Clathrin-independent carriers form a high capacity endocytic sorting system at the leading edge of migrating cells

Endocytosis provides a crucial and dynamic interface between the extracellular milieu and the interior of the cell. This interface is paramount for fluid and nutrient uptake, signaling regulation, lipid homeostasis, plasma membrane (PM) remodeling, synaptic vesicle recycling, and cellular motility (Grande-García et al., 2007; Mayor and Pagano, 2007; Idone et al., 2008b; Sigismund et al., 2008). Correspondingly, the endocytic system entails a rich diversity of mechanisms coordinated across several pathways. Mounting evidence suggests that eukaryotic cells maintain up to five distinct, constitutive pinocytic pathways (excluding phagocytosis and macropinocytosis; Doherty and McMahon, 2009). These are clathrin-mediated endocytosis (CME), caveolae, and three noncaveolar clathrin-independent endocytic (CIE) routes, which are defined based on specific regulation by RhoA (Lamaze et al., 2001), Arf6 (Radhakrishna and Donaldson, 1997), or Cdc42 (Sabharanjak et al., 2002). Many of the cargo and regulators of the classical CME and caveolae pathways have been identified and the dynamics of these pathways are the best understood (for reviews see Conner and Schmid, 2003; Parton and Simons, 2007). Comparatively, very little is known about the mechanisms governing CIE, although recent work has emphasized the crucial importance of these pathways (Mayor and Pagano, 2007).

The existence of CIE pathways was first proposed from the study of plant and bacterial toxin internalization, in combination with ultrastructural methods or with tools that inhibited CME (Montesano et al., 1982; Sandvig et al., 1987; Ricci et al., 2000). The plant toxin ricin was used to first describe the clathrin-independent, apical endocytic pathway in endothelial cells (Sandvig and van Deurs, 1994). Bacterial toxins, such as the Helicobacter pylori VacA toxin (Gauthier et al., 2007) and cholera toxin (Torgersen et al., 2001) provided further evidence for clathrin-independent endocytic pathways. However, a lack of specific tools to study these noncanonical routes has made it difficult to identify their guiding parameters.

Ultrastructural work on the CIE pathway regulated by Cdc42 has identified the morphologically distinct clathrin-independent carriers (CLICs) as the primary carriers involved in uptake within this route (Kirkham et al., 2005). Parameters currently identified for the CLIC pathway include dynamin-independent PM scission (Sabharanjak et al., 2002), enrichment in GPI-anchored proteins (Sabharanjak et al., 2002), reliance on Arf1 activity (Sabharanjak et al., 2002; Kumari and Mayor, 2008), sensitivity to cholesterol depletion (Kirkham et al., 2005; Chadda et al., 2007), and contribution to a significant fraction of cholera toxin B subunit (CTxB) and fluid internalization (Kirkham et al., 2005). CLICs arise directly from the PM and mature into the GPI-enriched early endosomal compartment (GEEC), as they acquire Rab5 and EEA-1 and, consequently, merge with early endosomes (EEs; Kalia et al., 2006). Ongoing work is linking cargo and regulators to these carriers, such as GTPase regulator associated with FAK (GRAF1), dysferlin, and Snx9 (Yarar et al., 2007; Hernández-Deviez et al., 2008; Lundmark et al., 2008). The basic parameters of these carriers, however, remain unknown.

Here we characterize key parameters of the CLIC pathway. A quantitative analysis of the volume and surface area of CLICs identifies them as the major constitutive pathway involved in bulk endocytosis within fibroblasts. High resolution electron tomography shows the complexity of these primary endocytic carriers and, in conjunction with recruitment of classical trafficking regulators, they are realized to be intricate sorting compartments. Identification of novel CLIC cargo suggests links to cellular adhesion turnover and plasma membrane repair. Consistent with this, CLICs become polarized in migrating fibroblasts, turning over adhesion complex components, such as Thy-1 and CD44, at the leading edge. Inhibition of the CLIC pathway subsequently impairs the capacity of fibroblasts to migrate.

Quantitation of the volume and surface area internalizing through the CLIC pathway is vital for understanding the contribution of this route to total endocytosis and PM homeostasis. Generic endocytic markers that enter via CME, caveolae, and CIE routes, such as HRP or CTxB, have been used in conjunction with ultrastructural morphology and dynamin independence to define CLICs (Kirkham et al., 2005). Therefore, we first used HRP-conjugated CTxB (CTxB-HRP) as an endocytic membrane marker with a previously characterized diaminobenzidine (DAB) incubation protocol for EM visualization (Kirkham et al., 2005). This method reveals only internalized carriers and can be used to label such carriers within seconds of scission. Although CTxB-HRP is an excellent membrane marker, a multivalent toxin could theoretically affect endocytosis, particularly when combined with an incubation step at low temperature (Römer et al., 2007). Therefore, quantitative measurements of the very early uptake of the fluid phase marker, HRP (at 20 mg.ml⁻¹), were compared with CTxB-HRP and assessed in the presence or absence of unlabeled CTxB (Fig. 1, A and B) to determine the relative contribution of CME and CLICs in both wild-type (WT) and caveolin1-null (Cav1^−/−) mouse embryonic fibroblasts (MEFs).These experiments allowed us to eliminate a cold step and showed that HRP and CTxB-HRP labeled identical compartments, whereas CTxB did not quantitatively affect endocytosis (Fig. 1, C–E).

A defining characteristic of the CLIC pathway is independence of dynamin for PM scission (Damke et al., 1994; Parton et al., 1994; Henley et al., 1998; Lamaze et al., 2001; Pelkmans et al., 2002; Sabharanjak et al., 2002; Kirkham et al., 2005). Therefore, we used a well-characterized small molecule inhibitor of dynamin, dynasore (unpublished data), as well as a more potent, hydroxylated analogue of dynasore, Dyngo4a (unpublished data) to inhibit CME, caveolar endocytosis, and dynamin-dependent CIE pathways. The activity of these inhibitors in WT and Cav1^−/− MEFs was confirmed by assessing the fluorescence uptake of the CME cargo protein transferrin (Tf-647) and the generic endocytic marker CTxB-555 (Fig. 1 F), as well as visualization of CTxB-HRP–labeled carriers by EM (Fig. 1 G). Although Tf-647 and CTxB-555 were both efficiently internalized in untreated control cells, after treatment with 30 µM Dyngo4a (or 80 µM dynasore) only CTxB-555 internalization occurred. In agreement with this observation, in untreated control cells, CTxB-HRP labeling was found in both CLICs (Fig. 1 G, arrows) and clathrin-coated vesicles (CCVs; Fig. 1 G, double arrowheads), visualized by EM. However, in the presence of Dyngo4a, only CLICs, defined by ultrastructure, were active (Fig. 1 G, arrows). Unlabeled clathrin-coated pits (CCPs) were readily observed but were surface connected (Fig. 1 G, arrowheads). Together, these data show that CTxB is a useful marker for CLICs and, when combined with acute dynamin inhibition using Dyngo4a, CLICs remain the only detectably active endocytic pathway.

Having established the constitutive nature of the pathway and lack of detectable perturbation by the used markers, we used EM-based stereology to calculate the proportion of total cytoplasmic volume (V(v)) contributed by the CLIC pathway. We used both Cav1^−/− and WT MEFs to compare cells with and without caveolae, as well as a fibroblast cell line, NIH3T3 (Fig. 2 A). In untreated Cav1^−/− MEFs, total CTxB-labeled structures occupied 0.10 ± 0.01% of the total cytoplasmic volume after 15 s of internalization, 0.19 ± 0.01% after 1 min, and 0.50 ± 0.04% after 2 min (Fig. 2 B, Untreated). WT MEFs internalized similar volumes as Cav1^−/− MEFs, indicating that caveolae contributed an insignificant proportion of internalization at this time. Intriguingly, when treated with Dyngo4a, under conditions where uptake of Tf is completely blocked and CLICs remain the only detectably active pathway, internalized CTxB-HRP still accounted for 0.09 ± 0.01% of cytoplasmic volume after 15 s, 0.16 ± 0.02% after 1 min, and 0.37 ± 0.02% after 2 min (Fig. 2 B, Dyngo4a). This indicates that 90.0% of the total endocytic volume is contributed by the CLIC pathway after 15 s of uptake, 85.9% after 1 min, and 74.0% after 2 min.

Confirmation of such high contribution of CLICs to total internalization was provided using strict ultrastructural criteria to compare the volume of tubular/ring-shaped structures (CLICs) and coated/vesicular structures (CCVs) in NIH3T3 cells that were not treated with any inhibitor (Fig. 2 C). Note that previous studies have shown that at the earliest times used here, Tf-HRP only labels CCVs as defined morphologically (Kirkham et al., 2005) but CTxB-HRP labels CCVs, caveolae, and CLICs. NIH3T3 cells internalized similar volumes to primary MEFs, with total CTxB-labeled structures after 2 min accounting for 0.47 ± 0.014% of cytoplasmic volume and tubular/ring elements (CLICs) contributing 78.7% (0.37 ± 0.014%) of that volume (Fig. 2 C). This alternative method avoids the need for inhibition of the dynamin-dependent pathways and suggests that under these conditions CLICs are not a compensatory pathway induced after acute dynamin inhibition (Damke et al., 1995). To confirm the striking contribution of the CLIC pathway to cellular internalization, V(v) measurements were next collected using free HRP, which cannot bind to membranes and induce clustering of microdomains (Fig. 2 D). Again, CLICs, morphologically defined as tubules/rings, constituted ∼74% of all HRP internalization after 2 min (V(v) 0.32 ± 0.05%).

To further confirm these results using an alternative method for assessment of endocytic contribution, CTxB was biotinylated and internalized in Cav1^−/− MEFs, in the presence or absence of Dyngo4a. Surface-accessible biotin was cleaved and the remaining, internalized biotin measured by Western blot using streptavidin-HRP. Using this alternative biochemcial method CTxB-biotin levels in Dyngo4a-treated cells were equivalent to 76.0 ± 8.4% of that in untreated control cells after 2 min of uptake (Fig. 2 E). Agreement between morphometric and biochemcial approaches (and with V(v) measurements using a specific marker of the CLIC pathway: see below) provides important confirmation of the validity of the methods used and the significant magnitude of the CLIC pathway.

We next estimated the contribution of this pathway to total plasma membrane surface endocytosis. By measuring the surface-to-volume ratio (S(v); see Materials and methods) of a typical CLIC (Fig. 2 F) and the absolute volume of an average Cav1^−/− MEF (2,346.7 ± 289.3 µm³; see Materials and methods), the compiled membrane surface area of CLICs was calculated after 15 s of internalization to be 97.1 ± 10.9 µm². Given that the estimated average PM surface area of a Cav1^−/− MEF was found to be 4,375.7 µm² (see Materials and methods), the equivalent of the total PM surface area could internalize through CLICs in less than 12 min (or 17 min after most extreme correction factor; see Materials and methods). Together these data indicate that CLICs provide a substantial capacity of internalization, more than any other constituite endocytic pathway in fibroblast cells.

Aside from the 2D morphological description of tubular CLICs, there is currently no data describing the 3D attributes of the primary carriers. To comprehensively visualize the ultrastructural properties of CLICs, high resolution electron tomography was performed on CTxB-HRP–laden structures (Fig. 3 A; Video 1). As a membrane marker, CTxB-HRP provided superior delineation of membrane profiles compared with free HRP at the concentrations and very early time points used here. CTxB-HRP was internalized for only 15 s to capture the earliest morphology of the carriers. The DAB reaction was performed on live cells to stabilize the morphology of the structures before glutaraldehyde fixation, simultaneously providing the necessary contrast for high resolution tomography and maintaining a close-to-native morphology. Morphology of DAB-labeled structures was similar in cells fixed before DAB reaction (not depicted). Images of 3D rendering from electron densities are shown (Fig. 3 A). There is clear evidence of vesicular membrane invaginations (Fig. 3 A, arrows), excluded from CTxB-HRP labeling, indicating that CLICs may have a capacity to bud internal membranes. Additionally, the tubular structure forms a complete ring, a facet previously overlooked (Fig. 3 A, arrowhead). As chemical fixatives can alter the morphology of endosomes (Murk et al., 2003), it was also important to confirm these observations without the use of aldehyde fixation. Therefore, CTxB-HRP was internalized for 15 s, DAB reaction performed, and the samples were high pressure frozen (HPF) as described previously (Richter et al., 2008). HPF CTxB-HRP–laden CLICs showed identical features to chemically fixed samples (Fig. 3 B), with vesicular invaginations from the limiting membrane and tubular ring-shaped extensions. Rather than simple tubules or vesicles, these primary carriers have a surprising degree of complexity.

To identify CLIC associated components, we next sought to enrich CLICs away from other cellular material by density gradient fractionation (see Materials and methods). To increase the number of CLICs within a population of cells, the PI₃K inhibitor wortmannin was used within a previously developed system that inhibits the fusion of CLICs with EEs (Kalia et al., 2006; Fig. S1). Using anti-GFP in GFP-GPI–expressing NIH3T3 cells (3T3-GPIs) as a marker for CLICs, we optimized conditions to concentrate GFP-GPI–positive membranes (Fraction 1.2; Fig. 4 A). CCVs (identified by clathrin heavy chain; CHC), EEs and recycling endosomes (REs; identified by TfR), and Golgi (identified by GM130) were below the detection limits within this fraction. However, this fraction also contained Cav1 (caveolae) and Grp78 (ER). The 10% fraction (Fraction 1.2) was, therefore, subjected to a second, continuous Nycodenz gradient, which provided a reproducible enrichment of anti-GFP–positive membranes (Fraction 2.8; Fig. 4 A). To further characterize the enriched fraction, we examined the membranes in detail by whole-mount EM (Fig. 4 B). Consistent with CTxB-HRP–laden CLICs visualized by EM in intact cells (Kirkham et al., 2005), the predominant structures within Fraction 2.8 have the striking morphology of CLICs. A low magnification image shows the consistent morphology of the structures enriched within this fraction (Fig. S2 A). At high magnification (Fig. 4 C, arrows), the structures within Fraction 2.8 were also consistent with tomography and HPF data (Fig. 3). Other structures displayed a clear connection between tubular extensions and spherical bulbs (Fig. 4 C, arrowheads). There was no evidence of contamination by other cellular compartments, such as mitochondria. Other fractions were enriched in structures distinct from those found in Fraction 2.8 (Fig. S2 C). This system also gave optimal, but incomplete, separation of GFP-GPI–positive carriers from Cav1 and Grp78. To quantify the membrane enrichment in Fraction 2.8 we analyzed internalized CTxB-HRP within the fraction after 5 min of internalization (Fig. 4 D, arrows). CTxB-HRP–labeled structures constituted 70.0 ± 0.9% of the membranes in the CLIC-enriched fraction as calculated from stereology (Fig. S2 B).

The validity of this method was further examined by analyzing membranes within Fraction 2.8 after inhibition of CLIC endocytosis. Using two independent methods of CLIC perturbation, expression of a Cdc42 dominant-negative mutant (Cdc42-DN; Sabharanjak et al., 2002) and cholesterol depletion with methyl-β-cyclodextrin (mβCD) under conditions where CME is unaffected (Kirkham et al., 2005; Glebov et al., 2006), there was a substantial reduction of CTxB and anti-GFP–positive membranes within Fraction 2.8 (Fig. 4 E; 59.7% reduction in CTxB levels in cells transfected with Cdc42-DN compared with Cdc42-WT; 53.6% reduction in anti-GFP antibody levels after treatment with mβCD). Thus, this approach shows concentration of CLIC marker positive membranes within Fraction 2.8 and reduced accumulation of these membranes after treatments known to inhibit CLIC endocytosis.

Unlike CHC and TfR, there was a small proportion of dynamin specifically concentrated in Fraction 2.8 (Fig. 4 F). Although dynamin is not necessary for the formation of nascent CLICs (Sabharanjak et al., 2002; Kirkham et al., 2005), it has been previously noted that expression of the DynK44A mutant reduces CTxB trafficking to the Golgi (Le and Nabi, 2003; Kirkham et al., 2005), independently from clathrin (Torgersen et al., 2001), suggesting a role in clathrin-independent CTxB uptake. As CLICs rapidly mature, however, it is difficult to assess a possible transient recruitment of dynamin under steady-state conditions. Therefore, colocalization between dynamin and anti-GFP in GFP-GPI–expressing cells was assessed after treatment with wortmannin (Fig. S3). Colocalization after 10 min was significantly increased in wortmannin-treated cells compared with untreated controls (13.1 ± 1.1% untreated; 39.2 ± 2.0% 100 nM wortmannin; Fig. S3, A and B), indicating that dynamin is recruited to CLICs rather than localizing to them from the PM. By EM, dynamin was also found on a small proportion of CTxB-HRP containing tubular/ring-shaped structures (Fig. S3 C). Although the majority of dynamin coincides with the CHC profile, the small but specific concentration in Fraction 2.8, as well as colocalization with GFP-GPI and CTxB, is consistent with the hypothesis that dynamin is recruited to CLICs, after PM scission.

To verify that membranes in Fraction 2.8 represent bona fide CLICs, 3T3-GPI cells were transfected with one of two markers for the CLIC pathway, Flot1-HA and GRAF1-myc. By Western blot analysis both endogenous and transfected Flot1 and GRAF1 were concentrated in Fraction 2.8 (Fig. 4 F). When fractions from transfected cells were labeled with anti-HA or anti-myc antibodies, a significant proportion of the structures within Fraction 2.8 were labeled (Fig. 4 G, arrows; 31.5 ± 2.5% Flot-HA and 41.3 ± 6.4% Graf1-myc positive). Comparatively, when Fraction 2.8 was labeled for endogenous Cav1, very few of the structures were positive (5.5 ± 0.8%) and, notably, these smaller vesicular structures were distinct from the CLIC-like structures (Fig. 4 G, arrowheads). Endogenous dynamin labeled a significant population of structures within Fraction 2.8 (Fig. 4 G, arrows; Fig. S3 D), supporting the notion that dynamin can associate with CLICs. We next used this fractionation approach to identify novel CLIC components.

The protein complement of Fraction 2.8 was examined after separation using 5–15% SDS-PAGE and the entire lane was cut into 2-mm slices, digested with trypsin, and analyzed by liquid chromatography mass spectrometry (LC-MS/MS; see Materials and methods). This identified 80 proteins with two or more peptide hits of 99% confidence (key targets and links to endocytosis in Table I, full list in Table S1). A significant cluster of proteins identified play a role in cellular adhesion and interaction with the ECM. Thy-1, a glycosylphosphatidylinositol-anchored protein (GPI-AP), is a critical component of focal adhesion disassembly through FAK phosphorylation (Rege et al., 2006). This interaction is dependent on the GPI anchor of Thy-1, implicating the GPI-AP–enriched CLIC pathway as a mediator of FAK activation. Integrin-β₁ is also a key component of focal adhesions and galectin3 is directly involved in clathrin-independent integrin-β₁ endocytosis (Furtak et al., 2001). Although a role for CME during integrin-β₁ endocytosis and cellular motility has been established (Palamidessi et al., 2008), it has also been found that integrin-β₁ uptake can be independent from clathrin and is mediated by lipid rafts (Vassilieva et al., 2008). Rap1 transmits outside-in integrin signals, leading to cell polarization, and its activation at the leading edge is necessary for chemokine-induced cellular migration (Shimonaka et al., 2003). CD98 also mediates integrin signaling during cell spreading and is localized to cell–cell contacts (Feral et al., 2005), where it activates Rap1 (Suga et al., 2001). Activation of Rap1 also leads to recruitment of CD44 to the leading edge (Shimonaka et al., 2003), while CD44 is linked to integrin-β₁ signaling by forming specialized lipid raft microdomains (Lee et al., 2008). CD44 is also an important component of extracellular contact, involved in the degradation of ECM, and is involved in maintenance of polarization and directionally persistent migration (Mrass et al., 2008). This also implicates CLICs as a mediator of cell adhesion and, given the proposed ability for CLICs to recycle directly to the PM, presents a rapid recycling mechanism during cellular adhesion turnover and motility. Supporting this, moesin directs actin-based rearrangements of lipid rafts directly through CD44 to orchestrate the topology of cell surface structures, such as filopodia (Bretscher et al., 1997), and the CD44-ERM (ezrin-radixin-moesin) association is a critical component of directional cellular motility (Legg et al., 2002). We next sought to verify some key targets as novel cargo for the CLIC pathway.

Preliminary verification of some key targets was performed by comparison to internalized CTxB-555 and Tf-647 (Fig. 5 A).Anti–Thy-1, anti-CD44, and myoferlin colocalized with CTxB-555 but not Tf-647 in NIH3T3 cells after 2 min of uptake, indicating they are specific cargo for the CLIC pathway. Interestingly, although only a low level of colocalization was seen between anti-GFP and Rab11 after 10 min (17.07 ± 1.49%) in untreated 3T3-GPI cells (Fig. 5, B and C), after treatment with wortmannin there was a significant increase in their colocalization after 10 min (41.12 ± 2.79%; Fig. 5, B and C). As wortmannin inhibits the maturation of CLICs, this localization is not likely to occur within REs. Additionally, there was no detectable TfR in the CLIC-enriched fraction, arguing against significant contamination of RE membranes within this fraction.

Additional evidence for the sorting capability of CLICs was identified by assessment of the endocytic maturation of CD44. Endocytosis of CD44 was consistently Tf negative (Fig. 5 D), while it colocalized with CTxB after 2 min, with less overlap after 10 min. Persistent colocalization was also seen between CD44 and dysferlin from 2 to 40 min (Fig. 5 D). As dysferlin also remains distinct from the classical endocytic network (not depicted), this suggests that CLICs are directly sorting cargo rather than representing distinct populations of similar mechanisms. Confirming this, the ultrastructure of CD44-positive carriers, as visualized using anti–CD44-HRP, showed identical morphology to that of CTxB-HRP–laden CLICs (Fig. 5 E, arrows), whereas no label was found in the classical endosomal network. In some instances, intriguing tubular networks were seen with labeling for anti–CD44-HRP after 40 min of internalization (Fig. 5 E, arrowheads), possibly representing a mature compartment arisen from the CLIC pathway. The presence of anti-CD44 antibodies had no effect on the number or morphology of CTxB-HRP–laden CLICs between 2 and 40 min (not depicted), indicating that the CD44 antibodies do not influence the kinetics or ultrastructure of the pathway.

Using CD44 as a novel and specific marker for CLICs, it was also possible to confirm stereology calculations using CTxB-HRP in the presence of dynamin inhibition. Anti-CD44-HRP–conjugated antibodies or Tf-HRP was internalized in Cav1^−/− MEFs and the DAB reaction was performed (Fig. 5, F and G). This method has the advantage of relying on the specificity of CD44 for the CLIC pathway and Tf-HRP for CME. We found that CD44-positive CLICs accounted for 0.09 ± 0.02% of the total cytoplasmic volume after 15 s, 0.14 ± 0.03% after 1 min, and 0.34 ± 0.01% after 2 min (Fig. 5 G). These volume estimates are in excellent agreement with those obtained using morphological criteria or using specific inhibitors of dynamin-dependent endocytosis (see above). These data identify a range of novel CLIC cargo, confirm the significant magnitude of the pathway, and provide evidence for the direct recruitment of dynamin and Rab11 to CLICs and the ability to sort cargo to distinct destinations.

Based on the association of novel CLIC cargo, such as Thy-1 and CD44 with extracellular attachment, we sought to visualize CLICs within migrating cells. As regulation of adhesion components is a fundamental process during cellular migration the CLIC pathway may be either up-regulated or spatially restricted within motile cells. A scratch-wound was applied to a confluent monolayer of fibroblasts and cells were allowed to migrate into the space provided by the wound. Early endocytosis of CTxB and Tf was assessed in the presence of either internalizing anti-CD44 or anti–Thy-1 antibodies or was followed by labeling for endogenous Cav1 (Fig. 6 A). Colocalization was seen between CTxB-555 and both CD44 and Thy-1 at the leading edge of migrating cells (arrows). In cells expressing GFP-GPI, anti-GFP antibody endocytosis also preferentially occurred at the leading edge (Fig. 6 A), indicating polarization of CLICs rather than polarized uptake of specific GPI-APs involved in migration, such as Thy-1. There was a clear spatial distinction between CTxB-555, Cav1, and Tf-647 in these cells, which was confirmed by assessing average pixel intensity across a selected field (Fig. 6 B). It has been well documented that Cav1 localizes to the trailing edge during 2D migration (Grande-García et al., 2007). We also found that CTxB binding after fixation was uniform over the entire cell, excluding the possibility that GM1 was polarized within these cells (not depicted).

To verify the polarization of CLIC endocytosis in migrating fibroblasts, ultrastructure of CTxB-HRP–laden carriers was assessed after a 2-min uptake in migrating MEFs (Fig. 5 C, large arrow indicates direction of migration). Labeled CLICs were predominantly found at the leading edge (Fig. 6 C; 1, 2). Conversely, an accumulation of unlabeled, surface-connected caveolae was seen at the trailing edge (Fig. 6 C; 3, 4 insets). Quantitation of the number of endocytic and surface-connected structures at the leading and trailing edges confirmed this polarized distribution (Fig. 6 D). Both surface-connected CCPs and internalized CCVs showed no preference for the leading or trailing edge. To our knowledge, this is the first evidence for a striking spatial distinction between the CLIC pathway, caveolae, and CME, and is suggestive of a role for CLICs during dynamic adhesion turnover specifically from the leading edge.

As the CLIC pathway is polarized during migration and internalizes key adhesion components at the leading edge, we speculated that this pathway could play a critical role during cell migration. Thus, a system was developed in which CLIC-dependent uptake could be inhibited without affecting CME (see Materials and methods). To optimize conditions for transient ablation of the CLIC pathway, we internalized CTxB-HRP for various short times, performed DAB-mediated ablation of the labeled compartments, and then analyzed the uptake of a subsequent pulse of CTxB-555, Tf-647, and antibodies to CD44. 2 min of CTxB-HRP internalization followed by the DAB reaction on living cells was found to give the strongest inhibition of CD44 and CTxB-555 endocytosis without any effect on Tf-647 uptake (Fig. 6 E). Under these conditions Tf-647 endocytosis and maturation was unaffected, even after a 40-min pulse, indicating the endosomal network was still functional (not depicted). The inhibition of CD44 and CTxB-555 internalization following CTxB-HRP–based ablation persisted for at least 4 h after the ablation step, indicating this assay could be used to assess possible effects of CLIC inhibition on migration. Quantitation of fluorescence intensity shows that CTxB-HRP inactivation of the CLIC pathway resulted in an 81.8 ± 7.2% reduction in CTxB-555 and a 93.8 ± 2.6% reduction in CD44 endocytosis, whereas Tf-647 internalization was not reduced (increased by 28.6 ± 24.8%) after a 4-h incubation (Fig. 6 F).

To test the requirement for CLIC endocytosis during migration, this inactivation assay was applied to migrating fibroblasts. The capacity of cell monolayers to migrate into a scratched wound was assessed by measuring the total distance between two cell fronts of a wound after CLIC inactivation. When CTxB-HRP was used to inactivate CLICs, without effect on other endocytic routes, cell monolayers had a reduced capacity to migrate into a wound after 4 h (15.8 ± 1.7% distance of wound closed; Fig. 6 G) compared with DAB only–treated cells (33.4 ± 3.7% closed). A similar effect was observed in Cav1^−/− fibroblasts, ruling out the possibility that the migration defect was due to inactivation of caveolae. Together with previous studies (Grande-García et al., 2007; Nishimura and Kaibuchi, 2007), these findings suggest that three major endocytic routes in eukaryotic cells, CME, caveolae, and the CLIC pathway, are all important for normal directional cell migration.

Here we identify the CLIC pathway as a high capacity route that accounts for the vast bulk of constitutive fluid uptake in fibroblasts, implicating CLICs as key mediators of PM remodeling. Previous estimations of endosomal tubule volume using stereology have noted that tubules close to section thickness (55 nm) can lead to overprojection effects (Griffiths et al., 1989). When correction factors are applied to volume estimates CLICs would still account for over 70% of internalization. Bulk phase early endocytosis in rat fetal fibroblasts has also been shown to be independent from clathrin and occurs via a constitutive, noncaveolar pathway, which rapidly regurgitates substantial amounts of endocytic volume (Cupers et al., 1994). We have previously shown that up to 40% of endocytosed fluid, mainly internalized through the CLIC/GEEC pathway, is regurgitated in 5 min (Chadda et al., 2007). Together, this suggests that the CLIC pathway internalizes and returns significant portions of the PM quickly, rapidly turning over membrane during key cellular processes, such as PM repair and homeostasis. It also provides the endocytic system with a mechanism to remodel features of the PM quickly, critical for extracellular interactions during various processes ranging from adhesion to signaling. Such vast membrane flow and recycling has recently been observed from a GPI-AP–positive CIE compartment, which provides a signficant fraction of the required membrane during cellular spreading (Gauthier et al., 2009).

Calculation of fluid uptake via clathrin-independent endocytosis has generally estimated the contribution of these routes to be 40–60% of total fluid internalization (Sandvig et al., 1987; Sabharanjak et al., 2002; Cheng et al., 2006). The higher contribution of 74%, calculated here, may be explained through differences in the resolution, both spatial and temporal, of experimental procedures used. By calculating volume internalized based on ultrastructure rather than fluorescence this work provides a higher resolution analysis than previous studies, which were limited by the optical resolution of fluorescence microscopy.

The 3D structure of CLICs identifies them as pleiomorphic, complex, multicomponent carriers. Although it is recognized that the bulk of CLIC-internalized fluid and membrane fuses with EEs, discrepancies between the ability of CLIC cargos, such as dysferlin/myoferlin and CD44 compared with GFP-GPI and CTxB, to fuse with Tf-positive endosomes strengthens the argument that some domain of CLICs traffics distinctly from the bulk (Fig. 7 A). Recruitment of dynamin to CLICs, after budding from the PM, further demonstrates the sorting complexity of the route. Although dynamin is not necessary for scission of CLICs from the PM (Kirkham et al., 2005; Sabharanjak et al., 2002), such recruitment of dynamin to maturing CLICs, after the initial scission, provides important machinery for CLICs to compartmentalize. Clearly, however, dynamin is not required for the generation of CLIC architecture, as the morphology of CLICs is identical in the presence or absence of dynamin inhibition. Recruitment of dynamin to CLICs as they mature would reconcile the direct interaction between GRAF1 and dynamin within this dynamin-independent pathway (Hansen and Nichols, 2009).

Myoferlin, a novel CLIC-associated target identified here, is a member of the ferlin family of proteins implicated in membrane fusion (Davis et al., 2000). We have previously shown that another ferlin family member, dysferlin, associates with the CLIC pathway (Hernández-Deviez et al., 2008). Myoferlin, like dysferlin, is involved in calcium-dependent membrane fusion after mechanical disruption to the PM (Bernatchez et al., 2007). Recently, a clathrin-independent, GPI-enriched endocytic compartment has been implicated in membrane repair after PM shearing (Idone et al., 2008a,b). This raises the exciting possibility of links between the role of dysferlin/myoferlin in this process and the ability of CLICs to provide the necessary membrane for repair of PM lesions. Supporting this, two targets identified here, AHNAK and Annexin A2, have been shown to bind dysferlin and myoferlin during membrane repair (Huang et al., 2007). Localization of dysferlin/myoferlin to a subdomain of CLICs that does not merge with Tf-positive EEs highlights that this pathway has the capability to rapidly recycle membrane back to the PM under specific conditions, such as PM repair (Fig. 7 B).

Evidence for an association between CLICs and cellular migration is provided by the distinct polarization of CLICs in motile cells, as well as the impairment of migration after transient ablation of the pathway. Specific internalization of Thy-1 and CD44 through the CLIC pathway at the leading edge provides an explicit role for CLICs in rapid turnover of adhesion receptors within a spatially restricted mechanism. How CLICs become polarized during migration remains an open question; however, one possibility arises from activity of the initial regulator of the CLIC/GEEC pathway, Cdc42. Cdc42 was originally described as directing cell polarity during yeast budding (Johnson and Pringle, 1990) and plays an assortment of fundamental roles during eukaryotic cell polarization (Etienne-Manneville, 2004). After wounding of cell monolayers Cdc42 activity is specifically directed to the front of the cell, resultant from signaling between contacts and the ECM (Etienne-Manneville and Hall, 2001). As Cdc42 activity is necessary for formation of CLICs, this would lead to specific localization of CLICs at the leading edge of wounded monolayers. The clathrin adaptor, Numb, is also specifically localized to the leading edge where it mediates integrin-β₁ endocytosis (Nishimura and Kaibuchi, 2007). It appears surprising that Numb-positive CCVs and CLICs both polarize to the leading edge during migration. One possible explanation for this is the types of cargo internalizing through each pathway. Adhesion components without a CCP sorting motif, such as CD44 and Thy-1, still need to be internalized for efficient migration and this could occur via CLICs at the leading edge. Consistent with such a hypothesis, a screen for genes involved in epithelial cell migration identified that knock-down of GRAF1, a regulator of the CLIC pathway (Lundmark et al., 2008), reduced the capacity of these cells to migrate (Simpson et al., 2008).

This study has shown the significant magnitude of traffic through this largely uncharacterized sorting compartment. Rather than acting as passive carriers that shuttle from the PM to EEs, we propose that CLICs arise from the plasma membrane and rapidly mature into a high capacity sorting station. We have shown that CLICs are a high capacity, dynamic, and complex sorting system and propose that they have the ability to sort subcompartments to distinct destinations for critical cellular processes such as membrane repair, migration, and maintenance of the PM.

Cells on 12-mm coverslips or 35-mm dishes were serum starved for 4 h in serum-free DME (Invitrogen) before two times 2-min wash in precooled CO₂-independent media (Invitrogen) on ice and placed on 50-µl drops of 10 µg.ml⁻¹ CTxB-555 (Invitrogen), 5 µg.ml⁻¹ Tf-488 (Invitrogen), and/or anti-GFP as desired for 20 min. For stereology, 35-mm dishes were not incubated on ice but were subjected to constitutive uptake at 37°C, 5% CO₂ for desired times. Coverslips were moved to prewarmed serum-free DME and incubated at 37°C, 5% CO₂ for desired times. Coverslips were placed back on ice in precooled CO₂-independent media. Two times 30-s 0.5 M glycine (pH 2.2) acid stripping was routinely performed before fixation in 2% paraformaldehyde. For immunocytochemistry, cells were permeabilized in freshly made 0.1% saponin for 10 min and blocked in 0.2% fish skin gelatin/0.2% BSA in PBS. Coverslips were incubated with primary antibodies for 1 h at room temperature, followed by three times 10-min wash in PBS before incubation with secondary Alexa Fluor–conjugated antibodies at 1:400 for 1 h.

Fluorescence micrographs were captured on a confocal laser-scanning microscope (Axiovert 200 m SP LSM 510 META; Carl Zeiss, Inc.). Images were captured under oil with a 63x Plan-Apochromat objective, at excitation wavelengths of 488, 543, and 633 nm for GFP-tagged constructs or Alexa Fluor 488–, Alexa Fluor 555–, or Alexa Fluor 647/660–conjugated antibodies, respectively. Band-pass and meta filters were applied as appropriate for respective fluorophores and images were captured from photomultiplier tube detectors (Carl Zeiss, Inc.). Colocalization was quantitated using Volocity v3.7. Images of individual cells were split into RGB channels, automatic thresholding applied, and colocalization coefficients calculated on percentage of overlapping voxels. Antibodies used include Thy-1 and CD44 (Abcam); CHC, GM130, Grp78, Cav1, and HA (BD); myc (Sigma-Aldrich); and TfR (Invitrogen). HRP-conjugated anti-CD44 was generated with Lightning-Link HRP conjugation kit (Innova Bioscience) as per the manufacturer’s instructions.

Cells were serum starved for 4 h in serum-free DME before incubation for 20 min with 30 µM Dyngo4a, 80 µM dynasore (Sigma-Aldrich), or 100 nM wortmannin (Roche). Cells were then incubated with an endocytic marker of interest in the presence of Dyngo4a and wortmannin for indicated periods at 37°C.

After constitutive internalization of 10 µg.ml⁻¹ CTxB-HRP (Invitrogen) at 37°C, cells were placed on ice and washed two times for 1 min in ice cold PBS, followed by 10 min in 1 mg.ml⁻¹ DAB (Sigma-Aldrich) ± 50 µM ascorbic acid (AA) in PBS. Cells were then incubated with DAB ± 50 µM AA with 0.012% H₂O₂ for 20 min, followed by fixation in 2.5% glutaraldehyde (ProSciTech) in PBS for 1 h at room temperature. For plastic sections, cells were contrasted with 1% osmium tetroxide and 4% uranyl acetate. Cells were dehydrated in successive washes of 70%, 90% and 100% ethanol before embedding in LX-112 resin. For frozen sections, cells were scraped and embedded in 10% gelatin and gelatin blocks were infused with 2.3 M sucrose. Blocks were frozen onto mounting pins and 55-nm sections were collected using a UC6 ultramicrotome (Leica). Tomograms of 300-nm plastic sections were captured in 1° increments, tilted ±60° on a transmission electron microscope (model F30; Tecnai) using UltraScan 4000 semi-automated software. High pressure frozen samples were essentially processed as described previously (Richter et al., 2008). In brief, CTxB-HRP was added directly to cells grown on sapphire discs and DAB reaction was performed on ice before discs were frozen in a Balzers HPM010 high pressure freezer (BAL-TEC AG) and stored under liquid nitrogen. Freeze substitution was performed in acetone with 0.7% uranyl acetate; 1% osmium tetroxide at −90°C before warming to 0°C over 3 d and infiltrated with Embed 812-Araldite resin.

NIH3T3–GFP-GPI cells were serum starved and treated for 20 min with 100 nM wortmannin and 30 µM Dyngo4a. Anti-GFP antibodies were directly added to treated cells in the presence of wortmannin and Dyngo4a for 5 min at 37°C. Cells were placed on ice and harvested by scraping in TES buffer (10 mM tricine, 1 mM EDTA, and 250 mM sucrose, pH 7.4) with dissolved complete protease inhibitor cocktail (Roche). Cells were homogenized by passaging 10 times each through 21- and 27-gauge needles. A 17,000-g spin at 4°C for 10 min was applied to sediment nonlysed cells and nuclei. Post-nuclear supernatants were mixed with an equal volume 70% Nycodenz in TES and placed on the bottom of a four-step gradient, consisting of 25%, 10%, and 0% Nycodenz steps. Gradients were spun at 160,000 g for 2 h. The 10% fraction (excluding the interface) was collected and diluted 1/3 in TES with protease inhibitors and placed on top of a 5–20% continuous Nycodenz gradient. This was spun at 206,000 g for 2 h. Fractions were collected from the top of the gradient and TCA-precipitable samples analyzed by SDS-PAGE and LC/MS-MS. For visualization of Nycodenz fraction morphology by EM, 10 µl of fraction of interest was mixed with 10 µl 5% glutaraldehyde and a 100 mesh copper EM grid was placed on top for 30 min at room temperature. Grids were washed three times 5 min with PBS, followed by five times 2 min H₂O before staining with 1:9 (by volume) of 4% uranyl acetate to 2% methyl cellulose for 10 min on ice.

CTxB-555 was biotinylated with 2 mg.ml⁻¹ EZ-link Sulfo-NHS-SS-biotin (Thermo Fisher Scientific) for 10 min at 37°C and biotin quenched with 1:1 volume of 200 mM glycine for 5 min. CTxB-biotin was added to Cav1^−/− MEFs either treated with 30 µM Dyngo4a or left untreated for the times indicated. After CTxB-biotin internalization, cells were placed on ice and stripped with three times 5-min washes with 100 mM MesNa (Sigma-Aldrich). MesNa was quenched with three times 5-min washes with 54 mM iodoacetamide (Sigma-Aldrich). Cells were harvested in TNE (20 mM Tris pH 7.5, 150 mM NaCl, and 5 mM EDTA) with complete protease inhibitors (Roche) and solubilized with 1% Triton X-100. BCA protein assays (Thermo Fisher Scientific) were routinely performed to accommodate equivalent loading. Western blots were probed with streptavidin-HRP (Invitrogen) in 5% skim milk, 0.1% Tween 20. Intensity of developed HRP chemiluminescent bands was determined for multiple exposures per sample in ImageJ using Gel Analyzer.

Serum-starved cells were incubated directly with 10 µg.ml⁻¹ CTxB-HRP, 20 mg.ml⁻¹ HRP (Sigma-Aldrich), 10 µg.ml⁻¹ anti-CD44-HRP antibodies or 20 µg.ml⁻¹ Tf-HRP (Invitrogen) for desired times before being placed on ice, washed with precooled CO₂-independent media, and processed for the DAB reaction and EPON embedding. Vertical sections, visualized at 20,000x on a transmission electron microscope (model 1011; JEOL), were overlayed with a 2,000-nm square lattice grid using iTEM software. The number of intersections that fell on top of the cytoplasm, excluding the nucleus, were recorded across 20–25 cells and three independent areas. A 200-nm square lattice grid was also applied and intersections that fell on top of DAB-labeled structures were recorded as above. Number of points lying over DAB-labeled structures multiplied by the grid size was compared, as a percentage, to points over cytoplasm by grid size to give the percentage of labeled structures to cytoplasmic volume (V(v)). Overprojection effects on endosomal tubules can lead to volume overestimation as described previously (Griffiths et al., 1989). Based on the section thickness, average diameter, and length of CLICs, a correction factor of 0.65 can be applied to the V(v) measurements, based on graphs previously developed (Weibel, 1979). Correction factors are approximate only and give an estimate of possible error. Based on tomography data (Fig. 3), however, CLICs display a depth greater than section thickness (entire volume is still within 300-nm section, far thicker than 55-nm sections used for volume calculation), in which case overprojection effects become negligible.

Surface density, or surface volume fraction (S(v)), was calculated by placing a cycloid grid over high resolution images of CTxB-HRP–labeled structures and using the formula: S(v) = 2I/P•l(p), where I = total intersections of cycloid grid with structure of interest, P = total points of cycloid grid lying over structure of interest, and l(p) = length of the cyloid in µm, which was an absolute distance of 75 nm in this study.

Total cellular volume of a Cav1^−/− MEF was calculated by binding CTxB-555 to fixed cells. Confocal Z-stacks were obtained from five representative cells and total average cellular volume was calculated using Volocity v3.7. Surface area of Cav1^−/− MEF was calculated by finding the S(v) of plasma membrane across 25 cells, from low magnification electron micrographs, multiplied by the total average volume.

To capture the average volume of single CTxB-HRP–positive CLIC, high resolution electron micrographs from whole-mount profiles were used. As CTxB-HRP label (as well as HRP and anti-CD44-HRP label) was only found within the tubular ring of CLICs, we assumed that the entire volume of CLICs is comprised solely within the tubule space. Therefore, the volume of a sphere based on the diameter of the inner membrane was subtracted from the volume of a sphere based on the diameter of the outer membrane of the rings. This calculated that an average CLIC had an estimated volume of 0.00263 ± 0.00048 µm³. In comparison, the volume of a single CCV was estimated to be five times smaller (0.0005 µm³ assuming sphere based on diameter of 100 nm).

An Agilent 1100 Binary HPLC system (Agilent Technologies) was used to perform reversed phase separation of the in-gel digests using a Vydac MS C18 300A column (150 mm × 2 mm) with a particle size of 5 µm (Vydac). Mass spectrometry experiments were performed on a hybrid quadrupole/linear ion trap 4000 QTRAP MS/MS system (Applied Biosystems). All analyses were performed using information-dependent acquisition and the linear ion trap (LIT) acquisition modes. Analyst 1.4.1 software was used for data analysis.

Confluent WT MEFs grown on coverslips were scratched with a 200-µl pipette tip. Cells were washed and allowed to recover for 4 h before internalization assays. Average pixel intensity from leading to trailing edges within a selected area was calculated using the Plot Profile option in ImageJ. Plot profiles show a representative image of 5–10 cells across three independent experiments.

10 µg.ml⁻¹ CTxB-HRP was added to cell monolayers for 2 min at 37°C. Cells were then incubated with 10 mg.ml⁻¹ DAB, 50 µM AA in DME for 2 min before incubation with 10 mg.ml⁻¹ DAB, 50 µM AA, 0.012% H₂O₂ in DME for a further 2 min. Cells were washed five times with DME before incubation in normal growth media for 4 h. CTxB-555, Tf-647, and anti-CD44 antibodies were then directly added to cells for 2 min before acid stripping, fixation, permeabilization, and labeling for anti-CD44 antibodies. For quantitation of distance migrated, monolayers were imaged after DAB inactivation (time zero) and again after 4 h. Distance between front edges of monolayers on both sides of the wound was measured using Adobe Photoshop CS2. Anti-CD44-HRP and Tf-HRP failed to inhibit either clathrin-independent or -dependent endocytosis at concentrations of 10–100 µg.ml⁻¹ as used at early times in this assay.

Fig. S1 shows that wortmannin treatment of MEFs inhibits the fusion of CLICs with early endosomes, leading to a threefold increase in the number of CLICs per cell. Wortmannin treatment does not noticeably affect the protein profiles of the CLIC-enriched fraction. Fig. S2 shows the consistent nature of membrane profiles in the CLIC-enriched fraction and that CTxB-HRP labels a majority of these profiles when added to cells before the fractionation. Fig. S3 shows the localization of dynamin to CLICs after wortmannin treatment by immunofluorescence and by ultrastructure. Dynamin also labels the neck of profiles in the CLIC-enriched fraction. Video 1 shows the tilt series of CTxB-HRP–laden CLICs after 15 s of internalization. Table S1 lists all the peptides identified from fraction 2.8 with two or more 99% confidence hits from the LC-MS/MS.

The Institute for Molecular Bioscience is a Special Research Centre of the Australian Research Council. Confocal microscopy was performed at the Australian Cancer Research Foundation (ACRF)/Institute for Molecular Bioscience Dynamic Imaging Facility for Cancer Biology, which was established with the support of the ACRF. The authors acknowledge the facilities as well as scientific and technical assistance from staff in the Australian Microscopy and Microanalysis Facility (AMMRF) at the Centre for Microscopy and Microanalysis at The University of Queensland. We thank Jean Gruenberg for critical appraisal of the manuscript.

This work was supported by grants from the National Health and Medical Research Council of Australia (511005; R.G. Parton and J.F. Hancock) and the Human Frontier Science Program.