The protein composition of the apical and basal-lateral plasma membrane domains varies between specialized epithelial cells, and defines the transport function of the epithelium (e.g., absorptive or secretory). The apical and the lateral cell surfaces are separated by the tight junction, which forms a tight, gasket-like seal around the top of each cell. The tight junction controls paracellular transport of ions and solutes between cells, and prevents intermixing of apical and basal-lateral plasma membrane lipids and membrane proteins.

Sorting of plasma membrane proteins to their target plasma membrane domain is crucial for the maintenance of functional transporting epithelia; increased or decreased amounts of, or incorrect sorting of specific plasma membrane proteins are common in human metabolic diseases and cancers. These abnormalities include several disorders associated with defects in water and salt homeostasis including untreated diabetes, acute and chronic renal failure, and urethral obstruction, all of which have altered levels of normal, functional proteins at the plasma membrane(1–7). Gene mutations in transport proteins also cause a wide range of metabolic diseases. For example, mutations in the vasopressin regulated renal water channel aquaporin-2 (AQP2) results in a severe urinary concentration defect with a daily urine production of up to 20 liters (Nephrogenic Diabetes Insipidus); mutations in the cystic fibrosis transmembrane conductance regulator (CFTR) block transcellular Cl⁻ movement resulting in an abnormally high NaCl concentration in the airway surface liquid and increased bronchial mucus accumulation (8, 9).

A slower means of regulating protein content on the plasma membrane is by adaptation to long-term external stimuli, which alter plasma membrane protein composition through changes in transcription and/or translation of specific proteins. AQP2 is also regulated by these mechanisms. Thirst increases transcription, translation, and targeting of AQP2 to the apical plasma membrane, whereas increased water intake ultimately leads to decreased transcription and translation of AQP2 (13–14).

The spatial regulation of proteins to either the apical or basal-lateral plasma membrane domains is regulated through specific sorting stations in the trans-Golgi Network (TGN) and (recycling) endosomes, and delivery pathways between these stations and target plasma membrane (s). In general, newly synthesized apical and basal-lateral proteins are sorted from each other in the TGN of polarized epithelial cells. Since this also occurs in non-polarized fibroblasts, which do not have cell surface polarity like that of transporting epithelial cells, it is thought that protein sorting in the TGN is a constitutive mechanism in many cell-types (15).

Sorting of apical and basal-lateral proteins is mediated by distinct intrinsic motifs. Protein sorting to the apical plasma membrane is dependant on sorting motifs exposed to the lumen of the TGN that include N- or O-linked carbohydrate moieties on the extracellular domain, and a glycosylphosphoinositol (GPI) lipid anchor (16, 17). It is not known, however, if glycosylation is directly involved in protein sorting per se or for correct protein folding or oligomerization which exposes a sorting motif (18). Some newly synthesized apical proteins are transported from the TGN to the apical plasma membrane via lipid rafts, and others may be sorted into the basal-lateral plasma membrane pathway and then re-sorted to the apical plasma membrane following endocytosis from the basal-lateral membrane; it has been suggested, however, that this pathway may be an effect of experimental conditions (18–20).

Basal-lateral sorting motifs are located in the cytoplasmic domain of proteins and frequently contain either a tyrosine residue in the amino acid sequence NPXY or YXXØ (X is any amino acid, and Ø is an amino acid with bulky hydrophobic group), or a dihydrophobic dileucine motif (21). Basal-lateral sorting signals interact with clathrin adaptor complexes AP1, AP2, AP3 or AQP4. AP1B, which is localized in the recycling endosome, recognizes tyrosine-based basal-lateral sorting motifs through its μ1B subunit, and absence of APμ1B in LLC-PK1 cells results in mis-localization of the basal-lateral LDL-receptor to the apical plasma membrane (22, 23). Following exit from the TGN, some proteins such as E-cadherin, vesicular stomatitis virus glycoprotein (VSV-G), CD147 and Transferrin Receptor (TfR) pass through the recycling endosome en route to the plasma membrane in a μ1B dependent manner (24–27). On the other hand, LDL-R is delivered from the TGN directly to the plasma membrane, but is dependent on APμ1B in the recycling endosome for recycling (25). In contrast to these basal-lateral membrane proteins, Na/K-ATPase accumulates in the basal-lateral membrane independently of AP1, and may be regulated by interactions with the membrane-cytoskeleton (23, 28). Hence, there appears to be multiple routes and regulatory pathways from the TGN to the basal-lateral plasma membrane.

Recycling endosomes contain both apical and basal-lateral membrane proteins but are able to separate proteins into different sub-domains, presumably based on differences in their sorting motifs, to ensure delivery to the correct plasma membrane domain (29). Interestingly, the biosynthetic route of the μ1B dependent basal-lateral proteins changes upon cell polarization. Newly synthesized VSVG and TfR pass through the recycling endosomes in a μ1B dependent manner in non-polarized cells, but seem to be independent of μ1B containing recycling endosomes in polarized cell monolayers (26). Hence, μ1B containing recycling endosomes may be less important in the biosynthetic pathway in polarized epithelia, and more important in recycling.

Vesicle delivery to different membrane domains appears to involve microtubule-based trafficking. Vesicles carrying apical and basal-lateral cargo appear to exit the TGN along microtubule tracks composed of polyglutamylated tubulin (30). Septin 2 binding to polyglutaminated tubulin inhibits binding of the microtubule-associated protein 4 to those microtubules, thereby generating fast tracks for vesicle delivery to the cell surface (Figure 2). In the absence of Septin 2, vesicle delivery to the cell surface is decreased which results in a loss of cell surface polarity, but not tight junction function, and inhibits morphogenesis of a columnar epithelium (30). Microtubule plus ends have been shown to extend towards adherens junctions in a dynein-dependent manner, where they appear to function as tracks for vesicles carrying newly synthesized basal-lateral cargo from the TGN to the plasma membrane (Figure 2) (31, 32). Microtubule plus ends have also been shown to extend into the apical plasma membrane domain, and the kinesin KIF5B is involved in transport of the apical protein p75 to the apical plasma membrane domain in polarized MDCK cells (33). Interestingly, apical transport of p75 in non-polarized MDCK cells is KIF5B independent indicating that cell polarization induces changes in mechanisms involved in both apical and basal-lateral membrane protein trafficking in the biosynthetic pathway (33).

At the plasma membrane, vesicles dock and fuse through specific apical and basal-lateral complexes of soluble N-ethylmaleimide-sensitive factor activating protein receptor (SNARE) proteins (Figure 2) (for review see (34)). Prior to fusion with the plasma membrane, basal-lateral cargo vesicles are thought to be initially tethered to the lateral plasma membrane by a multiprotein complex termed the Exocyst complex, also named the Sec6/8 complex (Figure 2). Different t-SNAREs have been localized to the apical (syntaxin-3) and basal-lateral membranes (syntaxin-4) (35). Significantly, mis-targeting of syntaxin-3 to the lateral plasma membrane leads to compensatory mis-targeting of apical proteins to that surface, disruption of tight junction formation and inability to form a polarized monolayer (36). It is unclear how the t-SNARE complex is localized to the basal-lateral membrane domain. The mammalian homolog of lethal giant larvae (Mlgl) has been shown to interact with both syntaxin-4 and SNAP23 in contact naive MDCK cells (37). In yeast, the Mlgl homolog is essential for the fusion of exocytic vesicles with the plasma membrane (38). Moreover, the yeast homologs of Lgl, Scro7p and Scro77p, directly interact with the yeast Exocyst component Exo84, and therefore Mlgl may serve as a link between the Exocyst and the SNARE complex (39).

In polarized cells, the Exocyst is localized to the apex of the lateral plasma membrane in the region of the adherens junction, and in an intracellular compartment which could be recycling endosomes (40–42). In non-polarized single epithelial cells, the Exocyst is localized in the cytoplasm, and to the leading edge of migrating cells (43, 44). The Exocyst complex is an eight subunit complex (Sec3, Sec5, Sec6, Sec8, Sec10, Sec15, Exo70 and Exo84) that is highly conserved between different organisms (45). Quick-freeze/deep-etch electron microscopy of the purified rat brain Exocyst complex showed that it comprises two “arms” approximately 6 nm in width and 15 nm in length that radiate from a more globular body that is 30 nm long and 13 nm wide; the “arm” morphology and the large size of the complex suggest that it may form a bridge between the plasma membrane or cytoskeleton and transport vesicles, which have an average diameter of 84nm (VSV-G); interestingly, Exo70 is an extended rod, 155 Å, 35 Å wide, principally comprised of α-helices (45–47). Inhibition of the Exocyst with function-blocking antibodies resulted in a decrease in basal-lateral, but not apical protein delivery to the plasma membrane in permeabilized polarized MDCK cells; this result indicates that the Exocyst is specific for basal-lateral cargo vesicles, and that it is up-stream of the SNARE complex in vesicle docking and fusion with the plasma membrane (40).

The Exocyst co-immunoprecipitates with microtubules, and may contribute to regulating microtubule organization at the adherens junctions. Also, the Exocyst co-immunoprecipitates with four different Septins, including Septin 2 (also known as Nedd5) (45). Septin 2 organization at the adherens junction is regulated by IQGAP, which also co-localizes with the Exocyst and Cdc42 (48, 49).

Contact between two single epithelial cells is initiated by lamellipodia that extend from the cells and involves opposing cell surfaces interactions between the Ca++-dependent cell-cell adhesion protein, E-cadherin (Figure 3B; (52)). In fully polarized cells, E-cadherin is localized on the lateral plasma membrane, and in some cell-types is concentrated at the apex of the lateral membrane below the tight junction (adherens junction). E-cadherin mediated cell-cell adhesion is essential for normal development of epithelia. Genetic deletion of E-cadherin inhibits formation of the trophectoderm, a transporting epithelium that surrounds the blastocoel and inner cell mass, and results in preimplantation death (53).

Upon initial cell-cell contact formation, E-cadherin rapidly accumulates at contact points (52). As cell-cell adhesion progresses, E-cadherin localize along the contact with the highest concentrations at the edges (52). Although the cortical actin bundle is disrupted and reorganized underneath cell-cell contacts, the remainder of the bundle around the non-contacting region of the plasma membrane appears to extend into the edges of the contacts; acto-myosin contraction drives the completion of contact formation and the highest levels of activated myosin II localize to the edges of the expanding cell-cell contact (54). The intracellular domain of E-cadherin binds to β-catenin and p120-catenin. β-catenin binds dynamically to monomeric α-catenin; in contrast, α-catenin dimers bind poorly to the E-cadherin-β-catenin complex, but strongly to the actin cytoskeleton (55, 56). It is thought that E-cadherin accumulation at the cell-cell contact increases the local concentration of α-catenin such that α-catenin monomers that dissociate from the E-cadherin/β-catenin complex dimerize in the cytoplasm and bind actin filaments, thereby stabilizing actin at the lateral plasma membrane (55, 56). p120-catenin is involved in regulating E-cadherin endocytosis and stability (for review see (57)).

It is unknown how the Exocyst is recruited to the initial contacts between two single cells, or localized to the apical junctional complex in polarized cells (40). In MDCK cell monolayers, Exocyst components are recruited to cell-cell contacts with a mixture of junctional proteins; E-cadherin and nectin-2α, a member of the Ig-superfamily of adhesion proteins, are in a complex with the Exocyst in polarized cells and may be directly involved in the recruitment of the Exocyst to the plasma membrane (58). Following maturation of cell-cell adhesions, the Exocyst may be sorted to the apex of the lateral membrane with components of tight junction and nectin complexes (58). In single cells prior to cell-cell contact, the Exocyst component Sec6 and ZO-1 co-localize in the cytosol (LN Nejsum, unpublished results) (Figure 3A). Moreover, Sec6 and ZO-1 co-localize at all stages of initial cell-cell adhesion between two single cells (LN Nejsum, unpublished results). Hence, Exocyst components and components of the tight junction complex may travel to the plasma membrane by the same transport mechanism. Studies have shown that the Exo70 subunit of the Exocyst is involved in this recruitment and anchoring of the Exocyst to the plasma membrane. The Rho protein TC10 promotes the recruitment of Exo70 to the plasma membrane in 3T3 adipocytes, and may involve a direct interaction of the Exo70 subunit with phosphatidylinositol 4,5-biphosphates mediated by the positively charged C-terminus of Exo70 (59, 60).

Even less is known about how the Exocyst is anchored to the plasma membrane. The Exo70 subunit of the Exocyst may be involved as expression of a dominant-negative form of Exo70 mislocalized Sec8 from the plasma membrane localization to an intracellular localization in HeLa cells (60). Alternatively, an Arf-like3 GTPase (dnd) may play a role as expression of a dnd mutant resulted in mislocalization of Sec5 in Drosophila embryonic fusion cells (61). In yeast, the Rab GTPase Sec4 is involved in recruitment of the Exocyst to the plasma membrane through interaction with the Sec15 subunit, and GTP-bound forms of Rho1 and Cdc42 directly interact with, and are essential for correct localization of the Exocyst (62–64). Also in yeast, Sec3 directly interacts with the GTP-bound form of Cdc42 and phosphatidylinositol 4,5-biphosphate, and this is essential for exocytosis and Exocyst localization (65). It is unknown whether these mechanisms are important in regulating Exocyst distribution in mammalian epithelial cells.

In single MDCK cells, both the apical (syntaxin-3) and basal-lateral (syntaxin-4) t-SNAREs localize to the plasma membrane on both the free (apical) surface and plasma membrane adhering to the extracellular matrix (Figure 3A) (66). Syntaxin-3 and syntaxin-4 assemble into separate microdomains distributed over the whole plasma membrane, and SNAP-23 (a component of both the apical and basal-lateral SNARE complexes) localized to these microclusters (66). It is unknown how and when SNARE complexes are transported to and maintained at their respective plasma membrane domain upon initial cell-cell adhesion. Syntaxin- 4 co-localized with E-cadherin at initial cell-cell contacts (Figure 3C), and this localization appeared spatio-temporally similar to that of E-cadherin at different stages of cell-cell adhesion (43). Moreover, disruption of the SNARE complex by injection of tetanus toxin, which cleaves the vesicular SNARE VAMP3/Cellubrevin, abolished AQP3 insertion into the plasma membrane at the forming contact, showing that the lateral SNARE complex, like the Exocyst, is functional at the forming contact (43). That the Exocyst complex and syntaxin 4 have different distributions in single cells and differ in their arrival at the initial cell-cell contact, indicates that these two components of the lateral targeting patch have different mechanisms for localization to the site of initial cell-cell contact. This was confirmed in studies in which one of the two complexes was disrupted and the effect on the other monitored: syntaxin-4 was recruited to the lateral plasma membrane in cells in which Exocyst distribution had been disrupted by micro-injection of a function-blocking antibody against the Exocyst subunit Sec8, and disruption of the basal-lateral SNARE complex by injection of tetanus toxin, which cleaves the v-SNARE cellubrevin/VAMP3, did not affect localization of Exocyst components to the lateral plasma membrane (43). In addition, both the Exocyst and SNARE complexes localized normally to the lateral plasma membrane in the presence of nocodazole, which disrupts microtubules (43). Taken together, these results show that components of the lateral targeting patch localize independently of each other to the lateral plasma membrane, but are required to function sequentially as a ‘holo-complex’ to mediated AQP3 (basal-lateral) vesicle delivery, tethering and fusion with the plasma membrane at cell-cell contacts (Figure 1C).

The integrity of the apical and basal-lateral SNARE clusters seems to depend on different mechanisms, as syntaxin-3 clusters are microtubule-dependent whereas syntaxin-4 clusters are dependent on intact actin filaments (66). Microtubule disruption also interferes with syntaxin-3 localization in polarized cells, but disruption of microtubules did not interfere with syntaxin-4 localization to the forming cell-cell contact (43, 67). Since syntaxin-4 interacts with actin filaments in vitro, the finding that syntaxin-4 clustering in single non-polarized cells is dependent on intact actin filaments may indicate that F-actin is involved in the anchoring of syntaxin-4 to the lateral plasma membrane in polarized epithelia and to the initial cell-cell contact of adhering cells. However, it remains unknown how syntaxin-4 is targeted to the site of initial cell-cell adhesion (68).