ReviewCell Biology

Plasma membrane–associated platforms: Dynamic scaffolds that organize membrane-associated events

See allHide authors and affiliations

Science Signaling  10 Mar 2015:
Vol. 8, Issue 367, pp. re1
DOI: 10.1126/scisignal.aaa3312


Specialized regions of the plasma membrane dedicated to diverse cellular processes, such as vesicle exocytosis, extracellular matrix remodeling, and cell migration, share a few cytosolic scaffold proteins that associate to form large plasma membrane–associated platforms (PMAPs). PMAPs organize signaling events and trafficking of membranes and molecules at specific membrane domains. On the basis of the intrinsic disorder of the proteins constituting the core of these PMAPs and of the dynamics of these structures at the periphery of motile cells, we propose a working model for the assembly and turnover of these platforms.


The cell membrane includes specialized areas that are required to perform specific functions. These areas are formed by the interaction of integral components of the membrane with extracellular and intracellular components. Examples of these specialized areas include the presynaptic active zones for neurotransmitter release (1), various types of invadosomes for the degradation and remodeling of the extracellular matrix (2), and the leading edge of motile cells (3). These specialized membrane-associated regions coordinate the complex structural and signaling events required for specific processes by engaging proteins that mediate adhesion to extracellular substrates, signaling receptors, adaptor proteins, and cytoskeletal assemblies, and by guiding the traffic of proteins and membranes from and to the specialized membrane areas (4). In this Review, evidence from the molecular and functional characterization of some of these specialized regions indicates that they share a core complex, which includes a limited number of scaffold and adaptor proteins. Scaffold proteins act as platforms to create a protein network, and adaptors are usually smaller proteins that function as a bridge between two proteins. The core proteins interact with each other to organize stable or dynamic supramolecular assemblies at specific membrane locations. We refer to these specialized regions at the interface between the cytosol and plasma membrane as plasma membrane–associated platforms (PMAPs). We describe evidence indicating the involvement of similar core complexes in the formation of PMAPs present in functionally diverse membrane specializations, such as presynaptic nerve terminals, podosomes in the postsynaptic membrane of the neuromuscular junction, and the leading edge of migrating normal or tumor cells.

Partitioning of specific cytosolic proteins by phase transition may represent a fundamental mechanism for the molecular organization of the cytoplasm (5). Membrane-less organelles or hydrogels enriched in specific proteins may form spontaneously in the cytoplasm or nucleoplasm as a consequence of liquid-liquid demixing that results from a phase transition (6). Intrinsically disordered proteins (IDPs) or intrinsically disordered regions (IDRs) within proteins (7, 8) are present in a large fraction of the human proteome (9) and are important for the assembly by phase transition of membrane-less organelles and other cytoplasmic or nucleoplasmic structures. The low complexity of IDRs provides the structural flexibility enabling the interaction with distinct partners and favors the formation of large assemblies driven by phase transition (10, 11). Because the proteins of the core complex of the PMAPs include extended IDRs and on the basis of the available limited amount of experimental data, we propose a model in which phase transitions within the cytoplasm may play a central role in regulating the assembly and disassembly of these specialized regions.

Common Scaffold Proteins of PMAPs

The molecular organization of the specialized regions of the plasma membrane devoted to complex cellular processes, such as neurotransmitter release, cell migration, extracellular matrix remodeling, and cell invasion, coordinates in space and time the events that are needed to perform these processes. Many of the morphological changes associated with these processes involve dynamic regulation of the actin cytoskeleton and structures associated with the cell cortex, which is the actin-rich layer beneath the plasma membrane. Moreover, these processes often involve the delivery of vesicles to specific membrane sites and endocytosis from the same sites (4). These trafficking events involve different types of proteins, including the guanosine triphosphatases (GTPases) of the Rab family (12) as well as microtubules (13, 14). In cell migration, for example, the protruding edge of the motile cell requires precise coordination of adhesion and de-adhesion to the extracellular environment, actin filaments treadmilling to drive protrusion, trafficking of membranes to the growing edge, and delivery of signaling molecules to regulate these processes (4, 15). In the presynaptic nerve terminals, synaptic vesicles fuse at the active zone, a site specialized for neurotransmitter release, where a complex network of adaptor proteins forms a cytoplasmic matrix anchored to the plasma membrane (1).

PMAPs are assembled at or near membrane domains defined by the presence of specific transmembrane components. At synapses between neurons, the clusters of transsynaptic adhesion molecules identify the sites for assembly of presynaptic PMAPs (1); at the neuromuscular junction, the clusters of acetylcholine receptors (AChRs) in the plasma membrane of the muscle fiber define the site for assembly of podosomes involved in maturation of the postsynaptic membrane (16); in nonneuronal cells, PMAPs may form near focal adhesions (17, 18), the integrin-based adhesions that link the extracellular matrix to the cytoskeleton (15).

Despite their different functions and dynamics, these PMAPs contain common scaffold proteins that are important for their assembly. These proteins include the scaffolds liprin-α and ELKS (protein rich in the amino acids E, L, K, and S), the phosphoinositide-binding protein LL5, and the microtubule-binding protein CLASP (CLIP-associated protein) (1922). We propose that these four proteins—liprin-α, ELKS, LL5, and CLASP—form part of a core complex (Fig. 1A) of PMAPs in different cells, where they function in organizing sites of exocytosis, sites of extracellular remodeling, or sites of dynamic protrusion during migration.

Fig. 1 Interactions and domain organization of the core components of PMAPs.

(A) Scheme of the interactions among the four identified core proteins (light blue), and with their binding partners regulating exocytosis or migration and cell polarity. Full lines indicate direct binding, and dotted lines indicate indirect interactions. (B to E) Schematic diagrams of the domains and structures of the core PMAP proteins. The N-terminal portion of liprin-α1 contains the region responsible for homodimerization (not shown). The PDZ-binding motif (Val-Arg-Thr-Tyr-Ser-Cys peptide) is absent in some alternative splice variants of liprin-α1. ELKS1β and ELKS2 each have a C-terminal PDZ-binding motif that mediates the interaction with the presynaptic protein RIM. The PH domain of LL5β mediates the interaction with PI(3,4,5)P3. The Ser-x-Ile-Pro (SxIP) motif in CLASP mediates microtubule plus tip tracking through EB1, and the C-terminal coiled-coil region binds LL5β. The white lines along each protein define the binding regions for the indicated core proteins. Binding partners mentioned are listed next to the protein diagram.


Liprin-α1 is part of the liprin family of scaffolds with conserved structure (Fig. 1B) that were identified as interacting partners of the tyrosine phosphatase LAR (23), a transmembrane protein that binds various extracellular ligands. The four mammalian liprin-α proteins, together with the two liprin-β and the liprin-related protein liprin-γ (also known as kazrin), form the liprin family (24). The N-terminal part of liprin-α proteins is predicted to form coiled coils and includes two highly conserved sequences referred to as liprin homology domains 1 and 2 (25). The N-terminal region of liprin-α proteins mediates the formation of homodimers (26) and interacts with several partners that are involved in regulation of vesicular trafficking or cytoskeleton, including the adaptors of the ELKS family (27, 28); the multidomain Arf GTPase-activating proteins (GAPs) of the GIT family (29) that regulate cell adhesion, migration, and membrane trafficking (30); the synaptic protein Rab3-interacting molecule (RIM) that participates in vesicle release at the presynaptic active zone (25); the Rho GTPase effector mDiaphanous (31) that is involved in cytoskeleton remodeling (32); and the motor KIF1A that takes part in the transport of membranous organelles and molecular complexes along microtubules (33). The C-terminal region includes three sterile α motifs (SAMs) (34) and binds to liprin-β proteins, the transmembrane tyrosine phosphatase LAR (23), which is a receptor involved in axon guidance (35) and synaptogenesis (36), and the multidomain protein kinase CASK, which organizes the presynaptic vesicle pools (37).

ELKS proteins

The mammalian ELKS family consists of two members: ELKS1 (also known as ELKS, ERC1, CAST2, and Rab6IP2) and ELKS2 (also known as ERC2 and CAST1), and each occurs in several splice variants (38, 39). ELKS1 (Fig. 1C) was first identified in papillary thyroid carcinoma as a protein fused upstream to the oncogene RET that is constitutively activated by the region involved in the homodimerization of ELKS (40). This protein was subsequently identified as Rab6-interacting protein2 (Rab6IP2) (41) and as an active zone protein called CAST or ERC (38, 42). ELKS proteins are predicted to be largely made of coiled coils with no defined domains. Among various direct binding partners of ELKS are the adaptors liprin-α (27) and LL5β (17), the presynaptic active zone protein RIM (38), and the GTPase Rab6 that regulates the transport and targeting of exocytotic carriers (43).

LL5 proteins

The LL5 family members LL5α (also known as PHLDB1) and LL5β (also known as PHLDB2) have a C-terminal pleckstrin homology (PH) domain that can interact with phosphoinositides, including phosphatidylinositol 4,5-bisphosphate [PI(4,5)P2] and phosphatidylinositol 3,4,5-trisphosphate [PI(3,4,5)P3], which may direct the recruitment of the cytosolic LL5 proteins to the plasma membrane (44, 45). LL5β is the best characterized of this family (Fig. 1D) and interacts with ELKS, the actin filament cross-linking protein filamin (45), and the CLASP family of microtubule-binding proteins (17).

CLASP proteins

CLASP proteins bind to the plus tip and the side of microtubules, where they regulate microtubule dynamics by promoting microtubule stabilization. CLASPs are multidomain proteins with three tumor overexpressed gene (TOG)–like domains, a Ser-x-Ile-Pro (SxIP, where x is any amino acids) motif, and a C-terminal coiled-coil region that mediates homodimerization (22, 46) (Fig. 1E). CLASP proteins interact directly with LL5β and with the microtubule plus end-tracking proteins EB1 and CLIP-170, both of which regulate microtubule dynamics (47).

PMAPs for Vesicle Exocytosis

Exocytosis in neurons

The presynaptic active zone is an organized structure that serves as the site of vesicle docking, priming, and fusion to release neurotransmitter in response to neuronal activity (1, 48). These events are tightly organized (49). We propose that the core components liprin-α and ELKS are part of PMAPs that organize the presynaptic active zone (Fig. 2A). In the nematode Caenorhabditis elegans, the liprin-α homolog SYD-2 regulates the morphology of the presynaptic terminals for the efficient delivery of synaptic vesicles to the active zone (50). In both Drosophila melanogaster and C. elegans, the homologs of liprin-α interact with the LAR family receptor protein tyrosine phosphatases to define the presynaptic active zone (51, 52). In mammals, liprin-α2 promotes the turnover of scaffold complexes in synapses and facilitates the release of synaptic vesicle (53). Moreover, in D. melanogaster, the interaction of liprin-α with liprin-β promotes synapse formation, whereas liprin-γ counteracts this function (54).

Fig. 2 Models for PMAPs associated with exocytosis.

(A) Diagram of the PMAP at the presynaptic active zone. Piccolo, Bassoon, and RIM are neuronal-specific adaptor proteins. Nidogen is a glycoprotein of the extracellular matrix that binds LAR. (B) Diagram of the PMAP associated with sites of exocytosis in nonneuronal cells. What specifies the site of LL5 association with the exocytic site is unknown.

ELKS is implicated in the presynaptic localization of liprin-α (27). In C. elegans, a mutant in SYD-2 with increased binding to ELKS enhances synapse formation (28). SYD-2 also promotes the assembly of electron-dense projections that dock synaptic vesicles at the active zone (55). Similarly, mutation of the Drosophila ELKS-related protein bruchpilot causes the loss of active zones and associated electron-dense projections (T-bars) and reduces evoked vesicle release (56, 57). Knocking out ELKS2 in mice reduces the size of the active zone in rod photoreceptors (58) and decreases action potential–triggered calcium influx in presynaptic terminals of hippocampal interneurons (39). In the ELKS2-knockout mice, the ultrastructure of the synapses was preserved (59), which may indicate that ELKS1 can partially compensate for the loss of ELKS2.

These data indicate that liprin-α and ELKS assemble the presynaptic PMAPs that create the correct environment for efficient neurotransmitter release. To perform this duty, liprin-α and ELKS need to cooperate with other proteins that are known to be involved in the organization and function of the presynaptic active zone, such as the small GTPase Rab3, the Rab3 scaffolding effector RIM (60, 61), the priming factor Munc13 (62, 63), and the adaptors piccolo or bassoon (48).

Exocytosis in nonneuronal cells

ELKS is involved also in exocytosis by pancreatic β cells (64), mast cells (65, 66), PC12 cells (67), and HeLa cells (43). In pancreatic β cells, insulin granules dock and fuse at the sites of stable ELKS1-positive clusters that colocalize with syntaxin 1, a plasma membrane–associated soluble N-ethylmaleimide–sensitive factor attachment protein receptor (SNARE) for vesicular fusion (64). Depletion of ELKS1 in HeLa cells reduces the secretion of Golgi-derived Rab6-positive vesicles, resulting in their accumulation at the cell periphery (43). These studies suggest that ELKS is necessary for the recruitment of vesicles to specific PMAPs at the plasma membrane where exocytosis preferentially occurs (Fig. 2B). In HeLa cells, ELKS1 colocalizes with LL5β to patches that function as sites for preferential docking and fusion of the Rab6-positive vesicles (43). These vesicles are delivered from the Golgi to the plasma membrane by traveling along microtubules. Transport of the vesicles from the Golgi to the plasma membrane is ELKS-independent, but ELKS reduces the delay between docking and fusion of the Rab6-positive vesicles with the plasma membrane. Rab8A colocalizes with Rab6 on the Golgi-derived vesicles and targets them to the cell periphery (68). MICAL3 is a monooxygenase that interacts with Rab8A and ELKS1 and colocalizes with ELKS1- and LL5β-positive cortical patches (68). Thus, MICAL3 may mediate the docking of the Rab8A- and Rab6-positive vesicles at the PMAP, where the monooxygenase activity of MICAL3 may promote the disassembly of cortical actin (69, 70) to facilitate the fusion of the vesicles with the membrane (Fig. 2B).

As with synaptic vesicle exocytosis, depletion of ELKS1 does not abolish secretion in nonneuronal cells, indicating that the role of neuronal and nonneuronal PMAPs is to spatially confine and improve the efficiency of the fusion process.

PMAPs at Sites of Extracellular Matrix Remodeling Between Cells

Podosomes are specialized F-actin–rich adhesive structures that mediate extracellular matrix remodeling and cellular invasion (71). Synapse formation requires the clustering of neurotransmitter receptors in the postsynaptic membrane facing the presynaptic nerve terminal. The developing postsynaptic site of the mammalian neuromuscular junction has podosomes that promote the remodeling of the extracellular matrix and of the immature plaque-shaped clusters of AChRs into mature branched AChR-positive junctions (16) (Fig. 3).

Fig. 3 PMAPs at podosomes of the neuromuscular junction.

Diagram of the PMAP associated with the podosomes at the postsynaptic membrane of the developing neuromuscular junction.

LL5β is required for AChR clustering in the postsynaptic membrane of myotubes (72). LL5β binds ELKS and the actin filament cross-linking protein filamin (17, 45). Postsynaptic podosomes have an actin-rich core where Asef2, a Rac and Cdc42 guanine nucleotide exchange factor (GEF), and the actin-binding proteins cortactin, filamin, Amotl2, and Flii are localized (73). The PH domain of LL5β concentrates LL5β and ELKS in the PMAP around the actin-rich core, in association with the focal adhesion proteins integrin β1, paxillin, talin, and vinculin (16). Maintenance of mature AChR clusters also requires the attachment of microtubules to the PMAP through CLASP (74).

The PMAP components LL5β and ELKS are enriched adjacent to the podosomes of macrophages (73). Cancerous cells make podosome-related structures called invadopodia that degrade the extracellular matrix, and LL5β and ELKS are also enriched in the region around these structures in Src-transformed NIH3T3 fibroblasts (73). As at the neuromuscular junction PMAP, Asef2, filamin, and Amotl2 colocalize with the central F-actin core of these structures (73). Thus, ELKS and LL5β are part of PMAPs associated with the organization of various types of podosomes and invadopodia. Because LL5β can bind CLASP, this PMAP may serve to anchor CLASP and thus microtubules to this site to enable trafficking of the podosome-promoting kinesin KIF1C (75). How the PMAPs associated with podosomes and invadopodia influence the formation and dynamics of these membrane structures requires further studies.

PMAPs in Cell-Substrate Interactions in Stationary and Motile Cells

PMAP components near focal adhesions

The four core proteins liprin-α1, ELKS1, LL5, and CLASP are constituents of stable PMAPs at the cortex of nonmotile HeLa cells (17, 76) (Fig. 4A). At these sites, LL5β anchors the plus ends of microtubules to the cell edge by interacting with the microtubule-stabilizing protein CLASP (17). ELKS, LL5β, and CLASP colocalize at the plasma membrane facing the extracellular matrix, near to, but distinct from, peripheral focal adhesions. The scaffolding proteins liprin-α1 and liprin-β1, the kinesin KIF21A that inhibits microtubule growth at these sites, and its binding partner KANK1, a protein with coiled-coil domains and ankyrin repeats (77), are also components of these PMAPs (76).

Fig. 4 Focal adhesion–associated PMAPs in stationary and motile cells.

(A) Diagram of a stable PMAP adjacent to a focal adhesion in a stationary cell. (B) Diagram of the PMAP adjacent to a dynamic focal adhesion in a moving cell.

How are the PMAPs at the cell cortex assembled? LL5β is required for the localization of ELKS and CLASP at the cell cortex, but not for the localization of CLASP at the plus ends of microtubules (17). The cortical localization of LL5β and ELKS1 is unaffected by microtubule depolymerization. Accordingly, CLASP depletion, which results in the loss of cortically associated microtubule tips from the cell margin (78), does not affect the localization of LL5β-positive clusters at cortical PMAPs (17). A hierarchy of interactions has been suggested: (i) LL5β is recruited to the plasma membrane by the interaction of its PH domain with PI(3,4,5)P3, which is produced in response to signals that activate phosphatidylinositol 3-kinase (PI3K) (45), and (ii) LL5β then recruits ELKS and CLASP to the cell cortex (Fig. 4A). However, the finding that PI3K inhibitors diminish, but do not abolish, the cortical localization of LL5β indicates that factors other than PI(3,4,5)P3 are required to recruit this protein to the cytosolic face of the membrane (17).

The assembly of PMAPs is linked to integral membrane components. Specific laminin receptors are needed for the organization of PMAPs at the ventral, extracellular matrix facing membrane of polarized human mammary MCF-10A epithelial cells. In these polarized cells, LL5α and LL5β localize near the basal extracellular matrix protein laminin-5 and its integrin receptors α3β1 and α6β4 (79). Activation of integrin α3 triggers the formation of the LL5-positive patches at the basal side (79). The LL5-positive PMAPs mediate the laminin and integrin-induced attachment of microtubule plus ends to the basal cortex of epithelial cells (79). A similar role of LL5 has been identified in vivo: during gastrulation, LL5 and CLASP are required for anchoring microtubules to the basal cortex of epiblast cells in the embryo. Anchoring is mediated by the transmembrane dystroglycan that links the cytoskeleton to the extracellular basement membrane (80).

Regulation of focal adhesion turnover by PMAPs at the leading edge

Liprin-α1, ELKS1, LL5, and CLASP proteins are also part of a molecular network that promotes the protrusive activity at the front of migrating cells (Fig. 4B). During integrin-mediated migration on extracellular matrix, liprin-α1 promotes the turnover of the focal adhesions at the protruding cell edge (81). Liprin-α1, together with ELKS1 and LL5 proteins, are part of dynamic and polarized structures near the protruding front of tumor cells in culture and contribute to migration and invasion of tumor cell lines in Matrigel invasion assays (18, 82). Similar dynamic LL5-positive structures are observed at the protrusive edge of human mammary MCF-10A epithelial cells (79) and at the front of migrating human keratinocytes, where LL5β colocalizes with CLASP near peripheral focal adhesions (83). The PMAPs observed in nonmotile cells and the dynamic PMAPs in migrating cells may represent two functional states of the same type of supramolecular assemblies.

Turnover of integrin-mediated adhesion to the extracellular matrix, consisting in the disassembly of the adhesive structures called focal adhesions and their reformation, is essential for cell migration. On the basis of the localization of PMAPs near focal adhesions in migrating cells, we hypothesize a role for PMAPs in the turnover of such adhesive structures (17, 18, 23, 79). Increasing the abundance of either liprin-α1 or ELKS1 induces a higher turnover of focal adhesions, whereas liprin-α1 knockdown reduces adhesion turnover (18, 81). Integrin endocytosis and recycling back to the cell surface is required for focal adhesion turnover during cell locomotion (84), and silencing liprin-α1, ELKS1, or LL5 inhibits the internalization of active β1 integrins (18). In keratinocytes, the LL5β-mediated clustering of microtubule-associated CLASP around focal adhesions is temporally correlated with focal adhesion turnover. LL5β or CLASP depletion enhances focal adhesion lifetime and results in a sharp reduction of directional migration (83).

How can PMAPs affect focal adhesion dynamics? Signaling by motogenic receptors causes the transient polarized production of PI(3,4,5)P3 at the leading edge of migrating cells. PI(3,4,5)P3-induced recruitment of LL5β at the cell edge may drive the assembly of dynamic PMAPs that include ELKS1, liprin-α1, and CLASP (Fig. 4B). Microtubules targeted to the leading edge by CLASP through the interaction with LL5β (17) would then promote the turnover of focal adhesions during migration (85).

Stabilization of cellular protrusions by PMAPs at the leading edge

Depletion of either liprin-α1, ELKS1, or LL5 decreases the stability of lamellipodia and the migration of tumor cell lines in two- and three-dimensional matrices (18). Moreover, LL5β is required for epithelial growth factor–induced large lamellipodia and migration of COS7 cells (86). One mechanism that may underlie the positive effects of these proteins on the stability of lamellipodia is the recruitment of filamin near the membrane through an interaction with LL5β (45, 87). Filamin cross-links actin filaments (88) and is necessary for the LL5β-mediated formation of large lamellipodia (86). Moreover, filamin can bind the inositol 5-phosphatase SHIP2, which hydrolyzes PI(3,4,5)P3 to PI(3,4)P2 (89) and may therefore regulate the dynamics of lamellipodia by decreasing PI(3,4,5)P3 abundance at the protruding cell edge (86). These data support a model in which liprin-α, ELKS1, and LL5β are constituents of dynamic PMAPs at the protruding edge of migrating cells, where they play an important role of coordination of membrane dynamics by modulating actin and microtubule networks through the interactions of LL5β with filamin and CLASP, respectively.

PMAP Partner Proteins at Sites of Cell-Cell Contact

PMAPs may be important for organizing membrane domains associated with cell-cell interactions, such as those between a T cell and an antigen-presenting cell or those between epithelial cells in an epithelial sheet. Although direct evidence for the implication of the four components of the core complex at these sites is not available, many of their partner proteins (Fig. 1A) are located at these specialized membrane domains.

For example, the liprin-α1–interacting protein GIT1 is recruited together with the Rac and Cdc42 GEF PIX and the kinase Pak1, which is activated by Rac and Cdc42, to the immunological synapse at the T cell–antigen-presenting cell contact site (90). The LL5-interacting protein filamin is also recruited to immunological synapses (91). PMAP-associated proteins have also been implicated in the regulation of different types of cell-cell junctions. In keratinocytes, liprin-γ localizes to desmosomes and associates with stabilized microtubules, which undergo extensive reorganization during keratinocyte terminal differentiation. Thus, liprin-γ may contribute to desmosome formation and epidermal differentiation by participating in the reorganization of microtubules (92). The cadherin-catenin complex is part of adherens junctions and colocalizes with liprin-α1 partner LAR at these cell-cell contact sites. LAR associates with the adherens junction proteins β-catenin and γ-catenin (also known as plakoglobin) and inhibits epithelial cell migration (93).

Dynamic Behavior of PMAPs

Characteristics of static and dynamic PMAPs

The available data indicate that a limited number of scaffolds and adaptors play a central role in the formation of PMAPs at diverse cellular locations in different cell types. It appears that distinct PMAPs are built by linking the core proteins to distinct sets of additional ligands. Not only do these PMAPs have different partners at the different structures, but there is a clear kinetic difference between the stable PMAPs at the presynaptic active zone in neurons or at the periphery of stationary cells and the dynamic PMAPs at the front of migrating cells. In the presynaptic active zone, the core scaffold proteins liprin-α and ELKS assemble a neuron-specific network linked to receptors and trans-synaptic adhesion molecules that bind postsynaptic ligands (94), which may provide the basis for long-lasting PMAPs. In nonneuronal cells, the cortical peripheral PMAPs in stationary and migrating cells share several components and features, which suggest that they represent distinct functional states of the same type of PMAP. In both cases, liprin-α and ELKS1 are connected to LL5β, which in turn binds to microtubules through CLASP. Static and dynamic PMAPs are closely apposed to peripheral focal adhesions, the transmembrane connections between cytoskeleton and extracellular matrix that are stable in nonmotile cells and highly dynamic in migrating cells. The appearance of dynamic PMAPs is linked to active protrusions at the front of migrating tumor or epithelial cells. Moreover, PMAP dynamics are similar in both epithelial and tumor cells: Small, punctate particles that are positive for PMAP core proteins are released from focal adhesion–associated sites and undergo retrograde flow toward the center of the cell before slowly disappearing (18, 83).

Assembly and disassembly of PMAPs: Implication of IDPs and phase transitions

On the basis of the biochemical properties of the core components of PMAPs and on the findings related to their dynamic behavior, we propose a working model in which the dynamic PMAPs that form at the cell periphery and at the front of migrating cells originate from the partitioning of the cytosolic PMAP core proteins by phase transition. Phase transitions occur when molecules convert from one configuration to a different one. Water can undergo phase transitions from vapor, to liquid, to solid ice. Also proteins, lipids, or other macromolecules can undergo phase transitions (10, 95). In the case of proteins, they can undergo phase transitions from a soluble monomeric state to an insoluble large assembly form that can still exhibit liquid-like behavior, such as the liquid droplets observed with concentrated protein solutions in vitro (8) or in cells (96). Partitioning of specific cytosolic proteins by phase transition may represent a fundamental mechanism for the organization of the cytoplasm or nucleoplasm (5, 96, 97). Examples of such assemblies are the P granules of C. elegans embryos or the stress granules in the cytoplasm (98, 99) and the Cajal bodies or the nuclear pores in the nucleus (100, 101).

IDRs and IDPs can participate in the formation of large protein assemblies driven by phase transitions (7, 8, 10, 11). IDRs can be identified by sequence analysis for low complexity and low content in hydrophobic residues that prevents the folding into domains (102). Upon binding to each other, to specific partners, or both, IDRs often undergo a disorder-to-order transition and fold into defined structures that help the formation of large molecular assemblies by phase separation (103). IDRs of more than 30 amino acids are present in 44% of the human proteins (9), with signaling proteins having a greater tendency to intrinsic disorder than other types of proteins (104).

Analysis with the D2P2 database (9) to predict disorder indicated that the four identified core constituents of the dynamic PMAPs—liprin-α1, ELKS1, LL5β, and CLASP1—have extended segments of IDRs (Fig. 5). These predicted disordered regions represent a major part of the coiled-coil regions of liprin-α1, ELKS1, LL5β, and the segments between the TOG and TOGL domains in CLASP1 (Fig. 1, B to E). The IDRs of the four proteins are enriched in segments called molecular recognition features (MoRFs) that are predicted to promote disorder-to-order transitions during protein-protein interactions (105, 106), as well as phosphorylation sites (PTM sites), which were predicted from PhosphoSitePlus (107, 108). In addition to IDRs, LL5β, liprin-α, and CLASP have structured domains that mediate protein or lipid binding. Moreover, liprin-α, ELKS, and CLASP form homodimers (22, 26, 40), and liprin-α1 may be regulated by an autoinhibitory mechanism (109). The structural and sequence characteristics of these partially or largely disordered PMAP components suggest the possibility of their combinatorial regulation by autoinhibition, homo- and hetero-oligomerization, as well as posttranslational modifications. Moreover, their structural flexibility may enable these disordered proteins to interact with distinct sets of partners in different PMAPs, resulting in assemblies with different dynamics and functions (Figs. 2 to 4).

Fig. 5 The predicted disorder of the four PMAP core proteins.

For each protein, the upper scheme shows the predicted domains (no domains are predicted for ELKS) and core protein interacting sites; the lower scheme includes the predicted disordered regions. In light green are the disordered regions predicted by the PrDOS program (121); in dark green are the more stringent predictions based on meta-analysis with the D2P2 database (9). Only dark green regions of at least 20 consecutive amino acids are shown in the schemes; shorter sequences highlighted by this analysis have been omitted.

Cues for localized phase transition–induced PMAPs

Phase separation inside cells occurs when the interacting components reach a critical concentration and is mediated by the weak homotypic or heterotypic interactions among IDPs or IDRs (6, 11). Although the spontaneous nucleation of a hydrogel is a rare event, nucleation may be triggered by the assembly of some of the molecular components at a preexisting site. In the case of dynamic PMAPs, transiently produced molecular cues, such as phosphoinositides or other unknown membrane components, may increase the local concentration of IDPs required for the assembly of a PMAP (Fig. 6A). One example is the plasma membrane–triggered phase transition that occurs during the assembly of myelin. The association of the cytosolic intrinsically disordered myelin basic protein (MBP) with the inner leaflet of the plasma membrane induces a phase transition in a membrane-associated MBP network that drives protein segregation from the cytoplasmic leaflets of two opposing membranes involved in the formation of the myelin sheet (110).

Fig. 6 A working model for phase transition–induced PMAP assembly and disassembly.

(A) Model for the assembly and disassembly of large molecular aggregates by phase transition: the recruitment of cytosolic proteins to the cytoplasmic face of the plasma membrane (1) induces demixing of the cytosolic components by phase transition (2). Upon changes in the local environment, the aggregates detach and move into the cytosol as membrane-less protein-rich structures (3), where they eventually dissolve (4). The soluble cytosolic proteins become available to form new aggregates (5). (B) Working model for the dynamic turnover of PMAPs at the edge of migrating cells. In migrating cells, particles containing the PMAP core proteins are formed by plasma membrane–triggered phase transition. These particles form close to the dynamic focal adhesions at the actively protrusive front of the cells. Soon after forming, membrane-less particles detach from the cytoplasmic side of the plasma membrane and move retrogradely from the focal adhesion–associated sites toward the cell center. These particles then disassemble and the constituents can reform the PMAP at a new site.

As for MBP, we propose that the recruitment of one or more PMAP core proteins at specific sites of the plasma membrane increases their local concentration and, by helping the interaction with the other PMAP core partners, facilitates the formation by phase transition of a membrane-associated network that includes some or all the cytosolic core components (Fig. 6A). Consistent with this model, liprin-α1 associates with the cytoplasmic surface of the plasma membrane (81, 111) and also with transmembrane proteins, such as LAR, which may enable liprin-α1 to initiate the phase separation, leading to the formation of PMAPs. In dynamic PMAPs, the local production of PI(3,4,5)P3 may recruit LL5β through its PH domain, and then the IDRs of LL5β may drive the concentration of itself and other disordered partners, thus facilitating the formation of a hydrogel that constitutes a membrane-bound signaling platform.

Reversibility of phase transition–induced PMAPs

The assemblies induced by the partitioning of IDRs by phase transition are reversible and may be disassembled by dilution of their components or by posttranslational modifications, such as phosphorylation (112) (Fig. 6A). In a dynamic situation such as a migrating cell, once the local conditions at the membrane change (for example, by local hydrolysis of active phosphoinositides), the PMAPs may lose their association with the membrane and move into the cytoplasm, where they move away from the membrane through flow mediated by actomyosin contraction or motor-driven transport along microtubules or a combination of these (Fig. 6B). One of the mechanisms that may underlie the release of PMAPs from the membrane is the recruitment of the lipid phosphatase SHIP2 by filamin-LL5β complexes (86). This would decrease the concentration of PI(3,4,5)P3 at the protruding membrane, leading to the detachment of the multimolecular complexes as free cytoplasmic protein particles containing the four disordered core components. The PMAP-derived particles would eventually be disassembled because of the dilution of the components as they move toward the center of the cell or because of posttranslational modifications affecting the homo- and heterotypic interactions among the core components, or a combination of these mechanisms (Fig. 6, A and B).

The proposed reversible switch between the active and inactive forms of liprin-α1 may also contribute to the dynamics of PMAPs (109). In the active form, liprin-α1 can interact with several binding partners, and the dimeric nature of this protein and of some of its binding partners may contribute to the assembly of very large aggregates to which other structural and signaling proteins may be recruited (19, 113). The switch of liprin-α to the inactive state may cause fast disassembly of the large aggregates.

PMAPs as proteinaceous liquid droplets

The working model presented here is speculative, and the evidence in support of the hypothesis that dynamic PMAPs are reversible membrane-triggered aggregates formed by phase separation is mainly circumstantial. The PMAP components liprin-α1, ELKS1, LL5β, and CLASP behave differently from focal adhesion components (18) and from nearby trafficking vesicles (83). Liprin-α1–, ELKS1-, and LL5-positive structures do not colocalize with any of many markers for cytoplasmic vesicular exo- or endocytic compartments (18). Moreover, a PH-defective LL5β mutant unable to bind PI(3,4,5)P3 accumulates together with ELKS in cytoplasmic structures that may be the result of a phase transition into membrane-less aggregates, rather than the accumulation at an undefined membrane compartment (18, 45).

PMAP behavior resembles that of naturally occurring (98) or experimentally induced liquid droplets (8). The punctate ELKS1-positive structures that move by retrograde flow from the edge toward the center of migrating cells tend to merge into larger dynamic structures (18), which is a characteristic of liquid droplets as they get closer to each other (8, 98). The shape of these ELKS1- or LL5β-positive structures is dynamic, similar to the behavior of liquid droplets that deform under shear flow (98).

Future analysis with approaches that can explore the biophysical properties of these large molecular assemblies, such as fluorescence recovery after photobleaching (FRAP) (98) and electron microscopy (114), may provide further information on PMAPs. FRAP analysis has been used to analyze the behavior of germline P granules, a type of liquid droplets that concentrate to the posterior of the one-cell embryo of C. elegans by controlled dissolution and condensation of the constituents in specific regions of the cell (98). This analysis revealed the rapid diffusion of proteins within P granules and indicated that the viscosity of these structures is comparable to that in colloidal liquids (115). Bleaching experiments may also be used to address the dynamics of the structures under study and to establish their equilibrium with soluble components. Three-dimensional particle tracking enables analysis of the events of condensation and dissolution of liquid droplets (98). Sharp liquid-liquid demixing phase separations can be induced by the interactions between diverse multidomain proteins, resulting in the generation of micrometer-sized liquid droplets in aqueous solution in vitro (8). Supported lipid bilayers have been used to study in a cell-free system the role of membrane-induced phase transition of MBP (110), which is a protein involved in myelin assembly.

Function and Dysfunction of PMAPs

The dynamic PMAPs assembled at the cell front may serve different purposes. On the one hand, they may guide the accumulation of the machinery regulating the turnover of focal adhesions and actin dynamics to increase the efficiency of protrusion. Consistent with this, the speed of N-WASP–stimulated formation of Arp2/3-induced actin filaments is increased when tested in vitro with assembled protein droplets (8). As an alternative or parallel function, dynamic PMAPs may act as barriers to avoid the inappropriate recruitment of molecules that regulate adhesion and cytoskeleton. With regard to directional migration, the identified PMAPs positioned just behind the protrusive front may prevent the random localization of signaling molecules at sites other than the very edge, and thus polarize and stabilize the protrusive activity. This hypothesis is supported by the finding that depleting each of three core components of the PMAP does not inhibit the capacity of the cell to make lamellipodia but negatively affects the stability of the lamellipodia, thus decreasing cell polarity and the efficiency of migration (18). A similar barrier function has been attributed to the plasma membrane–associated hydrogel of MBP formed by phase transition during myelin development, which excludes larger proteins from the cytoplasm of the myelin sheets (110). Another example is the hydrogel-based barrier identified at the nuclear pore to prevent the free nucleocytoplasmic transit of proteins (101). Because the disordered core proteins appear to be involved in the recruitment of vesicles at specific sites of the neuronal and nonneuronal plasma membrane, in migrating cells, the assembly of PMAPs may also be involved in targeting vesicular trafficking for delivery or removal of membrane and signaling components at the leading edge. The function of PMAPs in the recruitment of microtubules at the different plasma membrane domains is consistent with this function, because these anchored microtubules may direct trafficking of membranes and signaling components.

The inherent danger of phase transition is the formation of nonspecific or toxic protein aggregates. Although functional IDR-dependent aggregates are reversible, several IDR-containing proteins are implicated in the formation of irreversible pathological aggregates, such as those observed in neurodegenerative diseases (116, 117). Among the proteins involved in these dysfunctional aggregates are also components of PMAPs, such as the ArfGAP GIT1, that directly binds to liprin-α1 (29, 118) or the motor protein KIF21A that is present at the cortical PMAPs that anchor microtubules to the cell edge (76). Both GIT1 and KIF21A include extended IDRs and are regulated by intramolecular autoinhibitory mechanisms (76, 119). Mutations or structural alterations affecting the intramolecular autoinhibitory mechanism of these proteins promote the pathological accumulation of the activated proteins in neurodevelopmental disorders (76, 120). The mechanism of assembly and disassembly of PMAPs should be better understood as more information is determined about the structural organization and the regulatory mechanisms specific for the IDR-containing proteins that form the PMAP core. The more we learn about the structural and dynamic organization of the IDPs that constitute the PMAPs, the better we will understand how these proteins can combine functionally into PMAPs, and how they avoid making dangerous irreversible liaisons.


Acknowledgments: We thank R. Fesce (University of Insubria, Italy) for critical reading of the manuscript. Funding: This work was supported by Telethon Foundation Italy grant GGP12126, AIRC (Associazione Italiana per la Ricerca sul Cancro) grant IG15530, and the Italian Ministry for Research grant PRIN-20108MXN2J.
View Abstract

Stay Connected to Science Signaling

Navigate This Article