Abstract
αVβ3 integrin recognizes multiple extracellular matrix proteins, including vitronectin (Vn) and fibronectin (Fn). However, cell experiments are frequently performed on homogenously coated substrates with only one integrin ligand present. Here, we employed binary-choice substrates of Fn and Vn to dissect αVβ3 integrin-mediated binding to both ligands on the subcellular scale. Superresolution imaging revealed that αVβ3 integrin preferred binding to Vn under various conditions. In contrast, binding to Fn required mechanical load on αVβ3 integrin. Integrin mutations, structural analysis, and molecular dynamics simulations established a model where the extended-closed conformation of αVβ3 integrin binds Vn but not Fn. Force-mediated hybrid domain swing-out characterizes the extended-open conformation needed for efficient Fn binding. Thus, force-dependent conformational changes in αVβ3 integrin increase the number of available ligands and therefore the ligand promiscuity of this integrin. These findings for αVβ3 integrin were shown to regulate cell migration and mechanotransduction differentially on Fn compared to Vn and therefore to regulate cell behavior.
- List of abbreviations
- Fbg
- fibrinogen
- Fn
- fibronectin
- FnIII10
- 10th domain of fibronectin
- MEF
- mouse embryonic fibroblast
- MEF vcl -/-
- mouse embryonic fibroblasts derived from vinculin knockout mice
- MD
- molecular dynamics
- NMIIA
- nonmuscle myosin IIA
- Opn
- osteopontin
- Tsp
- thrombospondin
- Vn
- vitronectin
Introduction
Integrins, an important class of cell-matrix adhesion receptors, consist of α- and β-subunits forming transmembrane heterodimers (Campbell and Humphries, 2011). While the extracellular part binds to proteins of the extracellular matrix (ECM), the intracellular part is connected to actin via multiple adapter and signaling proteins that make up the so-called adhesome (Hytönen and Wehrle-Haller, 2014, Zaidel-Bar et al., 2007). The functional definition of integrin activation is based on the capacity to bind ligands. The structural definition associates the active structure with conformational changes from a bent, to an extended-closed, and finally to an extended-open state (Campbell and Humphries, 2011, Zhu et al., 2008, Su et al., 2016). Concerning the β-integrin subunit, the extended-open conformation is characterized by the swing-out of the hybrid domain in respect to the βA domain (Zhu et al., 2013). Molecular dynamics (MD) simulations indicate that force supports this hybrid domain swing-out and thereby leads to full integrin activation (Zhu et al., 2008, Puklin-Faucher et al., 2006, Dong et al., 2017). These theoretical studies are supported by multiple examples where mechanical forces regulate integrin functions, structural elements of the adhesome, and thus integrin-mediated adhesions (Balaban et al., 2001, Stricker et al., 2011, Hytonen and Wehrle-Haller, 2016, Nordenfelt et al., 2016, Rahikainen et al., 2017, Kuo et al., 2011, Schiller et al., 2011).
The best-studied class of integrins is the group of RGD-binding integrins sharing the ability to interact with an exposed Arg-Gly-Asp peptide in their ligands (Hynes, 2002). Promiscuity between ligands and receptors is especially common for these RGD-integrins. For example, αVβ3 integrin is reported to have the ability to recognize at least 12 different ligands (Humphries et al., 2006). Initially described as the vitronectin (Vn) receptor (Pytela et al., 1985), αVβ3 integrin is now widely accepted and analyzed in its function as a fibronectin (Fn) receptor (Roca-Cusachs et al., 2009, Schiller et al., 2013, Cao et al., 2017, Danen et al., 2002, Balcioglu et al., 2015). However, in these studies αVβ3 integrin was challenged with one ligand at a time. The ECM, in contrast, is more complex and might offer different ligands in close distance to each other potentially changing integrin behavior dependent on the ligand.
To address this issue of substrate heterogeneity, we have recently developed a method to produce microstructured, differential Fn/Vn substrates that enable us to analyze the influence of Fn and Vn coated surfaces on a subcellular scale (Pinon et al., 2014, Rahikainen et al., 2017). Here, we combined this method with super-resolution microscopy and super-resolution live cell imaging. We revealed a force-dependent ligand binding of αVβ3 integrin to Fn, while this integrin binds to Vn already under low mechanical load. Further experiments explain this behavior based on different integrin conformations. Finally, we found that Vn, together with osteopontin (Opn), belongs to a class of high-affinity ligands for αVβ3 integrin contrasting with the low-affinity, force-induced ligands Fn and fibrinogen (Fbg).
Results
Vitronectin is the preferred ligand for αVβ3 integrin
To study how the simultaneous presentation of two ECM ligands influences binding of αVβ3 integrin, we produced binary Fn/Vn substrates with subcellular structural resolution (Pinon et al., 2014). 2×2 μm squares of Fn separated by 1 μm gaps were stamped onto a coverslip and the remaining surface was covered with Vn leading to a geometrical coverage of about equal contribution (Pinon et al., 2014). The quality of substrates was analyzed by fluorescence and atomic force microscopy (Fig. 1a, b). To reveal integrin specific ECM ligand binding, β3-wt integrin GFP was transfected into NIH 3T3 cells expressing low levels of endogenous β3 integrins (Pinon et al., 2014), and cultured on the patterned surfaces for 2 h. In these cells, αV integrin is the only subunit pairing with β3 integrin. Thus, our results obtained with β3-wt integrin GFP are synonymous for αVβ3 integrin. Subsequent fixation and immunostaining for paxillin allowed detecting all integrin-mediated cell-matrix adhesions. Imaging with super-resolution structured illumination microscopy (SR-SIM) revealed extended overlap of αVβ3 integrin and anti-paxillin staining (Fig. 1c). However, some paxillin-clusters on Fn were free of αVβ3 integrin, indicating recruitment to endogenous Fn-bound α5β1 integrin (Supplementary Fig. 1b, g). Therefore, these substrates enabled us to analyze the influence of Fn and Vn on αVβ3 integrin adhesion formation at a micrometer scale within single cells. Surprisingly, we found a clear preference of αVβ3 integrin-positive adhesions for Vn with 83.5 % colocalization (Fig. 1d). This preference for Vn was independent of the size of the αVβ3 integrin-mediated adhesion (Fig. 1e), the presence of α5β1 integrin (Supplementary Fig. 1a, f), or the stamping procedure (Supplementary Fig. 1d, j). Moreover, combining Fn/Vn patterns with hydrogels confirmed this preference of αVβ3 integrin for Vn over a wide stiffness range (Supplementary Fig. 1k-m). Homogenous substrates supported these findings showing about twofold reduced αVβ3 integrin-mediated adhesion area and recruitment into adhesions on Fn compared to Vn (Supplementary Fig. 2).
To study the dynamics of αVβ3 integrin-mediated adhesion formation, we applied SR-SIM live cell imaging. During spreading, cells initiated numerous nascent adhesions (Fig. 1g and Supplementary Video 1 and 2). These adhesions almost exclusively appeared on Vn (Fig. 1f) while some of them were later translocated towards Fn in a centripetal direction towards the cell center (yellow arrows in Fig. 1g and Supplementary Video 2). In control experiments where Fn was replaced by albumin, αVβ3 integrin-mediated adhesions remained restricted to Vn (Supplementary Fig. 1c), indicating that ligand binding was necessary for localization and stabilization of αVβ3 integrins on Fn-coated squares. Next, we asked whether the preference for Vn could be explained by different affinities of αVβ3 integrin for these two ligands. Using biolayer interferometry, we analyzed the dissociation constant of purified αVβ3 integrins for Fn and Vn (Fig. 1h, i). Interestingly, Vn showed measurable dissociation from αVβ3 integrins only in the presence of the αVβ3 integrin inhibitor cilengitide (w/o cilengitide: KD < 1 pM; + 20 μM cilengitide: KD = 302.6 nM; Supplementary Fig. 3i-k), which was similar to previous reports showing non-dissociable binding to Vn (Orlando and Cheresh, 1991). Fn, in contrast, dissociated rapidly (KD = 53.98 nM), proposing that the higher affinity of the extracellular domain of αVβ3 integrin for Vn explains the observed Vn-preference. Accordingly, a chimeric integrin with extracellular β3 and intracellular β1 domains (β3/β1 chimera) showed the same preference for Vn as β3-wt integrin on the micropatterned Fn/Vn surfaces (Supplementary Fig. 1e, h). In addition, we performed atomic force microscopy (AFM)-based single-cell force spectroscopy (Langhe et al., 2016, Dao et al., 2012) on substrates with adjacent Fn and Vn areas. A living cell attached to the AFM cantilever was alternatingly brought in contact with and retracted from two homogenously coated but adjacent Fn and Vn areas. With respect to different contact times (typically ten force cycles per cell and contact time), the detachment force was determined from force-distance curves collected during cell retraction. Starting after 30 s of adhesion time, a significantly higher force was needed to detach cells from Vn compared to Fn substrates (Fig. 1j).
Together, our experiments revealed that initial αVβ3 integrin-mediated adhesions formed exclusively on Vn, potentially explained by the higher binding affinity of αVβ3 integrin to Vn. However, the question remained why the colocalization of αVβ3 integrin-mediated adhesions with Fn was specific for maturing focal adhesions.
Actomyosin contractility regulates the ligand preference of αVβ3 integrin
αVβ3 integrin-mediated adhesions translocated onto Fn squares strictly towards the cell center in the direction of retrograde actin flow. To test whether actomyosin forces are involved in the Fn/Vn selection process of αVβ3 integrin, we reduced actomyosin contractility with blebbistatin or Y27632. The latter compound inhibits the Rho/ROCK pathway and thereby reduces myosin activity, while blebbistatin inhibits myosins directly. As previously shown (Choi et al., 2008), both inhibitors increased the number of small, round nascent adhesions in the cell periphery (Fig. 2a, b). In addition, however, treatment with both inhibitors led to a significant decrease in colocalization of β3-wt GFP integrin with Fn (Fig. 2g and Supplementary Fig. 3a). Culturing β3-wt GFP expressing cells on Fn/BSA or Vn/BSA binary choice substrates confirmed that reduction of contractility affected αVβ3 integrin more on Fn (Supplementary Fig. 3e, f) compared to Vn (Supplementary Fig. 3g, h). Vinculin is well established as an important part of the molecular clutch to transmit forces from actin to the integrin-ECM bond (Humphries et al., 2007, Thievessen et al., 2013, Rahikainen et al., 2017). Therefore, we analyzed the localization of β3-wt GFP integrin on Fn/Vn substrates for mouse embryonic fibroblasts derived from vinculin knockout mice (MEF Vcl -/-). Indeed, the absence of vinculin led to a decrease in Fn binding of αVβ3 GFP integrin (Fig. 2e, h) that was comparable to the results obtained with the contractility inhibitors (Fig. 2g). Control experiments using wt MEFs (Fig. 2d) and Vcl -/- MEFs re-expressing vinculin mCherry (Fig. 2f) demonstrated that the localization of αVβ3 GFP integrin on Fn can be rescued by vinculin expression (Fig. 2h). To increase the mechanical load on integrins, we overexpressed non-muscle myosin IIA mApple (NMIIA) in NIH3T3 cells (Fig. 2c). Adding additional NMIIA caused a significant increase of αVβ3 GFP integrin localization on Fn (Fig. 2g). Taken together, these findings indicate that αVβ3 integrin binds to Vn irrespective of the mechanical load whereas Fn binding is fostered by intracellular force and vinculin.
Hybrid domain swing-out is required for Fn binding of αVβ3 integrin
Next, we asked whether increased activation of αVβ3 integrin might substitute force in the process of Fn binding. Therefore, we employed Mn2+ activation of αVβ3 integrin (Fig. 3a) or established integrin mutations (Fig. 3b-d): (i) Mn2+ treatment increases the affinity of the integrin headpiece for the ligand (Zhu et al., 2013), (ii) the β3-VE mutant increases the affinity for talin 20-fold leading to increased integrin activation (Pinon et al., 2014), (iii) β3-D723A disrupts the inhibitory salt-bridge at the inner membrane clasp between the αV- and β3-subunits (Saltel et al., 2009), and (iv) β3-N305T is reported to cause a constitutive hybrid domain swing-out and slower integrin dynamics (Luo et al., 2003, Cluzel et al., 2005). Surprisingly, on Fn/Vn binary choice substrates, only β3-N305T showed a significant increase of colocalization with Fn (Fig. 3f), while Mn2+ treatment and the intracellular activating mutations (β3-VE and β3-D723A) caused no significant difference. Next, we tested whether the conformational changes caused by the β3-N305T mutation are accompanied by the ability to initiate adhesions on Fn by using live cell SR-SIM (Fig. 3j and Supplementary Video 3, 4) and compared this to the β3-VE mutation (Supplementary Video 5, 6). In both cases, spreading cells initiated most αVβ3 integrin-mediated adhesions on Vn. However, cells expressing β3-N305T were able to initiate numerous adhesions on Fn as well (Fig. 3k).
Additionally, we made a peculiar observation with all activating conditions. They caused central clusters of αVβ3 integrin with irregular shapes compared to classical adhesions (Fig. 3 a-d, zoom-in 2). Interestingly, these clusters localized almost exclusively on Vn (Supplementary Fig. 3o). Analysis of integrin adapter proteins showed talin recruitment but no association of these αVβ3 integrin clusters with paxillin, vinculin, or actin stress fibers (Fig. 3a-d, Supplementary Fig. 3l-n). This indicates that they are not mechanically coupled and thus are under low intracellular force (compare to MEF Vcl -/- cells; Fig. 2e, h). Similar integrin clusters have been reported before to appear within minutes after Mn2+ addition (Cluzel et al., 2005, Saltel et al., 2009). The exclusive localization of these low-force αVβ3 integrins on Vn complements our experiments using contractility inhibitors and vinculin -/- cells and emphasizes the requirement of mechanical load on αVβ3 integrin to bind to Fn.
Combined, these experiments showed that only αVβ3 integrin activation by the N305T mutation increases Fn binding and allows initiating of adhesions on Fn. Surprisingly, other established activating conditions failed to initiate adhesion on Fn (β3-VE; Fig. 3k) and to increase Fn binding (β3-D732A, β3-VE, and Mn2+ treatment; Fig. 3f).
Complete force-dependent hybrid domain swing-out is necessary for Fn binding
The β3-N305T mutation leads to increased ligand-binding (Luo et al., 2003) and reduced integrin dynamics (Cluzel et al., 2005) which is attributed to the creation of a glycosylation site between the βA and the hybrid domain at Asn 303 (N303). This is proposed to cause a constitutive hybrid domain swing-out and thereby full integrin activation. The unique ability of β3-N305T to increase Fn binding (Fig. 3f) motivated us to study this mutation in more detail. Specifically, we set out to dissect the steric effect of N303-glycosylation in β3-N305T from force-induced hybrid domain swing-out (Zhu et al., 2008, Puklin-Faucher et al., 2006). To this end, we reduced actomyosin forces by adding blebbistatin to β3-N305T GFP transfected cells cultured on Fn/Vn substrates (Fig. 3e). β3-N305T mediated adhesions appeared less affected by blebbistatin compared to β3 wt forming many nascent adhesions under this condition (Fig. 2a). However, the colocalization of β3-N305T GFP with Fn was clearly reduced in the absence of force (Fig. 3g). We observed the same effect of reduced Fn-binding for all other activating conditions when combined with a blebbistatin treatment (Supplementary Fig. 3b). Thus, all integrin activating conditions relied on force for Fn binding by αVβ3 integrin. Even constitutive hybrid domain swing-out as reported for the β3-N305T mutation was not sufficient for efficient Fn binding. Seemingly, force caused additional conformational changes needed for Fn binding.
To understand the impact of force and glycosylation on αVβ3 integrin conformation we employed molecular dynamics (MD) simulations for a αVβ3 integrin structure that was glycosylated at N303. Zhu and colleagues published headpiece opening of αIIbβ3 integrin in eight steps (Zhu et al., 2013). We used a Fn-bound structure of αVβ3 integrin (PDB: 4MMX) and arranged a hybrid domain swing-out by superimposition with step seven (PDB: 3ZE1) in the activation cascade described by Zhu and colleagues. This structure was modified by adding a glycosylation at N303 and equilibrated for 100 ns. The same structure without glycosylation at N303 was used as a control. Our MD simulations showed that hybrid domains in both structures swing out to a similar angle with the glycosylated form appearing more stable (Fig. 3h, i). Accordingly, glycosylation at N303 stabilizes αVβ3 integrin in a conformation close to full activation. The final activation step (step 8 according to Zhu et al., 2013, PBD: 3ZE2), in contrast, is characterized by an increased hybrid domain swing-out (“Fully activated 1” in Fig. 3l). Another published structure of active β3 integrin (PBD: 3FCU) confirmed this correlation of full integrin activation and increased hybrid domain swing-out (“Fully activated 2” in Fig. 3l). Thus, comparison of the glycosylated to fully activated structures revealed that N303 glycosylation alone is not sufficient for full hybrid domain swing-out. Combining β3-N305T with Mn2+ slightly decreased Fn-binding (Fig. 3g) as it was the case for adding Mn2+ to β3-wt (Fig. 3f) potentially indicating that the effect of Mn2+ on αVβ3 integrin conformation is limited.
We conclude that our MD simulations, performed without mechanical load on the integrin, reflected the structure of β3-N305T integrin in experiments with contractility inhibition (Fig 3e, g). Thus, experiments and simulation imply that the final activation step with maximal hybrid domain swing-out stabilizes Fn binding and that this step is only achieved for αVβ3 integrin under mechanical load.
Vn binding is initiated for αVβ3 integrin in the extended-closed conformation
Experiments and MD simulations indicated that stable Fn-binding by αVβ3 integrin requires force-dependent hybrid domain swing-out. In contrast, αVβ3 integrin was able to bind Vn in experiments where cell contractility was reduced. Accordingly, it is tempting to speculate that αVβ3 integrin can bind Vn already in the extended-closed conformation. To test this, we set out to develop an integrin mutation that locks αVβ3 integrin in the extended-closed conformation. Therefore, we created a disulfide bridge between the βA and the hybrid domain to limit the degree of the hybrid domain swing-out (β3-V80C/D241C). Structural analysis supported our rationale for this mutation (Fig. 4a). We prepared a model, where cysteine mutations were introduced into extended-closed conformation of αVβ3 integrin (PDB: 4MMX) using PyMOL and energy minimization of the model. The disulphide bridge caused only minimal distortion for the protein; the distance between Cα atoms of V80C and D241C after introducing a disulphide bond did not change compared to the wildtype situation (both structures: d = 6.3 Ä). In contrast, αVβ3 integrin in the extended-open conformation showed an increased distance by a factor of three (d = 19.4 Ä) between Cα atoms of V80 and D241, indicating that a V80C-D241C disulphide bridge can block the transition to the extended-open conformation. The structure of the extended-open form was based on the closed integrin and opening was prepared by superimposition of βA and hybrid domains with the crystal structure of open αIIbβ3 integrin (PDB: 3FCU).
Next, we expressed this β3-V80C/D241C GFP in NIH 3T3 cells and cultured them on homogenous Fn or Vn (Fig. 4b, c). The GFP signal revealed clustering of β3-V80C/D241C into cell adhesions on Vn but no clustering at all on Fn. Opening disulfide bridges with dithiothreitol (DTT) allowed clustering into adhesions similar to β3-wt (Supplementary Fig. 4a, b, g). Additionally, incubating cells with Mn2+ failed to induce clustering of β3-V80C/D241C on Fn while it increased clustering on Vn (Fig. 4d, e). On Fn/Vn substrates, β3-V80C/D241C GFP showed also very dim adhesions that we could not quantify reliably due to high background fluoresence in the plasma membrane (Fig. 4f). However, a restriction of β3-V80C/D241C GFP localization on Vn was obvious. Incubating β3-V80C/D241C GFP expressing cells with Mn2+ increased the clustering of the mutated αVβ3 integrin while preserving the restriction to Vn (Fig. 4g, h). The increased fluorescent signal in focal adhesions compared to the plasma membrane in this experiment allowed a reliable analysis of β3-V80C/D241C GFP localization. The quantification revealed that the β3-V80C/D241C mutation has clear defects in Fn-compared to Vn-binding while our structural analysis confirmed that this mutation locks αVβ3 integrin in the extended-closed conformation. Additionally, adding the integrin inactivator Ca2+ to β3-wt GFP expressing cells reduced Fn-binding of αVβ3 integrin (Supplementary Fig. 3c, d) to a similar extent as contractility inhibition or the β3-V80C/D241C mutant. Therefore, we propose that αVβ3 integrin binds Vn initially in the extended-closed conformation, whereas αVβ3 integrin binds Fn only in the extended-open conformation.
We also compared the fluorescence recovery after photobleaching (FRAP) of β3-wt, β3-N305T, and β3-V80C/D241C, in order to understand the influence of these mutations on αVβ3 integrin turnover in adhesions (Fig. 4i, j). Interestingly, β3-V80C/D241C GFP showed very fast turnover, while β3-N305T GFP showed slow integrin exchange compared to β3-wt GFP as reported previously (Cluzel et al., 2005). Combined with our experiments on Fn/Vn substrates, this revealed that Fn binding of αVβ3 integrin correlates with reduced integrin turnover on Vn substrates. Control experiments with overexpression of FAK wt vs. FAK Y397F confirmed this correlation: Despite proper focal adhesion recruitment, FAK Y397F fails to recruit Src kinases causing increased focal adhesion size and reduced adhesion dynamics compared to FAK wt (Swaminathan et al., 2016, Hamadi et al., 2005, Schaller et al., 1994). At the same time, cells cultured on Fn/Vn substrates revealed increased colocalization of αVβ3 integrin with Fn when overexpressing FAK Y397F compared to FAK wt (Supplementary Fig. 4e, f, i). Thus, regulation of the exchange dynamics of αVβ3 integrin and αVβ3 integrin-mediated adhesions, emerged as an additional option to regulate Fn-binding of αVβ3 integrin.
MD simulation supported conformation-dependent Fn binding
The experiments with β3-N305T emphasized the influence of the extended-open conformation for Fn binding of αVβ3 integrin. To better understand the influence of different integrin conformations on Fn and Vn binding at a single protein level, we employed MD simulations. Because no suitable structural data for Vn was available, we used a CRGDC peptide as a proxy for a high-affinity ligand of αVβ3 integrin. We investigated the binding of the RGD-containing 10th domain of Fn (FnIII10; Supplementary Fig. 5a) and of the CRGDC peptide (Supplementary Fig. 5c) to extended conformations of αVβ3 integrin exhibiting either the open or closed headpiece. Interaction energies between αVβ3 integrin and the CRGDC peptide were stable and similar over the simulation time for the extended-closed integrin (Supplementary Fig. 5f); the extended-open conformation showed similar but less stable interaction energies (Supplementary Fig. 5g). In contrast, interaction energies for FnIII10 showed stable interaction with αVβ3 integrin only for the extended-open conformation (Supplementary Fig. 5b). Thus, these results support the model where FnIII10 - αVβ3 integrin binding is less stable in the extended-closed conformation, and becomes more stable once the integrin is in the extended-open form upon full hybrid domain swing-out. Binding of CRGDC, in contrast, appeared stable already for the extended-closed conformation of αVβ3 integrin. Additionally, structural analysis indicated that full-length Fn, compared to FnIII10, might show even higher preference for the extended-open conformation due to lower sterical hindrance compared to the extended-closed conformation (Supplementary Fig. 5d, e).
Preference for Vn influences cell migration and mechanotransduction
Our results so far revealed a mechanism enabling αVβ3 integrin to differentiate between Fn and Vn based on the degree of the force-dependent hybrid domain swing-out. However, stable binding to both ligands will most likely result in the fully active extended-open conformation of αVβ3 integrin irrespective of the actual ligand present. Therefore, it is possible that the preference of αVβ3 integrin for Vn compared to Fn is compensated on a cellular level when only one ligand is present. Thus, we performed additional experiments to test this hypothesis. First, we analyzed cell migration of GD25 cells (no β1 integrin expression and therefore relying on αVβ3 integrin only) on cover slips coated with 10 μg/ml Fn or Vn with live cell imaging. Cell tracking revealed that cells on Fn migrated almost two times faster (vFn = 12.0 ± 3.08 μm/h) compared to cells migrating on Vn (vVn = 6.7 ± 0.39 μm/h; Supplementary Video 7). To understand how cell migration is influenced when GD25 cells can choose between Vn and Fn, we produced a Vn/Fn stripe assay (Vn: 20 μm width; Fn: 40 μm width). Live cell imaging over 12 h on these Fn/Vn stripe patterns revealed a consistent movement of cells from Fn towards Vn (Fig. 5a and Supplementary Video 8). To quantify this behavior, we measured the area of single cells covering Fn coated surfaces for cell cultures after different time points (Fig. 5b). Half an hour after seeding, cells covered Fn and Vn coated surfaces according to the geometrical coverage of the pattern indicating random distribution (1/3 Vn, 2/3 Fn, cell/Fn colocalization: 67.5%). With increasing time, more cells adhered to Vn (cell/Fn colocalization after 8 h: 28.4%; 24 h:14.6%).
Next, we investigated cellular behavior that have been described to be regulated by force-dependent mechanisms. First, we analyzed the subcellular localization of αVβ3 integrin. For cells cultured on Fn, it is well known that αVβ3 integrin dominates in peripheral, high-contractile focal adhesions whereas α5β1 integrin establishes central, low-contractile fibrillar adhesions (Zamir et al., 2000). Based on our data, it is straightforward to propose that this integrin distribution is caused by an interplay of extracellular ligands and mechanical load on the integrin: αVβ3 integrin fails to bind Fn in the cell center due to the low force levels in fibrillar adhesions. On Vn, in contrast, we expected that low mechanical load on αVβ3 integrin in the cell center is still compatible with Vn binding. We tested this by culturing β3-wt GFP transfected NIH 3T3 cells on homogenous Fn or Vn substrates for 6 h and analyzed the subcellular distribution of αVβ3 integrin. We frequently observed central αVβ3 integrin adhesions for cells cultured on Vn but rarely on Fn (Fig. 5c, d). In addition, we analyzed the β3-N305T mutation in this assay. The almost complete extended-open conformation of this mutation might allow αVβ3 integrin to bind Fn in the cell center despite the reduced mechanical load compared to the cell periphery. Indeed, β3-N305T GFP expressed in cells cultured on Fn now localized in central adhesions (Fig. 5c, d).
In a second set of experiments, we asked whether mechanotransduction of a cell is affected by the force-dependent ligand binding of αVβ3 integrin. We used again GD25 cells, seeded them on Fn or Vn coated hydrogels of different stiffnesses, and analyzed cell area and adhesion maturation as indicators of cellular mechanotransduction. Both ligands caused a similar sigmoidal increase of cell area and adhesion length with increasing gel stiffness (Fig. 5f, g). Cells on Vn showed adhesion maturation and increased cell spreading already at 6.7 kPa. Cells on Fn reached similar plateau values for both parameters, however, at substrates stiffer than 6.7 kPa (Fig. 5e).
To summarize, we observed that cellular behavior is regulated by the extracellular ligand of αVβ3 integrin. Cell migration experiments indicated that ligand preferences of αVβ3 integrin guide cells towards the high-affinity ligand Vn. More importantly, mechanotransduction is differently affected on Vn and Fn implying that force-dependent ligand-binding by αVβ3 integrin is not compensated on a cellular level.
Ligands of αVβ3 integrin belong to two different affinity regimes
Finally, we asked whether our findings for the interaction of αVβ3 integrin with Fn and Vn are also valid for other ligand combinations. We performed additional affinity measurements of αVβ3 integrin and its ligand fibrinogen (Fbg) revealing slightly higher affinity compared to Fn (Fbg: KD = 8.15 ± 4.82 nM; Supplementary Fig. 3i, k). In addition, we produced binary choice substrates (Fig. 6g) to challenge αVβ3 integrin with Vn and Fbg (Fig. 6a), with Vn and osteopontin (Opn; Fig. 6b), or with Vn and thrombospondin (Tsp, Fig. 6c). Interestingly, αVβ3 integrin colocalized with Fbg (Fig. 6d) to a similar amount as it did with Fn on Fn/Vn substrates (colocalization to Fn: 16.5% on Fn/Vn, colocalization to Fbg 20.2% on Vn/Fbg). Opn caused a very different distribution with almost equal localization of αVβ3 integrin on both ligands (Fig. 6e, colocalization to Opn 50.4% on Vn/Opn). On Vn/Tsp, in contrast, αVβ3 integrin localized almost only on Vn (Fig. 6f, colocalization to Tsp 6.7% on Vn/Tsp) indicating that Tsp is no proper ligand for αVβ3 integrin in this context. Thus, Vn and Opn may present a class of high-affinity ligands for αVβ3 integrin compared to the low-affinity ligands Fn and Fbg.
Additionally, we asked whether the closely related αVβ5 integrin also shows a force-dependent binding to Fn. We transfected NIH 3T3 cells with β5-wt GFP and cultured these cells on Fn/Vn substrates. Comparing β5-wt GFP with anti-paxillin staining revealed two classes of adhesions. The majority of αVβ5 integrin clustered in the cell center without recruiting paxillin (Supplementary Fig. 4c). Only a subset of αVβ5 integrin colocalized with paxillin in the cell periphery. We quantified the colocalization of these peripheral αVβ5 integrin adhesions with Fn for control conditions and blebbistatin treated cells (Supplementary Fig. 4c, d, h). These measurements also showed a force-dependent binding of αVβ5 integrin to Fn (αVβ5 integrin + DMSO: (36.5 ± 13.47)%; αVβ5 integrin + 5 μM blebbistatin: (16.1 ± 4.84)%). Based on these findings it appears that force-dependent ligand binding is a general mechanism regulating the interaction of integrins with ECM-ligands.
Discussion
Here we have analyzed the interaction of αVβ3 integrin with different ligands on sub-cellular binary choice substrates. We found that αVβ3 integrin prefers binding to Vn under a wide range of conditions. Binding to Fn, in contrast, required mechanical load on the integrin. Combining experiments with integrin mutations and MD simulations revealed that this differential ligand binding is coupled to different integrin conformations (Fig. 6h). The extended-closed conformation of αVβ3 integrin binds to Vn but not efficiently to Fn. Only the extended-open conformation due to a force-mediated complete hybrid domain swing-out enabled αVβ3 integrin to bind Fn. Thus, force-mediated conformational changes regulate the ligand promiscuity of αVβ3 integrin. We demonstrated that these findings have consequences for cellular behavior during migration and mechanotransduction.
Fn vs. Vn preference of αVβ3 integrin
The preference of αVβ3 integrin for Vn compared to Fn is apparent in all our experiments. This might appear trivial given the measured in vitro affinities of αVβ3 integrin for Vn compared to Fn. However, this explanation fails to explain in cellula results like the exclusive initiation of αVβ3 integrin-mediated adhesions on Vn while αVβ3 integrin binds Fn in maturing focal adhesions. The reduced Fn-binding after contractility inhibition also points to a more complex regulation of αVβ3 integrin ligand-binding. We interpret these results as a consequence of force-dependent increase in Fn binding of αVβ3 integrin. We had no access to a single-protein method that would have allowed us to measure force-dependent affinity changes for single αVβ3 integrins. Instead, our experimental results are based on ensemble measurements revealing equilibrium changes of a population of integrins. Accordingly, a lack of adhesion initiation by αVβ3 integrin on Fn is still compatible with transient binding of αVβ3 integrins to Fn on spatial and temporal scales below the resolution of the methods we used. Thus, our interpretation of ligand-αVβ3 integrin interactions are based on the binding of many integrins over seconds to their ligands and not on affinity measurements of single integrins and ligands. Yet, we think that a force-dependent change in affinity of single αVβ3 integrins is the most plausible explanation for our results. Avidity instead of affinity changes or differential recruitment of adapter proteins could be possible alternative explanations. In fact, we showed that vinculin recruitment is needed for increased Fn-binding of αVβ3 integrin. However, to date vinculin is best characterized as a transmitter of force from actin to the talin-integrin axis (Humphries et al., 2007, Rahikainen et al., 2017, Elosegui-Artola et al., 2016). Therefore, we conclude that the effect of vinculin in our experiments is best explained by its role as a force-transmitter. Additionally, recruitment of adhesome proteins to focal adhesions seemed rather unaffected by a vinculin knockout (Thievessen et al., 2013). Avidity regulation on the other hand, is also not easily compatible with our data. Nascent adhesions are the smallest clusters of αVβ3 integrin we could resolve. Single localization techniques estimated 40-50 integrins in such clusters (Shroff et al., 2008, Changede et al., 2015). Therefore, the differential ligand binding of αVβ3 integrin that we observed during adhesion maturation occurred on a range of 40 up to probably hundreds of integrins in focal adhesions. In contrast, affinity vs. avidity regulation is rather discussed for a range of one vs. ten integrins (Abrams et al., 1994). These pitfalls of alternative explanations let us favor a hypothesis of force-regulated differential affinity change of αVβ3 integrin (Fig. 6h). More importantly, our data obtained with different αVβ3 integrin mutations, structural analysis, and comparisons with MD simulations offer a mechanistic explanation supporting this hypothesis.
The active αVβ3 integrin for binding Vn is not the active integrin for binding Fn
Live cell imaging and contractility inhibition indicated that retrograde actin flow and cellular contractility might influence the ligand binding of αVβ3 integrin. In fact, MD simulations predicted that mechanical load on an integrin, imitating retrograde actin flow, should favor the hybrid domain swing-out and therefore the extended-open conformation (Zhu et al., 2008, Puklin-Faucher et al., 2006). Zhu and colleagues concluded that mechanical forces activate integrins (Zhu et al., 2008). Integrin activation can be defined on a structural level and on a functional level. The latter definition is straightforward with ligand binding indicating an active integrin while the inactive integrin fails to bind a ligand. For the structural definition, it is probably not controversial to define a bent integrin as the conformation of the inactive and the extended conformation as the active integrin (Campbell and Humphries, 2011). These structures are also described as low-affinity (bent integrin) and high-affinity (extended integrin) conformations indicating the probability of ligand binding (Kong et al., 2009, Zhu et al., 2008). In this way, the link between structural and functional definition can be established.
However, some integrins show extended conformations that can be subdivided into extended-closed and extended-open. This is, for example, the case for αVβ3 integrin. Whether this is also the case for α5β1 integrin is still debated (Su et al., 2016, Miyazaki et al., 2018). These subclasses of αVβ3 integrin make it less clear whether extended-closed or extended-open αVβ3 integrin should be considered as the active conformation. Our data offered a surprising answer to this question by showing that the active αVβ3 integrin (functional definition) has different conformations for different ligands. Fn binding only to the extended-open αVβ3 integrin defines this conformation as the active integrin structure. In contrast, Vn binds already to the extended-closed conformation rendering the inactive conformation for Fn-binding to the active conformation for Vn-binding. This rather complex and situation-dependent structure-function relationship of αVβ3 integrin might also be relevant for the development of high-affinity inhibitors that should bind without activating the integrin. For such a purpose, it will also be necessary to better understand the molecular detail how ligands bind, or do not bind, to the binding pocket for different conformations of αVβ3 integrin. Our data indicate that the influence of mechanical forces at the binding pocket can alter the affinity in a ligand specific manner. An interesting aspect of this issue was recently postulated by Cormier and colleagues who argued that integrin activation could also be about improved accessibility of the RGD peptide of a ligand to the binding pocket in the αVβ3 integrin headpiece (Cormier et al., 2018). Interestingly, Fn presents its RGD peptide in a rather short loop implying that accessibility to the αVβ3 integrin binding pocket could indeed be a limiting step for this ligand.
Such a scenario would also explain the limited effects of Mn2+ treatment on Fn-binding by αVβ3 integrin assuming that Mn2+ increases the affinity for the RGD peptide but does not change the accessibility of Fn to the αVβ3 integrin binding pocket (Fig. 6h). Also, it was reported that Mn2+ in absence of a ligand failed to cause a conformational shift to active conformations for the closely related αIIbβ3 integrin (Zhu et al., 2013).
The vitronectin receptor and the cell
Cell migration relies on stable adhesion to the substrate involving an extension phase and generation of new adhesions underneath lamellipodia or filopodia while subsequent actomyosin forces pull on these anchors to push the cell body forward compared to the substrate (Giannone et al., 2007). Therefore, adhesions have to switch from a short lived exploratory mode (nascent adhesions) to a more stable sessile mode (focal adhesions) (Choi et al., 2008). αVβ3 integrin (Elosegui-Artola et al., 2016) and α5β1 integrin (Kong et al., 2009, Friedland et al., 2009) support this process by increasing the lifetime of their bond to Fn when force is applied (i.e. a catch-bond). However, adhesions also have to lose their contact to the substrate to allow the cell body to move forward. This can occur by weakening of adhesions during translocation to the cell center (Zamir et al., 2000, Sun et al., 2016) and by disassembly of adhesions (Wehrle-Haller, 2012). Our Fn/Vn stripe assay (Fig. 5a) and our analysis of subcellular localization of αVβ3 integrin revealed that the respective ligand of an integrin can be decisive for adhesion organization and cell migration. Faster off-rates of the αVβ3 integrin-Fn bond (Fig. 1h) allow a faster cycle of binding and unbinding compared to Vn. Additionally, αVβ3 integrin builds less adhesions in total (Supplementary Fig. 2e) and specifically lower numbers of central adhesions on Fn (Fig. 5d). This is accompanied by less adhesive forces on Fn compared to Vn (Fig. 1j). In total, this leads to a more exploratory, faster migrating cell on Fn (Supplementary Video 7, 8). On Vn, in contrast, αVβ3 integrin causes higher adhesive forces, more adhesion area per cell, and sticks to Vn even under low mechanical load explaining why cells moved from Fn away towards Vn. This raises the interesting option that cell migration towards Vn from the blood plasma contributes to wound healing by attracting αVβ3 integrin-expressing cells to a site of coagulated serum proteins. We believe it might also be relevant for physiological settings that the Fn-binding capacity of αVβ3 integrin is not only regulated by forces but also by adhesion dynamics (Fig. 6h). Therefore, an αVβ3 integrin expressing cell would not be at the mercy of the surrounding tissue stiffness for binding, or not binding, to Fn. Instead, adhesion dynamics, regulated by multiple signaling cascades, offer an opportunity to tune the binding of αVβ3 integrin to Fn and to integrate mechanical and biochemical signals. On the other hand, assuming that the bond between a soluble ligand and an integrin experiences low mechanical load, our results can also explain the inability of αVβ3 integrin to bind soluble Fn (Huveneers et al., 2008) avoiding the clogging of surface exposed αVβ3 integrin by Fn from the blood plasma. Experiments with cells cultured on hydrogels of different stiffnesses revealed a mechanotransduction on a sigmoidal basis, i.e. an on-off mechanoswitch, as reported by others (Elosegui-Artola et al., 2016). Interestingly, GD25 cells on Vn coated gels exhibited increased αVβ3 integrin-mediated adhesion length at an intermediate gel stiffness (E = 6.73 kPa) where cells cultured on Fn coated gels still failed to establish mature adhesions. We showed that αVβ3 integrin establishes bonds to Vn, but not to Fn, under low-contractility conditions. Thus, assuming that softer gels lead to lower mechanical load on αVβ3 integrin, our findings are able to explain the differential mechanotransduction of cells on two different ligands.
In total, we showed that the same integrin, αVβ3, can behave differently and cause different cellular phenotypes dependent on the ligand it binds to. On the other side, intra- and extracellular parameters can affect the ability of αVβ3 integrin to select between different ligands. Interestingly, αVβ5 integrin was recently shown to contribute to non-canonical integrin functions either in so called reticular adhesions or by associating with clathrin containing structures (Zuidema et al., 2018, Lock et al., 2018, Baschieri et al., 2018). The switch between these structures and focal adhesions is regulated by different intra- and extracellular parameters including actomyosin-contractility. Moreover, a recent report showed that α4β7 integrin selects between VCAM-1 and MAdCAM-1 dependent on the integrin conformation induced by different chemokines (Wang et al., 2018). This indicates that at least some integrins can change their ligand selectivity and/or their function based on their conformation regulated by environmental factors. This proposes an additional, emerging way to control integrin behavior.
Author contributions
M. Bachmann, B. Wehrle-Haller, and M. Bastmeyer conceived the study. M. Bachmann, M. Schäfer, V. Hytönen, B. Wehrle-Haller, and M. Bastmeyer designed the experiments, M. Bachmann, M. Schäfer, V. Mykuliak, M. Ripamonti, L. Heiser, K. Weißenbruch, S. Krübel, and C. M. Franz performed and analyzed the experiments. M. Bachmann, M. Schäfer, B. Wehrle-Haller, and M. Bastmeyer wrote the paper. All authors discussed the results and implications and commented the manuscript drafts.
Acknowledgements
We thank Marc Hippler for help with atomic force microscopy measurements, Melanie Merkel for help with the analysis of de novo clusters, Deepthy Kavungal for help with cell migration assays, and the Bioimaging core facility at CMU, University of Geneva. This work was supported by the German Research Foundation (Karlsruhe School of Optics and Photonics (KSOP) to M. Bachmann) and by the Baden-Württemberg Stiftung (State Scholarship to M. Schäfer). We thank Academy of Finland for financial support (grant no. 290506 to V. Hytönen) and acknowledge CSC – IT center for science for computational resources and EDUFI (former CIMO) for postdoctoral fellowship for V. Mykuliak. Swiss National Science Foundation supported the work of M. Ripamonti and B. Wehrle-Haller (31003A_166384). The authors declare no competing financial interests.