Research Articles

Tubular clathrin/AP-2 lattices pinch collagen fibers to support 3D cell migration

See allHide authors and affiliations

Science  16 Jun 2017:
Vol. 356, Issue 6343, eaal4713
DOI: 10.1126/science.aal4713

Helping a cell to migrate in 3D space

Clathrin-coated pits are well known to be involved in receptor-mediated endocytosis. Independent of their role in endocytosis, Elkhatib et al. observed that clathrin-coated structures strongly accumulated along collagen fibers in migrating cells. Clathrin-coated structures assembled on and then partially wrapped around and pinched the fibers. In a three-dimensional (3D) network, this mechanism provided multiple anchoring points along cellular protrusions. In the absence of clathrin-coated structures, protrusions were shorter and migration was impaired. This mode of adhesion may cooperate with classical focal adhesions to help cancer cells move in a 3D environment.

Science, this issue p. eaal4713

Structured Abstract

INTRODUCTION

Migrating cells need to adhere to their environment in order to pull themselves forward. Adhesion to the extracellular matrix (ECM), including collagen fibers, is achieved by integrins that cluster in focal adhesions (FAs). Integrin clustering is a key process during migration because it increases the cell’s avidity for the ECM. The physical connection between the cellular traction machinery and the ECM that occurs at FAs allows the cell to pull on the substratum, which generates forces that are required for locomotion both in two-dimensional (2D) and 3D environments.

RATIONALE

FAs are numerous and prominent when cells migrate on 2D, stiff substrata. However, cells migrating on softer 2D or in 3D environments show fewer and/or seemingly weaker FAs. In addition to FAs, integrins have been shown to accumulate in other structures of the plasma membrane such as clathrin-coated structures (CCSs). CCSs are professional cell surface receptor–clustering machineries that also bend the membrane to progressively pack these receptors into endocytic clathrin-coated vesicles that bud into the cytosol. Although the structure, dynamics, and functions of CCSs have been extensively described in cells seeded on 2D environments, little is known about their features in more physiological soft, 3D conditions. We analyzed the dynamics of CCSs in cells migrating in a 3D environment of collagen fibers.

RESULTS

We observed that CCSs accumulated along collagen fibers in the 3D environment. This was the result of an increased nucleation rate as well as a prolonged lifetime of CCSs at cell–collagen fiber contact sites. Both local plasma membrane bending, resulting from the physical contact with collagen fibers, and local integrin engagement triggered CCS accumulation on fibers. Electron microscopy analyses revealed that the CCSs engaged with collagen fibers adopted a distinct, tubular morphology to wrap around and pinch the fiber. Surprisingly, clathrin was not required for the coat of clathrin-adaptors, which includes adaptor protein 2 (AP-2) and Dab2, to accumulate on and wrap around fibers. We named these particular type of CCSs tubular clathrin/AP-2 lattices (TCALs). Cell adhesion and the cell’s capacity to grab collagen fibers were inhibited by disruption of TCALs or by inhibition of integrin accumulation at TCALs by using AP-2 and Dab2 siRNAs, respectively. However, in agreement with our morphological analysis, clathrin itself was not required for cells to adhere to fibers, demonstrating that this process is independent of CCSs’ role in endocytosis. FAs were mostly found at both extremities of elongated cells migrating in the 3D environment, whereas CCSs were distributed all over the plasma membrane of cellular protrusions. By laser ablating the tip of individual cellular extensions, we demonstrated that TCALs stabilize protrusions and thereby promote cell migration, in an endocytosis-independent manner.

CONCLUSION

Thus, in soft, 3D environments, CCSs promote cell migration not only through the regulation of endocytosis but also by providing anchoring points to collagen fibers. This is necessary for the cell to cope with the tension exerted at FAs, which allows cells to develop long protrusions and to migrate efficiently. We conclude that TCALs represent a mode of cell adhesion that, in coordination with FAs, supports cell migration in 3D.

Structural and functional analysis of TCALs at cell–collagen fiber contact sites.

Cells migrating in a 3D network of collagen fibers (red) adopt an elongated morphology as they extend protrusions in between fibers in order to move forward. CCSs (green) accumulate on and pinch collagen fibers, thus providing local anchoring points to the environment. This allows the cell to stabilize long extensions that are needed to migrate efficiently.

Abstract

Migrating cells often use focal adhesions in order to move. Focal adhesions are less prominent in cells migrating in three-dimensional (3D) as compared with 2D environments. We looked for alternative adhesion structures supporting cell migration. We analyzed the dynamics of clathrin-coated pits in cells migrating in a 3D environment of collagen fibers. Both topological cues and local engagement of integrins triggered the accumulation of clathrin-coated structures on fibers. Clathrin/adaptor protein 2 (AP-2) lattices pinched collagen fibers by adopting a tube-like morphology and regulated adhesion to fibers in an endocytosis-independent manner. During migration, tubular clathrin/AP-2 lattices stabilized cellular protrusions by providing anchoring points to collagen fibers. Thus, tubular clathrin/AP-2 lattices promote cell adhesion that, in coordination with focal adhesions, supports cell migration in 3D.

Clathrin-mediated endocytosis is a fundamental process that controls a wide variety of cell functions, including cytokinesis (1), cell migration, and cell invasion (2). Clathrin-coated structures (CCSs) allow migrating cells to adapt to their environment by selectively controlling the uptake of specific cargos (3), regulating processes such as cell protrusion (4) and adhesion dynamics (5, 6). CCS recruitment at focal adhesions (FAs) regulates integrin endocytosis and leads to adhesion disassembly and cell movement (5, 6). Conversely, CCS dynamics is locally modulated by cellular adhesions (7, 8). However, whether this reflects a physical engagement of integrin-rich CCSs with the extracellular matrix (ECM) or an indirect regulation by adhesion sites is unclear. In addition, integrin-positive flat clathrin structures (9) establish tight contacts with planar substrates (10), suggesting that they could participate directly in cell adhesion (11). The extracellular environment impinges on adhesion dynamics, and FAs are reduced in size and number when cells are cultured on physiologically soft substratum or in a soft three-dimensional (3D) network of collagen fibers (1216). However, although clathrin-mediated endocytosis plays a role in cancer cell migration in a 3D network of collagen fibers (4, 17), CCS dynamics, structure, and functions in physiologically relevant environments are unknown. We set out to investigate CCS dynamics in cells facing a 3D environment composed of collagen fibers.

CCSs accumulate along collagen fibers

We first observed that CCSs of breast cancer–derived MDA-MB-231 cells marked with the α-adaptin subunit of the clathrin adaptor protein 2 (AP-2) were often located along collagen fibers in the 3D network (Fig. 1A). Visualization of 3D reconstructions of cells in the collagen network (movie S1), followed by automatic detection of CCSs and fibers, revealed that 63.3 ± 6.6% (SEM) of CCSs were in contact with fibers [as compared with 43.5 ± 7.9% (SEM) when the position of collagen fibers were randomized; P < 0.01] (Materials and methods). Electron microscopy analyses confirmed that CCSs established intimate contacts with collagen fibers [64.1 ± 17.5% (SEM) of detectable CCSs were contacting collagen fibers] (fig. S1, A to C). Although the ultrathin sectioning of the sample did not allow us to fully visualize the morphology of CCSs in contact with fibers, we occasionally observed collagen fibers apparently laying in the cleft of invaginated CCSs (Fig. 1B and fig. S1, A to C). The strong accumulation of CCSs along collagen fibers suggested that either the lifetime of CCSs and/or their nucleation rate at cell/collagen fibers contact sites were increased. Live cell imaging of genome-edited MDA-MB-231 cells expressing a green fluorescent protein (GFP)–tagged endogenous μ2-adaptin subunit of AP-2 (18) revealed that CCSs tended to nucleate preferentially on collagen fibers (Fig. 1C, fig. S1D, and movie S2) and that fiber-contacting CCSs were longer-lived as compared with CCSs located in other regions of the plasma membrane (Fig. 1D and movie S2). These observations explain why CCSs accumulate on collagen fibers and suggest that collagen fibers control CCS recruitment.

Fig. 1 CCSs accumulate along collagen fibers.

(A) MDA-MB-231 cells were seeded in a 3D network of collagen fibers (red) and fixed 24 hours later before being stained for α-adaptin (green). A single optical section is shown. Scale bar, 10 μm. (Insets) Higher magnifications of boxed regions. Scale bars, 3 μm (B) Electron microscopy micrograph of a MDA-MB-231 cell in the 3D collagen network depicting a collagen fiber in the cleft of a CCS. Scale bar, 100 nm. (C) Kymograph showing CCS (green) dynamics and collagen fibers (red) in a genome-edited MDA-MB-231 cell expressing endogenous GFP-tagged μ2-adaptin, imaged with spinning disk microscopy every 5 s for 5 min in the 3D collagen network. (D) Quantification of the lifetime of CCSs located on fibers or not, as indicated in genome-edited MDA-MB-231 cells in the 3D network (mean ± SEM; n ≥ 125 CCSs, *P < 0.01, Student’s t test). (E) MDA-MB-231 cells spreading on a thin layer of collagen fibers (red) polymerized on glass were fixed at the indicated time point and stained for α-adaptin (green). (F) Distribution of the average fluorescence intensity of α-adaptin staining per collagen fibers as a function of cell spread area. For linear regression, correlation coefficient (r) = 0.6475; P < 0.0001, ANOVA test. (G) Average fluorescence intensity of α-adaptin staining on collagen fibers 15 min after plating in control or hypotonic medium as indicated (mean ± SEM; n = 3 experiments, *P < 0.01, Student’s t test).

To study the underlying mechanisms of this process, we used a 2D system by polymerizing a limited number of collagen fibers on glass coverslips. Cells plated on such a substratum contact both collagen fibers and the collagen-coated glass coverslip. Fifteen min after plating cells, CCSs strongly aligned along collagen fibers (Fig. 1E). Similar accumulation of CCSs along collagen fibers was also visible in HeLa cells (fig. S2A). Vinculin-positive focal and/or nascent adhesions were often found on collagen fibers as well but were more peripherally distributed as compared with CCSs, and CCSs did not colocalize with them nor with cortactin stretches that mark ECM degradation structures termed linear invadopodia (fig. S2, B and C) (19, 20). In addition, clathrin heavy chain (CHC) and dynamin-2 were not required for AP-2–positive structure recruitment on collagen fibers (fig. S2, D to G). When cells were allowed to spread for more than 15 min, the number of CCSs colocalizing with collagen fibers dropped, and by 1 hour, no obvious alignment was detected anymore (Fig. 1E). The loss of CCSs from collagen fibers during cell spreading on 2D differs from the 3D condition in which CCSs were consistently seen associated with fibers long after embedding the cells in the collagen network. We noticed that cell spreading area was inversely correlated with the degree of CCSs alignment along fibers at any given time point after cell plating on 2D (Fig. 1F). Inhibiting cell spreading by inactivating the FA-associated protein talin increased alignment of CCSs along collagen fibers, also at late time points (fig. S2, G to I). Because membrane tension increases during cell spreading in two dimensions (21, 22) and has been proposed to be weaker in softer 3D environments (23), we speculated that it could be an important parameter regulating CCS accumulation on collagen fibers. Indeed, increasing membrane tension by exposing cells to hypotonic medium led to a partial loss of CCSs from fibers (Fig. 1G). Thus, high membrane tension resulting from cell spreading on a stiff 2D substrate prevents CCS accumulation on collagen fibers.

Tubular clathrin/AP-2 lattices pinch collagen fibers

We next took advantage of the 2D system to analyze the morphology of collagen fiber–contacting CCSs in a more comprehensive manner than was possible in the 3D condition. For this, we performed metal-replica electron microscopy analysis of the adherent plasma membrane of unroofed cells after a 15-min plating period onto collagen fibers. The electron density of collagen fibers made them easily traceable, even when located below the plasma membrane (Fig. 2, A to F). This feature allowed us to unambiguously analyze “en face” the many CCSs that were detected on collagen fibers, which is in agreement with our immunofluorescence analyses (Fig. 2, A to D). Clathrin coats on fibers were composed of continuous and interconnected lattices that elongated along the long axis of the fiber, seemingly wrapping around and pinching it (Fig. 2, B to D, and fig. S3, A and B). 3D visualization confirmed that these structures assembled as semitubes around collagen fibers (Fig. 2, C and D, and fig. S3, A and B, insets), demonstrating the plasticity of clathrin assemblies that were able to adopt the topology of the plasma membrane. These tubes could span the size of three to four canonical clathrin-coated pits and may represent a curved version of the flat clathrin lattices observed in areas of the plasma membrane far from collagen fibers (Fig. 2C and fig. S3B). It is also possible that clathrin-coated tubes are an intermediate arrangement of maturing CCSs on fibers. Furthermore, we confirmed through immunogold labeling of cells completely lacking any detectable clathrin coats at the plasma membrane (Materials and methods) that CHC was dispensable for the accumulation of α-adaptin on collagen fibers (Fig. 2E and fig. S3C). Close inspection of metal replicas revealed the presence of characteristic patches of proteinaceous material, which corresponded to the coat of clathrin-adaptors (Fig. 2, G to I, and fig. S3, D and E). These structures were also found to accumulate on collagen fibers, confirming that clathrin itself was not required for patches of adaptors to accumulate on and to wrap around fibers (Fig. 2F). Thus, we named the CCSs that accumulated on collagen fibers tubular clathrin/AP-2 lattices (TCALs).

Fig. 2 Ultrastructural characterization of TCALs.

(A to I) Survey view of the cytoplasmic surface of the plasma membrane in unroofed control [(A) to (D) and (G)] or CHC-depleted [(E), (F), (H), and (I)] MDA-MB-231 cells plated for 15 min on a thin layer of collagen fibers polymerized on glass. For (B) to (F), use view glasses for 3D viewing of anaglyphs (left eye, red). In (E), (G), and (H), α-adaptin was immunolabeled with gold particles (yellow dots). Arrows indicate collagen fibers. Pictures in (G) to (I) show areas of the plasma membrane in contact with the flat, collagen-coated glass coverslips and depict flat clathrin-coated lattice (G) and characteristic patches of proteinaceous material [(H) and (I)] positive for α-adaptin (H) and thus corresponding to the coat of clathrin-adaptors. Scale bars, (A) 1 μm, (B) to (H) 500 nm, and (I) 200 nm.

In agreement with our immunofluorescence analysis, we observed fewer clathrin coats on collagen fibers after a 30-min plating period (fig. S4A). In contrast, and also confirming our immunofluorescence analysis, TCALs were abundant on collagen fibers in talin-depleted cells after a 30-min spreading period (fig. S4, B to D). Flat clathrin lattices in regions away from fibers were rare in control cells at this time point (fig. S4A). This latter observation is in agreement with the measured lifetime of CCSs in cells plated on a planar, collagen-coated substratum and analyzed during different time windows upon plating [69.2 ± 28.8 s (SEM) during the first 15 min of spreading versus 43.2 ± 13.9 s (SEM) between 15 and 30 min of spreading], suggesting that clathrin structures are longer-lived early after plating because they remain flat for a prolonged period of time.

Mechanism of CCS recruitment on fibers

Both CCS lifetime and nucleation rate are increased on collagen fibers in the 3D network, explaining why CCSs accumulate on fibers (Fig. 1, C and D). However, in the 2D system we did not measure a significant difference in lifetime between CCSs located on fibers or in other areas of the plasma membrane (Fig. 3A). This may have been a consequence of CCSs’ engagement with the ECM because in this assay, the cell contacted both collagen fibers and the collagen-coated glass. Nevertheless, CCS nucleation rate was increased on collagen fibers in the 2D system, which is in agreement with our observations in the 3D situation (Fig. 3, B and C). We reasoned that the plasma membrane of cells in contact with collagen fibers is likely to be locally curved. Indeed, adaptor-dependent clathrin-assembly is facilitated by membrane curvature in reconstituted in vitro systems (24). Thus, we decided to analyze whether the local topology of the plasma membrane could account for an increased CCS nucleation rate. To test this hypothesis, we plated cells onto noncoated fluorescent beads spotted onto collagen-coated glass coverslips. We used 200-nm-diameter beads to mimic the deformation imposed by collagen fibers that have a comparable diameter. At 15 min after plating, CCSs often colocalized with fluorescent beads (Fig. 3, D and E). The number of beads colocalizing with CCSs was higher than expected by chance because randomizing the position of beads dramatically reduced the occurrence of colocalization events (Fig. 3E). Live cell imaging analysis showed that beads were stably bound to the coverslip and were not internalized by cells (Fig. 3F). In addition, multiple CCSs sequentially assembled on the same beads, confirming that local membrane curvature triggers CCS nucleation (Fig. 3F). When using collagen-coated fluorescent beads, the percentage of beads colocalizing with CCSs increased to nearly 45% (Fig. 3E). This suggested an active and local role of the ECM in CCS accumulation at curved sites. Because membrane curvature is key in controlling local CCS nucleation, and membrane tension inhibits CCS accumulation on collagen fibers (Fig. 1G), it is possible that high tension prevents the membrane to bend sufficiently at fiber/cell contact sites, thus preventing an increased CCS nucleation rate. Accordingly, we observed a slight but reproducible defect of CCS accumulation on beads when membrane tension was increased upon incubation in hypotonic medium (Fig. 3G).

Fig. 3 Membrane curvature and collagen control TCAL accumulation on fibers.

(A) Quantification of the lifetime of CCSs located on fibers or not, as indicated, in genome-edited MDA-MB-231 cells imaged for 10 min between 10 and 20 min after plating on a thin layer of collagen fibers polymerized on glass (mean ± SEM; n ≥ 125 CCSs, *P < 0.01, Student’s t test). (B) Kymograph of a genome-edited MDA-MB-231 cell that was plated on a thin layer of collagen (red) fibers and imaged with spinning disk microscopy every 5 s for 10 min between 10 and 20 min after plating. CCSs are marked with the endogenous GFP-tagged μ2-adaptin (green). (C) Quantification of CCS nucleation rate index on collagen fibers or after shifting the position of collagen fibers by 5 pixels to the right (Offset) (mean ± SEM; n = 3 experiments, *P < 0.01, Student’s t test). (D) MDA-MB-231 cells spreading on 200-nm-diameter beads (red) spotted on glass were fixed 15 min after plating and stained for α-adaptin (green). A merge picture is shown. Scale bar, 10 μm. (E) Quantification of the percentage of uncoated or collagen-coated beads found below cells and colocalizing with CCSs. Positions of beads were shifted by 5 pixels to the right in the offset condition (mean ± SEM; n = 3 experiments, *P < 0.01, Kruskal-Wallis one-way ANOVA). (F) Kymograph of a genome-edited MDA-MB-231 cell that was plated on 200-nm-diameter beads (red) spotted on a glass-bottom dish and imaged with spinning disk microscopy every 5 s for 20 min immediately after plating. CCSs are marked with the endogenous GFP-tagged μ2-adaptin (green). (G) Quantification of the percentage of collagen-coated beads found below cells and colocalizing with CCSs 15 min after spreading in isotonic or hypotonic media as indicated (mean ± SEM; n = 3 experiments, *P < 0.01, Student’s t test).

β1-integrin (the β-subunit of all the human collagen-binding integrin heterodimeric receptors) strongly accumulated along the entire length of collagen fibers in contact with the cell in the 2D assay, which is in agreement with recent findings in 3D collagen networks (Fig. 4A) (25). However, close inspection of β1-integrin staining showed that it was enriched at CCSs (Fig. 4B and fig. S5A). In addition, β1-integrin was also enriched at a subset of CCSs in cells plated on a planar, collagen-coated coverslip (fig. S5B). The alternative clathrin-adaptor Dab2 has been proposed to link β1-integrins to CCSs (5, 2628), and we observed that the vast majority of CCSs, including those located on collagen fibers, contained Dab2 (fig. S5C). Dab2 silencing inhibited β1-integrin enrichment in CCSs located on fibers, suggesting that it is necessary to cluster integrins at CCSs (Fig. 4C). In addition, cells depleted for Dab2 did not display the strong accumulation of CCSs along collagen fibers that is observed in control cells after a 15-min plating period (Fig. 4D and fig. S6, A to D). It was not possible to directly assess the role of β1-integrin in CCSs recruitment on collagen fibers in the 2D system because its inhibition prevents cell adhesion. However, inhibiting β1-integrins by using a blocking antibody (4B4) significantly reduced the lifetime of fibers-associated CCSs in the 3D network, without affecting the lifetime of CCSs in other region of the plasma membrane (Fig. 4E). Similarly, Dab2 depletion reduced the lifetime of CCSs on collagen fibers (Fig. 4E). Thus, CCSs on collagen fibers experience “frustrated” endocytosis because budding forces may be balanced by the Dab2- and β1-integrin–dependent engagement of CCSs with fibers.

Fig. 4 TCALs regulate binding to collagen fibers.

(A) MDA-MB-231 cells spreading on a layer of collagen fibers (left) were fixed and stained for activated β1-integrin (right) 15 min after plating. Scale bar, 10 μm. (B and C) Average fluorescence distribution along collagen fibers of β1-integrin and α-adaptin in (B) control or (C) Dab2-depleted cells. Error bars indicate mean ± SEM. (D) Average fluorescence intensity of α-adaptin staining per collagen fibers measured 15 min after plating in control or Dab2-depleted cells. (E) Average lifetime of CCSs in contact (black bars) or not (gray bars) with collagen fibers in the 3D network in cells treated with indicated siRNAs and with or without 4B4 blocking antibody, as indicated. (F) Calcein-loaded MDA-MB-231 cells treated with the indicated siRNAs were plated on collagen-coated 5-kPa acrylamide gel for 15 min before washing unattached cells. Cell-associated fluorescence was measured and expressed as a percentage ± SEM of control cells. (G) Quantification of collagen fibers movement measured during the first 15 min of spreading of MDA-MB-231 cells treated with the indicated siRNAs. For (D) to (G), data are expressed as mean ± SEM; n = 3 experiments, *P < 0.01, Kruskal-Wallis one-way ANOVA.

TCALs regulate cell adhesion to collagen fibers

Next, we investigated whether the tight interaction we uncovered between CCSs/TCALs and ECM could play a role in cell adhesion. As expected, adhesion of MDA-MB-231 cells to a collagen-coated substratum drastically depended on β1-integrin because virtually no cells could adhere upon incubation with a β1-integrin blocking antibody (fig. S7A). We next measured cell adhesion at 15 min after plating in cells depleted of AP-2 subunits, CHC, or dynamin-2. Cell adhesion to collagen-coated plastic dishes was not significantly affected upon inhibition of either of these proteins (fig. S7BC). However, plastic represents a rigid environment on which cell contractility is high (12), and FAs were readily detectable 15 min after plating (fig. S7D). On the other hand, cells showed very few detectable FAs 15 min after plating onto a 5-kilopascals (kPa) collagen-coated gel (fig. S7D). We reasoned that rapid FA formation on the nonphysiological hard substratum may hinder a potential role for CCSs in cell adhesion. Indeed, AP-2 was required for optimal cell adhesion on the soft substratum (Fig. 4F). However, neither CHC nor dynamin-2 depletion inhibited cell adhesion on soft collagen (Fig. 4F). Because AP-2–positive structures are still visible at the plasma membrane in the absence of CHC or dynamin-2 (fig. S2, D and E) (29, 30), our results indicate that AP-2–dependent protein clustering at the plasma membrane, but not endocytosis per se, is required for optimal adhesion. Along this line, AP-2–positive structures were still able to cluster β1-integrin on fibers in the absence of CHC (fig. S7E). We also assessed the role of the integrin-binding, phosphotyrosine binding domain (PTB)–containing clathrin adaptors Numb, ARH, and Dab2 in the adhesion assays. Only Dab2 was required for cell adhesion on the soft substratum (Fig. 4F and fig. S7, B and F). We next investigated the possibility that adhesive TCALs may allow the cell to “grab” collagen fibers. When cells are plated on a thin layer of collagen fibers, they face a mixed environment because the glass coverslip is rigid but the network of collagen fibers shows some degree of flexibility and can be rearranged by cells (fig. S7G). Inhibiting the formation of CCSs using AP-2–specific small interfering RNAs (siRNAs) impaired collagen fiber rearrangement (Fig. 4G). Dab-2 was similarly required for fiber rearrangement (Fig. 4G). However, we did not detect rearrangement defects with CHC- or dynamin-2–silenced cells, indicating that cells can grab collagen fibers in an endocytosis-independent manner (Fig. 4G). We obtained similar results with cells that underwent two rounds of CHC depletion (Materials and methods) (fig. S7H). In addition, expression of a wild-type Dab2 construct was able to correct the remodeling defect seen in Dab2-depleted cells, but a PTB-deleted mutant, unable to recruit β1-integrins to clathrin structures (27), was not (fig. S7I). Both constructs were correctly localized (fig. S7J), suggesting that integrin recruitment by Dab2 in CCSs is crucial to grab and move collagen fibers. Thus, CCSs regulate cell adhesion to collagen fibers, which depends on CCSs’ capacity to locally cluster integrins at the cell surface.

TCALs stabilize cellular protrusions

If TCALs are adhesive structures pinching and grabbing collagen fibers, they may help cells to migrate in the 3D network. Cell velocity in the 3D environment was reduced in AP-2– and Dab2-silenced cells and to a lesser extent in CHC- and dynamin-2–depleted cells (Fig. 5A). Cells treated with AP-2 or Dab2 siRNAs had a tendency to form more protrusions per unit of time as compared with that in controls, although this was not statistically significant (fig. S8A). In addition, AP-2 and Dab2, but not CHC, were required for optimal elongation of cellular protrusions (Fig. 5B). Similar differences were observed when measuring only the leading protrusion of actively migrating cells (fig. S8B). Thus, CCSs are required for the extension of long protrusions in an endocytosis-independent manner and suggest that CCSs’ roles in adhesion and endocytosis are two distinct functions that both participate in cell migration. Furthermore, CCS lifetime may not be a critical parameter in this process because CHC-depleted cells harbored very long-lived structures on collagen fibers [152.8 ± 43.8 s (SEM)] without any detectable consequence on protrusion dynamics and collagen remodeling capacity (Figs. 4G and 5B).

Fig. 5 TCALs regulate migration by stabilizing cell protrusions.

(A) Average velocity of MDA-MB-231 cells treated with the indicated siRNA and migrating in the 3D collagen network (mean ± SEM; n = 3 experiments, *P < 0.05, **P < 0.01, Kruskal-Wallis one-way ANOVA). (B) Average maximum elongation of all visible protrusions in MDA-MB-231 cells treated with the indicated siRNA and migrating in the 3D collagen network (mean ± SEM; n = 3 experiments, *P < 0.01, Kruskal-Wallis one-way ANOVA). (C) The extremity of a MDA-MB-231 cell migrating in a 3D collagen network (top) was laser ablated. (Top middle and bottom middle) Collagen fibers corresponding to boxed area at top, 1 s before and after laser ablation, as indicated. The dashed line shows where ablation was performed. (Bottom) The calculated displacement field based on collagen fiber pictures 1 s before and after laser ablation, to illustrate collagen fiber movements upon relaxation. Color-coded bar of displacement intensity is shown. Scale bar, 10 μm. (D) Representative galleries of protrusion retraction over 140 s after laser ablation of the extremity of MDA-MB-231 cells treated with the indicated siRNAs. Green arrows show the extremity of protrusions. Dashed lines show where laser ablation was performed. White arrows show the free extremity of retracting protrusions after ablation. Scale bar, 10 μm. (E) Average distance of protrusion retraction over 3 min after laser ablation of their extremity in MDA-MB-231 cells treated with the indicated siRNA (mean ± SEM; n = 3 experiments, *P < 0.01, Kruskal-Wallis one-way ANOVA).

Our data suggest that TCALs may help the cell to migrate by anchoring protrusions to collagen fibers (movie S3). FAs are mostly observed at the extremities of elongated cells in the 3D network (fig. S8C) (13), whereas CCSs are distributed all over the cell’s plasma membrane (Fig. 1A). We hypothesized that in addition to FAs, TCALs pinching collagen fibers may generate friction, helping the cell to build long protrusions and to migrate. To test this possibility, we laser ablated the FA-rich leading extremity of the main cell protrusion and monitored the consequences for its stability. Immediately after ablation, we observed a relaxation of the collagen network around the edge of the protrusion (Fig. 5C). Particle image velocimetry (PIV) analysis showed that relaxation occurred bidirectionally along the long axis of the protrusion (Fig. 5C). Fiber relaxation away from the leading edge demonstrated that the cell exerts traction forces on the network. Relaxation in the direction of the cell body is most likely a consequence of protrusion retraction that occurred immediately after ablation of the extremity (Fig. 5, D and E). This suggested that the protrusion was under tension before ablation—most likely as a result of pulling forces exerted by the cell at FAs. Silencing AP-2 or Dab2, but not CHC, led to a more pronounced retraction of protrusions as compared with that in control cells (Fig. 5, D and E, and movie S4). Thus, CCSs stabilize cellular protrusions in an endocytosis-independent manner. Together, our data suggest that TCALs anchor the protrusion to collagen fibers and that this is critical to sustain high tension across the cell and to form long protrusions.

Outlook

Overall, we have found that TCALs are recruited at contact sites with collagen fibers and control cell adhesion to these fibers, powering cell migration in 3D environments. We propose that this mode of adhesion is critical for the mesenchymal mode of migration, in which TCALs provide several and dynamic anchoring points to collagen fibers, allowing the cell to cope with FA-associated high tension across protrusions. Because ECM fibers are abundant in complex organisms, this mode of adhesion could play a general and universal role in helping cells to interact with their environment.

Materials and methods

Cell lines and constructs

MDA-MB-231 cells (a gift from P. Chavrier, Institut Curie, Paris, France) or genome edited MDA-MB-231 cells engineered to expressed an endogenous GFP-tagged μ2 subunit (a gift from D. Drubin, University of California-Berkeley, California, USA) were grown in DMEM Glutamax supplemented with 10% foetal calf serum at 37°C in 5% CO2. DNA sequences encoding residues 1-766 (full length) or 172-766 (ΔPTB) of Dab2 were obtained by PCR by using cDNA of mouse Dab2 (a kind gift from M. Mettlen, University of Texas Southwestern Medical Center, Dallas, USA) as a template. PCR fragments with engineered flanking restriction sites were subcloned into the multi-cloning sites of pEGFP-C2 (Clontech) to encode in-frame fusion proteins with the amino-terminal EGFP-tag. MDA-MB-231 lines stably expressing EGFP, EGFP-Dab2 or EGFP-Dab2ΔPTB were generated upon transfection of cells with corresponding plasmids followed by selection using G418. These MDA-MB-231 lines were used for rescue experiments in the collagen fibers remodelling assay (see below) upon depletion of endogenous Dab2.

Antibodies and drugs

Mouse monoclonal anti-clathrin heavy chain (CHC) antibody and mouse monoclonal anti-α-adaptin antibody were obtained from BD Transduction Laboratories (Becton Dickinson France SAS, Le Pont-De-Claix, France). Rabbit polyclonal anti-α-adaptin antibodies (M300) were purchased from Santa Cruz Biotechnology Inc. (Santa Cruz, CA, USA). Cortactin was obtained from Millipore and Vinculin antibodie was a kind gift from Dr. M. Glukhova. HRP-conjugated anti-mouse and anti-rabbit antibodies for Western blot and Cy3-conjugated anti-rabbit antibodies were from Jackson ImmunoResearch Laboratories (West Grove, PA, USA). Alexa488-conjugated anti-mouse antibodies and Alexa545-labeled phalloidin were from Molecular Probes (Invitrogen). 4B4 antibody was obtained from Beckman coulter. Blebbistatin was purchased from Sigma and used a final concentration of 10 μM.

RNA interference

For siRNA depletion, MDA-MB-231 cells were plated at 50% confluence and treated with the indicated siRNA (30 nM) using RNAimax (Invitrogen, Carlsbad, CA) according to the manufacturer's instruction. Protein depletion was maximal after 72 hours of siRNA treatment as shown by immunoblotting analysis with specific antibodies. For some experiments a modified protocol was used (see below). Equal loading of the cell lysates was verified by immunoblotting with anti-tubulin antibodies. The following siRNAs were used: α-adaptin, 5′-AUGGCGGUGGUGUCGGCUCTT-3′; μ2-adaptin, 5′-AAGUGGAUGCCUUUCGGGUCA-3′; Clathrin heavy chain (CHC), 5′GCUGGGAAAACUCUUCAGATT-3′; Dynamin-2 SMARTpool siRNAs (Dharmacon); Talin, 5′-ACAAGAUGGAUGAAUCAAUUUU-3′; Dab2-1, 5′-GGUUGGCCUUAGUAGUCAATT-3′; Dab2-2, 5′-GAGCAUGAACAUCCAGAUAATT-3′; ARH-1, 5′-GAUACAGCUUGGCACUUUATT-3′; ARH-2, 5-'CAGACAAGAUGCACGACAATT-3′; Numb-1, 5′-GGACCUCAUAGUUGACCAGTT-3′; Numb-2, 5′-GAUAGUCGUUGGUUCAUCATT-3′; non-targeting siRNAs (siControl), ON-TARGETplus Non-Targeting SMARTpool siRNAs (Dharmacon).

Fluorescent beads

Red fluorescent (580/605) 200 nm diameter carboxylate-modified fluospheres (Thermofisher) were incubated or not overnight with 50 μg/ml collagen in 0.02M acetic acid before being sonicated and spotted on collagen-coated glass coverslips or glass-bottom fluodishes (World Precision Instruments) for 30 min at 37°C. 300,000 parental or genome edited MDA-MB-231 cells were seeded in complete medium in 6 well plates or fluodishes and imaged immediately by spinning disk microscopy (see below) or fixed in ice-cold methanol at indicated time after plating and processed for immunofluorescence using anti-α-adaptin antibodies. At least 20 cells per conditions and per experiments in 3 independent experiments were analyzed.

Collagen networks

For preparation of a layer of thin collagen fibers, a 30 μl droplet containing a 10:1 ratio of unlabeled and Alexa548-labeled collagen type I (rat tail acid extracted, BD bioscience) respectively, at a final concentration of 2.2 mg/ml was polymerized on glass coverslips or on glass-bottom dishes (Fluorodish, World Precision Instruments) at room temperature for 3 min before the excess of non-polymerized collagen being gently washed with PBS. The preparation was used immediately and immerged in complete medium seeded with 100,000 cells/ml. Cells were imaged during the spreading period by spinning disk microscopy (see below) or fixed with ice-cold methanol at the indicated time after plating and stained with indicated antibodies. For detection of β1-integrin, cells were incubated at 4°C for 1 hour in the presence of the anti-activated β1-integrin antibody (clone TS2/16, Santa Cruz) before to be fixed and processed for secondary and other antibodies labeling. At least 20 cells per conditions and per experiments in 3 independent experiments were analyzed.

For 3D cell migration assays, a 50 μl droplet of collagen I at a final concentration of 2.2 mg/ml and seeded with 10,000 MDA-MB-231 cells was polymerized on glass-bottom dishes at room temperature for 30 min before to immerge the setup in pre-warmed complete medium. The next day, cells were imaged by phase contrast microscopy at 100 ms exposure every 20 min for 30 hours for migration assays or every 2 min for 4 hours for protrusion dynamics measurements. For CCSs dynamics analysis in the 3D environment, the droplet of collagen contained a 10:1 ratio of unlabeled and Alexa548-labeled collagen, respectively, and cells were imaged the next day by spinning disk microscopy (see below). In all cases, a single optical section was selected and imaged. Manual tracking of cell migration and measurement of protrusion or CCSs dynamics parameters were performed using Metamorph software. At least 10 cells per conditions and per experiments from 3 independent experiments were analyzed.

Indirect immunofluorescence microscopy and fluorescence quantification

MDA-MB-231 cells plated on the top of a thin layer of collagen fibers on coverslips or on the top of fluorescent beads spotted on collagen-coated coverlips were fixed in ice-cold methanol and processed for immunofluorescence microscopy by using the indicated antibodies. Cells were imaged through a 100× 1.40NA UPlanSApo objective lens of a wide-field IX73 microscope (Olympus) equipped with an Orca-Flash2.8 CMOS camera (Hamamatsu) and steered by CellSens Dimension software (Olympus).

For anti-vinculin and β1-integrin staining, cells were briefly extracted for 1 min using 0.1% Triton prior to fixation.

For calculating the degree of CCSs alignment along collagen fibers, collagen fibers were segmented using Fiji software and the average fluorescence intensity of the anti-α-adaptin staining in fibers area was measured for each individual cells and normalized to the area occupied by collagen fibers. At least 20 cells per conditions and per experiments in 3 independent experiments were analyzed. Data are expressed as a mean percentage of control situation ± SEM.

For 3D immunofluorescence analyses, cells were fixed with 4% PFA 24 hours after embedding in the collagen network containing a 10:1 ratio of unlabeled and Alexa548-labeled collagen. Cells were stained using anti-α-adaptin antibodies, (AP6 clone, ThermoFisher). Stacks of images were collected along the z-axis with a 0.2 μm interval between optical sections and cells were reconstructed in 3D using Imaris software. Distance between each individual α-adaptin-positive dots and the most adjacent collagen fibers were automatically measured upon segmentation of both structures using the Spots-Close-To-Surface XTension tool of Imaris software. Parameters used to segment objects were: 0.8 μm XY and 1.5 μm Z diameters for spot detection in the green channel (AP-2 staining) and 0.8 μm spherical diameter for the Surfaces objects detection module in the red channel (collagen fibers). A distance of less than 0.5 μm was considered a positive hit (see movie S1). To assess the specificity of colocalization, the same process was applied to the same cells matched with irrelevant collagen network regions. 12 cells per experiments in 2 independents experiments were analyzed.

Electron microscopy of unroofed cells

Adherent PM from MDA-MB-231 cells plated for 15 or 30 min on glass coverslips coated with a thin layer of collagen were disrupted by sonication as described previously (31). Glutaraldehyde/paraformaldehyde-fixed cells were further sequentially treated with OsO4, tannic acid and uranyl acetate prior to dehydration and Hexamethyldisilazane drying (HMDS, Sigma-Aldrich). Dried samples were then rotary-shadowed with platinum and carbon with a high vacuum sputter coater (Leica). Platinum replicas were floated off the glass by angled immersion into hydrofluoric acid, washed several times by floatation on distilled water, and picked up on 200 mesh formvar/carbon-coated EM grids. The grids were mounted in a eucentric side-entry goniometer stage of a transmission electron microscope operated at 80 kV (Philips, model CM120) and images were recorded with a Morada digital camera (Olympus). Images were processed in Adobe Photoshop to adjust brightness and contrast and presented in inverted contrast. Anaglyphs were made by converting the -10° tilt image to red and the +10° tilt image to cyan (blue/green), layering them on top of each other using the screen blending mode in Adobe Photoshop, and aligning them to each other.

For analyzes of CHC-depleted cells, cells were transfected once with CHC-specific siRNAs and then a second time, 48 hours later, with the same siRNAs to completely eliminate any trace of clathrin at the plasma membrane of most cells as visible on EM metal replica. Cells were analyzed 96 hours after the first transfection. Of note, this protocol did not produce different functional results as compared to a single round of siRNA transfection as confirmed using the collagen remodeling assay (fig. S8G).

Electron microscopy of cells in 3D

For electron microscopy of MDA-MB-231 cells cultured in a 3D network of collagen fibers, cells were fixed with 2.5% glutaraldehyde in 0.1 M cacodylate buffer (pH 7.4) and then washed with PBS and incubated for 1 hour with 1% osmium tetroxide. The samples were then dehydrated and processed for Epon resin embedding and ultrathin sections and then contrasted with uranyl acetate and lead citrate. Samples were observed under an electron microscope (Philips CM120; FEI Company, Eindoven, The Netherlands) and digital acquisitions were made with a numeric camera (Keen View; Soft Imaging System, Germany).

Spinning disk microscopy

Genome edited MDA-MB-231 cells were imaged for exposure times of 200 ms at 5 s intervals for the indicated time using a spinning disk microscope (Andor) based on a CSU-W1 Yokogawa head mounted on the lateral port of an inverted IX-83 Olympus microscope equipped with a 60x 1.35NA UPLSAPO objective lens and a laser combiner system, which included 491 and 561 nm 100 mW DPSS lasers (Andor). Images were acquired with a Zyla sCMOS camera (Andor). The system was steered by IQ3 software (Andor). For genome edited MDA-MB-231 cells cultured in 3D collagen network, a z-stack of 3-4 frames at 0.25 μm interval were acquired every 5 s. Subsequent quantifications were performed only on one selected optical section.

For calculating the nucleation index of CCSs on collagen fibers, collagen fibers were segmented using Metamorph software. 2-3 unambiguously individual fibers were selected per cell and the number of μ2-adaptin-GFP spots appearing along these fibers in a 10 min time window was measured. Nucleation index was calculated by dividing this number by time (in min) and by the length of the collagen fibers (in μm). As a control, the same analysis was performed after shifting the position of the fiber by approximately 10 pixels so that the virtual fiber remained in the cell and did not overlap with an actual fiber. 15 cells from 2 independent experiments were analyzed. Data are expressed as a mean percentage of control situation ± SEM.

For measuring the lifetime of CCSs in 2D or 3D conditions, time periods between first and last detection of individual μ2-adaptin-GFP spots on fibers or in other area of the cell were measured manually using Metamorph software in 2 different 60x60 pixels large regions per cell over a 5 min period. At least 8 cells from 2 independent experiments were analyzed for a total of at least 125 CCSs per conditions.

For measuring the movement of collagen fibers, MDA-MB-231 spreading on the top of a thin layer of collagen fibers were imaged every 2 min for 30 min by spinning disk microscopy. To calculate the movement index, fibers were segmented using Metamorph software and the cumulative area occupied by fibers below individual cells during the 30 min acquisition window was measured and divided by the area occupied by the same fibers in the first frame of the movie. At least 30 cells per conditions from 3 independent experiments were analyzed. Data are expressed as a mean percentage of control situation ± SEM.

TIRF microscopy

For total internal reflection fluorescent microscopy (TIRF), MDA-MB-231 cells seeded onto collagen-fibers-coated glass coverslips were fixed in ice-cold methanol and stained with indicated antibodies before being imaged through a 100x 1.49 NA TIRF objective lens on a Nikon TE2000 (Nikon France SAS, Champigny sur Marne, France) inverted microscope equipped with a QuantEM EMCCD camera (Roper Scientific SAS, Evry, France / Photometrics, AZ, USA), a dual output laser launch, which included 491 and 561 nm 50 mW DPSS lasers (Roper Scientific), and driven by Metamorph 7 software (MDS Analytical Technologies, Sunnyvale, CA, USA). A motorized device driven by Metamorph allowed the accurate positioning of the illumination light for evanescent wave excitation.

Laser ablation

Laser ablation was performed on a Leica Sp8 confocal microscope equipped with a Pecon incubation chamber to maintain the cells at 37°C and 5% CO2. The FRAP wizard of the Leica software was used to laser ablate cell protrusions. Briefly, wide-field images of MDA-MB-231 cells migrating in the 3D collagen network were acquired 1s before laser ablation. Laser ablation was then performed in two-photon mode with 2 consecutives 1s laser pulses at 820 nm and 100% of power focused on the specimen through a 63× 1.4 NA objective. Cells were imaged immediately after ablation every 1s for 3 min. Displacement field of collagen fibers was calculated using the PIV plugin running under ImageJ (32) based on fluorescence pictures of Cy-3-labeled collagen fibers acquired 1s before and 1s after laser ablation.

Adhesion assays

MDA-MB-231 cells treated with indicated siRNAs were incubated with 5 μM calcein (Sigma) for 30 min in DMEM to label cells. Cells were then washed in PBS, harvested using Versene (PBS, 0.02% EDTA) for 5 min at 37°C and resuspended in DMEM 10% FCS. 24 wells polystyrene plates (Falcon) or 5 kPa 24 wells soft plates (ExCellness) were incubated for 60 min at 37°C with 50 μg/ml collagen in 0.02M acetic acid. 400,000 cells were seeded per well and allowed to adhere on the collagen-coated polystyrene or soft plates for 15 min at 37°C. Cell-associated calcein fluorescence was then immediately measured using a FLUOstar Optima plate reader (BMG Labtech) before and after washing thoroughly wells 3 times with PBS. Ratio between fluorescence levels before and after washing was used to calculate the adhesion index. Experiments were performed 3-5 times in triplicate. Results are expressed as a mean percentage of control cells (siControl) ± SEM.

Statistical analyses

Statistical analyses in Figs. 3E; , D to G; and 5, A, B, and E; and figs. S2I; S7, C, F, and I; and S8, A and B have been performed using Kruskal-Wallis One Way Analysis of Variance (ANOVA) followed by an All Pairwise Multiple Comparison Procedure (Tukey Test). Data in Figs. 1, D and G; and 3, C and G; and figs. S1D and S7, A and H, have been tested using Student’s t test. Statistical analysis of Imaris-extracted data on the percentage of CCSs contacting collagen fibers in the 3D environment was performed using Student’s t test. All statistical analyses were performed using SigmaStat software.

Supplementary Materials

References and Notes

  1. Acknowledgments: The authors thank M. Piel, B. Ladoux, P. Chavrier, and J. Ivaska for critical comments on the manuscript. We thank the Gustave Roussy imaging facility for help with image acquisition and the Electron Microscopy Facility of the Institut de Biologie Paris–Seine (IBPS). Core funding for this work was provided by the Gustave Roussy Institute and INSERM, and additional support was provided by grants from ATIP/Avenir Program and la Ligue Nationale Contre le Cancer and from the Agence Nationale de la Recherche (ANR-15-CE15-0005-03) to G.M. and a young researcher grant (ANR-14-CE12-0001-01) to S.V. N.E. designed and performed experiments, analyzed results, and wrote the manuscript. E.B., F.B., and A.L.R. performed experiments. G.V.N. designed and performed electron microscopy analyses in the 3D environment. S.V. designed and performed metal-replica electron microscopy analyses and wrote the manuscript. G.M. supervised the study, designed experiments, and wrote the manuscript. Data described can be found in the main figures and supplementary materials. The authors declare no competing financial interests.
View Abstract

Navigate This Article