ABSTRACT
Integrin-mediated adhesions are convergence points for multiple signaling pathways. Their inner structure and diverse functions can be studied with super-resolution microscopy. Here, we examined the spatial organization within focal adhesions by analyzing several adhesion proteins with structured illumination microscopy (SIM). Paxillin (Pax) serves as a scaffold protein and signaling hub in focal adhesions, and focal adhesion kinase (FAK, also known as PTK2) regulates the dynamics of adhesions. We found that their phosphorylated forms, pPax and pFAK, form spot-like, spatially defined clusters within adhesions in several cell lines and confirmed these findings with additional super-resolution techniques. These clusters showed a more regular separation from each other compared with more randomly distributed signals for FAK or paxillin. Mutational analysis indicated that the active (open) FAK conformation is a prerequisite for the pattern formation of pFAK. Live-cell super-resolution imaging revealed that organization in clusters is preserved over time for FAK constructs; however, distance between clusters is dynamic for FAK, while paxillin is more stable. Combined, these data introduce spatial clusters of pPax and pFAK as substructures in adhesions and highlight the relevance of paxillin–FAK binding for establishing a regular substructure in focal adhesions.
INTRODUCTION
Integrin-mediated adhesions between cells and the surrounding extracellular matrix are not only important for physical anchorage, but are also points of convergence for different intracellular and extracellular signals (Bachmann et al., 2019; Conway and Jacquemet, 2019; Green and Brown, 2019). These adhesions consist of a plethora of structural and signaling proteins establishing the ‘adhesome’ (Byron et al., 2011; Kuo et al., 2011; Schiller et al., 2011). Based on adhesome studies, a meta-analysis curated a consensus adhesome in which the members form functional clusters connected with each other into a complex network (Horton et al., 2015). The advent of super-resolution microscopy confirmed these functional clusters by showing that members of a given cluster are organized within a specific functional layer in the axial direction (Kanchanawong et al., 2010). In lateral directions, pointillistic super-resolution methods and stimulated emission depletion (STED) microscopy revealed the existence of a substructure of small individual clusters within adhesions (Bachmann et al., 2016; Changede et al., 2015; Diez-Ahedo et al., 2009; Shroff et al., 2008, 2007; Xu et al., 2018). Some studies successfully attributed functional aspects to these lateral clusters (Spiess et al., 2018), but the structure-function relationship along lateral directions in focal adhesions remains less studied.
One of the most important proteins for the structure and function of integrin-mediated adhesions is paxillin (Pax or PXN). Paxillin is part of the LIM-domain protein family and is recruited to adhesions by integrins and their activators, talin and kindlin (Pinon et al., 2014; Theodosiou et al., 2016), potentially facilitated by membrane interaction of the LIM4 domain of paxillin (Ripamonti et al., 2021). Integrin activation and integrin-mediated cell adhesion can occur in the absence of paxillin but cell spreading and proliferation relies on paxillin recruitment to integrin adhesions (Pinon et al., 2014; Soto-Ribeiro et al., 2019). Structurally, paxillin serves as a scaffold protein and its multiple binding partners are regulated by paxillin phosphorylation (Deakin and Turner, 2008). Tyrosine phosphorylation at Y31 (pPax-Y31) and Y118 (pPax-Y118) by a complex of focal adhesion kinase (FAK, also known as PTK2) and Src has been intensively studied and recruits activators and inhibitors for Rho-GTPases (Deakin and Turner, 2008; Petit et al., 2000; Schaller and Parsons, 1995; Thomas et al., 1999). These phospho-tyrosine sites are also important for adhesion dynamics (Ripamonti et al., 2021; Zaidel-Bar et al., 2007) and vinculin recruitment (Pasapera et al., 2010). The adhesome studies mentioned above cluster FAK and paxillin within the same functional group (Horton et al., 2015). Moreover, paxillin and FAK localize in the same z-layer (Kanchanawong et al., 2010) and form a complex, preferentially when paxillin is phosphorylated at Y31 and Y118 (Choi et al., 2011; Digman et al., 2008).
FAK contains an N-terminal FERM domain, followed by a kinase, an unstructured region and a focal adhesion targeting (FAT) domain. In the inactive state, the FERM and kinase domains bind to each other, and this inhibition is released by trans-autophosphorylation of FAK-Y397 (Acebrón et al., 2020). FAT binding to paxillin is necessary and sufficient for focal adhesion localization of FAK (Deramaudt et al., 2014). Simultaneous binding of FAK to membrane lipids and to paxillin is proposed to put FAK under mechanical strain, and to increase activation (Seong et al., 2013; Torsoni et al., 2003; Wong et al., 2011) and kinase activity of FAK (Bauer et al., 2019). Phosphorylated FAK-Y397 not only activates FAK but also recruits Src, highlighting the scaffolding function of FAK that might be more relevant than its kinase function (Arold, 2011). Accordingly, blocking the kinase activity of Src, but not of FAK, largely inhibits paxillin phosphorylation at Y118 (Horton et al., 2016), highlighting again the relevance of Src for the actual phosphorylation of focal adhesion proteins.
Here we studied the lateral organization of several adhesome proteins within integrin-mediated adhesions using different optical microscopy super-resolution techniques. We analyzed proteins belonging to different functional clusters or axial (z-) layers and found different patterns of spatial distribution of adhesome proteins within the same adhesion. Most proteins, including paxillin, were distributed continuously throughout an adhesion, while phosphorylated paxillin (pPax) and activated FAK showed enrichment in distinct clusters. Moreover, we detected these spots in different cell lines and showed with super-resolution live-cell imaging that the lateral mobility of spots within adhesions depends on FAK conformation. Finally, we demonstrated that this spacing is modulated by contractility and vinculin.
RESULTS
Substructure of adhesome proteins in focal adhesions
Rat embryonic fibroblast (REF) cells are an established cell line for the analysis of adhesions and their composition (Cavalcanti-Adam et al., 2007; Franz and Müller, 2005; Gudzenko and Franz, 2015; Hoffmann et al., 2014). We cultured these cells on fibronectin-coated coverslips and transfected fluorescently tagged proteins of interest or performed secondary immunolabeling to analyze several consensus adhesome proteins. We tested: (i) two different integrins [GFP-tagged β3 (ITGB3) and BMB5 antibody against α5 (ITGA5), staining αVβ3 and α5β1, respectively]; (ii) talin-1 and kindlin-2 (also known as TLN1 and FERMT2, respectively; integrin-activating adapters); (iii) paxillin, FAK and pPax-Y118 (signaling adapters); (iv) vinculin and zyxin (Vin or VCL, and ZYX, respectively; actin binders); and (v) actin itself. Imaging was performed with SIM, allowing us to analyze the organization of focal adhesions with twice the resolution compared with diffraction-limited microscopic methods (Bachmann et al., 2016). Admittedly, SIM does not resolve single fluorophores. Thus, structures visible in SIM represent local agglomerations (clusters) of several fluorophores, while structures labeled with one or very few fluorophores might be disregarded as background in contrast to more sensitive microscopic methods.
To analyze the spatial organization of adhesome proteins inside single cell–matrix adhesions, we focused on focal adhesions (Gardel et al., 2010). With the experimental conditions we used, we rarely observed fibrillar adhesions, while nascent adhesions or focal complexes were too small to manifest a sub-structural organization that could be detected with SIM. Using SIM, we could confirm earlier observations that focal adhesions are split into parallel ‘stripes’ (as indicated, for example, by the paxillin staining in Fig. 1), compared with a more homogenous organization observed with diffraction-limited microscopy (Hu et al., 2015; Young and Higgs, 2018). We found that these stripes were well separated from each other, and could be considered as individual focal adhesions in our experimental setup. A closer analysis of protein localization in these focal adhesions revealed that some adhesome proteins were organized in substructures that were well separated from each other. Intensity profiles along the long axis of focal adhesions gave the impression that, for example, pPax-Y118 was segregated into separate clusters within the same adhesion. Moreover, pPax-Y118 spots showed a remarkable regularity concerning the spacing of these clusters. In contrast, labeling intensities of paxillin, vinculin, zyxin and actin varied along the focal adhesions, but were almost continuously present throughout their whole lengths. Spatially separated clusters were only visible occasionally and not as regularly as observed for pPax-Y118. Fluorescent labeling of the remaining proteins – α5β1 and αVβ3 integrin, talin-1 and kindlin-2 – revealed a behavior somewhat between the two cases mentioned above. This is highlighted by the magnified panel for talin-1 in Fig. 1 which shows focal adhesions with continuous and spot-like organizations of talin-1 next to each other.
Thus, SIM gave the impression of spatial distributions within focal adhesions that differed between different adhesome proteins but also depended on the phosphorylation status, as shown for pPax-Y118 in comparison to paxillin.
Phosphorylated paxillin organizes in clusters with regular spacing to each other
Surprised by the difference in the paxillin and pPax-Y118 staining patterns, we decided to analyze their spatial organization in a quantitative manner. We labeled endogenous paxillin in REFs, as well as pPax-Y118 (Fig. 2A). Magnifications of single focal adhesions confirmed our earlier observation that paxillin is organized rather homogenously throughout adhesions, while pPax-Y118 localized in discrete clusters (Fig. 2A′–A‴). To analyze differences between spatial distribution of paxillin and pPax-Y118, we applied two independent methods to detect intensity peaks and to measure distances between these peaks. We used a published ImageJ-based plugin (NanoJ Core, Laine et al., 2019) to measure nearest-neighbor (NN) distances and independently developed a custom-written MATLAB routine that measures the center-to-center distance of the intensity maxima of labeled proteins (Fig. S1A). Using both methods, we analyzed our SIM images of pPax-Y118 and performed an analysis of paxillin labeling using the NN method as reference. From these distance values, we created histograms and plotted the average distance distribution based on histograms from independent experiments (Fig. 2B). These distance distribution plots revealed a narrower distance distribution of pPax-Y118 intensity peaks compared with that of paxillin and a peak distance that was shifted towards shorter distances for pPax-Y118. A quantitative comparison confirmed the impression of shorter distances between pPax-Y118 maxima (Fig. 2C; 555 nm with NN measurements and 469 nm with our MATLAB-based algorithm) compared with paxillin (616 nm). The distance distribution of pPax-Y118 also appeared more centered around the peak indicating less variation in distances between intensity maxima. We quantified the full-width half-maxima (FWHM) of distance distributions (Fig. 2D), which indeed revealed significantly smaller FWHM for pPax-Y118 compared with paxillin. Our MATLAB code also allowed us to measure the diameters of detected spots (Fig. 2E; paxillin, 332 nm; pPax-Y118, 267 nm) and to plot the number of detected spots within an adhesion against the length of that adhesion (Fig. 2F). Additionally, we tested the spatial distribution of vinculin (Fig. S1B–D) and found that the distances between vinculin maxima were broadly distributed, similar to those of paxillin.
Overall, these data and their analyses showed that pPax-Y118 forms ‘spots’ or spatially constrained clusters. These clusters are closer to each other and have a more regular spacing compared with clusters analyzed for paxillin. We believe that these analyses correspond well with the visual impressions from Fig. 1 and Fig. 2A, and show that pPax-Y118 has a stronger tendency to be constrained in clusters whereas paxillin and vinculin seem to lack a well-defined pattern within adhesions.
Additionally, to confirm that the differences in spatial distribution were not induced by labeling artefacts, we performed titration experiments of primary antibodies for paxillin (Fig. S2E–H) and pPax-Y118 (Fig. S2A–D). We could not detect a concentration-dependent effect of antibodies on the rather homogenous appearance of paxillin compared with the spot-like pPax-Y118 staining. We also tested a primary antibody for pPax-Y118 from another supplier and an antibody for pPax-Y31 (Fig. S2A–D). Both antibodies against pPax-Y118 confirmed the observation that pPax-Y118 organizes within separable spatial clusters, as does pPax-Y31.
pPax-Y118 organizes in spatially separated clusters in several cell lines
Next, we tested whether the observation of pPax-Y118 confinement in spatially separated clusters can be extended to different cell lines. We compared other established fibroblast cell lines (mouse embryonic fibroblasts, MEFs; mouse fibroblasts, NIH3T3; primary human foreskin fibroblasts, HFFs), a cancer cell line (HeLa) and an epithelial cell line (normal rat kidney cells, NRKs) (Fig. 3). In all cell lines, pPax-Y118 showed a distribution in separated clusters in contrast to the more homogenous paxillin staining (Fig. 3A). Quantitative analyses with our MATLAB workflow revealed that pPax-Y118 cluster spacing is overall conserved between different cell lines (Fig. 3B). This was also true for HeLa and HFF cells that differed the most regarding their maximum adhesion lengths (Fig. 3A,C). A histogram plot of distance distributions also confirmed that distances of pPax-Y118 clusters did not vary significantly between these cell lines and REF cells (Fig. 3D).
Phosphorylated tyrosines in focal adhesions organize in clusters
Paxillin is phosphorylated at Y31 and Y118 by a FAK–Src complex that forms after the activation of FAK by Y397 auto-phosphorylation (Arold et al., 2001; Sulzmaier et al., 2014; Thomas et al., 1998). We decided to test whether pFAK shows a spatial behavior that is similar to pPax-Y118 and paxillin. First, we labeled all combinations of these proteins in REF cells, imaged them with AiryScan microscopy (Jacquemet et al., 2020), and analyzed their intensity correlation and spatial co-occurrence (Pearson correlation and Mander coefficients; see also Aaron et al., 2018). These analyses confirmed high correlation and high co-occurrence of both FAK and paxillin with their phosphorylated versions (Fig. S3), and thereby a close interaction between both proteins. Next, we labeled focal adhesions in REF cells for pFAK (Fig. 4B) as well as for pPax-Y118 as a control (Fig. 4A) and imaged them with SIM. These images revealed that pFAK (Fig. 4B) is spatially organized in clusters as seen before for pPax-Y118. We then labeled focal adhesions with primary antibodies that bind promiscuously to phosphorylated tyrosines (pTyr; Fig. 4C), which also revealed a spatial pattern of separated clusters. Finally, we used these data and performed a quantitative NN analysis as done before for pPax-Y118 and paxillin. Histograms of NN distances showed spacings of pFAK and pTyr that closely followed that of pPax (Fig. 4D), albeit slightly shifted to smaller distances (Fig. 4E; pFAK, 483 nm; pTyr, 530 nm). Measurements of FWHM for pTyr and pFAK (Fig. 4F) confirmed the impression from distance distribution plots that the distances between clusters of these phosphorylated proteins followed a more regular distribution as observed before for pPax-Y118, in contrast to the more homogenous paxillin labeling.
Finally, we tested pFAK organization in NIH3T3 cells with STED microscopy (Fig. S4). We used different intensities of the depletion laser to mimic confocal microscopy (0% depletion laser) and SIM microscopy (5% depletion laser) or used a higher intensity to detect pFAK clusters with a better resolution than we did before (15% depletion laser). The distances between pFAK clusters were comparable to those measured before, but decreased slightly with increasing STED resolution (Fig. S4B,C; pFAK distances were 466 nm, 415 nm, 348 nm for 0%, 5%, 15% depletion laser, respectively). Upon visual inspection, it seemed that a higher resolution resolved comparably dim structures that were lost as background in the case of SIM imaging or STED at a lower resolution (Fig. S4A,D; example indicated with arrow). We tested this by creating an adhesion mask based on pFAK images resolved at 0% STED and applied this mask to 0%, 5% and 15% STED images (Fig. S4E). The fluorescent intensity found within this mask was comparable and showed that the improved resolution did not resolve relevant additional amounts of labeled pFAK in this setting. This also fit to intensity profiles as shown in Fig. S4D where the highest intensity maxima fully overlapped for all three STED imaging conditions whereas only the dimer regions differed depending on the resolution limit. Based on the STED results, we propose that spatial distribution of pFAK is not strictly binary as it might appear from images at lower resolution. However, clusters resolved at lower resolution contained the majority of the fluorescent intensity and thereby labeled protein. This also implies that the differences in distance distribution that we observed are representative of the majority of the respective proteins in adhesions.
FAK conformation regulates spacing of FAK clusters
Next, we wanted to study FAK and pFAK in more detail and specifically if we could identify protein domains of FAK or conformations of FAK that are linked to the specific spatial distribution of pFAK. To do so, we transfected NIH3T3 cells with plasmids encoding FAK wild type (wt) or different mutations of FAK and imaged them with AiryScan microscopy (Fig. 5A). Additionally, we labeled these cells for endogenous paxillin and for pFAK. We tested the following mutations of FAK: (i) a dominant-negative form without the kinase domain (FAK dKin), (ii) FAK that is not activatable (FAK Y397F), (iii) a double mutant combining FAK dKin with Y397F (FAK dKin Y397F), (iv) the isolated FAT domain (FAK FAT) or (v) a chimera of a membrane-targeted fluorophore, connected with a flexible linker to FAT (mFAT). This chimera connects to the lipid membrane within focal adhesions but lacks kinase activity, Src recruitment and specificity for phosphatidylinositol-4,5-bisphosphate (PIP2). Images of cells expressing these FAK constructs showed that all FAK constructs localized in focal adhesions (Fig. 5A, second row) together with endogenous paxillin (Fig. 5A, first row). Pearson correlation analysis showed that the intensities of all FAK versions correlated to similar degrees with the intensity of endogenous paxillin (Fig. 5B). The Pearson correlation with pFAK (Fig. 5C) was low for mutants lacking a functional FAK Y397 phosphorylation site (FAK Y397F, FAK dKin Y397F, FAT and mFAT) and high for those with the Y397 site (FAK wt and FAK dKin). This result is in line with the significant reduction of pFAK intensity for FAK dKin Y397F, FAK FAT and mFAT (Fig. 5D).
Next, we again used NN analysis and analyzed the spatial distributions of all labeled protein species. This analysis revealed shorter NN distances for pFAK compared with those of paxillin and FAK wt (Fig. 5E; paxillin, 501 nm; pFAK, 359 nm; FAK wt, 467 nm). These distances for paxillin and pFAK are shorter than observed before in REF cells imaged with SIM (Fig. 4E). However, the trend of pFAK cluster showing shorter distances between each other remained the same. For the different FAK constructs, NN distances of FAK dKin, FAK dKin Y397F and mFAT showed slightly lower values, and FAK Y397F and FAK FAT showed slightly higher values compared with FAK wt (FAK dKin, 435 nm; FAK Y397F, 508 nm; FAK dKin Y397F, 425 nm; FAK FAT, 519 nm; mFAT, 462 nm). These differences were not very pronounced, but it is worth mentioning that FAK dKin and FAK dKin Y397F had significantly shorter NN distances compared with FAK Y397F (Fig. 5E). Also, of all the FAK constructs, both FAK dKin and FAK dKin Y397F showed a distance distribution that differed the strongest from that of paxillin (Fig. 5H,J). The deletion of the kinase domain in both FAK dKin constructs also abolishes FAK autoinhibition, potentially linking the open FAK conformation to its spatial distribution within adhesions. Additionally, we plotted distance distributions for all FAK constructs and endogenous paxillin (Fig. 5G–L) as well as FWHM values for paxillin and pFAK labeling from cells expressing FAK wt and all FAK mutants (Fig. 5F). This analysis confirmed that pFAK clusters, like those of pPax-Y118, were closer to each other and showed a more defined spacing compared with labeling of FAK and paxillin.
We also tested whether PIP2-dependent recruitment of FAK to the lipid membrane via the FAK FERM domain could explain the spatial distribution of pFAK clusters. We expressed FAK FERM, a sensor for PIP2 (PLCd PH GFP) or a sensor for phosphatidylinositol-(3,4,5)-trisphosphate (PIP3; sensor AKT PH GFP) in NIH3T3 cells and labeled endogenous paxillin and pFAK (Fig. S5A). AKT PH GFP showed lower Pearson correlation with paxillin and pFAK compared with FAK FERM and PLCd PH GFP (Fig. S5B,C). However, analysis of NN distances (Fig. S5D,E) indicated no differences among the lipid sensors, making it unlikely that the specificity of FAK FERM for PIP2 explains the spatial distribution of pFAK compared with FAK. We also tested constructs for FAK wt, FAK dKin, FAK Y397F and FAK dKin Y397F in mouse fibroblasts with knockout of endogenous FAK (MEF FAK−/−). These experiments confirmed the differences observed before for paxillin and FAK wt compared with pFAK (Fig. S5F–H; NN distances for paxillin, 535 nm; FAK wt, 540 nm, pFAK, 363 nm). Among the different FAK constructs (Fig. S5F,H–K), FAK dKin and FAK dKin Y397F again showed the lowest NN distances and mimicked pFAK distribution the most. However, in contrast to experiments in NIH3T3 cells, NN distances for FAK Y397F were shorter than those of FAK wt (FAK dKin, 427 nm; FAK dKin Y397F, 413 nm; FAK Y397F, 451 nm).
Overall, this analysis confirmed that pFAK follows a spatial distribution that is different from paxillin and total FAK; distances between pFAK clusters were shorter and followed a preferred distance for the spacing of pFAK clusters. Mutational analysis indicated that the spatial distribution of FAK could be influenced by its conformational status. In addition, experiments with FAK FERM showed that FAK recruitment to the lipid membrane did not explain the spatial distribution of pFAK. The open conformation of FAK appears to be required for implementing the spatial distribution of pFAK.
Live-cell imaging shows dynamic clusters of FAK
To understand the creation and regulation of spatial clusters in focal adhesions in more detail, we set out to analyze their temporal evolution. First, we cultured REF cells on coverslips and fixed them at different time points after cell seeding, followed by immunolabeling of paxillin and pPax-Y118 (Fig. S6A,B). Starting with culturing cells for 2 h, we observed focal adhesions with the same appearance as described above; paxillin was spread throughout adhesions while pPax-Y118 was confined in distinct spatial clusters. Quantitative analysis with our MATLAB code revealed that the spacing between neighbored pPax-Y118 clusters was preserved over time and remained constant (Fig. S6C). To get more insights into the spatial dynamics of clusters in focal adhesions of living cells, we used NanoJ-SRRF (Gustafsson et al., 2016) and imaged FAK wt, FAK dKin, mFAT and paxillin (Fig. 6). We imaged living NIH3T3 cells expressing these proteins on a total internal reflection fluorescence (TIRF) microscope and calculated super-resolution radial fluctuation (SRRF) images (25 images taken at 20 Hz used for one SRRF image, temporal resolution 1.25 s/SRRF frame). According to NanoJ SQUIRREL (Culley et al., 2018), we achieved a local resolution with these settings of around 100 nm and lower (Fig. 6A″). Analyzing single adhesions from these movies with kymographs indicated spatial clusters for all FAK constructs, but also identified different mobility behaviors (Fig. 6B). While FAK wt clusters showed dynamic cluster movements, including merging and splitting, mFAT clusters appeared more stable. Paxillin, in contrast, showed a lower tendency to cluster and was more spread throughout adhesions. Kymographs of FAK dKin showed long-lasting tracks, similar to mFAT. In contrast to all FAK constructs, tracks of paxillin were harder to identify, potentially reflecting the broader distribution of paxillin in distance distribution plots (Fig. 2B). We calculated mean square displacement (MSD) values based on tracks of clusters to describe dynamic behavior (Fig. 6D). We also imaged fixed cells expressing FAK wt and calculated MSD values as a reference for immobile tracks. This information was used to define a mobile fraction in living cells and to calculate corresponding MSD values (Fig. 6E). All MSD values (FAK wt, 0.00675 µm2/s; FAK dKin, 0.00752 µm2/s; mFAT, 0.00376 µm2/s; paxillin, 0.0042 µm2/s) were in a similar range to earlier reports of nanoclusters of αLβ2 integrin binding to ICAM-1 (0.0078 µm2/s, Diez-Ahedo et al., 2009). Overall, FAK wt and FAK dKin showed the highest MSDs, however, with more variation in the case of FAK wt. Additionally, we calculated cross-correlation values between every fifth temporal frame (Fig. 6C). Based on this calculation, high cross-correlation indicated little spatial change within five timeframes (5×1.25 s) and vice versa. FAK wt showed low cross-correlation corroborating higher mobility of clusters compared to all other conditions (Fig. 6F). Thus, cross-correlation and MSD values both indicate higher dynamics of FAK wt clusters. The FAK mutant in a constitutively open conformation, FAK dKin, showed similar MSD but less abrupt changes judging from cross-correlation values. Paxillin and mFAT both showed reduced mobility compared with that of FAK dKin.
Super-resolution live-cell imaging with SRRF confirmed the presence of spatial clusters of FAK constructs within focal adhesions. Moreover, it became clear that the spatial dynamics of these clusters differ between paxillin and FAK, showing higher dynamics for the latter. Reduced dynamics of mFAT might reflect its binding to the more stable paxillin, and also the lack of kinase activity of FAK and/or of Src.
Spacing of pPax clusters is mechanosensitive and depends on vinculin
Finally, we wondered whether it is possible to interfere with the spacing between clusters of phosphorylated proteins depending on cell contractility. Previous publications have indicated that actomyosin contractility increases paxillin phosphorylation (Pasapera et al., 2010; Zaidel-Bar et al., 2007). At the same time, several reports highlighted a force-dependent activity of FAK (Bauer et al., 2019; Torsoni et al., 2003; Wong et al., 2011). Thus, we treated REF cells with the myosin inhibitor blebbistatin or with Y27632 (Rho-ROCK pathway inhibitor) and labeled paxillin and pPax-Y31 (Fig. 7A–C; pPax-Y31 showed no differences from pPax-Y118; Fig. S2C,D). Analysis with our MATLAB workflow indicated a reduction in pPax-Y31 cluster spacing (Fig. 7D) and of pPax-31 localization in adhesions (Fig. 7E) for both inhibitors. Thus, reduced contractility caused a higher density of pPax clusters within focal adhesions but a reduction in the total amount of pPax-Y31.
Paxillin phosphorylation has also been linked to vinculin recruitment (Pasapera et al., 2010). To test this, we cultured mouse embryonic fibroblasts from vinculin knockout mice (MEF Vin−/−) and analyzed pPax-Y118 in these cells compared with MEF wt cells (Fig. 8). Quantitative analysis of the pPax-Y118 cluster distance revealed an increase in the cluster distance in the absence of vinculin (Fig. 8E). Additionally, we tested the impact of cellular contractility in MEF cells (Fig. 8C,D) as done before in REF cells. These experiments confirmed that reduced cellular contractility decreases pPax-Y118 spacing (Fig. 8E) as observed before for pPax-Y31 in REF cells.
To conclude, actomyosin contractility affected the spacing of pPax clusters within focal adhesions in REF cells and in MEF wt cells. Moreover, we showed that vinculin expression impacted pPax-Y118 cluster spacing.
DISCUSSION
We set out to study whether adhesome proteins show different lateral organization within integrin-mediated adhesions. We observed that several adhesome proteins appeared to localize homogenously in focal adhesions when observed with SIM. This observation was in striking contrast to pPax and pFAK localization, for which we detected clusters that were spatially separated from each other using super-resolution light microscopy. Additionally, we found that these clusters have a more defined spacing between them compared with the intensity maxima of labeling for paxillin and FAK. Control experiments and quantifications confirmed this impression irrespective of the antibodies used and for different microscopic methods. However, STED imaging at higher resolution indicated that protein localization is not binary and that a minority of pFAK can also localize outside of these clusters. Live-cell imaging of different FAK constructs and paxillin indicated that all FAK constructs had a tendency to localize in clusters compared with paxillin. Finally, experiments with contractility inhibitors pointed to a connection of the distribution of pPax clusters to actin-mediated forces on adhesions.
Focal adhesions have been a popular target of super-resolution microscopy from the beginning of the development of these techniques (Betzig et al., 2006). The work by Betzig and colleagues already demonstrated that vinculin is organized with fluctuating density in focal adhesions, creating more and less dense areas of vinculin. Since these earliest super-resolution reports, the pointillistic nature of focal adhesions has been confirmed many times when observed with high resolution microscopy (Bachmann et al., 2016; Changede et al., 2015; Diez-Ahedo et al., 2009; Shroff et al., 2008, 2007; Spiess et al., 2018; Xu et al., 2018). In the work presented here, we mostly used super-resolution methods that have limited resolution (typically around 100 nm) compared with the studies mentioned before. These reports using higher resolution methods found diameters of spatially confined clusters of integrins within focal adhesions to be in a range from 40 nm (Shroff et al., 2008; Spiess et al., 2018) to 100 nm (Changede et al., 2015). At the same time, cryo-electron microscopy showed even smaller structures at adhesion sites (Patla et al., 2010; donut-shaped structures connected to actin with a diameter around 20 nm and nearest neighbor distance of 15–60 nm). In fact, our STED images at higher resolution (15% depletion laser, resolution around 68 nm) also indicated substructures within some clusters that were not resolved at low resolution (Fig. S4D; for example, first maxima 0% versus 5% depletion laser). Moreover, higher-resolution imaging also revealed some pFAK labels that were not detected at lower resolution (arrow in Fig. S4A,D). These dim pFAK species imply that the spatial distribution of pFAK (and most likely other phosphorylated adhesome proteins) is not binary as it appeared from SIM images. A non-binary spatial organization is also in line with reports showing clusters of different adhesome proteins throughout adhesions (Bachmann et al., 2016; Changede et al., 2015; Diez-Ahedo et al., 2009; Shroff et al., 2007, 2008; Spiess et al., 2018; Xu et al., 2018). However, it appears that such dim pFAK species account for a minority of labeled protein. Thus, the spatial organization of adhesome phosphoproteins might be preferentially, but not exclusively, in clusters. Such a preferred localization of phosphoproteins in ‘hotspots’ or functional clusters has also been described by Ballestrem and colleagues using Förster resonance energy transfer (FRET) (Ballestrem et al., 2006). Importantly, the clusters we described here are well above the resolution limit, even of conventional microscopy (Fig. 2E, diameter of Pax-Y118 clusters around 267 nm). In fact, published images of pPax or pFAK already show ‘pointy’ or ‘spotty’ localization (see, for example, Horton et al., 2016; Pasapera et al., 2010; Zaidel-Bar et al., 2007). These arguments further strengthen the point for the existence of pPax and pFAK clusters.
The question remains how these differences in spatial distribution between pFAK and pPax, on the one hand, and FAK and paxillin, on the other hand, are implemented on the molecular level, and why they are organized in this particular manner. FAK is recruited to the lipid membrane and focal adhesions in an inactive closed conformation before auto-phosphorylation of Y397 activates FAK by releasing the kinase domain from inhibitory binding to FAK FERM. Thus, the distance distributions of FAK and pFAK indicate differences in the spatial distribution of the conformations of FAK, i.e. an equilibrium of FAK conformations in the case of FAK, versus only the open FAK conformation in the case of pFAK. Of the FAK constructs we tested, FAK Y397F can be considered to be locked in a closed conformation. At the same time, the constructs FAK dKin and FAK dKin Y397F that lack the kinase domain can be considered to be locked in an open conformation. Indeed, NN distances of both FAK dKin and FAK dKin Y397F were significantly shorter (and thereby closer to that of pFAK) compared with NN distances of FAK Y397F (Fig. 5E). Thus, an open FAK conformation might be needed to achieve the particular spatial distance distribution observed for pFAK. Studies in FAK−/− cells (Fig. S5F–K) also showed shorter distances for FAK constructs lacking a kinase domain compared with FAK wt. However, NN distances of FAK Y397F were similar to both FAK dKin constructs and differed in this regard from experiments in the presence of endogenous FAK. An additional contributing factor for the spatial distribution of pFAK could be the binding of Src to the active FAK complex. In this regard, it is noteworthy that FAK dKin has a phosphorylated Src-binding site that is lost in FAK dKin Y397F due to the additional Y397F mutation. Nevertheless, both mutations show similar NN distances and a similar distance distribution implying that Src binding is not essential for the spatial distribution of pFAK.
Live-cell imaging of different FAK constructs and of paxillin (Fig. 6) showed clustered localization of all FAK constructs but a less clustered distribution of paxillin. This difference indicates that distance distributions of FAK wt and pFAK do not differ because of an inability of FAK wt to cluster, but rather because of an inability to keep a defined distance between clusters. Indeed, tracks of FAK wt (Fig. 6B) showed abrupt changes while FAK dKin clusters followed more consistent tracks. This difference could be related to FAK dKin being locked in an open conformation compared with FAK wt undergoing conformational changes. Tracks of mFAT appeared to be more clustered than was the case for paxillin. This was surprising for us, given that mFAT localizes to adhesions only via FAT, the paxillin-binding domain of FAK. Thus, clustering of FAK constructs might be linked to FAK–paxillin binding via the FAT domain. At the same time, there are many reports highlighting the relevance of FAT–paxillin binding for FAK localization in adhesions and for FAK activity (Deramaudt et al., 2011, 2014; Mousson et al., 2021). This motivated us to study in more detail how FAK and paxillin interact. FAK FAT is reported to bind paxillin via its LD2 and LD4 motifs (Hoellerer et al., 2003). At the same time, several reports point to the relevance of paxillin phosphorylation at Y31 and Y118 for paxillin–FAK interaction (Choi et al., 2011; Digman et al., 2009). However, Y31 and Y118 are clearly located outside of the LD2/LD4 domains, raising the question of how pY31 and pY118 are supposed to impact FAT–paxillin binding. No protein structure of full-length paxillin is currently available and, therefore, we studied AlphaFold (Jumper et al., 2021; Mirdita et al., 2021preprint) predictions of paxillin (Fig. S7). All five predictions of paxillin suggested that paxillin LIM domains are organized as a rigid rod while N-terminal LD domains and unstructured regions are flexible with no clear spatial arrangement between the different domains (Fig. S7A,C). The only two exceptions are two models (confidence rank 1 and rank 3) where Y31 and Y118 localize on the surface of LIM domains in areas with negative surface charges (Fig. S7B–E). Importantly, these models show the lowest error for the spatial arrangement between Y31/Y118 and LIM domains (Fig. S7D), indicating that these models are more likely compared with models where Y31/Y118 do not bind LIM domains. Binding of Y31/Y118 on the LIM surface will lead to LD motifs being wrapped around the LIM rod (Fig. S7A), thereby limiting steric access of FAK FAT to paxillin LD2 and LD4 motifs. However, phosphorylation of Y31 and Y118 will cause electrostatic repulsion to the negative surface of LIM domains. This will free not only Y31/Y118 but also LD motifs, release steric constraints, and eventually allow FAT–LD2/LD4 binding. In this scenario, FAT binding to phosphorylated paxillin will bring the FAK–Src complex in close proximity to paxillin. The FAK–Src complex is able to phosphorylate paxillin Y31 and Y118, in turn stabilizing FAK–paxillin interaction. Eventually, this establishes a positive feedback loop that can spawn local clusters of pFAK and pPax and explain their similar spatial distributions that we observed here. Such a positive feedback loop could be limited by FAK switching back to its closed conformation (dephosphorylation of pY379) and/or by reducing the FAT–paxillin interaction that affects FAK recruitment to adhesions and signaling (Deramaudt et al., 2011, 2014; Mousson et al., 2021). In this respect, a phosphorylation site in FAK FAT (Y925) was shown to prevent paxillin binding when phosphorylated by Src (Kadaré et al., 2015). Thus, FAK–Src binding could be part of the positive feedback loop described above, while, at the same time, it could apply a limit to the progression of the feedback loop.
Several reports showed that FAK activation and kinase activity is increased by mechanical forces (Bauer et al., 2019; Seong et al., 2013; Torsoni et al., 2003; Wong et al., 2011) and that contractility inhibition decreases paxillin phosphorylation at Y31/Y118 (Fig. 7E, (Zaidel-Bar et al., 2007)) and also FAK phosphorylation at Y397 (Pasapera et al., 2010). The fact that both FAK-Y397 phosphorylation and paxillin phosphorylation depend in the same manner on contractility is in line with the hypothesis of a positive feedback mechanism between these two events. For clusters of pPax, we found in two different cell lines that contractility reduction causes a small but significant reduction in cluster spacing. Thus, reduced pPax levels are connected to reduced spacing of pPax clusters, suggesting that less pPax leads to a shorter distancing between neighbored pPax clusters. This finding fits a model of lateral inhibition in which the local amount of pPax reduces at the same time as does the concentration of pPax in its vicinity. It will be interesting to test in more detail whether this effect indicates a patterning mechanism according to Turing (reviewed in Howard et al., 2011) for pPax spacing. Such a model is not mutually exclusive with a role of paxillin LIM domains in sensing actin fibers under mechanical stress (Sun et al., 2020; Winkelman et al., 2020) and/or sensing their contractility-dependent angle towards adhesions and the lipid membrane (Liu et al., 2015).
Vinculin knockout caused a small increase in pPax spacing compared with that in wild-type cells. Vinculin knockout leads to reduced traction stress (Thievessen et al., 2013), probably linked to the function of vinculin as a mechanical talin-actin linker (Carisey et al., 2013; Grashoff et al., 2010; Humphries et al., 2007). Based on these findings, we expected that vinculin knockout phenocopies contractility inhibition. Instead, we found the opposite. A possible explanation for this contradiction could be that the relevance of vinculin for mechanical force transmission, on the one hand, and for paxillin binding, on the other, cause opposing effects for pPax cluster spacing that require further analysis.
We also wondered about the evolutionary advantage of clustering pFAK/pPax in adhesions. It might be interesting to consider alternative spatial organizations: Why are pPax/pFAK not spread more homogenously throughout focal adhesions? One possibility might be that pPax/pFAK signals have to be detected in the noisy environment of protein phosphorylation. In this case, it might be advantageous to locally concentrate a pPax/pFAK signal so that it may exceed the background noise; this would not be the case if pPax/pFAK were homogenously spread in focal adhesions. A mechanism of local enrichment to increase signal strength is, to our knowledge, hypothetical for focal adhesions. However, the general idea of local signal enhancement by concentrating the signal also applies to other cell signaling events. Lipid rafts concentrate transmembrane proteins, myelinated axons organize localized depolarization at myelin-sheath gaps, and filopodia concentrate signaling pathways in a limited volume compared to the complete cell body. Thus, it might not be surprising if signaling hubs like focal adhesions also implement the same strategy.
In summary, we have presented new data for the structural implementation of paxillin and FAK and their phosphorylated forms in focal adhesions. pPax and pFAK preferentially localize in clusters with a distance distribution that differs from that of paxillin and FAK that are organized more randomly. This suggests a regulated spacing mechanism for such pPax/pFAK signaling hubs, thereby connecting the function of focal adhesions with their lateral structure. Mutational analyses of FAK point to the open conformation of FAK to be necessary for the regularity of pFAK cluster spacing, whereas paxillin–FAT binding appears to regulate FAK clustering itself. Further work will address the dynamics of the FAK–paxillin interaction in more detail and the connection of FAK clusters to other components of the adhesome.
MATERIALS AND METHODS
Cell culture and plasmids
Cells were cultured in Dulbecco's Modified Eagle Medium (DMEM, Pan-Biotech, Germany) supplemented with 10% fetal calf serum (FCS, HyClone, USA) at 5% CO2 and 37°C. NIH3T3 cells, HeLa cells and NRK cells were obtained from American Type Culture Collection (ATCC, USA). Rat embryonic fibroblasts were a gift from B. Geiger and human foreskin fibroblasts (HFFs) were obtained from PromoCell (Germany). MEF wt cells and MEFs from vinculin knockout mice (MEF Vin−/−) were a gift from W.H. Ziegler (Mierke et al., 2010). MEFs from FAK knockout mice (MEF FAK−/−) were a gift from S. Aizawa (Ilić et al., 1995) and P. Rondé. Transfections were carried out with Lipofectamine 2000 (ThermoFisher) according to the manufacturer's instructions. cDNA encoding the full-length mouse β3-wt–GFP integrin expressed in a cytomegalovirus promoter-driven pcDNA3/EGFP vector has been previously described (Ballestrem et al., 2001), zyxin–RFP was deposited by A. Huttenlocher (Addgene #26720; Bhatt et al., 2002) and vinculin wt GFP was a gift from C. Ballestrem (Humphries et al., 2007). Paxillin–GFP was prepared as described before (Ripamonti et al., 2021), as well as PLCd PH GFP (Saltel et al., 2009), AKT PH GFP (Ivarsson et al., 2013) and talin1–GFP (Saltel et al., 2009). The coding sequence for mouse FAK was a gift from David Schlaepfer (UCSD, USA). Fluorophores were exchanged to tagRFP and mutations were introduced with site-directed mutagenesis followed by digestion and ligation using unique restriction sites. FAK dKin Y397F was created by ClaI+XbaI digestion of FAK dKin and FAK Y397F, followed by replacement of the kinase-containing part of FAK Y397F with the kinase-deleted part from FAK dKin. mFAT was created by replacing FAK FERM and the kinase domain from tagRFP FAK wt with Lck-mScarletI (deposited by Dorus Gadella; Addgene #98821; Chertkova et al., 2020preprint) by NheI and XhoI digestion and ligation. FAK sequence in mFAT starts with Ser840.
Antibodies and reagents
Inhibition experiments were performed with blebbistatin (Sigma-Aldrich, USA) or with the ROCK inhibitor Y27632 (Sigma-Aldrich, USA) at concentrations as indicated. Immunostaining was performed after fixation of cells with 4% paraformaldehyde (Sigma-Aldrich, USA) in PBS. Reagents used for immunostaining were mouse antibodies against paxillin (BD Biosciences, #610051, 1:500 or 1:100, 1:500, 1:1000, 1:5000 in Fig. S2E-H) and vinculin (Sigma-Aldrich, #V-9131, 1:100), rat antibodies against α5 integrin (BMB5 clone, Merck, #MAB2514, 1:100) and rabbit antibodies against pPax-Y118 (Cell Signaling, #2541S, 1:500 or 1:100, 1:500, 1:1000 for Fig. S3: ThermoFisher, #44722G, 1:100, 1:500, 1:1000 for Fig. S3), against pPax-Y31 (ThermoFisher, #44720G, 1:100, 1:500, 1:1000 for Fig. S3 and 1:500 for Fig. 7), against pTyr (Sigma-Aldrich, #T1325, 1:500) and against pFAK-Y397 (ThermoFisher, #700255, 1:1000). Primary antibody staining was followed by washing steps and incubation with antibodies against mouse antibodies labeled with Cy3 (Jackson Immunoresearch, #115-165-146, 1:500), against rabbit antibodies labeled with Alexa Fluor 488 (ThermoFisher, #A11070, 1:500) or Cy3 (1:500, Dianova, #111-165-144, 1:500), or with phalloidin coupled to Alexa Fluor 647 (1:200, ThermoFisher, #A22287). Primary rat antibodies were visualized with Alexa Fluor 488-labeled secondary antibodies (ThermoFisher, #A11006, 1:500).
Microscopy
SIM imaging was performed on a non-serial Zeiss Elyra PS.1 microscope with a 63×/1.4NA oil immersion objective and an Andor iXon EMCCD camera. The grid for SIM was rotated three times and shifted five times leading to 15 frames of raw data, out of which a final SIM image was calculated with the structured illumination package of ZEN software (Zeiss). Values for calculation were selected for best resolution without causing image artifacts. Channels were aligned using a correction file that was generated by measuring channel misalignment of fluorescent tetraspecs (ThermoFisher, #T7280). Airyscan imaging was performed on a Zeiss LSM 800 with a 63×/1.4NA oil immersion objective and in super-resolution AiryScan mode. Data for NanoJ-SRRF (Gustafsson et al., 2016) were acquired in TIRF mode on a Nikon Eclipse Ti with perfect focus system, 100×/1.49 NA oil objective and an Andor-EMCCD iXon897 camera. STED images were acquired on a Leica TCS SP8 STED 3× in gated STED mode and with Lightning deconvolution. Images were acquired with a 100×/1.4NA oil immersion objective. Depletion laser intensity and pixel size are indicated in the text. Secondary immunolabeling of pFAK was done with anti-rabbit Abberior Star Red (abberior, #STRED-1002-500UG, 1:1000). Live cell imaging was performed on a heated stage at 37°C and 5% CO2. Cells were cultured on fibronectin-coated (10 µg/ml, 1 h at room temperature) glass-bottom dishes in DMEM+10% FCS for 5 h. Medium was changed to F12 buffer (ThermoFisher, USA) with 1% FCS 1 h before imaging. Cells were imaged for 2–3 min with 50 ms framerate. NanoJ-SRRF was used to calculate super-resolution frames consisting of 25 raw frames, i.e. 1 SRRF frame/1.25 s. Quality and resolution of SRRF movies was analyzed with NanoJ-SQUIRREL (Culley et al., 2018).
Image analysis
Images were prepared and analyzed (intensity profiles) using the Fiji software package (Schindelin et al., 2015). Line profiles were measured by averaging over the width of single adhesions in order to consider the full width of focal adhesion stripes. For analysis of spot-to-spot distance, we used a custom-written software, utilizing MATLAB Image Processing Toolbox (The MathWorks Inc., USA) and a nearest neighbor method in ImageJ (see below). The workflow for MATLAB-based process was as follows: two images, one serving as a focal adhesion mask and one being the staining of interest, were needed. Labeled paxillin or vinculin was used as mask in our case (Fig. S1A, Mask) while pPax, pFAK or pTyr were labels of interest (Fig. S1A, Image of interest). The mask image was thresholded and structures below an area limit of 1–3 µm2 were excluded in order to limit the analysis to focal adhesions. Optionally, focal adhesions within a mask could be slightly broadened by adding the desired number of pixels (typically around one) around their periphery to ensure them covering most of the labels of interest within focal adhesions. Then, all focal adhesions within the mask were indexed and fitted as ellipses to extract their positions, lengths and widths (Fig. S1A, Characterize). Based on these features, labeled areas within corresponding focal adhesions were isolated and an intensity line profile was created for a peak-to-peak distance calculation (Fig. S1A, Intensity profile). Size of individual spots was also extracted from intensity line profiles as width of each peak at its half prominence. Images were pre-processed in similar ways for nearest neighbor analysis and for measuring correlation and co-occurrence: the paxillin channel was background subtracted (sliding paraboloid, rolling ball radius: 1 pixel); mask for focal adhesions was created (threshold Otsu, create selection, enlarge selection −0.1 µm followed by +0.1 µm to remove structures smaller than focal adhesions); this selection was transferred to unprocessed images of labeled proteins of interest to be analyzed. Measurements of nearest neighbor distances were performed with the ImageJ plugin NanoJ-Core (Laine et al., 2019). Tolerance levels were chosen in order to faithfully detect peaks of intensity (SIM images, 1000; AiryScan, 100). Distance values (>100 nm and <1000 nm) obtained either with the NN method in NanoJ-Core or with our MATLAB workflow were used to create histograms with a bin width of 50 nm. For FWHM analysis, we counted the number of bins with normalized frequency ≥0.5 and multiplied this number with 50 nm as a measure for FWHM. FWHM plots show these values as small dots and averages per independent replicate as big dots. For distance distribution plots, measured distances were pooled for one independent experiment and histograms were created per independent experiment. These histogram data were used to plot distance distribution with PlotTwist (Goedhart, 2020). Correlation and co-occurrence were calculated with the ImageJ plugin JACoP (Bolte and Cordelieres, 2006). Correlation between pFAK and lipid sensors was calculated after using masks based on pFAK staining, not paxillin staining. Kymographs were created, tracks identified and analyzed (MSD) with the plugin KymographTracker (version 1.1.0.0; Chenouard et al., 2010) in Icy (de Chaumont et al., 2012). Cross-correlation of kymographs was calculated with a custom-written MATLAB code. We calculated the average normalized cross-correlation between every fifth-time slice (i.e. first slice in cross-correlation plot=cross-correlation of first and fifth slice in kymograph; second slice cross-correlation=cross-correlation of second and sixth slice in kymograph, and so on). This approach was chosen because changes between consecutive slices was high for all conditions, masking dynamic changes occurring over more than one time frame (=1.25 s). The scale of normalized cross-correlation is 0 to 1, where 1 indicates perfect correlation and, therefore, spatial consistency across the intensities during the observed time period. x-axis in cross-correlation plots is a relative spatial value and does not directly correlate to x-values in kymographs.
Statistics and software
In Figs 3,7,8, Figs S2, S6, data in box plots are represented as follows: dot within the box representing mean value, box outline representing the 25th and 75th percentiles, whiskers representing standard deviation, and upper and lower bars representing the 5th and 95th percentiles. Preparation of graphs and statistical significance testing were done with OriginPro 2017 software (OriginLab Corp., USA). In Figs 2–7, Figs S3–S5, figures were prepared with PlotTwist (Goedhart, 2020) for histogram-based distance distribution plots and SuperPlotsOfData (Goedhart, 2021) for bar charts. Distance distribution plots show mean (indicated as thick line) and 95% confidence intervals (shade) calculated from independent experiments. Bar charts show technical replicates as small dots and mean from independent experiments as big dots. Black lines indicate mean from all independent experiments and upper and lower black line show standard deviation. The numbers of independent experiments are mentioned in figure legend as ‘n’. Statistical tests were performed by SuperPlotsOfData and PlotTwist, respectively. When tests for normal distribution (Shapiro–Wilk test) indicated non-normal distribution for the majority of conditions (applied to Fig. 6D,E), we calculated Mann–Whitney U-test instead (Jupyter notebook implemented code, stats package from scipy; code available, see Data availability section) Paxillin models were created with AlphaFold (Jumper et al., 2021) and ColabFold (Mirdita et al., 2021 preprint); MSA alignment was performed with Jackhmmer. Visualization of models of paxillin structure was done in Chimera X (Goddard et al., 2018; Pettersen et al., 2021).
Acknowledgements
We are grateful to Monica Julio Barreto and Tatiana Fomekong for help with molecular cloning and to Nikolaos Athanasopoulos and Christopher Henry for their help with data analysis. We thank the bioimaging core facility of University of Geneva for their help and support and Philippe Rondé (Université de Strasbourg, France) for help with reagents. We acknowledge UCSF ChimeraX, developed by the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco, with support from National Institutes of Health (R01-GM129325) and the Office of Cyber Infrastructure and Computational Biology, National Institute of Allergy and Infectious Diseases.
Footnotes
Author contributions
Conceptualization: M. Bachmann, A.S., B.W.-H., M. Bastmeyer; Methodology: A.S., M. Bachmann; Software: A.S.; Formal analysis: M. Bachmann, A.S.; Resources: A.S., B.W.-H., M. Bastmeyer; Data curation: M. Bachmann, A.S., K.W.; Writing - original draft: M. Bachmann, A.S.; Writing - review & editing: M. Bachmann, A.S., B.W.-H., M. Bastmeyer; Visualization: M. Bachmann, A.S.; Supervision: M. Bachmann, B.W.-H., M. Bastmeyer; Project administration: M. Bachmann, M. Bastmeyer; Funding acquisition: M. Bastmeyer, B.W.-H.
Funding
M. Bachmann acknowledges funding by Deutsche Forschungsgemeinschaft (DFG, German Research Foundation; BA 6471/1-1). A.S. acknowledges funding from the Erasmus Mundus Europhotonics program. The work of M. Bastmeyer is supported by the Deutsche Forschungsgemeinschaft under Die Exzellenzstrategie des Bundes und der Länder (Germany's Excellence Strategy) through EXC 2082/1-390761711 (3DMM2O Excellence Cluster by the Karlsruhe Institute of Technology and Heidelberg University). The work of B.W.-H. is supported by the Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung (Swiss National Science Foundation, grant 310030_185261 and 310030L_170112).
Data availability
Microscopy images analyzed for this study, analyzed data, MATLAB code and ImageJ macro codes are accessible on Yareta (the archive portal of University of Geneva) at doi:10.26037/yareta:lapi22rvsfddrgtmip4ffi4p2i. All code is accessible via Github (https://github.com/Mitchzw/pPax-and-pFAK-in-adhesions).
Peer review history
The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.258764.
References
Competing interests
The authors declare no competing or financial interests.