Reticular Adhesion Formation is Mediated by Flat Clathrin Lattices and Opposed by Fibrillar Adhesions

Reticular adhesions (RAs) consist of integrin αvβ5 and harbor flat clathrin lattices (FCLs), long-lasting structures with similar molecular composition to clathrin mediated endocytosis (CME) carriers. Why FCLs and RAs colocalize is not known. Here, we show that FCLs assemble RAs in a process controlled by fibronectin (FN) and its receptor, integrin αvβ5. We observed that cells on FN-rich matrices displayed fewer FCLs and RAs. CME machinery inhibition abolished RAs and live-cell imaging showed that RA establishment requires FCL co-assembly. The inhibitory activity of FN was mediated by the activation of integrin α5β1 at Tensin1-positive fibrillar adhesions. Conventionally, endocytosis disassembles cellular adhesions by internalization of their components. Our results present a novel paradigm in the relationship between these two processes by showing that endocytic proteins can actively function in the assembly of cell adhesions. Furthermore, we show this novel adhesion assembly mechanism is coupled to cell migration via a unique crosstalk between cell matrix adhesions.


Introduction
Integrins are nonenzymatic dimeric transmembrane receptors which recognize extracellular matrix (ECM) components. These mechanosensory proteins govern cell adhesion to the ECM maintaining correct tissue development and function, with elaborate connections to cellular homeostasis and disease (Kanchanawong and Calderwood, 2023). Ligand availability, and biochemical and physical properties of the ECM, determine integrin activation status, integrin clustering, and, ultimately, the formation of cellular adhesion structures (Kechagia et al., 2019).
Cells can form a variety of integrin-based adhesions. Small Integrin clusters engaged to the ECM, called nascent adhesions, form on the cell periphery and establish their connection to the actin cytoskeleton via adaptor proteins such as Talin. A balancing act of traction forces and signaling molecules determine whether nascent adhesions mature into the larger and molecularly more complex focal adhesions (FAs) (Wehrle-Haller, 2012). In migrating cells and in the presence of the ECM component fibronectin (FN), FAs can serve as platforms for the formation of fibrillar adhesions (FB), where FNbound α5β1 integrins "slide" from FAs to form FN fibrils (Georgiadou and Ivaska, 2017). In common to all these types of cell adhesions, their disassembly is mediated by the removal of integrin molecules from adhesion sites via endocytosis (Kechagia et al., 2019).
Recently, a novel type of integrin-based cell adhesion was discovered (Lock et al., 2018). Called reticular adhesions (RAs), these structures contain integrin αvβ5, lack the typical markers for the other adhesion types, such as Talin1 or Paxillin, and are not connected to actin stress fibers. RAs can occupy a significant portion of the substrate-facing surface of cells in culture and can significantly outlast FAs. Their physiological function is, however, not clear.
Intriguingly, RAs colocalize with large, persistent forms of clathrin structures at the cell membrane called Flat 1 Clathrin Lattices (FCLs) (also referred to as clathrin plaques) (Grove et al., 2014). The structure containing FCLs and RAs is called clathrin-containing adhesion complexes (CCAC) (Lock et al., 2019;Zuidema et al., 2020). FCLs were previously considered as stalled endocytic events of the Clathrin-Mediated Endocytosis (CME) pathway. However, recent studies have changed this view and support the idea that FCLs are signaling platforms (Leyton-Puig et al., 2017;Grove et al., 2014;Alfonzo-Méndez et al., 2022). In vivo, FCLs localize to adhesive structures between bone and osteoclasts (Akisaka et al., 2008) and are required for the organization of sarcomeres (Vassilopoulos et al., 2014).
The functional relationship between FCLs and RAs is not clear. A confounding factor in this relationship lies in the fact that, although FCLs always localize to RAs, the opposite is not true. RAs can occur as large structures with FCLs covering only a fraction of their area. Moreover, integrin αvβ5 can localize to both RAs and FAs. Although details on the factors mediating integrin αvβ5 localization to FCLs are becoming clearer (Zuidema 2018 and, why these structures co-exist, what their function is and how cells control their formation remain a mystery.
In this study, we show that FCLs are required for the establishment of RAs. Moreover, we found that a FN-rich ECM acts as an inhibitor of FCL-mediated RA formation. This inhibitory role of FN is mediated by the activation of integrin α5β1 localized at fibrillar adhesions. Furthermore, we show that the transition from a static to a migratory state is mirrored by the disappearance of FCLs and RAs.

Fibronectin inhibits the formation of FCLs
While studying CME dynamics, we serendipitously observed that cells on fibronectin (FN) appear to display fewer FCLs when compared to cells plated on noncoated glass dishes. To confirm if ECM proteins in general can influence CME, we assessed the effect of several major ECM components as well as non-ECM coatings and non-coated surfaces on the amount of FCLs. For that, dishes were coated for 16-24 h after which cells were let to attach for 16-20 h in serumcontaining medium before imaging. For quantifications, we established the metric FCL proportion, which defines the average fraction of FCLs per frame among all clathrin coated structures detected in a 5-minute movie (see methods for details). These experiments were performed using U2Os cells with an endogenously GFP-tagged alpha adaptin 2 sigma subunit (AP2S1, hereafter referred to simply as AP2). AP2 is a widely used CME marker which faithfully mirrors clathrin dynamics (Almeida-Souza et al., 2018;Ehrlich et al., 2004;Rappoport and Simon, 2008). We used endogenously tagged cell lines throughout this study as the expression level of the AP2 complex was shown to modulate the amount of FCLs (Dambournet et al., 2018).
U2Os cells on non-coated dishes presented typical and abundant FCLs (i.e. bright, long-lived AP2-GFP marked structures) ( Figure 1A, B, S1A and Supplementary video 1), similar to what has been found in many cell lines (Zuidema et al., 2022;Moulay et al., 2020;Sochacki et al., 2021;Saffarian et al., 2009). Similarly, U2Os cells plated on dishes coated with the non-ECM proteins bovine serum albumin (BSA) and poly-L-lysine (PLL) also presented high FCL proportions ( Figure 1B, S1A and Supplementary video 1). Out of the major ECM proteins tested, FN, collagen IV (Col IV) and laminin-111 (LN111) reduced FCL proportion significantly. The integrin αvβ5 ligand vitronectin (VTN) did not increase or decrease the FCL proportion when compared to non-coated dishes ( Figure 1B, S1A) (see discussion). Similarly, and in line with a recent study (Baschieri et al., 2018), collagen I (Col I) did not reduce FCLs ( Figure 1B, S1A and Supplementary video 1). Different concentrations of FN used for coating (10 or 20µg/ml) did not show significant differences ( Figure 1B).
Recently, it was described that SCC-9 cells produce more FN when plated on Col IV or LN111 (Lu et al., 2020). To probe if this is also the case for our cells, we stained FN from U2Os cells plated directly onto non-coated dishes, or plated on FN, VTN, Col IV, Col I or LN111. While U2Os cells produce little FN overnight, cells plated on Col IV produced a striking amount of FN, which assembled into elongated fibrils ( Figure 1C and 1D). LN111 coating also induced FN production, but less strikingly than Col IV. Col I and VTN coating were unable to stimulate FN production ( Figure 1C, D). These results suggest that FN is the main ECM component inhibiting FCL formation.
For many cell lines, it is common to find considerable variability in the amount of FCLs in culture. We thus decided to test if this variability is due to differential FN production within the culture. Confirming this hypothesis (and bearing in mind that U2Os secrete FN modestly, see below), we found that cells plated on non-coated dishes displaying fewer FCLs were predominantly lying on top of a FN-rich region of the culture ( Figure S1B).
Next, we asked if the reduction in FCL proportions observed in FN-coated samples are a cell-wide effect or specific to cellular regions in direct contact with the extracellular substrate. For that, we used patterned dishes containing FN-coated regions interspersed with uncoated regions, where single U2Os-AP2-GFP cells could adhere simultaneously to both a FN-and a noncoated region. In line with a contact-dependent effect, low FCL proportions were observed in cellular regions in contact with FN whereas FCL proportion was high in cellular regions contacting non-coated surface ( Figure  1E-G and Supplementary video 2).
effects we observe are also due to changes in clathrin splicing, but found no difference when comparing cells plated on non-coated or FN-coated dishes ( Figure S1C). Thus, these results show that FN is a potent inhibitor of FCLs. Moreover, FN inhibits FCLs in a contact-dependent manner locally within a single cell ( Figure 1H).

Fibronectin inhibits the formation of RAs in a similar manner as FCLs.
As discussed in the introduction, FCLs localize to RAs. To check how ECM composition affects these structures, U2Os AP2-GFP cells plated on FN, VTN, Col IV, Col I or LN111 and stained with the RA component integrin αvβ5 and ̶ to be able to distinguish integrin αvβ5 on RAs or FA s ̶ w e r e a l s o s t a i n e d w i t h a n FA m a r k e r (phosphorylated paxillin, p-PAX Y118). Cells plated overnight without coating formed abundant RAs ( Figure   2A, B). On FN-coated dishes, big RAs were largely absent but small "dot-like" nascent RAs were present in a few cells. Similarly, on Col IV and LN111 coatings (which stimulated FN production) ( Figure 1C), cells formed significantly fewer RAs than on non-coated dishes ( Figure  2A, B). Coating with VTN, the ligand for integrin αvβ5 present at RAs and FAs alike ( Figure S1C) (Lock et al), did not result in more RAs (Figure 2A, B). Different coatings also changed the total amount of integrin αvβ5 on the bottom surface of cells ( Figure S1D). However, they did not follow a clear relationship with the amount of RAs. To quantify differences in RA amounts in cells we developed a metric called RA coverage, which measures the fraction of the area of the cell covered by integrin αvβ5 signal (excluding FAs). RA coverage serves as a good metric to distinguish between large and nascent RAs and, crucially, it shows a clear correspondence between RA content and 4 both FCL proportion and FN abundance in the ECM (see figures Figure 1B, 1D and 2B).
Next, we used our substrate patterning strategy to check if the local FN effects on FCLs were also similar for RAs. Strikingly, cells plated on patterned FN revealed that RAs, akin to FCLs, were completely inhibited on cellular regions in contact with FN. Cellular regions in contact with non-coated surfaces displayed many FCLs colocalizing to RAs while regions in contact with FN presented no RAs or FCLs ( Figure 2C). Interestingly, in these patterned substrates, most of the integrin αvβ5 signal segregated to non-coated regions forming typical RAs ( Figure 2C,D). This contrasts with cells plated in fully coated dishes (Figure 2A), where integrin αvβ5 can be seen in both RAs and FAs. Hence, the inhibitory effects of FN on FCLs affects RAs in a similar manner ( Figure 2E).

The effect of fibronectin on FCLs and RAs is clear in various cell lines.
Next, we checked if the effects we see in U2Os cells are also true for other cell lines. To avoid problems of 5 overexpression, we endogenously tagged AP2 with either Halo tag or GFP in various human cell lines: HeLa (epithelial, cervical carcinoma), MCF7 (epithelial, breast cancer), HDF (dermal fibroblast, noncancerous), Caco2 (epithelial, colon carcinoma) and hMEC (Human mammary epithelial cells). These cell lines presented a large variation in the amount of FCLs and the morphology of RAs. Importantly, these cells could be divided into two groups in terms of endogenous FN secretion, and this division clearly correlated with the amount of FCLs and RAs ( Figure 3A, B). U2Os, HeLa and MCF7 composed the group of low FN-secretion cells. U2Os form large RAs on non-coated dishes, whereas HeLa formed multiple dot-like nascent RAs (which colocalized with FCLs) with bigger RAs found more seldom ( Figure 3A). MCF7 cells formed many FCLs and large RAs covering almost the entire cell area ( Figure 3A). None of the high FN-secreting cell lines (HDF, Caco2 and hMEC) formed large RAs ( Figure 3A, B). In these high FNproducing cells, small FCL/RA dots were often found in areas with less deposited FN ( Figure 3A).
We next evaluated the response of these cell lines to FN pre-coating. In low FN-producing cell lines (U2Os, HeLa, and MCF7), RA coverage dropped significantly ( Figure  3C-E). Among the high FN-production cells lines, only Caco2 reduced its RA coverage on FN-coated dishes ( Figure 3C,D). As expected, HDF and hMEC, which had low RA coverage without coating, did not show a significant response to FN coating ( Figure 3C, E).
For all experiments so far, we used media supplemented with serum, which is known to contain ECM components, including FN. Given that our cells are left to attach overnight in this media, it would be reasonable to expect that the FN present in serum would coat the dishes and completely mask our results. To test why this does not seem to happen ( Figures 1C and 3A), we compared the amount of FN deposited on the glass surface in different conditions: dishes were coated for 24 h with 10, 5 and 1µg/ml of FN (diluted in PBS), 100% fetal bovine serum (FBS), media with 10% FBS and PBS as a control. After coating, U2Os cells were plated and left to attach for 16 h before being fixed and stained for FN. Surprisingly, our results revealed that very little FN was deposited on glass in dishes "coated" with full media or pure FBS ( Figure  S1F-G). These results are in line with similar experiments performed 30 years ago (Steele et al., 1992). We hypothesize that this phenomenon occurs due to the high concentrations of BSA in serum (40 mg/ml) which rapidly saturates the surface of culture dishes, thereby acting as a blocking agent for the binding of serum FN. in different cells lines ( Figure 3F), suggesting a common ̶ and general ̶ mechanism for the establishment of these structures.

The CME machinery is essential for RA formation
Next, we set out to dissect the relationship between the formation of FCLs and RAs. It has been shown that integrin αvβ5 is required for the establishment of FCLs (Baschieri et al., 2018;Zuidema et al., 2018). We confirmed this observation by silencing integrin β5 from U2Os AP2-GFP cells plated on non-coated dishes and, indeed, they displayed a significantly lower FCL proportion compared to control cells ( Figures S2A, B). Further, while integrin β5-silenced cells were unable to form RAs, they did form FAs ( Figure S2C, D). The dependency of integrin β5 on FCL formation was further confirmed using Cilengitide, the inhibitor for integrin αvβ5 (Desgrosellier and Cheresh, 2010), as the treatment led to a rapid disassembly of FCLs and RAs (Figures S2E-F).
While all FCLs colocalize to RAs, FCL-free areas of larger RAs are rather common (e.g. Figures 2A, 3A, 3C and S2D), which may give the impression that FCLs are formed on pre-existing RAs. Nevertheless, the fact that both structures are inhibited independently by FN suggests a deeper relationship and led us to ask if RAs can exist without the CME machinery. To answer this question, we quantified the RA coverage in U2Os-AP2-GFP cells silenced for the clathrin adaptor AP2 complex subunits alpha 1 (AP2A1) or sigma 1 (AP2S1) in cells plated on non-coated dishes, a condition where we observe large RAs. Consistent with an important role played by the CME machinery in RA formation, AP2A1or AP2S1-silenced cells (easily recognizable as cells with little to no AP2-GFP signal), did not display RAs. Instead, integrin αvβ5 localized to FAs ( Figure 4A, B).
To confirm these results, we expressed the AP180 Cterminal fragment (AP180ct), which acts as a strong dominant negative of CME (Ford et al., 2001). AP180ctpositive U2Os-AP2-GFP cells plated on non-coated dishes displayed low AP2 signal at the membrane and, akin to AP2-silenced cells, RAs were largely absent with integrin αvβ5 localized to FAs, whereas AP180ct-negative cells displayed typical FCLs and RAs (Figure 4 C, D). Thus, the CME machinery is required for the formation of RAs ( Figure 4E).
Next, we set out to visualize the dynamics of AP2 during RA formation. For that, we generated a double U2Os knock-in cell line ̶ AP2-GFP and Integrin β5 (ITGB5)-mScarlet. RAs are remarkably stable structures (Lock et al., 2018) and their de novo formation is rare, making it rather difficult to capture such events. To minimize this challenge, we optimized the conditions for Cilengitide treatment to disassemble RAs followed by a washout, when RAs could start reforming ( Figures S3A, B). Using these washout conditions, we were able to capture events showing that the formation and growth of ITGB5postive structures are accompanied by the formation of of the dot-like structures we see in many cells), we noticed that the establishment of an FCL was typically accompanied by an increase in ITGB5 fluorescence (Figures 5B,S3C and S3D and Supplementary video 3). Importantly, ITGB5-postive structures, which did not colocalise with an FCL, rapidly disappeared. In many cases, this disappearance was preceded by bona fide CME events (short lived AP2-GFP signals), likely representing CME-mediated ITGB5-cluster disassembly ( Figures 5B and S3D). Data are the mean ± SD, ns. non-significant p-value; * p-value<0.05, *** p-value < 0.001. Scalebars 10 µm, insets 5 µm.
Taken together, our results show that the relationship between FCLs and RAs is beyond a simple colocalization. In fact, our data reveals a strict co-dependency, where FCLs are required for the stabilization and growth of integrin αvβ5 clusters thereby establishing RAs ( Figure  5C).

The inhibitory effect of fibronectin on FCL and RA formation is mediated by integrin α5β1.
To understand the mechanism controlling the coassembly of FCLs and RAs, we turned our attention back to FN. While integrin αvβ5 binds to VTN at FAs and RAs, the major FN receptor is integrin α5β1 (Humphries et al., 2006). First, we acutely interfered with integrin β1 binding to FN using the function-blocking antibody mab13. U2Os-AP2-GFP cells seeded on FN-coated dishes were treated with mab13 and monitored for the acute formation of FCLs and RAs. Over the time course of 45 min, mab13 induced the relocalization of integrin αvβ5 from FAs to small, newly formed RAs ( Figure 6A, B). Further supporting the role of FCLs in RA assembly, these newly formed RAs completely colocalized with FCLs (bright AP2 signals) ( Figure 6A, C). A similar experiment followed by live-cell imaging confirmed these results and showed a gradual increase in FCL proportions after mab13 treatment ( Figure 6D In line with these results, integrin β1 silencing in U2Os-AP2-GFP cells plated on FN resulted in a high FCL proportion and large, prominent RAs ( Figure 7A-D and S4A-C). Despite the striking increase of integrin αvβ5 on the bottom surface of silenced cells, this increase was not reflected in expression levels, indicating that the stimulation of RA formation leads to a change in the trafficking of this integrin dimer ( Figure S4C). A significant increase in RAs was also seen in cells silenced for integrin a5, the alpha subunit which pairs with integrin β1 for FN binding ( Figure 7C, E and S4D, E). Taken together, these results show that the inhibitory activity of FN on RAs and FCLs occurs via the activation of integrin α5β1 ( Figure 7F).

Activation of integrin α5β1 at fibrillar adhesions controls RA and FCL formation
When bound to FN, Integrin α5β1 can slide centripetally on the cell membrane, translocating from FAs to form elongated structures called fibrillar adhesions (FB). This movement generates long FN fibrils in a process called FN fibrillogenesis and is mediated by the cytoskeleton scaffolding protein Tensin1 (Pankov et al., 2000). To determine which type of adhesion structure active integrin α5β1 localizes to under our experimental conditions, we plated U2Os-AP2-GFP cells on FN and non-coated dishes and stained them with an active Data are the mean ± SD,*** p-value < 0.001. Scalebars 10 µm, insets 5 µm.
integrin β1-specific antibody (12G10) and Tensin1 or p-Pax to mark FBs or FAs, respectively. The staining revealed that in the FN-coated dishes, active integrin β1 was colocalizing to FBs (Tensin1) ( Figure 8A). As expected, active integrin β1 and Tensin1-positive adhesions were largely absent in non-coated dishes ( Figure 8A).
Next, to determine which active integrin β1 pool is more important for the inhibition of FCLs and RAs, we silenced FAs and FB components on U2Os-AP2-GFP cells and plated them on FN. In accordance with the higher accumulation of active integrin β1 in FBs, silencing of Tensin1 led to a marked increase of RAs and FCLs accompanied by a reduction in the presence of active integrin β1 on the membrane (evidence by 12G10 antibody staining) ( Figure 8B, C and S5A-C). Silencing of the FA component Talin-1 also led to increased RAs and FCLs and a reduction of active integrin β1 on the membrane ( Figure S5D-F). Given the strong phenotype on Tensin1 knockdown, this result was expected as FAs are precursors of FBs.
FB formation indicates activated and migratory cell phenotypes. Indeed, active sliding of integrin α5β1 and Tensin1 bound to FN along central actin stress fibers increases traction forces (Georgiadou and Ivaska, 2017;Pankov et al., 2000) and is required in cell migration during development and cancer metastasis (Efthymiou et al., 2020;Schwarzbauer and DeSimone, 2011). If the extension of active integrin α5β1 into FBs is indeed required for the inhibition of FCLs and RAs, we hypothesized that physical confinement of cells -which inhibits cell migration -would also inhibit the sliding of FB from FAs. In turn, the absence of integrin α5β1 in FBs would favor FCLs and RAs, even if cells were plated on an FN-rich matrix. To test this possibility, we turned to single cell micropatterns. In contrast to the patterned coatings we used in Figures 1 and 2, these micropatterns do not allow cells to attach outside the defined areas on a coverslip. Given the small size of these areas (1100 µm 2 ), cells are laterally confined. U2Os-AP2-GFP cells were plated on slides with arrow-and H-shaped micropatterns either precoated with FN or not and stained for integrin αvβ5 and p-PAX and imaged to measure RA coverage. In addition, to measure integrin β1 activation, U2Os-AP2-GFP-ITGB5-mScarlet cells were plated similarly and stained for active integrin β1. Supporting our hypothesis, we could detect clear FCL and RAs in FN-coated micropatterns ( Figure 9A, B, S5K). On arrows, FCLs and RAs developed on the shaft of the pattern, rather than the arched area. In the H-patterns, FCLs and RA developed all over the pattern. Cells on non-coated patterns made large FCLs and RAs often extending throughout the pattern ( Figure 9A, B). Crucially, the RA coverage was not significantly different between coated or non-coated patterns ( Figure 9B). As expected, staining with active integrin β1 (12G10) showed a clear difference in signal between FN-coated and non-coated patterns ( Figure 9C, S5L). Importantly, further supporting a need for FB formation to inhibit FCLs and RAs, 12G10 signal was not organized as elongated, central FBs but rather confined at the cell periphery ( Figure S5K). Thus, the inhibitory role of FN on FCLs and RA formation occurs primarily via the activation of integrin α5β1 on Tensin1-positive fibrillar adhesions.

The disassembly of FCL/RA is coupled to cell migration
As physical restriction favored FCLs and RAs, we wondered if inducing migration will have the opposite effect. To test this hypothesis, we monitored FCLs and RAs in a classic wound healing assay. U2Os-AP2-GFP-ITGB5-mScarlet cells were plated on non-coated dishes and allowed to grow to full confluency for 2 days. Cultures were then wounded and cells were allowed to migrate. At 0 minutes (i.e. just after wounding), FCLs and RAs were abundant and equally distributed at the edge and away from the wound ( Figure 9D). Within 80 minutes, the cells at the migration front had lost most of their FCLs and RAs, while cells further away from the edge maintained their FCLs and RAs ( Figure 9D). At 4 hours, as the migratory front grew larger, the loss of FCL and RAs also extended away from the wound ( Figure 9D). In full accordance with the results we presented above, the disappearance of FCLs and RAs was preceded by the increase in FN secretion by the cells at the edge of the wound ( Figure 9E, F). Together, these results place the resolution of FCLs and RAs as an intrinsic part of the cascade of events triggering cell migration ( Figure 9G).

Discussion
The extracellular environment is a key regulator of cellular physiology with integrins playing a key role translating the chemical composition of the extracellular milieu into intracellular signals. Among various mechanisms controlling integrin function, integrin trafficking via endocytosis and exocytosis plays a major role (Moreno-Layseca et al., 2019). Thus far, the relationship between integrin-based matrix adhesions and endocytosis has been considered primarily antagonistic, with endocytosis playing a role in the disassembly of said adhesive structures (Ezratty et al., Here we provide evidence, for the first time, of a constructive relationship between the endocytic machinery and cellular adhesions, where the CME machinery, in the form of FCLs, is key for the formation of integrin αvβ5 RAs. Moreover, we show that FCL-mediated αvβ5 RA formation is counteracted by the activation of a distinct integrin heterodimer, α5β1, in distinct adhesion structures, FBs, revealing an interesting mechanism of inter-adhesion crosstalk. Our results support the idea that FCLs and RAs are two sides of the same structure (Lock et al., 2019;Zuidema et al., 2020). Previous studies have demonstrated the importance of integrin αvβ5 at RAs in the formation of FCLs (Zuidema et al., 2022(Zuidema et al., , 2018Lock et al., 2019). Here, we show that this relationship is also crucial in the other direction, with FCLs being required for the formation of integrin αvβ5 RAs. Therefore, we believe the previously suggested term clathrin containing adhesion complexes (or CCAC for short) is a more appropriate terminology to refer to these structures.

The mechanism of FCL-mediated RA formation
We observed that FCL-mediated RA formation events are rare, which led us to use non-physiological conditions ̶ a Cilengitide washout experiment ̶ to detect them. Therefore, the physiological trigger leading to the formation of FCLs and establishment of RAs remains to be understood. VTN, the ligand for integrin αvβ5 could be considered a good candidate. However, as we show in figure 2E and as reported by others (Zuidema et al., 2022), integrin αvβ5 binds VTN equally on FAs and RAs. While it is clear that the presence of VTN is important as an extracellular tether for the formation of integrin αvβ5 adhesions (Zuidema et al., 2022(Zuidema et al., , 2018Lock et al., 2018), the switch between these adhesion types is likely an inside-out mechanism. We did not detect an increase in integrin αvβ5 RAs on VTN-coated dishes ( Figure 1B). This could seem counterintuitive, but VTN ̶ which was initially called "serum spreading factor" (Hayman et al., 1983) ̶ is readily secreted by cells during attachment. Therefore, we advise caution when making conclusions based on the results of non-coated and VTN-coated dishes on the role of this ECM component on RA coverage. Further work is necessary to shed light on this issue.
Recent evidence showed that EGFR activation led to the enlargement of FCLs in an integrin β5 phosphorylationdependent manner (Alfonzo-Méndez et al., 2022), pointing to a possible mechanism for the initial coassembly of FCLs and RAs. This possibility is further reinforced by the fact that the relationship between growth factor receptors and integrins has been established in multiple contexts (Ivaska and Heino, 2011).
Another key unknown aspect of FCL-mediated RA formation concerns how these structures can be molecularly differentiated from canonical endocytic events. The connection between integrin αvβ5 located in RAs and FCLs occurs primarily via the endocytic adaptors ARH and NUMB (Zuidema et al., 2018). Importantly, these adaptors also participate in integrin endocytosis (Ezratty et al., 2009;Nishimura and Kaibuchi, 2007), suggesting that other mechanisms may be required to define the identity of FCLs.
Recently, a correlation was found between the presence of clathrin plaques and an alternatively spliced isoform of clathrin containing exon 31 in myotubes (Moulay et al., 2020). We did not detect any changes in clathrin splicing in our experimental system, which was not surprising given the effects we see are contact-dependent and could not be explained by transcriptional changes. In addition, we cannot ensure that the clathrin plaques detected in myotubes are equivalent to the FCLs we observe here. Nonetheless, it is possible that the abundance of the exon 31-positive clathrin isoform works as a dial that changes the probability, speed or efficiency by which cells form FCLs.

Another unusual function for the clathrin machinery
In addition to its endocytic function, the clathrin machinery has been shown to participate in other processes. For example, clathrin helps to stabilize the mitotic spindle by binding to microtubules (Royle, 2012) and, during E. coli infection, the CME machinery is coopted to form a clathrin-based actin-rich adhesive structure for the bacteria called a pedestal (Veiga et al., 2007). Furthermore, a clathrin/AP2 tubular lattice was recently described to envelop collagen fibers during cell migration (Elkhatib et al., 2017). The results we present here add to this list of non-endocytic functions of CME components with an important twist. FCLs can also be disassembled into individual endocytic events (Lampe et al., 2016;Tagiltsev et al., 2021;Maupin and Pollard, 1983), providing an elegant and efficient mechanism for cells to switch the same machinery from an adhesion assembly to an adhesion disassembly function.

Inhibition of clathrin-containing complexes (CCAC) and its relationship to cell migration
In addition to defining FCLs as key factors in the establishment of CCACs, our work has also revealed many interesting aspects on the inhibition and disassembly of these structures. We show that activation of integrin α5β1 by FN and the capacity of this integrin heterodimer to slide on the plasma membrane to form fibrillar adhesions are both essential conditions for the inhibition and disassembly of CCAC. In a classical wound healing assay, we observed that as cells start to migrate they secrete FN, leading to the disappearance of CCACs. However, using laterally confined cells ̶ which cannot form fibrillar adhesions ̶ we observed that the mere presence of FN is not enough to inhibit CCACs. A recent study showed that high levels of activated myosin light chain (p-MLC) correlated with integrin αvβ5 localizing to FAs over RAs (Zuidema et al., 2022). Moreover, overexpression of a constitutively active RhoA mutant in a cell line with low p-MLC levels promoted integrin αvβ5 localization to FAs (Zuidema et al. 2022). As Integrin α5β1-mediated FN fibrillogenesis is required for optimal activation of the RhoA-MLC pathway, which in turn increase actin stress fiber-based migration along fibrillar adhesions (Gagné et al., 2020;Huveneers et al., 2008;Danen et al., 2002), these findings perfectly complement our data. Together, these results suggest that the disappearance of CCACs is the result of a computation of multiple cellular signals occurring during the cell migration process. Whether the disassembly of CCACs occurs actively or is a mere consequence of a nonpermissive environment for the de novo formation of new adhesions is still unknown.
Given the fact that RAs are long-lasting cellular adhesion structures, it is tempting to hypothesize that these structures act as a "parking brake" for a cell. As the cell is triggered to migrate, this "brake" needs to be released for efficient cell movement. This process would be analogous to the loss of cell-cell contacts which happens during epithelial to mesenchymal transition (Kalluri and Weinberg, 2009), but instead of happening between cells, it would happen between the cell and the ECM. Therefore, we propose that disassembly of RAs is an intrinsic process during cells migration.
We showed that FN regulates CCAC assembly in all the cell lines we tested. However, how these in vitro findings will operate in vivo is still unknown. Even though the ECM composition in tissues is complex, the FN effect on CCAC formation is local and strictly contact dependent, which opens the possibility that, in vivo, tissues may use focal changes in ECM composition to control these structures.

FN patterning
To study local vs. global effects of FN, FN was mixed with 50 ng/ml of Alexa647-labelled BSA and used to precoat the imaging dishes overnight in +37°C. The coated surface was subsequently scratched with a needle to allow partial reappearance of non-coated surface.
After scratching, the dishes were heavily rinsed with PBS. 20 000 U2Os-AP2-GFP cells were seeded on patterned imaging dishes to ensure sufficient single-cell attachment to border areas.

Overexpression of mammalian proteins
The clathrin inhibitor AP180 c-terminal fragment (AP180ct; amino acids 516-898) cDNA, from rat origin, was described previously (Ford et al., 2001). This construct was cloned into Gateway compatible pCI vectors, containing an N-terminal monomeric EGFP using the Gateway system.
The donor template sequence was: where C terminal tagging with GFP is in green (codon-optimized) + short linker in purple. 150 bp homology arms (orange) were incorporated via PCR amplification from a synthesized (IDT), codon-optimized monomeric EGFP. dsPCR product was purified and 150 ng was used directly for transfection together with gRNAs. 70-80% confluent 24-well plates of U2Os cells were transfected with 2 µg PEI (1 µg/ml), 150 ng of plasmid and 150 ng of the PCR product. In addition, hMEC-AP2-GFP were treated with 1 μM DNA-PKc inhibitor NU7441 for 48 h post transfection. Two days after transfection cells were treated with puromycin (1 µg /ml) to enrich for successfully transfected cells. After expansion, GFP-positive cells were sorted by FACS, and single clones were expanded and genotyped.

Generating the U2Os-AP2-halo and HeLa-AP2-halo cell lines
U2Os-AP2-halo and HeLa-AP2-halo cell lines were generated with the same protocol as the U2Os-AP2-GFP cell line.
The donor template sequence was: where C terminal tagging with halo is in green (codon-optimized) and short linker in purple. 150 bp homology arms (orange) were incorporated via PCR amplification from a synthesized (IDT), codon-optimized monomeric halo tag.

Generating the HeLa-AP2-GFP, MCF7-AP2-halo, HDF-AP2-halo and CAco2-AP2-halo cell lines
The most effective gRNA (TGCTACAGTCCCTGGAGTGA) was ordered as sgRNA from Synthego and Cas9 protein (purified in the lab) was used instead of plasmid. The donor templates for these cell lines to insert either EGFP or Halo tag were same as above, respectively.
Cell lines were then treated with 1 µM DNA-PKc inhibitor NU7441 for 48 hours post nucleofection.

Generating the U2Os-AP2-GFP-ITGB5-mScarlet cell line
This cell line was produced by the same protocol as the U2Os-AP2-GFP cells with the following changes: The gRNA sequence was CAAATCCTACAATGGCACTG, and the donor template was: GGTTTGAGTGTGTGAGCTAACATGTGTCCTCATCCTCTTCCCCGCCGTGTTCTGTAGGCTTCAAATCCATTATACAGAAAGCCTATCTCCACGCACACTGTGGACTTCA  CCTTCAACAAGTTCAACAAATCATATAACGGCACTGTTGACGGAAGTGCATCTGGGAGCTCAGGCGCTAGTGGTTCAGCGAGCGGGGTGAGCAAGGGCGAGGC  AGTGATCAAGGAGTTCATGCGGTTCAAGGTGCACATGGAGGGCTCCATGAACGGCCACGAGTTCGAGATCGAGGGCGAGGGCGAGGGCCGCCCCTACGAGG  GCACCCAGACCGCCAAGCTGAAGGTGACCAAGGGTGGCCCCCTGCCCTTCTCCTGGGACATCCTGTCCCCTCAGTTCATGTACGGCTCCAGGGCCTTCATCAAG  CACCCCGCCGACATCCCCGACTACTATAAGCAGTCCTTCCCCGAGGGCTTCAAGTGGGAGCGCGTGATGAACTTCGAGGACGGCGGCGCCGTGACCGTGACC  CAGGACACCTCCCTGGAGGACGGCACCCTGATCTACAAGGTGAAGCTCCGCGGCACCAACTTCCCTCCTGACGGCCCCGTAATGCAGAAGAAGACAATGGGC  TGGGAAGCATCCACCGAGCGGTTGTACCCCGAGGACGGCGTGCTGAAGGGCGACATTAAGATGGCCCTGCGCCTGAAGGACGGCGGCCGCTACCTGGCGG  ACTTCAAGACCACCTACAAGGCCAAGAAGCCCGTGCAGATGCCCGGCGCCTACAACGTCGACCGCAAGTTGGACATCACCTCCCACAACGAGGACTACACCG  TGGTGGAACAGTACGAACGCTCCGAGGGCCGCCACTCCACCGGCGGCATGGACGAGCTGTACAAGTAATGTTTCCTTCTCCGAGGGGCTGGAGCGGGGATCT  GATGAAAAGGTCAGACTGAAACGCCTTGCACGGCTGCTCGGCTTGATCACAGCTCCCTAGGTAGGCACCACAGAGAAGACCTTCTAGTGAGCCTGGGCCAGGA  GCCCACAGTGCCT where A=silent mutations in 5' HA, linker region is in purple, and mScarlet is in orange.

Lentiviral shRNA production and transduction
Lentiviruses for shRNA production were produced using packaging plasmids pCMVR and pMD2.g and specific shRNAs in pLKO.1 vector as follows: 80% confluent HEK293T cells in DMEM supplemented with 10% FBS and 100U penicillin-streptomycin were transfected using PEI MAX transfection reagent. 5 hours later, the medium was changed to DMEM supplemented with 4% FBS and 25mM HEPES.
U2Os-AP2-GFP cells were transduced with lentiviral media expressing respective shRNAs in the presence of Polybren 8 µg/ml (Sigma-Aldrich, TR-1003) for 5 hours, and replaced with culture medium. 48 h later, puromycin (1 µg/ml) was added for 24 h to allow selection Microscopy All live videos and images from fixed samples were acquired with the ONI Nanoimager microscope equipped with 405, 488, 561 and 647 lasers, an Olympus 1.49NA 100x super achromatic objective and a Hamamatsu sCMOS Orca flash 4 V3 camera.
The ONI nanoimager microscope set to TIRF angle was used to acquire AP2 lifetimes at the cell membrane from 300 frames (1 frame/s) with the exposure time of 330 ms. Each video represents endocytic events from 2-3 cells (total field of view).

Acute manipulation of integrin activity
Integrin β1 blocking Acute modulation of ligand binding activity for integrin β1 was achieved using the function blocking antibody mab13 (0.3 µg/ml).
U2Os-AP2-GFP cells were plated on FN as explained above, and 16-20 h later subjected to live TIRF imaging. 0 min sample has no mab13 added, to control base-line FCL proportions. Immediately after mab13 addition, 5 min time lapses were continuously collected until 35 min, control videos (time point 0) had no mab13 added.

Integrin β5 blocking
To acutely induce the inhibition of integrin αvβ5 we used the small molecular inhibitor Cilengitide (MedChem Express HY-16141, 10 µM). U2Os-AP2-GFP cells plated on non-coated imaging dishes were treated with Cilengitide for 15 or 45 min, fixed, stained, and imaged with the ONI nanoimager microscope at TIRF angle, and analyzed for the resulting reduction of RA coverage.

Cilengitide washout
U2Os-AP2-GFP-ITGB5-mScarlet cells were plated on non-coated imaging dishes and 1 d later confluent monolayers were treated with 1 µM Cilengitide for 15-25 min, during which most FCLs and RAs were dissociated from the cell membrane. Samples were then washed twice and immediately subjected to live TIRF imaging to detect the de novo formation of FCLs and RAs. 1 h time-lapses were acquired with the ONI nanoimager microscope at TIRF angle, at 30 s intervals, with an exposure time of 100 ms for AP2 and 300 ms for integrin β5.

Immunofluorescent staining and imaging
For immunofluorescence experiments, cells were fixed with 4% paraformaldehyde-PBS for 15 min in a +37°C incubator, washed with PBS and blocked with 1% BSA-PBS. Primary antibodies diluted in 1% BSA-PBS were incubated for 1 h, samples were washed with PBS, and secondary antibodies diluted in 1% BSA-PBS were let to bind for 30 min. Samples were imaged with the ONI nanoimager microscope using TIRF angle and exposure times of 500 ms or 1000 ms.

CME lifetime analyses (FCL proportion)
To track CME events and measure lifetimes we used 'u-track 2.0' multiple-particle tracking MATLAB software at default settings (Jaqaman et al., 2008). For all experiments, "n" refers to a movie, which contained 2-4 cells. To determine the proportion of FCLs, we used the output from u-track to count the number of pits (events lasting longer than 20 s and shorter than 120 s) and the number of FCLs (events lasting longer than 120 seconds, as described in (Saffarian et al., 2009) in all frames. We took a conservative approach to identify CCPs, where events that were present at the start or lasted beyond the end of the movies were not counted as CCPs. This approach artificially led to higher FCL proportions in the first and final 120 movie frames. Therefore, FCL proportions are presented as the average FCL proportion from frames 120 to 175 for each movie.

Other analyses
With the exception of FCL proportions, all image analyses were performed using ImageJ. Simple fluorescence measurements were done manually. Others were performed using custom scripts as shown below:

RA coverage
Individual cells were marked and ITGB5 (or αvβ5) and p-Pax channels were segmented using the Robust Automatic Threshold Selection function. RAs were defined as ITGB5 (or αvβ5) signals not colocalizing with p-Pax. The area of RAs was then divided by the area of each marked cell to obtain RA coverage. Data is presented as percentage of the cell area covering RAs.
For the wound healing experiment, a line on the migration front of each image was manually drawn. This line was then used as a reference to automatically draw a box, 100 pixels in width (11.7 µm). RA coverage (as above) was calculated for this box, which was then moved inward in the culture in 50 pixels steps, where the RA coverage analysis was repeated. Values are normalized to the average RA coverage on the three innermost areas in the culture.

AP2-ITGB5 dynamics
Events showing the appearance of both AP2 and ITGB5 were identified by visual inspection of videos. For the generation of graphs, we selected only events where we could unambiguously ensure that significant FCLs and ITGB5 signals were not present in the region for at least 3 minutes. Time zero was defined as the frame where AP2 signal appeared and fluorescence Intensity from a 10 µm x 10 µm region around each event was measured for 3 min before (six frames) and 5 min after (10 frames). Fluorescence was normalised to the highest value in these frames.

AP2 intensity per colocalisation status
AP2, ITGB5 and p-Pax channels were segmented using the Robust Automatic Threshold Selection function. Each segmented AP2 spot had the fluorescence intensity measured from the original image and classified for its colocalisation with either marker (ITGB5 or p-Pax).
We used full images for these analyses.

RAs colocalising to AP2
Individual cells were marked and AP2, ITGB5 and p-Pax channels were segmented using the Robust Automatic Threshold Selection function. RAs were defined as ITGB5 signals not colocalizing with p-Pax. In the conditions used for these experiments, (FN + mab13) RAs 23 were primarily individual spots. The colocalisation of RAs to AP2 was classified by measuring the intensity of each RA region at the segmented AP2 channel.

FN intensity vs. AP2 intensity
AP2, ITGB5 and P-PAX channels were segmented using the Robust Automatic Threshold Selection function. Each segmented AP2 spot had the fluorescence intensity measured from the original image. A 3 µm x 3 µm region was drawn around each AP2 spot and used to measure the intensity of FN from the original image. Data is presented as the fluorescence for each AP2 spot.

Statistics
Figure legends state the exact n-values and individual repeats used in analyses. For multiple comparisons one-way ANOVA was performed followed by Tukey's multiple comparison. Pairwise comparisons were performed using two-tailed Student's t-test with equal variance. All graphs and statistical calculations were performed with GraphPad Prism 9.