 64, 383). Therefore, apical sorting likely involves more than one mechanism, but the details of these mechanisms are obscure.

Certain basolateral membrane proteins (e.g., Na⁺-K⁺-ATPase) may be sorted by exclusion from glycosphingolipid rafts (225). For this type of protein, efficient basolateral sorting may be achieved in the absence of an active basolateral sorting signal. We propose that sorting of many proteins for which basolateral sorting signals have been actively sought, but not identified, may occur by this mechanism (385, 388). Other basolateral proteins may be clustered into microdomains in the TGN by a cytosolic protein coat structure that assembles onto the membrane by recognizing a sorting signal in the cytoplasmic domain of proteins (134, 180, 244, 319). However, as mentioned above, proteins that interact with basolateral sorting signals have not yet been identified.

After apical and basolateral membrane proteins are sorted into distinct subdomains of the TGN, cargo must be packaged into transport vesicles. High-voltage electron microscopic studies have revealed that the TGN is composed of multiple distinct membrane tubules that possess either clathrin coats or a novel "lacelike" coat (188). Each tubule produces only one type of vesicle, which suggests that protein sorting and vesicle budding are distinct processes (188). However, vesicle budding mechanisms are not understood. One protein implicated in the budding of basolateral transport vesicles is p200 (reviewed in Ref. 243). Purification and sequencing have shown p200 to be identical to the heavy chain of nonmuscle myosin II (163, 242). P200/myosin II is recruited onto Golgi membranes upon activation of heterotrimeric G proteins (66, 163, 242) and is released from membranes by treatment with brefeldin A (249). P200/myosin II is present on non-clathrin-coated vesicles that are distinct from coatamer-coated vesicles (250). In MDCK cells, p200 is present on basolateral transport vesicles that contain VSV G and absent from apical transport vesicles containing influenza HA (242). Immunodepletion of p200 from semi-intact MDCK cells blocks vesicular release of VSV G, but not influenza HA (242). Vesicle budding is restored by readdition of myosin II to the system. However, p200 depletion had no affect on basolateral transport in SLO-permeabilized MDCK cells, although technical differences between this study and the vesicle budding experiment may explain different results (164). Experiments involving the p200 monoclonal antibody must be interpreted with caution, since the antibody appears to cross-react with other proteins (e.g., filamin and -COP; see Refs. 164, 324). Nonetheless, depletion of myosin II by sedimentation with phalloidin-polymerized actin fibers, and several independent lines of evidence (e.g., inhibition of myosin II ATPase activity and competition of actin-myosin interaction) also reduced basolateral vesicle budding (242). Therefore, p200/myosin II appears to be involved in the budding of basolateral, but not apical transport vesicles from the TGN.

Vesicle budding from the Golgi in vitro is dependent on an 85-kDa cytosolic complex composed of p62 and a small GTPase (172). This protein complex cycles on and off Golgi membranes in a phosphorylation-dependent manner. Purification and sequencing of p62 has revealed homology with the regulatory subunit of phosphatidylinositol 3-kinase (PI 3-kinase) (173). Upon recruitment to the membrane, p62 is thought to assemble with the 100-kDa catalytic subunit of PI 3-kinase to facilitate vesicle budding (173). The mechanism is not clear, but changes in lipid charge may promote membrane curvature required to form a vesicle (321).

Fission of transport vesicles likely requires the activity of a Golgi-associated dynamin (174). The function of dynamin has been most thoroughly characterized in endocytosis, where it has been shown to form rin