Functions and Regulation of Circular Dorsal Ruffles

Abstract

Cells construct a number of plasma membrane structures to meet a range of physiological demands. Driven by juxtamembrane actin machinery, these actin-based membrane protrusions are essential for the operation and maintenance of cellular life. They are required for diverse cellular functions, such as directed cell motility, cell spreading, adhesion, and substrate/matrix degradation. Circular dorsal ruffles (CDRs) are one class of such structures characterized as F-actin-rich membrane projections on the apical cell surface. CDRs commence their formation minutes after stimulation as flat, open, and immature ruffles and progressively develop into fully enclosed circular ruffles. These “rings” then mature and contract centrifugally before subsiding. Serving a critical function in receptor internalization and cell migration, CDRs are thus highly dynamic but transient formations. Here, we review the current state of knowledge concerning the regulation of circular dorsal ruffles. We focus specifically on the biochemical pathways leading to CDR formation in order to better define the roles and functions of these enigmatic structures.

INTRODUCTION

“Ruffles” are membrane structures that were first observed in fibroblasts under locomotion (2, 49). Initially described as the protrusive leading edge of a migrating cell (16), this term was later refined to refer more precisely to the nonadhesive structures that extend vertically from the cell periphery or dorsal surface. This distinguishes ruffles from the lamellipodium, which is the flat, broad protrusion of the cell's leading edge that lies along the substratum and is weakly adherent (2, 10, 16). Ruffles are further subdivided into two clearly distinct, independent types: peripheral ruffles (PRs) and circular dorsal ruffles (CDRs) (16).

PRs are upward-bending structures of the cell membrane at the leading edge that fold or crease backwards and sometimes (but rarely) are found to propagate toward the cell center (Fig. 1A). CDRs are characterized by their unique appearance on the dorsal surface of cells, erecting vertically to form enclosed, ring-shaped structures (93) (Fig. 1B). PRs and CDRs share many similarities in composition, type and mode of stimulation, and assembly factors. However, there are fundamental differences, with many experiments advocating the two being separate and independent events. PRs are active and persistent. They form immediately (about 1 min) after stimulation and thereafter persistently cycle between formation and disassembly. In contrast, CDRs are dynamic and transient, forming only once upon stimulation. They emerge, constrict, enclose, and subside within 5 to 30 min of stimulation (10). Topologically, PRs form in proximity to or, in some cases, at the cell periphery, whereas CDRs are formed not at the cell periphery but on the dorsal surfaces of the cells. Morphologically, PRs are mostly linear, whereas CDRs have a distinct circular architecture. WASP-family verprolin homologous protein 2 (WAVE2)-deficient mouse embryonic fibroblasts (MEFs) are impaired in PR formation, but CDR formation remains unaltered. Similarly, WAVE1 knockout inhibits CDR formation but not PR formation (10, 93). These findings suggest that that the two types of ruffles are independent.

Fig 1.

Cells exhibiting dorsal and peripheral ruffles. The images show phalloidin staining of NIH 3T3 cells expressing peripheral (A) and dorsal (B) ruffles. The arrowhead indicates a peripheral ruffle, and the arrow indicates a dorsal ruffle. Bar, 15 μm.

Circular dorsal ruffles, sometimes known as dorsal ruffles or waves, are membrane protrusions composed of actin-rich structures and are formed on the apical surfaces of cells. CDRs are induced in response to growth factors, such as epidermal growth factor (EGF), hepatocyte growth factor (HGF), and platelet-derived growth factor (PDGF) (10, 62, 72). PDGF was one of the earliest growth factors used to study CDRs. The phosphatidylinositol 3-kinase (PI3K) and the small GTPase Rac are the two major proteins implicated in the signaling events downstream of the PDGF receptor that drive CDR formation. However, the signaling mechanism involved in CDR formation has not been fully elucidated. It is also not clear how the actin cytoskeleton is organized into the CDR. Some proteins, which are associated with the actin cytoskeleton or the regulation of actin, such as cortactin, dynamin 2, N-WASP, and actin-related protein 2/3 (Arp 2/3), have been found colocalized with the actin ring of the CDR (47, 71, 72). Interestingly, clathrin, clathrin adaptors, and caveolin, which are normally associated with endocytosis, are not required for CDR formation (71).

Although first reported more than 5 decades ago, we still do not know much about CDRs. The fundamental aspects of these structures, such as their roles and functions, as well as the biochemical pathways leading up to their formation, remain obscure. This review primarily focuses on the cellular phenomenon of CDRs. We summarize the current knowledge and findings on these structures, paying special attention to their proposed functions and the factors affecting their formation. Principal signaling pathways and regulatory proteins leading to dorsal ruffling will also be covered, in an attempt to achieve a fuller understanding of these structures.

FUNCTIONS OF CDRs

Putative functions of CDRs can be broadly divided into two categories: (i) internalization (of substrates, receptors, and membrane) and (ii) cell motility. Although these can be seen as separate functional aspects, in many cases both aspects share an intimate cause-and-consequence relationship in that some cellular activities transpire from the cooperation between the two functions. For instance, locomotion would be impossible without the uptake and reorganization of necessary proteins, receptors, or substrates. Thus, the two aspects of CDR functions are linked.

RECEPTOR INTERNALIZATION

Following ligand binding and subsequent signal transduction, growth factor receptors are internalized, followed by either degradation or recycling of the internalized receptors (46, 91). Recent studies have uncovered a CDR-constituted, nonclathrin/caveolin pathway in the sequestration and internalization of receptor tyrosine kinases (RTKs), in particular the EGF receptor (EGFR) (71). These dorsal ruffle-mediated, ligand-receptor internalizations are highly specific, targeting only the RTKs that activated them (10, 71). It is indeed important to make a distinction between these CDR-driven internalizations and bulk fluid uptake, macropinocytosis (another implicated function of CDRs). The former transforms discrete local plasma membrane sites into tight tubular-vesicular invaginations into the cytoplasm, from which vesicles are observed to pinch off. The uptake is specific in the case of receptor internalization, whereas macropinocytosis involves large, spherical intracellular vesicles that become severed from the plasma membrane (71) in the rapid and nonspecific uptake of macrosolutes (27). Although dorsal ruffle-dependent internalization is a clathrin/caveolin-independent process, clathrin-dependent receptor uptake is not completely absent, but in fact works in parallel. However, since CDR-driven internalization contributes up to 50% of receptor intake, it is likely to be the major modulator of receptor downregulation (71).

PREPARATION FOR CELL MOTILITY

CDRs are also implicated in a cell's transition from the static to a motile state (47, 87). For the initiation of cell migration, the stress fibers and cell-cell/cell-substrate attachments mediated by focal adhesions must first be disassembled (10). Simultaneous extension of the lamellipodia toward a stimulus or chemoattractant must also take place. Finally, the cell's trailing edge should retract to aid the forward motion. These processes are accompanied by a massive relocalization of actin and the plasma membrane to the leading edge for membrane extension. Indeed, stress fibers undergo extensive disassembly and relocation to PRs and CDRs in PDGF-stimulated NIH 3T3 fibroblasts (20, 51, 93). The presence of cortactin and the Arp 2/3 complex in these structures suggests a possible dendritic branching of dissolved actin filaments. The proximity of these ruffle structures to the leading edge, coupled with the availability of branched actin filaments, capacitates the growth of a cortical meshwork in lamellipodia (10, 47, 71). Relocation of plasma membrane and integrins to the anterior is purportedly performed through macropinocytosis and endocytosis, which involves a rapid uptake of membrane and subsequent reinsertion at sites of membrane extensions (27, 50).

MACROPINOCYTOSIS

As mentioned above, macropinocytosis is the nonspecific fluid-phase endocytosis of macrosolutes, and it can either occur constitutively or be triggered by growth factor stimulation (27). Macropinocytosis functions in cellular events ranging from nutrient uptake and pathogenic invasion to the aforementioned membrane transfer (27). It has been hypothesized that macropinosomes form concordantly with the closure of CDRs (10, 28, 102). Accordingly, real-time phase microscopy revealed the presence of macropinosomes upon closure of dorsal ruffles in MEFs (51). Several proteins critical to CDR formation are also implicated in the process. Inhibition of PI3K, filamentous actin (F-actin), or p21-activating kinase (PAK1) with wortmannin and cytochalasin D blocks both the formation of CDRs and macropinocytosis (27). However, macropinocytosis does not seem to be an exclusive function of CDRs, as it is also linked to PRs (10, 51).

Some experimental evidence seems to contradict these notions. Fluorescent-labeled dextran uptake experiments suggested that CDRs are not the sites of macropinocytosis (93). Also, MEFs lacking WAVE1, which is required for CDR formation, could still induce PRs and macropinocytosis, while WAVE2^−/− fibroblasts with impaired PRs had no macropinosomes (10, 93). Put together, these findings suggest that PRs, and not CDRs, are necessary and sufficient for macropinocytosis. Overall, the validity of macropinocytosis as a downstream function of dorsal ruffling requires further verification. Future research should therefore focus on dissection of this relationship.

MESENCHYMAL MIGRATION

Apart from preparing the cell for movement, CDRs are also implicated in a specific mode of cell locomotion, namely, mesenchymal migration, the acquired ability of cells to traverse three-dimensionally within the extracellular matrix (ECM). In the absence of the ECM, MEFs execute amoeboid migration or general cell crawling with PRs. Migration in the ECM, however, encompasses both PR-dependent cell motility and CDR-dependent matrix invasion (93).

To permit movement in three-dimensional (3D) environments, mesenchymal cells must possess means to disrupt and remodel the matrix. It was found that matrix-degrading matrix metalloproteinase-2 (MMP-2) colocalizes with WAVE1 at the tips of CDRs (10, 93), advocating its role in these circular structures. Such invasive motility may be important in pathological events, such as tumor cell invasion, indicating a possible correlation between CDRs and cancer. Indeed, several similarities between CDRs and the invasive leading edge of cells have been identified. Molecules involved in both cell-ECM adhesion (paxillin and Pyk2) and invasion (MMP-2) localize to the dorsal ruffles and are important components of ECM invasion (10, 93). Interestingly, paxillin knockout fibroblast (pax^−/−) MEFs exhibited more CDRs when stimulated with PDGF than did pax^+/+ cells (87). This finding further supports that CDRs might correspond to invasive protrusions, which are formed when the cells lose focal adhesions. In addition, Missing-in-Metastasis (MIM), a protein which is largely absent in metastatic tumor cells, appears to exert some form of negative regulation on CDR formation (109), suggesting the relevance of these structures to cancer.

It has been shown that tumor cells generally possess fewer dorsal ruffles than nontransformed cells (72). The lower CDR count could be a result of defective receptor downregulation, leading to unregulated growth. However, this would also mean that fewer pseudoinvading structures could potentially participate in cancer metastasis (72). Therefore, it would be instructive to investigate the relevance and relationship, if any, between the CDR-mediated 3D migrations and cancer cell metastatic behaviors.

FORMATION AND REGULATION OF CIRCULAR DORSAL RUFFLES

It is well established that dorsal ruffles form in response to stimulation by a range of agonists, such as PDGF, EGF, HGF and vascular endothelial growth factor (VEGF), binding to their cognate receptors (1, 10, 28). Following stimulation, RTKs dimerize, and their cytoplasmic domains undergo autophosphorylation in trans, generating docking sites for Src homology 2 (SH2) domain-containing early signaling proteins (11, 94). This allows recruitment and phosphorylation of downstream protein cascades of additional signaling proteins, kinases, and adaptor proteins (10). These components congregate at a particular site on the plasma membrane and function in the regulation of actin dynamics and, eventually, the coordinated formation of CDRs (Fig. 2). This is merely a simplified and generalized description of the mechanics of CDR formation. In actuality, at least 20 proteins localize to the sites of dorsal ruffles (Table 1), with several candidates potentially participating at multiple stages in the same pathway, further complicating the mechanistic studies of CDR formation.

Fig 2.

The mechanistic pathways leading to CDR formation, as shown in this shematic diagram of the signaling pathways involved. Important upstream regulators include the Gab1, PI3K, Nck, and Rac1, and downstream multiprotein complexes of N-WASP, WAVE1, and Arp 2/3. In recent years, several novel protein regulators of CDRs have been discovered, including recruitment of SH3YL1-SHIP2 by the PH domain of PI3K, activation of SFKs by the S1P-EDG-G_i cassette, prevention of ubiquitination-related degradation of N-WASP by β1-integrins, and the roles of Crk, Tapp1, and mAbp1 in the formation of dorsal ruffles. For specific details of the signaling and kinase-substrate coupling, refer to the text.

Table 1.

List of proteins that localized to circular dorsal ruffles

Protein(s)	Reference(s)
Abl	93, 106
Arp 2/3	47
Cbl	83
Cortactin	47
Dynamin-2	47
IRSp53	65
MMP2	93
Nck	76
N-WASP	51
PAK1	26, 27
Paxillin, Pyk2	93
PI3K, PIP3 (nascent), PIP2 (mature)	41, 71
Rab5	50
Rac1, EPS8/Sos/Vav/SWAP70	11, 29, 63
Ras	50
Src	10
WAVE1, WAVE2	63, 93
WIP	20, 51
β1-Integrin	46
Crka	5, 31, 88
Gab1	45, 79
mAbp1	20
MIMb	109
Rab5/actinin-4/RN-tre	50
SH3YL	41
SHIP2	41
Tuba	74

^aCrk localizes to membrane ruffles.

^bIt is unclear if MIM localizes to CDRs, but it is highly likely, since it binds PIP₂ (enriched in CDRs).

FORMATION OF CDRs

Although little is known about the mechanism of CDR formation, attempts have been made to model the formation of CDRs based on the experimental findings (74, 110). Activation of the small GTPase Rac has been shown to be required for membrane ruffling and CDR formation (36, 75). It has also been observed that stress fibers disappear in the vicinity of CDRs (10, 66). This suggests an antagonistic relationship between stress fibers and CDR formation. Based on mathematical modeling, Zeng and colleagues have suggested that antagonism between RhoA and Rac signaling may tune the levels of actin available to regulate stress fiber and CDR formation (110). On the other hand, modeling work by Paleg and colleagues (74) implied that CDRs are initiated when the levels of actin polymerization induced by proteins containing the BAR and I-BAR domains have reached a threshold (74). These BAR and I-BAR domain-containing proteins form the curved activator complexes that aggregate at regions where the membrane shape matches their curvature. Actin bundling and branching proteins are then recruited by the curved activator complexes to promote the formation of branched actin bundles that result in membrane protrusions (Fig. 3).

Fig 3.

Role of membrane-deforming proteins in CDR formation. Curved activator complexes diffuse in the membrane, inducing a spontaneous curvature dependent on their local concentrations. The activators aggregate at regions where the membrane shape matches their curvature. Actin bundling and branching proteins are then recruited by the activator complexes, where they promote branched actin filament formation to give rise to protrusions.

Briefly, upon the stimulation of growth factors, activated receptors initiate a downstream signaling cascade that leads to the activation of Rac and the accumulation of phosphatidylinositol 4,5-bisphosphate (PIP₂) at the cell membrane. Aggregation of the curved activator complexes follows, resulting in the recruitment of actin nucleators as well as actin bundling and branching proteins. The actin required for CDR formation possibly comes from the disassembly of nearby stress fibers (47, 51) (Fig. 4).

Fig 4.

The pathway for formation of CDRs. Following stimulation by growth factors, activation of RTK leads to further activation of PI3K and Src, resulting in the accumulation of phospholipids at the cell membrane as well as the activation of small GTPases such as Rac, Ras and Rab5. Aggregation of BAR domain-containing proteins and the recruitment of the WASP/WAVE family proteins, Arp 2/3 complex, cortactin and dynamin lead to actin remodeling at the plasma membrane, resulting in CDR formation. The actin required for remodeling into CDRs might come from disassembly of stress fibers. Actin is recycled to form lamellipodia when the CDR disassembles.

How the ring sizes of the CDRs are regulated is not well understood. A recent report suggested that the ARAP1 protein is recruited to the plasma membrane following growth factor stimulation and is localized to CDRs (42). ARAP1 belongs to a family of ArfGAPs, which also contains a RhoGAP domain. Expression of ARAP1 induces larger CDR rings, whereas knockdown of ARAP1 results in smaller rings. The numbers of CDRs in ARAP1 knockdown cells are also markedly reduced. In addition, expression of dominant-negative Arf1 and Arf5 also results in larger CDR rings. In the same study, the long and short axes of the CDRs in control NIH 3T3 cells that were stimulated with PDGF (20 ng/ml for 5 min) were reported to be 15 and 7 μm, respectively, with an average perimeter of 27 μm (42). Overexpression of ARAP1 could increase the long axis to about 35 μm, with the perimeter increased to about 80 μm. However, in another study, the CDR ring size of NIH 3T3 cells stimulated with PDGF (30 ng/ml, for 5 min) appeared much larger, with an estimated long axis of about 60 μm and a short axis of about 40 μm (47). These findings imply that there is no fixed ring size for CDRs. The same cell type is able to display CDRs with a range of ring sizes, depending on the context. Further investigation is required to define the elements and factors that influence or determine the sizes of CDR rings.

WHY DO THE DORSAL RUFFLES FORM RINGS?

Perhaps the most prominent feature of CDRs is the centripetal, traversing wave movement of the plasma membrane during maturation and closure. The zone of actin polymerization must therefore propagate concentrically inward to generate these movements. Previous modeling work suggested that the combined activities of actin and myosin resulted in the propagation of membrane waves (90). In the case of CDRs, it can be considered propagation of a wave in an excitable medium (110). A recent study provided an explanation of how the membrane waves could be generated: Peleg and colleagues (74) proposed the localization of at least two membrane-deforming proteins (one for each curvature, i.e., convex and concave) to the sites of dorsal ruffling. Concomitant effects of positive and negative feedback loops in the convex and concave proteins result in the shifting of membranes laterally, thus explaining the wave-like movements in CDRs (74). Mechanistically, CDRs might be formed by two types of forces: (i) the primary active force of actin polymerization on the plasma membrane and (ii) a secondary aiding force that emanates from membrane deformation induced by the curved proteins (82, 109). CDR formation is preferred at sites where the shape of the local membrane matches the spontaneous curvature of the curved proteins and where the concentration of curved proteins is high (74). These curved proteins have dual functions in membrane deformation and activation of actin polymerization, and they are therefore also known as curved activators. For example, TUBA is a concave curved activator of N-WASP (74); the I-BAR domain-containing IRSp53 is a convex curved activator that recruits WAVE1 and is known to mediate Rac-WAVE binding during the formation of dorsal ruffles (3, 29, 74). Not surprisingly, the removal of any curved activator abolishes the formation of CDRs (74), underscoring their importance in the morphogenesis of CDRs.

Zeng and colleagues (110) proposed another model for the ring-like structure of CDRs. Rac activity is elevated at the cell membrane, where the PDGF receptor is activated. As a consequence, the WAVE-Arp 2/3 protein complex is recruited for remodeling of actin into the CDR. According to the mathematical modeling by Zeng and colleagues, a RacGAP that is linked to WAVE is required for the negative feedback that downregulates Rac and causes a dip in Arp 2/3-regulated CDR formation. Their simulations assumed that activated PDGF receptors are localized to a small area at the cell membrane. Activation of the PDGF receptor leads to rapid phosphorylation of PIP₂ to phosphatidylinositol 3,4,5-trisphosphate (PIP₃) by PI3K. PIP₃ is likely localized with a peak at the origin (site of PDGF receptor activation) due to the low diffusiveness of the plasma membrane. Signaling downstream of PIP₃ in turn leads to the activation of Rac. Activated Rac then forms two peaks that travel away from the origin and maintain a stable spatial location. The two peaks eventually move toward the origin when Rac-GTP is hydrolyzed to Rac-GDP by a proposed RacGAP, which is linked to the WAVE complex. The CDR actin behaves similarly, and the whole process resembles the generation of a wave from a point of origin into a ring. In this simulation, the RacGAP is important, because the negative feedback is crucial for the generation of the ring-like structure. A simulation performed without RacGAP suggested the formation of an actin patch instead of the ring structure. Although this model of negative feedback is yet to be demonstrated experimentally, the existence of RacGAPs, which bind to the WAVE protein complex, has been reported. An example of such a RacGAP is the 3BP-1 protein, which is linked to the WAVE complex through Abl (17).

CONTROL OF THE CDR LIFETIME

Compared to the PRs, CDRs are transient, with lifetimes ranging from 5 to 30 min. By plating fibroblasts on fibronectin-coated polymers with different stiffness levels, to modulate the amounts of stress fibers in cells, Zeng and colleagues (110) found that although the diameters and sizes of the CDRs were independent of the substrate stiffness, their lifetimes were dependent on substrate stiffness. The average lifetime of CDRs increases with substrate stiffness, and this can be attributed to the antagonism between Rac and RhoA signaling. Higher substrate stiffness generally results in more stress fibers. Since the amount of actin required for CDR formation is derived from the disassembly of stress fibers, more actin is available for CDR formation on stiffer substrates.

The rate of CDR turnover is also reported to be dependent on G₁₂ and G₁₃ G proteins (103). In G_α12^−/− G_α13^−/− fibroblasts, the lifetimes of CDRs after growth factor stimulation were substantially (∼60 min) longer than that of the wild-type cells. Since CDRs are regulated by Rac, Ras, and Rab5 GTPases, the G proteins signal to specific GAPs, which downregulate active Rac, Ras, or Rab5 and lead to CDR turnover. Indeed, the absence of G₁₂ and G₁₃ also prolonged the activated status of Rac compared to wild-type cells.

CDR RING CLOSURE

It has been documented that CDR rings constrict and close to form macropinosomes. Clearly, contractile activities are involved, but the mechanism of ring closure has not been well studied. CDR closure requires the activity of PI3K, as wortmannin (a PI3K inhibitor) treatment resulted in the receding of CDRs without closing into macropinosomes (7). Interestingly, the PIP₃ amount and Rac activity peak just after CDR closure (108). The role of myosin in CDR closure is not clear, although myosin I, II, and V localize to CDRs (30). CDR formation is blocked in cells treated with ML-7, an inhibitor to myosin light chain kinase, which indirectly inhibits myosin II, suggesting that myosin is required for CDR formation rather than the closure (6).

ESSENTIAL PROTEINS FOR CDR FORMATION

Although the formation of CDRs involves a highly complex interplay of regulators and cross talk, there is a well-characterized pathway involving the Ras superfamily of small GTPases (Ras, Rac1, and Rab5) and the SH2-SH3 adaptor proteins (Nck) (76, 78) that links extracellular signals to downstream actin dynamics (11, 43, 78). Actin polymerization induced by Arp 2/3 activation by the Wiskott-Aldrich syndrome (WAS) family of proteins (N-WASP and WAVE [16, 37]) and several other important regulatory proteins, such as PI3K, cortactin, and dynamin, make up the “classical” pathway of CDR formation.

The SH2/SH3 domain adaptor protein Nck has been shown to be essential for both PR and CDR formation. Nck can bind directly to the activated receptor or indirectly via the Crk-associated substrate (p130^cas) during signal transduction (76, 81). p130^cas participates in PDGF-mediated actin reorganization by recruiting early signal proteins, such as Crk and Nck. This is by virtue of 9 YDXP motifs, which have high affinity for the Crk/Nck SH2 domains (5, 31, 88). Such an indirect association was affirmed by Western blot analysis (76) and further supported by the attenuation of CDRs in p130^cas−/− fibroblasts (1, 76).

Upon activation, Nck-SH3 domains recruit and bind to a number of effectors (11, 76), such as PAK1, N-WASP, WAVE, and WASP-interacting protein (WIP) (78), leading to several downstream events. Nck targets and recruits PAK1 to the membrane for activation (58, 76). Meanwhile, Nck is fully capable of alleviating the inhibited states of N-WASP and WAVE by binding to the proline-rich domain of the former and Nck-associated protein 125 in the latter (29), thereby initiating Arp 2/3-mediated CDR formation. The presence of Nck is a requisite for dorsal ruffle formation (76), as Nck-deficient cells not only display a general lack of CDRs but also fail in localizing N-WASP/Arp 3A to ruffle sites in response to PDGF (81).

Besides Nck, the Ras superfamily of small GTPases are also the key molecular switches which regulate the formation of cellular protrusions (49). Indeed, several studies have reported an absolute requirement for Rac and Rab5 in the generation of CDR (50, 73). Disruption of any of these signals abrogates the formation of CDR structures. Among them, Rac1 remains one of the most implicated effectors in CDR regulation. Inhibition of PDGF-induced CDRs by a dominant-negative Rac1 mutant and the lack of dorsal ruffles in Rac1^−/− fibroblasts support a requirement for the GTPase (49). Early studies have proposed a model in which Rac activates type I PIP-5-kinase, leading to the formation of PIP₂. PIP₂ binds and removes the actin-severing gelsolin to promote elongation of filaments. Actin polymerizing machinery, such as the WAVE-profilin complex (64), then takes over for the addition of new actin monomers and formation of ruffle structures. A better-defined Rac-dependent dorsal ruffling pathway, however, involves the GTPase activation of WAVE1. Activated Rac1 recruits WAVE1 to membrane ruffles (29, 64), followed by the alleviation of trans inhibition and release of active WAVE complex, which mediates Arp 2/3-dependent actin reorganization (3, 29).

However, activation of Rac alone is insufficient for dorsal ruffling. It has been demonstrated that constitutively active Rac primarily induces PRs (93) and not CDRs (9, 50, 75) by targeting WAVE2, while dominant-negative Rac abolishes both types of ruffles (45). Generally, Rac participates in the formation of both CDRs and PRs. While Rac alone appears to be sufficient in inducing PRs (73, 93), additional signals parallel to or downstream of Rac are required for the generation of CDRs (93).

Rab5, another small GTPase, functions in both the regulation of endocytic/endosomal dynamics and also the formation of CDRs (50, 111). Ligand-bound RTKs activate Rab5 (50), leading to the trafficking of both Rac1 and the GEF T-cell lymphoma invasion and metastasis 1 (Tiam1) to early endosomal compartments via clathrin-mediated endocytosis. There, Rac1 undergoes activation and is subsequently rerouted back to the plasma membrane, where it participates in the formation of CDRs (50, 73). Meanwhile, Rab5 also functions in a separate, Rac1-independent pathway leading to dorsal ruffling. This pathway entails a tripartite association of Rab5 and actinin-4 (ACTN4) via the Rab5 GTPase-activating protein (GAP), RN-tre. ACTN4 binds and cross-links F-actin and preferentially modulates actin dynamics favorable for CDR formation. Accordingly, expression of dominant-negative Rab5 constructs or knockdown of RN-tre and ACTN4 in cells abrogates CDR formation (50). ACTN4 also presents as a molecular intersection between phospholipid signaling and downstream dorsal ruffling. Previous studies have shown that PIP₃ binds to α-ACTN via a phosphoinositide-binding domain (conserved in ACTN4) and regulates its activity (50). Future research can therefore be directed to affirm if such an association is relevant to the regulation of CDRs.

As alluded to above, the involvement of the kinase PI3K, along with several other proteins (e.g., Rac and Rab5), is strictly required for dorsal ruffling (41, 83). Wortmannin treatment annulled CDRs in PDGF-treated fibroblasts (107) and also abolished the effects of Rab5 on CDRs (73). However, the exact roles and mechanisms of phosphoinositide kinases and phospholipids in CDR formation are poorly defined. The involvement of PI3K/PIP₃/PIP₂ in multiple pathways and also at different stages of the pathway further complicates the study of these modulators. Recently, it was found that RTKs activate PI3K via a novel adaptor, GRB2-associated-binding protein 1 (Gab1). The signal is then relayed, in part by activating Rac1 (1, 22), through the binding of Rac GEFs such as Son of Sevenless (SOS), VAV, and SWAP-70 to PIP₃ (43, 89, 93). In addition, phospholipids target membrane-bending BIN-amphiphysin-RVS (BAR)/inverse BAR (I-BAR) proteins and recruit them to the cortical membranes (41, 99). These proteins induce membrane deformations, which, together with actin polymerization, lead to the formation of CDRs. Early studies have suggested a direct association of PIP₃/PIP₂ with actin-regulating proteins, such as profilin, cofilin, gelsolin, and α-actinin, to modulate their activities.

A different kinase, PAK1, has also been reported to play a role in dorsal ruffle formation (27). Rac-GTP binds the N-terminal GTPase-binding domain of PAK1 to relieve the autoinhibition exerted by an adjacent inhibitory domain (27, 113). Activated PAK1 then colocalizes with F-actin to immature dorsal ruffles, indicating a potential early regulatory role (26, 27, 50). Large numbers of CDRs are later observed in these cells, which eventually develop CDR-related macropinosomes.

Further downstream, cortactin, a cytoplasmic regulator of several cortical actin structures, mediates CDR formation. It forms a complex with the large endocytic GTPase dynamin 2 (Dyn2) (47) in mediating actin remodeling. In most instances, N-WASP is likely to be involved (10, 61, 105), since dominant-negative constructs and depletion of N-WASP disrupt the N-WASP–Arp 2/3–cortactin–Dyn complex (51). The cortactin-Dyn2 interaction is crucial, as it regulates the ability of cortactin to stimulate Arp 2/3-dependent actin polymerization (10, 20, 84). Accordingly, anti-Dyn2 antibodies, or expression of nonfunctional Dyn2 constructs which fail to bind cortactin, abrogate the dorsal ruffles (47). Within the complex, cortactin binds and cross-links preexisting F-actin, while simultaneously stabilizing the filament branch points. Meanwhile, the Arp 2/3 complex stimulates actin nucleation and polymerization. This N-WASP–Arp 2/3–cortactin–Dyn complex is often implicated as the minimal essential machinery for a number of actin structures (10).

There exist contrary views on whether WAVE or N-WASP is responsible for CDR formation. Although Suetsugu and colleagues (93) have shown that WAVE1 is responsible for CDR formation, work by Legg and colleagues (51) suggested that it is N-WASP that is required to form a complex with Arp 2/3 to direct CDR formation in MEFs. Neither WAVE1 nor WAVE2 is essential.

OTHER CDR PROTEINS

The pathway and proteins described above are representative of a general backbone in CDR formation. By itself, this backbone seems somewhat insufficient for the elucidation of the complete picture of CDR mechanics. Within the last decade, many groups have reported the identification of novel protein regulators of dorsal ruffles. Increased understanding of how these emerging players act and participate in the formation of CDRs may offer new perspectives on these ring-like structures and provide the missing links in the pathway.

Src FAMILY KINASES

Of late, a special pool of distinct, noncaveolae Src family kinases (SFKs) (102) have been implicated in the formation of CDRs. Activated by means of a spingosine-1-phosphate–G-protein–coupled EDG1 receptor pathway (S1P-EDG–G protein) (102, 104), this novel pool works upstream in the activation of the early effector kinase Src (10). Although their roles are not well understood, it is known that SFKs activate nonreceptor tyrosine kinases (Abl) and lipid kinases (PI3K), which are important regulators of CDRs (10). Located extremely upstream, SFKs either bind activated RTKs directly via an SH2-phosphotyrosine association, i.e., SFK binding sites on PDGFRβ (68, 100, 102), or through a phospholipase Cγ (PLCγ)/Ca²⁺ pathway linked to the SFKs, whereby tyrosine phosphorylation of the Gab1 adaptor by c-Met creates docking sites for PLCγ (80, 102). SFK inhibition, or expression of kinase-inactive constructs, abrogates the CDRs (102). Accordingly, disruption of any member upstream of the S1P-EDG–G protein cassette impairs Src activities and negates the associated dorsal ruffles (102). Phosphorylation of Abl by SFKs was found to be important for the activation of the Abl kinase in CDR formation (32, 102); Abl-induced phosphorylation may enhance cortactin or WAVE1 activity in stimulation of dorsal ruffles (10).

ROLE OF Gab1

Gab1 is a novel scaffolding protein with a central role in the upstream cell signaling events that lead to dorsal ruffle formation. As it is highly versatile, the protein is common and essential to a range of RTKs (c-Met, EGFR, PDGFR, and VEGFR-2) that lead to Nck-dependent CDR dynamics (1, 45). Following growth factor stimulation, Gab1 is recruited to the activated receptor Met via a direct association of the Met-binding motif with c-Met/EGFR (45, 56, 57) or through indirect binding of growth factor receptor-bound protein 2 (Grb2) to RTKs (1, 22). Tyrosine phosphorylation of Gab1 facilitates its interaction with SH2-containing partners (22, 45) and subsequent recruitment to membrane compartments via its pleckstrin-homology (PH) domain (39, 70). Gab1 binding partners downstream of the signaling pathway include PI3K and Nck. The Gab1 SH2 domain recruits the p85 subunit of PI3K, bringing about a localized activation (45, 77) and accumulation of PIP₃. PIP₃ reciprocally exerts positive feedback to recruit more PI3K via Gab1 to the membrane.

Consistent with this, Gab1 mutants lacking the PH domain fail to form CDRs (1, 45, 79). Gab1 also directly binds the adaptor protein Nck (1) and mediates dorsal ruffle formation by targeting N-WASP- and/or WAVE-dependent Arp 2/3 actin remodeling. Several experiments have demonstrated the importance of Gab1 as a regulator of CDR: (i) Gab1 localizes to HGF-induced CDRs (45, 79); (ii) expression of Gab1 small interfering RNA (siRNA) suppresses 75% of CDR formation, while Gab1^−/− MEFs are almost rid of their dorsal ruffles (1, 45). Inversely, (iii) overexpression of Gab1 more than doubles the number of CDRs, and (iv) ectopic expression of Gab1 in HeLa cells, which do not spontaneously form dorsal ruffles, induces growth factor-stimulated CDRs (1, 73).

Crk PROTEIN

Similar to Nck, Crk is a ubiquitous SH2-SH3 adaptor protein and is a link between activated RTKs and downstream CDR formation. Crk I, an alternatively spliced form, lacks the regulatory phosphorylation sites and C-terminal SH3 domain present in Crk II/Crk L (5, 96). PDGF stimulation induces the recruitment of Crk to activated receptors by p130^Cas as an early signal protein in CDR formation (5). The p130^Cas-Crk interaction is disrupted when Crk is phosphoarylated by Abl. Dorsal ruffling is impaired in Crk I/II/L knockdown cells. Both activation and termination of Crk activity represent critical steps in the formation and normal progression or maturation of CDRs. Phosphorylation of Crk regulatory sites (Y207 and Y221) by Abl/Abelson-related gene (Arg) (5) does not merely terminate Crk activity but ensures maturation and proper closure of CDRs. A Crk II mutant that cannot be phosphorylated displays neither centrifugal wave-like progression nor CDR closure. Intriguingly, the formation of CDRs in Crk II-/L knockdown cells transfected with a Crk II mutant lacking a functional C-terminal SH3 domain appeared normal, despite severely impaired Rac1 activation. It has been proposed that Crk I activates another small GTPase, Rap1, through its N-terminal SH3-binding partner C3G, which is a RapGEF (5, 38). The ankyrin repeat and PH domain-containing protein 3 (ARAP3), a downstream effector of Rap1, then promotes dorsal ruffling (5, 48). Whether this represents a constitutive pathway or a compensatory mechanism for CDR formation in the absence of Crk II/L warrants further examination. Since Abl is a target of Nck/Src, activation of these proteins can also, in part, participate in Crk regulation (5, 76).

SH3 DOMAIN-CONTAINING YSC84-LIKE PROTEIN

A recently discovered lipid-binding domain SYLF, the SHE domain-containing YSC84-like protein (SH3YL1), binds phospholipids (PIP₃) strongly and translocates the protein to dorsal ruffles upon growth factor stimulation. In addition, Src homology 2-containing inositol-5-phosphatase 2 (SHIP2), a binding partner of SH3YL1, displays highly enhanced interactions with SH3YL1 5 min after PDGF stimulation, coinciding temporally with CDR formation dynamics. These proteins are crucial for CDRs, since silencing of SH3YL1 or SHIP2 by siRNA negatively alters the number and appearance of CDRs, as well as macropinocytosis (41). Ectopic expression of SH3YL1 rescues the phenotype. The early localization of SH3YL1 to immature CDR sites (2 min after PDGF treatment), where short actin filaments are arranged in a circular fashion, suggests an early regulatory role (41). Notably, PIP₃ levels peak within the same time frame as SH3YL1 localization (41, 93), as SH3YL1 recruits PI3K through its SH3 domain. SH3YL1-bound SHIP2 subsequently exerts a lipid phosphatase activity, inducing the production of PIP₂ from PIP₃ (after 5 min of PDGF treatment), which is found enriched in mature CDRs and is deemed important in CDR formation.

INTEGRINS

β1-Integrin cross talk with growth receptors, such as EFGR and PDGFRβ, mediates an array of downstream signaling events, including dorsal ruffles (46, 85, 92). Endogenous β1-integrin localizes to dorsal ruffles, alongside PDGFRs and N-WASP, while cells lacking β1-integrin fail to form CDRs in response to PDGF (46). It is likely that β1-integrin promotes CDR formation by regulating the N-WASP–WIP association, consequently protecting N-WASP from ubiquitin-mediated degradation. Since ectopic expression of WIP in β1^−/− fibroblasts restores CDRs only partially, β1-integrin is likely to engage in an N-WASP–WIP–independent pathway to control dorsal ruffling. For instance, the engagement of integrin receptors can trigger focal adhesion kinase (FAK)/Src-mediated phosphorylation of p130^cas and its activation, leading to Nck-/Crk-induced CDRs (76). In addition, active β1 integrin is also responsible for localized activation of Cdc42, which can in turn activate N-WASP (46). On a different note, the enrichment of phospholipids at sites of integrin clustering potentiates the synergistic activation of N-WASP in dorsal ruffling (41, 46).

Besides β1-integrin, β3-integrin has also been observed to relocalize from the ventral surfaces to CDRs after PDGF stimulation (40). Integrins found at CDRs are then internalized by macropinocytosis and recycled to the focal adhesions. Interestingly, CDR formation appears to be coregulated by EGFR signaling and events downstream of β1-integrin and integrin-linked kinase, but not β3-integrin signaling (8).

CDR AND THE PODOSOME

CDR has often been compared with another actin-rich structure called the podosome. The podosome is also circular in shape and contains many proteins that are found in CDRs. Unlike CDRs, podosomes are formed at the ventral side of the cell, which contacts the substratum. Posodomes differ from focal adhesion complexes in the arrangement of constituent proteins; proteins related to the focal adhesion and actin cytoskeleton, such as vinculin and α-actinin, are arranged into clusters that form a circular shape. In fact, due to their appearance, podosomes were initially called rosettes in Rous sarcoma virus-transfected fibroblasts (23, 95). A related structure to the podosome is the invadopodium. The term “invadopodia” was first coined by W. T. Chen because it was discovered that the podosomes in the Rous sarcoma virus-transfected fibroblasts also contained high matrix degradation activities, suggesting that they could aid invasion into the ECM (15). The lifetime of a podosome is much shorter (2 to 20 min) than that of the invadopodia (over 1 h). In more recent studies, invadopodia were in cancer cell lines, and podosomes have also been reported to form in osteoclasts, macrophages, and other cell types. Hence, there have been proposals to call these structures invadopodia in cancer cells and podosomes in nontransformed cells (69).

Some of the proteins reported to localize to the podosomes are similar to those that localize to CDRs. They presumably exert similar functions in these two actin-rich structures. Examples are cortactin, N-WASP, dynamin, and Arp 2/3 (10). More recently, integrin, which has long been associated with podosomes, has also been reported to be required for CDR formation (24, 46). The formation of podosomes is also regulated by proteins from small GTPase family proteins, such as Ras, Rho, and Rab (101). Hence, the assembly of CDRs and podosomes may be regulated similarly. Further details are provided in Table 2 and also in a review by Buccione et al. (10).

Table 2.

Comparison between CDRs and podosomes

Comparison category	CDR characteristic (reference[s])	Podosome characteristic (reference[s])
Cell type	Multiple cell types, such as fibroblasts, differentiated epithelial cells, and smooth muscle cells (10)	Cells of monocytic lineage (12, 55, 60), smooth muscle cells (35), endothelial cells (67)
Structure	Ring-like protrusion (93)	Ring-like protrusion, occurring as individual puncta or circular clusters (23)
	Actin-rich core, together with actin-associated proteins, forming a ring-like structure (10, 93)	Actin-rich core (contains actin and actin-associated proteins) surrounded by adhesion proteins (54)
Location	Dorsal surface (16, 93)	Ventral surface (23)
Lifetime	Minutes (10)	Minutes (25)
Function	Receptor internalization (71)	Adhesion (52)
	Macropinocytosis (51)	ECM degradation (14, 33)
	Cytoplasmic remodeling (47)	Mechanosensor (4, 18)
	Cell motility (47, 87)	Cell migration (69)
Proteins involveda	Arp 2/3 (47)	Arp 2/3 (47)
	Cortactin (47)	Cortactin (47)
	WAVE1/2 (93)	Talin (59)
	WIP (20)	Vimentin (19)
	Rac (50)	Cdc42 (55)
	Rab5 (50)	Rab7 (112)
	Ras (50)	Tks5 (21)

^aFor a more comprehensive list of proteins involved in CDR and podosome formation, refer to reference 10.

Although there are many common proteins involved in the regulation and signaling network required for the formation of CDRs and podosomes, there are proteins that are specific for either structure. The adaptor proteins and Src substrates Tks4 and Tks5 appear to be required for podosome and invadopodia formation but have yet to be reported to be involved in CDR formation (13, 86). Microtubules have also been shown to be required for podosome formation in macrophages (53). However, the F-BAR domain protein Cdc42-interacting protein 4 (CIP4) has been implicated in both CDR and podosome formation. Since CIP4 also binds WASP and the microtubules, it is possible that microtubules also participate in the regulation of CDR formation through this connection (97, 98).

The circular rings of posodomes can range from 0.5 μm to several micrometers and are generally smaller than those of CDRs, but their lifetime is comparable (34). Because podosomes exhibit ECM adhesion and matrix metalloproteinase activities, the biological functions attributed to podosomes are cell adhesion, migration, and invasion. What stimulates the formation of podosomes is not clear. So far, an increase in tyrosine kinase activities and the inhibition of tyrosine dephosphorylation have been identified as enhancing podosome formation. It is possible that many upstream signaling pathways leading to the activation of Src can also contribute to podosome formation.

CONCLUSIONS

Although much information and experimental results have accumulated on the cellular phenomenon of CDRs, our understanding of their circular architecture is still rather limited. Fundamental questions regarding their roles and precise make-up remain to be addressed. Cells are known to generate a number of actin structures (16) to meet different physiological and cellular demands, and CDRs are but one of these structures. CDRs function in endocytic internalization for receptor uptake and downregulation, as well as the rapid reorganization of cellular components in preparing a cell for motility. Cells also employ CDR-related mesenchymal migration in 3-dimensional matrices, which has been recently associated with cancer metastasis. If such a connection is established, this might offer valuable information towards understanding the invasive behavior of metastatic cancer cells.

The formation of dorsal ruffles is, in actuality, a sophisticated and multifaceted event. More than 20 proteins have been found to localize, interlace, and function in these specialized structures. The engagement of a common downstream actin-polymerizing machinery (N-WASP–Arp 2/3–cortactin–Dyn) across several distinct structures (10) implies that the structural determinants rely on the different upstream mediators. Identification of these differentiating factors should remain the priority of future research, which might provide answers to how these morphologies differ. For instance, the fact that specific GTPases elicit different cell protrusions implies that some form of specificity may exist within the same kind of GTPase (49). This can either manifest in the form of spatial restriction (Rab5-mediated Rac activation) (73) or some other mechanism that is not yet fully understood (e.g., dominant-negative Cbl-induced activated Rac did not target p38 mitogen-activated protein kinase despite clearly being a downstream effector of Rac signaling) (83). In addition, participation of these CDR regulatory proteins is often a necessary prerequisite. However, the spatial and temporal aspects, as well as the nature of participation, are not yet clear (e.g., in the case of PI3K and Src). The existing controversies surrounding WAVE1/N-WASP concerning their involvement (or redundancy) in CDRs should be further investigated. Future explorations should focus on the dissection of the biochemical pathways that define the roles of these proteins in CDR assembly and dynamics.