|Home | About | Journals | Submit | Contact Us | Français|
The polarized distribution of functions in polarized cells requires the coordinated interaction of three machineries that modify the basic mechanisms of intracellular protein trafficking and distribution. First, intrinsic protein-sorting signals and cellular decoding machineries regulate protein trafficking to plasma membrane domains; second, intracellular signalling complexes define the plasma membrane domains to which proteins are delivered; and third, proteins that are involved in cell–cell and cell–substrate adhesion orientate the three-dimensional distribution of intracellular signalling complexes and, accordingly, the direction of membrane traffic. The integration of these mechanisms into a complex and dynamic network is crucial for normal tissue function and is often defective in disease states.
Cell polarity is a structural and functional specialization that is ubiquitous in biology. The commonality of polarity across the phyla reflects a fundamental requirement of individuals to localize different activities to distinct regions of cells, especially when individual cells come together to form complex multicellular tissues. The specialized domains of the plasma membrane that result from polarization determine cell orientation, function and fate. For example, polarization enables long-range communication by neurons and short-range communication in the immune system, vectorial transport of ions across epithelial cells and niche-specific orientation of stem-cell division, which specifies the developmental fate of daughter cells.
At first glance, the fact that different cell types exhibit diverse polarized phenotypes implies that a diverse array of specialized machineries has evolved. However, it seems that simple variations of common mechanistic themes result in the unique shapes, asymmetries and functions that characterize polarized cells and tissues. First, intrinsic protein-sorting codes are recognized and segregated by cytoplasmic adaptor complexes that regulate protein trafficking to plasma membrane domains. Second, signalling complexes and scaffolds become differentially associated with the cytosolic face of the membrane, where they define and stabilize the biochemical features of resulting domains. Third, adhesion receptors that detect neighbouring cells and the extracellular matrix (ECM) provide cues that orientate cells in three-dimensional (3D) space.
Two considerations support the idea that cell polarity is achieved through the integration of these three conserved molecular mechanisms. First, all eukaryotic cells share common cellular machineries for post-translational protein trafficking and compartmentalization1. Second, cells can adopt different shapes and functions in response to specific physiological contexts. For example, during embryogenesis, a single cell can change its shape and function as it migrates, according to morphogenetic gradients, and can then repolarize on detecting transcriptionally specified cell–cell interactions2. In disease states, such as cancer, epithelial cells lose polarity (through epithelial– mesenchymal transition (EMT)), disengage from multicellular interactions, migrate and then reintegrate into a second tissue, in which they undergo structural and functional reorganization to reside at the new site3,4. Thus, the dynamics and plasticity of the loss and re-establishment of polarity suggest that common machineries of membrane traffic are used at all times, but are deployed differently depending on the physiological context.
Here, we summarize the basic cellular machineries and biochemical rules that control the delivery of protein components to different plasma membrane domains in a generic polarized cell. We then describe the spatial cues and signalling pathways that organize these basic machineries to produce various cell shapes and functions. Finally, we consider how component asymmetry at the single-cell level is orientated in 3D space to define polarity in complex tissues. We emphasize how these different pathways generate plasticity in the forms of cell polarity, and how defects underlie important pathological states. Although much has been learnt about each of these three mechanisms in isolation, understanding how they are integrated and coordinated into a network remains a central challenge — one that is fundamental to understanding organogenesis, tissue function and various pathological states.
The generic post-translational pathway of membrane-protein trafficking to the plasma membrane involves the sequential transport of proteins between different membrane compartments (for example, the endoplasmic reticulum (ER), the Golgi complex, endosomes and the plasma membrane1). Membrane proteins are modified in these membrane compartments, usually through the assembly and processing of complex oligosaccharide chains (FIG. 1). Transport is mediated by vesicular intermediates that bud from one compartment and fuse with the next. Genetics and in vitro reconstitution experiments have identified three classes of protein complexes that form these vesicular intermediates5–7: coatomer protein complex-II (COPII) complexes are crucial for ER–Golgi trafficking; COPI complexes are crucial for Golgi–ER trafficking and intra-Golgi transport; and the adaptor protein (AP)–clathrin complex (the AP–clathrin complex) is crucial for transport between the Golgi, the plasma membrane and endosomes. Certain proteins of each complex recognize and selectively recruit different cargo proteins into transport vesicles (non-recognized proteins are excluded from vesicles), whereas other complex proteins deform and sculpt the membrane to produce the membrane vesicle or a tubule. The AP–clathrin complex also has accessory proteins that modulate the basic functions of the complex, increasing the complexity of cargo-selection strategies and affecting membrane curvature8. Regardless of how or where transport vesicles are formed, vesicle trafficking between compartments is mediated by the actin and microtubule cytoskeleton9, and vesicle fusion with the target plasma membrane is regulated by organelle-specific Rab GTPases, vesicle-tethering complexes10 and SNAREs5. These generic mechanisms are the core machineries of trafficking in all cells, and are modified in polarized cells to sort proteins into separate plasma membrane domains. Below, we focus on how machineries that are involved in the Golgi–plasma membrane–endosome pathways control membrane-protein sorting to different plasma membrane domains in polarized cells.
There are three major sites of protein sorting in the exocytic and endocytic pathways: the Golgi complex, the plasma membrane and endosomes (FIG. 1). At each site, proteins can be sorted into separate vesicle carriers on the basis of intrinsic sorting signals and the cellular machineries that recognize those signals. A major component of protein sorting at all three sites is provided by the AP-complex family, which not only recognizes protein-sorting signals but also regulates the assembly of clathrin scaffolds, which sculpt the membrane to form vesicles.
Four AP complexes (AP-1–4) have been identified, and these localize to different membrane compartments between the Golgi complex, the plasma membrane and endosomes11. Each AP complex is composed of two large subunits (α, γ, ε, δ or β1– β4; ~100 kDa each), one small subunit (σ1–σ4; ~20 kDa) and one medium subunit (μ1–μ4; ~50 kDa). Together with other proteins (such as GGA, AP180 (assembly protein of 180 kDa), epsin-1, epsin-2, EPS15, β-arrestin and ARH) that interact with clathrin, the AP subunits recognize and bind specific amino-acid motifs in the cytoplasmic domain of membrane proteins and cluster these proteins into patches on the membrane by assembling a clathrin cage8,11. One of the best-studied interactions is between the AP-2–clathrin complex and the transferrin receptor (TfnR), which occurs on the plasma membrane. The μ subunit of AP-2 (μ2) recognizes a degenerate tetrapeptide cytoplasmic domain sorting motif (YXXØ; in which Ø represents any hydrophobic amino-acid residue), resulting in the clustering of TfnR into clathrin-coated pits on the plasma membrane12. These coated pits then bud into the cytoplasm and are delivered to early endosomes. This sorting motif can be recognized in endosomes, probably by the AP-1–clathrin complex, and the receptor is recycled back to the plasma membrane by an endosome-derived vesicle population11 (FIG. 2).
In general, protein recycling from the endosome to the plasma membrane can be signal dependent or signal independent. It is difficult to distinguish these two mechanisms in non-polarized cells, which exhibit a single, non-differentiated plasma membrane. However, the situation is different in polarized cells. For example, in epithelial cells and neurons, TfnR and other well-characterized recycling receptors (such as low-density lipoprotein receptor (LDLR)) are localized to a specialized domain of the plasma membrane: the basolateral membrane in epithelia and the somatodendritic membrane in neurons13,14. In such cases, the recognition of specific targeting information in the cytoplasmic domain of TfnR is essential to maintain the polarized localization of the receptor to the appropriate domain.
In polarized cells, cytoplasmic domain-sorting signals are not only required for endocytosis, they are also required for the polarized delivery of newly synthesized receptors and the maintenance of recycling receptors in their appropriate membrane domains15,16 (FIG. 2). Canonical internalization signals have a role in polarized endocytosis and recycling, and they are assisted by additional sorting signals and decoding adaptor complexes. LDLR is a well-studied example. It has two Tyr-based sorting motifs — a membrane-proximal motif and a C-terminal motif — in its cytoplasmic domain, and these are required for post-Golgi delivery and the maintenance of LDLR to the basolateral or somatodendritic domains13,14. The membrane-proximal motif is important for LDLR endocytosis, and although the C-terminal motif is not required for endocytosis, it encodes a signal for sorting to the basolateral or somatodendritic domains13,14. Many other basolateral membrane proteins (TABLE 1) have Tyr-based targeting motifs16, although diLeu-based signals have also been identified (diLeu motifs can also be responsible for clathrin-mediated endocytosis)17. Other motifs, such as the sorting motif in TfnR18, seem to be devoid of crucial Tyr or Leu residues18,19.
The observation that many basolateral proteins have a Tyr- or diLeu-based motif for basolateral polarity has long suggested that AP–clathrin complexes are important in one or more steps in specifying protein trafficking in polarized cells. The only AP complex to be identified to date that is definitively associated with basolateral protein trafficking is an epithelial cell-specific variant of the clathrin-associated AP-1 complex (designated AP-1B), in which the ubiquitously expressed μ1A subunit is replaced by a closely related μ1B subunit20,21. AP-1B clearly functions to determine the polarity of Tyr-based signals (and some non-Tyr signals, such as that in TfnR), as the absence of μ1B leads to a non-polarized or apically localized protein, a similar effect to that of eliminating the sorting motif. However, it is unlikely that all such signals are decoded by AP-1B. DiLeu-based signals, for example, programme basolateral polarity, but in an AP-1B-independent manner20,21 (TABLE 1). A second AP complex, AP-4, also has a role in basolateral targeting, although its cargo and signal specificity remain poorly defined22.
AP-1B is not expressed in all polarized cells, indicating that other adaptors or protein-sorting pathways exist. Many neurons, for example, do not express μ1B, but protein transport to the somatodendritic plasma membrane domain involves the same targeting signals that are used for transport to the basolateral membrane in most epithelial cells14. Hepatocytes also do not express μ1B, but they use the same Tyr-based signals on basolateral proteins (such as LDLR) as μ1B-positive epithelial cells23. The identities of these alternative basolateral adaptors remain elusive. Additionally, do AP-1B or any alternative adaptor proteins work in conjuction with clathrin? This is possible, but the removal of clathrin by small interfering (si)RNA has a surprising range of effects on basolateral membrane-protein sorting24. Proteins such as VSVG (the major surface-coat protein of the stomatitis virus, which is a Tyr-dependent protein), NCAM (neural cell adhesion molecule, which is Tyr-independent) and CD147 (which is Leu dependent) are delivered in a 1:2 ratio to the basolateral and apical membranes, whereas others are delivered in a 2:1 ratio (TfnR, which is Tyr independent), a 3:1 ratio (epithelial (E)-cadherin, which is diLeu dependent) or are unaffected (Na+/K+-ATPase). By comparison, deletion of both basolateral sorting motifs in LDLR results in the almost complete mistargeting of mutant LDLR to the apical membrane13. The observation that some proteins are less affected than others by clathrin knockdown further indicates that there is a degree of redundancy and overlap in sorting signals and decoding elements, only some of which are likely to be clathrin dependent. For example, ankyrins are adaptor proteins that bind to a subclass of membrane proteins (such as E-cadherin and Na+/K+-ATPase) and to the cytoskeletal protein spectrin25. Importantly, decreased expression of ankyrin B or ankyrin G reduces E-cadherin trafficking from the Golgi to the basolateral membrane26. Further studies are required to uncover additional mechanisms and pathways of protein sorting and trafficking.
The sorting motifs and the cellular machineries that sort proteins to the apical membranes of epithelial cells are different from those that sort proteins to basolateral membranes. In contrast to the cytoplasm-orientated basolateral sorting motif, apical sorting motifs are localized in the extracellular or transmembrane domains of proteins27 (TABLE 1). Extracellular domain motifs contain N- and O-linked oligosaccharide chains (such as those found in p75 (REF. 28) and sucrase-isomaltase29), although it is unclear whether these oligosaccharide chains are directly recognized by sorting machinery or whether they control protein conformation and hence the display of a sorting motif in the amino-acid backbone of the protein.
The membrane-associated signal can be the transmembrane domain itself (as occurs in some viral glyco-proteins, such as haemagglutinin and neuraminidase), but the best described signal is the glycosyl phosphoinositol (GPI) lipid anchor30. GPI-anchored proteins are sorted into the apical pathway in the Golgi complex. This occurs by GPI-anchored protein oligomerization in lipid rafts31 that are enriched in glycosphingolipids, sphingomyelin and cholesterol30. Interestingly, large clusters of GPI-anchored proteins fail to form in a lectin-resistant Madin–Darby Canine Kidney (MDCK) cell line that missorts GPI-anchored proteins. This suggests that carbohydrates are involved in this clustering event and are necessary for apical sorting32. Several other proteins in lipid rafts might have roles in lipid-raft synthesis (such as FAPP2 (REF. 33)) and stabilization (such as MAL30), and in clustering of apical proteins (such as the lipid-raft protein galectin-4 (REF. 34)). It is unknown, however, how apical protein clustering, by lipid rafts or any other mechanism, leads to deformation of the membrane and to the formation of a transport vesicle that can deliver its cargo to the apical plasma membrane.
In general, the cytoplasmic domain of apical proteins is not thought to be important in protein trafficking. The protein rhodopsin, however, is targeted to the apical membrane of photoreceptors by a sorting determinant that is localized to its cytoplasmic C terminus35. This sorting determinant interacts with dynein36. Importantly, missorting of rhodopsin leads to retinitis pigmentosa (BOX 1).
Various disease states (see table) show that cells can dissolve or reorganize the cellular machinery that generates polarized distributions of proteins4. This process can occur at the three layers of regulation of cell polarity that are discussed in the main text.
Mutations in the intrinsic protein-sorting motifs that sort proteins to specific membrane domains are found in genetic disorders of major organs, such as the kidney, lungs, intestine and nervous system125,131. These mutations can affect protein transport through the secretory pathway, and can cause missorting of proteins to the wrong membrane domain or cause loss of protein localization at the plasma membrane. In these cases, overall polarity might not be affected, but the key role of the mutated protein in a biological process can result in the loss of cell function.
Mutations in key components that detect extrinsic orientational cues have strong effects on cell organization and function. Many of these proteins are tumour suppressors, including phosphatase and tensin homologue (PTEN)127, and many oncogenic mutations activate phosphoinositide 3-kinase (PI3K) or AKT/protein kinase B (PKB) directly. Changes in the distributions of phosphatidylinositol-3,4,5-trisph osphate (PtdIns(3,4,5)P3) and phosphatidylinositol-3,4-bisphosphate (PtdIns(3,4)P2) might have additional effects on the structure and function of cells. Scribble, lethal giant larvae (LGL) and discs large (DLG) were isolated as neoplastic tumour-suppressor mutations in Drosophila melanogaster, based on the distinctive ‘giant larvae’ phenotype of zygotic mutant animals132. Little is known about their roles in human disease, but the loss or decreased expression of human homologues of Scribble and LGL have been found in colorectal cancers and malignant melanoma128.
Changes in the orientational cues that specify cell polarity occur in a form of reversed cell polarity, termed epithelial–mesenchymal transition (EMT). EMT has an essential role in embryogenesis and homeostasis (during morphogenetic movements, epidermal wound healing and migration of neutrophils in chemotactic gradients) and in cancer-cell metastasis3. EMT is characterized by decreased expression of epithelial (E)-cadherin (which is mediated by the expression of transcriptional repressors such as SNAIL, ZEB and some basic helix–loop–helix factors133), the loss of cell–cell adhesion and epithelial characteristics, and altered cell–extracellular matrix (ECM) interactions, including ECM remodelling by secreted metalloproteinases and increased cell migration. It is conceivable that the ‘forward reaction’ can be re-initiated when a metastatic cell reaches the appropriate distal tissue site or niche, resulting in a partly differentiated tumour3.
|Bartter syndrome* (REF. 123)||ROMK||Missorted to apical membrane|
|ClCK||Missorted to basolateral membrane|
|Liddle syndrome* (REF. 124)||βENaC||Recycling on apical membrane|
|Cystic fibrosis125||CFTR (ΔF508)||ER exit|
|CFTR (C-terminal PDZ)||Golgi exit|
|Congenital sucrase-isomaltase deficiency126||Sucrase-isomaltase||Missorted to basolateral membrane|
|Familial hypercholesterolaemia23||LDLR (FH-Turku)||Missorted to the apical membrane|
|Polycystic kidney disease‡||Polycystin-1 and -2||Growth or mitotic defects; protein mislocalization (for example, Na+/K+-ATPase, EGFR)|
|Situs invertus‡||Left–right dynein||Cilium motility|
|Bardet–Biedl syndrome‡||BBS4, BBS5 and BBS8||Unknown (protein transport to cilium?)|
|Retinitis pigmentosa‡||Kinesin-2 (IFT)||Rhodopsin transport into cilium|
|Cardiac arrhythmia25||Ankyrin B||T-tubule organization of Na+/K+-ATPase, Na+/Ca2+ exchanger and IP3R complex; stress-induced sudden cardiac death|
|Brugada syndrome25||Voltage-gated Na+ channel (E1053K mutation in the ankyrin-G binding site)||Loss of plasma membrane localization, ventrical arrhythmia and sudden cardiac death|
|CRASH syndrome25||L1 cell adhesion molcule (S1224L and Y1229H mutations in the ankyrin-B binding site)||Loss of axon and dendrite localization, corpus callosum hypoplasia, retardation and aphasia|
|Oncogenesis||PTEN and PI3K127||Activates the AKT/PKB signalling pathway (promotes cell survival, proliferation and growth)|
|Scribble and LGL128||Decreased expression; function is unknown but the defect is found in colorectal cancer and malignant melanoma|
|Retinitis pigmentosa129||Rhodopsin||Missorting to photoreceptor apical membrane|
|Peutz–Jeghers syndrome130||LKB1 (PAR4)||Mutations; overgrowth of cells in the gastrointestinal tract (hamartomas)|
ER, endoplasmic reticulum; LDLR, low-density lipoprotein receptor; PAR4, partitioning-defective-4.
Basolateral (and somatodendritic) sorting signals are often ‘dominant’ over apical sorting signals. In other words, if both a functional basolateral and an apical signal are present, basolateral targeting will ensue. Transcytosis of membrane proteins from the basolateral to apical domains can occur if the basolateral signal is inactivated after newly synthesized proteins are included into basolaterally directed transport vesicles (FIG. 2); transcytosis involves endocytosis from one membrane domain, delivery to endosomes, resorting and trafficking to the other membrane domain. For example, the adhesion-signalling protein neuronal-glial (Ng)-CAM has a Tyr-containing, AP-1B-dependent basolateral targeting signal that is inactivated by phosphorylation on reaching the basolateral surface. In the absence of a functional basolateral signal, Ng-CAM is not captured by AP-1B in endosomes after internalization and is instead transcytosed to the apical surface37. Variations on this simple theme probably explain other examples of trancytosis (such as the extensively studied polymeric Ig receptor38).
The sorting of plasma membrane proteins to their appropriate destinations seems to be dependent on the general core mechanisms found in the endocytic and secretory pathways, but protein sorting is subject to slight modifications in polarized cells. In all cells, the endocytic and secretory pathways possess multiple sorting sites. In most cells, however, the sorting sites that are most relevant for generating and maintaining polarity — the trans-Golgi network (TGN) and recycling endosomes — are found in the perinuclear cytoplasm (FIG. 2). It is highly likely that signals for apical or basolateral sorting are recognized at either or both of these sites, with the TGN mediating the sorting of newly synthesized proteins in the secretory pathway and recycling endosomes responsible for polarized recycling or transcytosis after endocytosis37,38. It also seems likely that newly synthesized membrane proteins could pass through recycling endosomes after leaving the Golgi and before reaching the surface39. Thus, recycling endosomes might be a common site for protein sorting; perhaps adaptor complexes in these endosomes select cargo for inclusion in transport vesicles that are destined for specific plasma membrane domains.
Sorting proteins during exit from the Golgi or recycling endosomes is one level of adaptive machinery for generating different protein distributions in polarized cells. However, molecular sorting must be coupled to mechanisms that specify vesicle delivery and fusion to different plasma membrane domains. These mechanisms involve the cytoskeleton, for long-range vesicle delivery, and (at short range) protein complexes at target plasma membrane domains that specify the docking and fusion of vesicles.
Most polarized cells have specialized organizations of actin and microtubule cytoskeletons. Meshworks of short actin filaments are localized underneath the entire plasma membrane and might provide linkages through the ankyrin–spectrin complex to membrane proteins25. The actin cytoskeleton might have several roles in cell polarity and protein trafficking40 as it is regulated by CDC42. CDC42 is involved at several crucial control points that regulate cell polarity (see below), including myosin motor proteins, which could regulate vesicle trafficking, and the ankyrin–spectrin complex. Indeed, ankyrin knockdown results in the loss of lateral membrane surface area41, perhaps by decreasing protein trafficking from the Golgi (see above) or by structurally destabilizing the basolateral membrane in a manner that is analogous to spectrin deletion in erythrocytes.
Microtubule arrays are normally organized in fibroblasts with the dynamic plus ends at the cortex and the minus ends near the centrosomes, close to the nucleus. In polarized cells, microtubules are generally bundled along the long axis of the cell (apico–basal in epithelia, soma–axon in neurons) with all the plus ends at the base of the cell or end of the axon, respectively42,43. Microtubules seem to regulate the efficiency rather than the fidelity of vesicle delivery to the apical and basolateral membrane in epithelial cells44. Plus-end microtubule kinesin motors are involved in apical protein delivery (for example, the delivery of the p75 protein, but not of GPI-anchored proteins)45, and are generally involved in the targeting of basolateral46 and axonal membrane proteins47 (see the review by Li and Gunderson in this issue).
The fusion of vesicles with the correct membrane domains is a crucial step in the generation of polarity. Two interconnected mechanisms regulate the initial tethering of vesicles to the membrane and then the fusion of vesicles with the target-membrane domain. Members of the Rab family of small GTPases seem to control many stages of vesicle docking and fusion, especially by having a role in tethering vesicles to their target membranes10. Vesicle-tethering complexes are found at many stages of the exocytic pathway and are thought to increase the efficiency and specificity of vesicle delivery (FIG. 2). At the plasma membrane, the exocyst complex seems to regulate the docking of a subset of vesicles, including basolateral vesicles in epithelial cells48, neurite outgrowths and axonal synapse-assembly domains49. The exocyst is a large complex of at least six proteins, some of which are localized on the target plasma membrane and others of which are localized to the transport vesicle, along with a Rab family GTPase that helps to control tethering-complex assembly (RAB8 on basolateral vesicles of epithelia; RAB3 on synaptic vesicles)50,51. This suggests that vesicle docking at the plasma membrane might occur through the assembly of an exocyst holocomplex. Another type of vesicle-tethering complex comprises annexins and is present on other plasma membranes, such as the apical membrane of epithelial cells. Annexins bind to membranes in a phosphatidylinositol-3,4-bisphosphate (PtdIns(3,4)P2)-dependent and Ca 2+-dependent manner and self-aggregate. As with components of the exocyst complex, annexins might be present on both vesicles and target membranes52. The delivery of some apical proteins in epithelial cells (for example, the delivery of sucrase-isomaltase53) requires annexin II, but the full repertoire of proteins that require annexins for plasma membrane delivery is not known.
Fusion of vesicles with a target plasma membrane is mediated by the SNARE complex, which comprises vesicle (v)-SNAREs (such as vesicle-associated membrane protein (VAMP)) and target (t)-SNAREs (the syntaxins)5. In polarized epithelial cells, apical and basolateral vesicles contain different v-SNAREs (such as tetanus-insensitive (Ti)-VAMP and VAMP8, respectively54), and different t-SNAREs are localized to the apical (syntaxin-3) and basolateral (syntaxin-4) plasma membranes55,56 (FIG. 2). Loss-of-function or mislocalization of SNAREs leads to a concomitant disruption in the delivery of apical or basolateral vesicles to the plasma membrane54,57–59.
Polarized membrane-protein localization can also be achieved by asymmetric stabilization at a plasma membrane domain. Selective stabilization can occur in the absence of direct sorting in the exocytic or endocytic pathways60, or as a mechanism for further refining proteins that are sorted in these pathways61. Several classes of membrane proteins bind to the cytoplasmic scaffold complex of ankyrin–spectrin, including ion transporters and channels (such as the anion exchanger, Na+/K+-ATPase, voltage-gated Na+ channel and Na+/Ca2+-exchanger), receptors (such as Ins(1,4,5)P3 receptor, ryanodine receptor and N-methyl-D-aspartate receptor (NMDAR)), and cell-adhesion proteins (such as E-cadherin, L1 (Ng-CAM) and CD44)25. In epithelial cells and neurons, ankyrin–spectrin complexes are differentially localized to the basolateral membrane62, and to axons and dendrites, respectively63. Binding to these scaffolds can recruit membrane proteins that function in the same biological process (such as T-tubule clustering of Na+/K+-ATPase, IP3R and Na+/Ca2+-exchanger)25 and can sequester membrane proteins from the endocytic pathway, thereby increasing their residence time and accumulation60.
Another essential feature for the development and maintenance of polarity is a diffusion barrier that separates membrane domains that are otherwise part of a continuous lipid bilayer. In epithelial cells, this barrier is localized to the boundary between the apical and basolateral membrane domains and comprises the tight junction, which acts as a molecular fence to prevent the free diffusion of proteins from one domain to the other64; regulated vesicular transport (transcytosis) is therefore required to bypass the barrier. Diffusion barriers are not always dependent on cell–cell contact, and can be constructed in single cells by the deposition of cytoskeletal scaffolds that selectively recruit high concentrations of membrane proteins, thereby impeding the movement of freely diffusing membrane proteins. Such a barrier occurs at the axon initial segment in neurons and restricts the movement of axonal membrane proteins into the soma, and vice versa65. This molecular fence involves localized arrays of actin and ankyrin, and might form the basis of diffusion barriers in other cell types25. Other cytoskeletal arrays might restrict membrane-protein diffusion in other systems (such as in the immunological synapse, in the border between a nascent bud and a mother yeast cell, and in the membrane-associated cytosolic ‘necklace’ that is found at the base of cilia).
Although asymmetry in the localization of plasma membrane proteins is achieved by intrinsic sorting mechanisms, it remains unclear how these pathways are coupled to the overall orientation of cells in a multicellular context. Below, we discuss evidence that intracellular complexes (partitioning defective (PAR), Crumbs and Scribble complexes) provide orientation cues that specify overall cell polarity, shape and other functional specialization (FIGS 3,,44).
Polarization in single cells and in multicellular organisms can be induced by a pathway that involves the Ser/Thr kinase and tumour suppressor LKB1 (the mammalian orthologue of PAR-4, one of six PAR proteins that are required to establish polarity in early Caenorhabditis elegans development66; see below). Activation of LKB1 in epithelial cells results in apico–basal polarization of single cells, even in the absence of cellular adhesion67; inactivation of LKB1 in neurons inhibits axon formation in vitro and in vivo68. Here, we consider the roles of PAR-4 (LKB1) and other PAR proteins in cell polarity in multicellular contexts.
Among the substrates of LKB1 are members of the AMP-activated protein kinase (AMPK) family, which have multiple roles in cells69. In addition to being a glucose or energy sensor, AMPKα and AMPKβ phosphorylate non-muscle myosin light-chain kinase, an activity that is required for the formation of junctional complexes in Drosophila melanogaster70. AMPK members have several functional homologues: the kinase synapses of amphids defective (SAD), the Ser/Thr kinase PAR1 and the family of ELKL-motif kinases (EMKs; also known as microtubule affinity-regulating kinases (MARKs)), which destabilize microtubules71. These functional homologues have important roles in the orientation of polarity in neurons and epithelial cells69.
Overexpression of PAR1b (also known as EMK1 and MARK2) in MDCK cells results in the partial reorientation of the apical plasma membrane to intercellular lumens on the lateral membrane domain, similar to the orientation of the apical (bile canalicular) membrane in hepatocytes72. MDCK apical plasma membrane reorientation is accompanied by a change in microtubule organization, such that the minus ends of microtubules localize to the intercellular lumens instead of to the top of the cells. In addition, trafficking of post-Golgi transport vesicles to the apical surface is redirected into an indirect pathway (vesicles first appear at the basolateral domain and then transcytose to the apical domain), as also occurs in hepatocytes73. Because induction of the transcytotic pathway of apical vesicle transport occurs independently of microtubule reorganization, PAR1 probably affects apical protein sorting into vesicles or the delivery of vesicles to the plasma membrane. Interestingly, the PAR1 homologues in budding yeast, Kin1 and Kin2, interact with the machinery for vesicle tethering (the Rab proteins and the exocyst complex) and membrane fusion (the SNAREs)74. Thus, both PAR1 and PAR4 might regulate at least two key control points in post-Golgi vesicle delivery: microtubule remodelling and orientation and, hence, the directionality of post-Golgi vesicle transport; and the docking and fusion of vesicles at the plasma membrane. The actual mechanisms by which these kinases exert their effects are unclear, as D. melanogaster mutants that are defective in AMPK or LKB1 exhibit generalized defects in polarity that affect both the apical and basolateral domains70.
In addition to PAR1 and PAR4, the PAR complex comprises three scaffolding proteins with either multiple PDZ domains (PAR3 and PAR6) or 14-3-3 homology (PAR5), and a RING-finger protein75,76 (PAR2) (FIG. 3). Biochemical studies show that PAR3, PAR6 (which also binds active CDC42 (REF. 77), see below) and atypical protein kinase C (aPKC) form a subcomplex (FIG. 4) that localizes to the apical junctional complex in polarized epithelial cells78,79 and to the tips of neuron axons80. Genetic studies reveal that the PAR3–PAR6–aPKC subcomplex contributes to the establishment and maintenance of apico–basal polarity in embryonic epithelial cells79,81 and to axon formation in neurons80,82 (see the review by Iden and Collard in this issue).
The PAR3–PAR6–aPKC subcomplex interacts functionally with two other complexes that control cell polarity: the Scribble and Crumbs complexes (FIG. 3). The Scribble complex also comprises lethal giant larvae (LGL) and discs large (DLG), regulates the identity of the basolateral membrane, and is localized below the apical junctional complex and along the membrane at cell–cell contacts79,81. The Crumbs complex comprises the PDZ-domain-containing proteins PALS1 (protein associated with LIN-7)-1 (also known as stardust (STD)) and PATJ (PALS1-associated tight-junction protein)83. Crumbs regulates the identity of the apical membrane81 and localizes to the apical side of the junctional complex in polarized epithelial cells81. Loss of function of either the Crumbs or Scribble complexes results in defects in epithelial polarity that are due to the reduction of the surface area of the apical and basolateral plasma membrane domains, respectively79,81. How the Crumbs and Scribble complexes regulate the identity of the apical and basolateral membranes is unknown.
The PAR, Crumbs and Scribble complexes mutually regulate the localization and activity of each other (FIG. 3). Genetic analysis showed that aPKC is required in early D. melanogaster development to maintain the presence of the Crumbs complex at the apical membrane, perhaps by direct phosphorylation84 or by Crumbs binding, through its PDZ-interaction domain, to PAR6 (REF. 85). Later in development, Crumbs is required to stabilize the PAR3–PAR6–aPKC complex at the apical junctional complex79,81. aPKC can also phosphorylate PAR3, decreasing the affinity of PAR3 for aPKC and suggesting that the association of PAR3 with PAR6–aPKC is dynamic86. This is further supported by the finding that the AMPK family member PAR1 phosphorylates PAR3, causing PAR3 to bind PAR5 and thereby inhibiting the association of PAR3 at the basolateral membrane87 (FIG. 3). Interestingly, PAR1 itself is phosphorylated by aPKC, which causes PAR1 to bind to PAR5 and results in the inhibition of the association of PAR1 with the cortex at the apical membrane88. Finally, a member of the Scribble complex, LGL, is also a target for aPKC phosphorylation, and LGL phosphorylation inhibits its localization to the apical cortex89. The activities of aPKC, PAR1 and PAR5 have the overall effect of ‘protecting’ the identities of the apical and basolateral membrane domains. This is achieved by maintaining the apical activities of the Crumbs (and PAR) complex by blocking the intrusion of components of the Scribble complex (LGL), and vice versa. The positioning of the PAR3–PAR6–aPKC complex at the crucial apical–basolateral junction might allow it to arbitrate the phosphorylation of components, such as LGL (preventing its apical association) and PAR3, and prevents the PAR complex from encroaching too far into the basolateral domain.
The PAR complex and Crumbs-3, an isoform of Crumbs90,91, also colocalize in the primary cilium in vertebrate polarized epithelial cells. The primary cilium is involved in many vertebrate developmental pathways and in tissue homeostasis92. Crumbs-3 and the PAR complex are directly bound through PAR6, and PAR5 is also present in the complex90. Localization of the complex to the cilium requires association with the microtubule motor protein KIF3A and intact microtubules90. Deletion of Crumbs-3 inhibits ciliogenesis90, whereas deletion of PAR3 or inhibition of aPKC results in a decrease in cilium length91. Although it is not known whether localization of the Crumbs-3–PAR complexes to the primary cilium affects cell polarity, defects in genes for the assembly or function of the primary cilium have been associated with many genetic disorders that are characterized by defects in growth and polarity of epithelial cells, including polycystic kidney disease, left–right asymmetry defects and Bardet–Biedl syndrome92 (BOX 1).
Although the intrinsic cell mechanisms described above regulate sorting, targeting and separation of proteins to different membrane domains, it is important to note that these domains are not organized arbitrarily on the cell surface, but rather are topologically ordered in 3D space. Here, we consider how extracellular cues provide the template that orientates intracellular mechanisms of domain specification and protein sorting. Cell adhesion, both to the substratum (ECM) and to other cells, is important in establishing the polarized orientation of cells.
In general, individual epithelial cells that are grown in suspension — that is, in the absence of cell–cell and cell–ECM adhesion — do not develop polarity, but instead undergo programmed cell death (anoikis). When grown on a surface, however, even single epithelial cells develop an apico–basal axis of polarity. When this surface has biological relevance, such as a surface of laminin-containing ECM, neurons specifically form an axon93, single mammary epithelial cells selectively secrete β-casein from the apical surface94 and 3D epithelial cysts polarize correctly95. In this way, cell–substrate adhesion is coupled not only to post-translational control of membrane traffic and cytoskeletal reorganization, but also to the control of transcriptional activity that specifies the function of the resulting polarized cell types96.
Cell–cell contact, mediated by a range of cell-adhesion proteins, is intimately involved in cell polarization97. For example, epithelial cell–cell adhesion in suspension culture induces the segregation of basolateral membrane proteins to the cell–cell contacts and induces the trafficking of apical proteins to the free surface98. Perhaps the most important protein for cell–cell adhesion and cell polarization is E-cadherin, which is required for epithelial organization in early embryonic development99. Homotypic E-cadherin interactions are necessary to form the junctional complexes that maintain the diffusion barrier between the apical and basolateral domains of epithelial cells. They are also required for localization of the exocyst complex48 and to ensure that basolateral vesicles are delivered100 to sites of cell–cell adhesion and to the lateral membrane domain. Although cell–cell contacts can help control epithelial cell differentiation101, cell–cell adhesion is insufficient to trigger full apico–basal polarity in epithelia and, as discussed above, a set of signals must be supplied following adhesion to ECM.
We have discussed three main layers of regulation that control cell polarity. Intrinsic mechanisms sort membrane proteins into different vesicles (sorting motifs and decoding machinery), and deliver these vesicles to different membrane domains (cytoskeleton-mediated targeting, vesicle tethering through exocyst complexes and annexins) for membrane fusion (SNARE proteins). Protein complexes at the plasma membrane (the PAR, Crumbs and Scribble complexes) control the identity and distribution of functionally and structurally unique plasma membrane domains from the cytosolic face. Extrinsic cues provided by cell adhesion to the ECM and by other cells control the orientation of cell polarity. But how are these different levels of regulation of cell polarity integrated into a single network (FIG. 5)?
As mentioned above, PAR3–PAR6–aPKC complexes are localized at sites that are important for cell-polarity decisions: the boundary between the apical and basolateral membranes of epithelial cells (demarcated by the apical junctional complex), the cilium and the axon tip in neurons (FIG. 3). The localization of PAR3–PAR6–aPKC complexes to these sites could involve direct binding to cell-adhesion complexes or indirect recruitment by other complexes that are locally activated by adhesion at those sites (FIG. 5). Genetic studies in D. melanogaster suggest that PAR3 functions upstream of the cell–cell adhesion signal that is mediated by D. melanogaster E-cadherin102, but membrane proteins that recruit PAR3 to the apical junctional complex in D. melanogaster have not been identified. By contrast, in mammalian cells, PAR3 binds directly to the cell–cell adhesion proteins JAM-A103 and nectin104, both of which colocalize with E-cadherin at the apical junctional complex. Protein–protein interactions between the PDZ-domain-containing proteins of the Crumbs complex (PALS1 and PATJ) and proteins of the PAR complex (PAR3 and PAR6) provide links that in principle could allow co-recruitment of the PAR and Crumbs complexes to the pre-specified plasma membrane domain83 (FIG. 4a). PALS1 is also required for the trafficking of E-cadherin to the plasma membrane and thereby the localization of the exocyst complex to the plasma membrane105, indicating that complex feedback mechanisms are involved in the organization of these protein complexes at sites of cell–cell adhesion. Interestingly, disruption of one of the exocyst complex subunits, EXO84, results in the loss of Crumbs from the epithelial cell surface of early D. melanogaster embryos106. This suggests that the exocyst complex has a role in regulating the polarized organization of proteins at the apical junctional complex. Mechanisms involved in the localization of the Scribble complex to plasma membrane domains are poorly understood.
Local recruitment of the PAR and Scribble complexes to the plasma membrane occurs following cell-adhesion-mediated activation of CDC42 and RAC1 due to a complex regulatory network107 (FIG. 4a). CDC42 and RAC1 are members of the Rho family of small GTPases. Although the precise mechanisms are not clear, loss of activity of RAC1 and CDC42 inhibits the normal polarization of neurons and epithelial cells by causing a lack of axonal differentiation82 and the misorientation of the apico–basal axis95,108, respectively. This feature undoubtedly reflects the fact that components of the PAR complex are linked to the local regulation of CDC42 activity: CDC42 binds and recruits PAR6 to the cortex77,109; PAR3 binds the RAC1 guanine nucleotide-exchange factor (GEF) T-cell-lymphoma invasion and metastasis-1 (TIAM1)110; and loss of PAR3 inhibits CDC42-induced RAC1 activation in neurons82. Thus, the localization and activities of the PAR complex, CDC42 and RAC1, and perhaps the junctional complex itself, seem to be regulated by a positive-feedback loop (FIG. 4a).
Local asymmetries in the phospholipid content of plasma membrane domains also effect localization of CDC42 and cell polarity (FIG. 4a). Local changes in phosphoinositide synthesis are determined by the relative activities of phosphoinositide 3-kinase (PI3K) and phosphatase and tensin homologue (PTEN). PI3K generates phosphatidylinositol-3,4,5-trisphosphate phosphate (PtdIns(3,4,5)P3) and PTEN removes the 5′ from PtdIns(3,4,5)P3 to generate PtdIns(3,4)P2. Inhibition of PI3K activity affects the normal polarization of neurons, by causing a lack of axonal differentiation80, and of epithelial cells, by causing the misorientation of the apico–basal axis111. PtdIns(3,4,5)P3 is localized to the axon tip of neurons80. In polarized epithelial cells, PtdIns(3,4,5)P3 is enriched in the basolateral plasma membrane111 and is absent from the apical in which PtdIns(3,4)P2 membrane is enriched112 (FIG. 4a).
How might different phospholipids regulate the generation of cell polarity? Interestingly, annexin II, a putative scaffolding protein that is required for the delivery of sucrase-isomaltase to the apical membrane53, also binds PtdIns(3,4)P2 (REF. 113). Artificially induced accumulation of PtdIns(3,4)P2 on the basolateral membrane causes apical membrane proteins and annexin II to mislocalize to that membrane domain; concomitantly, the former apical membrane is disrupted112. Annexin II might in turn recruit other polarity-inducing proteins, such as CDC42 and aPKC (possibly through CDC42 as part of the PAR3–PAR6 complex). However, this cannot be the only mechanism involved in positioning CDC42 and aPKC. Active CDC42 might also be excluded from cell–cell contacts by a homologue of PAC-1, a Rho GTPase-activating protein (RhoGAP) that inactivates CDC-42 at cell–cell contacts in early C. elegans embryos114. It is interesting, however, that PAR3 also binds PTEN115,116 (FIG. 4a). This suggests a more complex local regulation of phospholipid synthesis on either side of the apical junctional complex.
How phosphoinositide asymmetry connects to the polarity of membrane proteins is not clear. For example, it is unknown whether differences in the distributions of PtdIns(3,4)P2 and PtdIns(3,4,5)P3 have a role in the fidelity of docking and fusion of Golgi and endosome-derived transport vesicles (FIG. 4a). CDC42 and RAC1 activate the ubiquitously distributed Wiskott–Aldrich syndrome protein neural isoform N-WASP and WASP family verprolin-homologous protein-1 (WAVE1), respectively, and together these proteins regulate actin polymerization by the actin-related protein-2/3 (ARP2/3) complex117. The mammalian exocyst subunit EXO70 associates with the plasma membrane by binding to PtdIns(3,4)P2 and PtdIns(3,4,5)P3. This membrane association is required for the recruitment of other exocyst components into the complex and the docking and fusion of post-Golgi basolateral transport vesicles with the plasma membrane118. Scribble also interacts with proteins that locally regulate CDC42 and RAC1 and regulate vesicle exocytosis. Scribble binds to the PAK-interacting exchange-factor-β–G protein-coupled receptor kinase-interactor (βPIX–GIT) complex119 (FIG. 4b), of which βPIX is a GEF for CDC42 and RAC1 (REF. 120), and GIT is a GAP121. The βPIX–GIT complex regulates Ca2+-dependent docking and fusion of vesicles with the plasma membrane, and βPIX–GIT function requires its GEF activity and membrane localization by Scribble119 (FIG. 4b). Importantly, Scribble mutants in the central nervous system result in the accumulation of vesicles at synaptic boutons and in decreased vesicle release, suggesting that Scribble is important for vesicle docking and fusion with the synaptic membrane122. Whether the Scribble–LGL–DLG complex directs local tethering and fusion of vesicles through local regulation of CDC42 and RAC1 activity, and through recruitment of cognate t-SNAREs, requires further study. Thus, control of vesicle trafficking by the exocyst, actin polymerization induced by CDC42 and complexes such as Scribble and βPIX–GIT could reinforce the overall polarized phenotype that is induced by the PAR complex.
Such a mechanism shows how intrinsic cellular machinery for specifying cell polarity is intimately coupled to extrinsic spatial cues; these cues provide a template that coordinates the development of complex tissue organization in 3D space. Indeed, it seems likely that even in single polarized cells, the same intrinsic machinery can be used for different purposes, depending on whether a cell is generating polarity de novo or is maintaining a previously formed polarized phenotype. This, in turn, might simply reflect the fact that the spatial cues detected during and following polarity induction are, themselves, different.
Three major layers of regulation that control cell polarity have been identified on the basis of genetic, biochemical and cell biological criteria (FIG. 5). Intrinsic mechanisms sort membrane proteins into different post-Golgi or endosomal vesicles and deliver those vesicles to different membrane domains for membrane fusion. Protein complexes at the plasma membrane (the PAR, Crumbs and Scribble complexes) control the identity of plasma membrane domains. Extrinsic cues that are provided by cell adhesion to the ECM and to other cells control the orientation of cell polarity.
Although the basic biochemical rules that govern protein sorting in the exocytic or endocytic pathway have been described, and important signalling complexes that control cell polarity have been identified, many of the details remain poorly understood. How are apical vesicles formed? What is the full repertoire of adaptor proteins that is required for basolateral protein sorting? How are intrinsic protein sorting and trafficking pathways integrated with functions of the Crumbs, PAR and Scribble complexes in specifying membrane-domain identity? How are extrinsic cell-adhesion orientational cues linked to the intracellular machineries that define membrane-domain organization? Answers to these questions will shed new light on many fundamental aspects of normal development and on the mechanisms that underlie the induction and progression of many diseases, such as cancer.
W.J.N. was supported by a grant from the National Institutes of Health (GM35527).
AP180 | CDC42 | EPS15 | GGA | LDLR | LKB1 | NMDAR | PAR1b | VAMP
ALL LINKS ARE ACTIVE IN THE ONLINE PDF