αV-class integrins exert dual roles on α5β1 integrins to strengthen adhesion to fibronectin

Upon binding to the extracellular matrix protein, fibronectin, αV-class and α5β1 integrins trigger the recruitment of large protein assemblies and strengthen cell adhesion. Both integrin classes have been functionally specified, however their specific roles in immediate phases of cell attachment remain uncharacterized. Here, we quantify the adhesion of αV-class and/or α5β1 integrins expressing fibroblasts initiating attachment to fibronectin (≤120 s) by single-cell force spectroscopy. Our data reveals that αV-class integrins outcompete α5β1 integrins. Once engaged, αV-class integrins signal to α5β1 integrins to establish additional adhesion sites to fibronectin, away from those formed by αV-class integrins. This crosstalk, which strengthens cell adhesion, induces α5β1 integrin clustering by RhoA/ROCK/myosin-II and Arp2/3-mediated signalling, whereas overall cell adhesion depends on formins. The dual role of both fibronectin-binding integrin classes commencing with an initial competition followed by a cooperative crosstalk appears to be a basic cellular mechanism in assembling focal adhesions to the extracellular matrix.

The maximum adhesion force is the key parameter considered here, which makes sense. Yet, the authors may want to clarify whether changes in the shape of the force profiles were observed from on condition to another, i.e. did all the curves look qualitatively like supplem fig 1a? Could it be interesting to report the work of adhesion to take into account the whole curve? Also, little is said about the adhesion probability. Could it be an interesting parameter to look at when changing conditions?
The clustering section is an important one. The TIRF data should be better presented and discussed for a broad audience, not necessarily expert in the technique. In particular, how can we proof clustering from the data, please explain in detail. Do we have an idea of the size of the clusters? The authors may want to discuss the limitations of TIRF and the potential, for future research, of higher resolution techniques (superresolution, single-molecule AFM imaging) to analyze nanoscale clusters.
Reviewer #2, an expert in integrins and biophysics (Remarks to the Author): In this work, Bhadradwaj et al. study the interplay between avb3 and a5b1 integrins in determining cell adhesion strength to fibronectin. The authors use an elegant single cell force spectroscopy setup in combination with a well controlled cell system of selective integrin expression. With this setup, they determine that whereas a5b1 integrins are more effective at withstanding forces, avb3 integrins have higher binding rates to fibronectin. This leads them to outcompete a5b1 integrins for fibronectin binding, and to reduce overall resistance to force. By using different approaches to interfere with avb3 localization and function, the authors further demonstrate the relevance of the interplay between the two integrins. The results are interesting and novel, and are carefully designed and controlled. Further, they add clarity and an explanation to previous work that had shown sometimes apparently contradicting results (see for instance Schiller et al. NCB 2013versus Balcioglu et al. JCS 2015. However, some important issues should be addressed before publication. 1. My main concern is with figures 4 and s7, and their interpretation. First, the authors should show the time course of the evolution of fluorescence, and not only one time point (which I assume corresponds to the end of the experiment). Second, they should provide a statistical analysis to compare the different conditions. Finally and most importantly, I don't understand the interpretation of the authors. The results show that paxillin is recruited only in the case of VNcoated cantilevers, which is interpreted to mean that VN binding of alpha-v class integrins induced a5b1 clustering. However, from my understanding, in the previous figures VN coating of cantilevers is used precisely to recruit alphav integrins away from the substrate, eliminating the competition with a5b1 integrins. Thus, the results seem more consistent with avb3 impairing the formation of paxillin clusters, not inducing it as the authors claim (at least for the very initial stages analyzed). To address this, the author should repeat the experiment after labelling fluorescently not paxillin, but avb3/a5b1 integrins. I know that this would likely alter the respective concentrations of integrins, but it would still allow to observe where and to what extent the different integrins localize as a function of the coating both of the substrate and of the cantilever. Carrying out not only tirf but also epifluorescence imaging would be useful to see what is happening at the cantilever/cell interface.
2. Relatedly, the abstract states that "once engaged, av class integrins activate a5b1 integrins to establish additional adhesion sites to fibronectin, dislocated from those formed by av class integrins". This is not direct evidence shown in this work, but rather a proposed mechanism based on previous literature. The way it is written, it seems as if those were results presented here. This should be corrected. 3. It is unclear why the effects of coating tips with VN or CiL are so different ( fig. 2). I agree with the authors that the different binding properties (functional binding in VN, but mere inhibition in CiL) may play a role. However, the main role of VN and CiL in this case is to sequester integrins away from the substrate, and in principle one would expect that potential differences in binding properties would apply to the site of binding (the cantilever) and not so much to the substrate, which is the one that determines the adhesion measurements. Why would the lack of signaling (in the case of CiL) reduce the effect of integrin sequestration to the cantilever with respect to VN? 4. The results of figure 3 are very interesting but somewhat confusing. For instance, it is very intriguing that some contractility inhibitors (the Y compound or blebbistatin) inhibit the effect of VN cantilever coating, but some others (ML-7) do not. In this respect, carrying out fluorescence experiments such as those in figure 4, by tagging fluorescently avb3 and a5b1 integrins, after the different inhibitions would be very useful. This would allow to see whether the different inhibitions alter the differential recruitment of the two integrin types at the substrate versus the cantilever, for instance. I realize that carrying out such experiments for all the conditions tested would represent an enormous amount of work, but doing it for the Y compound and ML7 for instance would add very valuable information on the differential effect of the two drugs.
Minor: 5. Given the rich previous literature on the topic of a5b1 versus avb3 integrins, a short discussion of how this work fits in with or reinterprets previous data would be useful. This is already done for some publications, but adding for instance Balcioglu et al. , jcs 2015, or some of the work by the spatz/cavalcanti groups would be useful.
6. An effect of differential on/off rates between alphav and a5b1 integrins was already described previously (Elosegui-Artola et al., nat. mater. 2014). This previous work does not affect the novelty of this submission since it involved different measurements and a different av integrin, but it should still be mentioned.
Reviewer #3, an expert in integrin crosstalk (Remarks to the Author): This manuscript describes a detailed analysis using single-cell force spectroscopy, of the differential interaction of aV-containing integrins, typically binding extracellular ligands such as vitroectin (VN), osteopontin and weaker binding to fibronectin (FN), with the classical FN-receptor a5b1. The experiments are very well controlled and use genetically engineered cells and extracellular ligands, to exclude any contamination with the other type of integrin receptors, or contaminations of ligand preparations. The result is an impressive study showing that av-integrins have the capacity to inhibit the recruitment and adhesion reinforcement of a5b1-integrins on the cell-binding fragment of FN. Although this rather surprising observation is not entirely new (Pinon et al, 2014, not cited!), new elements are added to explain the cross-talk between the aV and a5b1 integrin receptors. Notably, the authors show that the inhibition by aV-integrins on a5b1mediated FN binding can be prevented by physically separating the ligands (not surprising), but that this separation is creating a positive intracellular feedback, involving classical integrinsignaling pathways that enables efficient b1-reinforcement after previous binding and signaling of aV-integrins on VN. Although the work is interesting for specialist in the domain of integrins, the study is not really revealing the mechanistic basis for the observed competition and cross-talk between aV-and b1-integrins (while aV binds weaker to FN, it has a faster on rate than b1integrins). Therefore I suggest a couple of issues that should be evaluated: 1. Unfortunately a key question remains unanswered. What is the qualitative difference, between aV integrin versus a5b1 integrin binding, in respect to adapter recruitment, signaling, and differential adhesion to VN versus FN. Maybe avb3, originally termed the VN-receptor, is never intended to bind to FN, but competes with endogenous a5b1 because there is no better extracellular binding partner, implying that the observed difference is due to the extracellular ligand binding domain, and faster on-rate of aV-integrins. However an alternative explanation is that the observed competition between aV and a5b1 integrins is not due to the extracellular domain, but linked to differences at the level of the cytoplasmic domain. Is not talin a better binder for b3 than b1 (Anthis et al., 2010)? For example Pinon et al, (2014) has shown that avb3 can cluster on FN after 1 hour of spreading, preventing a5b1-integrin recruitment, very similar to what is shown here in this manuscript. However, when b3 is absent or mutated in its talin-binding motif, a5b1-is readily engaged on FN to induce adhesion and spreading. Thus, despite a normal extracellular domain, such a aVb3-integrin with a cytoplasmic talin-binding mutation, can no longer compete with a5b1 integrins. This can lead to two conclusions: (1) talin is required for aVb3-activation and binding to FN, or (2) b1-integrin is normally prevented from rapid binding because it has a lower affinity for talin, or because its cytoplasmic domain is interacting with inhibitory adapter proteins. It would therefore be important to test how the deletion of b1-integrin specific inhibitory adapters, such as ICAP-1 or filamin would affect the recruitment dynamic or the here called cross-talk between the analyzed integrins. A plausible explanation, consistent with the observed effects, could involve the signaling-dependent inactivation (e.g. by phosphorylation) of b1-specific inhibitors and thus rapid a5b1 activation and spreading as for example seen in ICAP-1 deleted fibroblasts (Millon-Fremillion).
2. Reinforcement of binding in ConA/b1-cells versus VN/av-b1-cells. In figure 1a (second column), the reinforcement of binding on FN is seen for cells expressing only b1-integrins. However, this type of binding is never compared with FN binding in av-b1 cells when bound to a ConA/VN cantilever. Is binding significantly enhanced in the latter condition? If so, this could explain the enhanced recruitment of paxillin for the latter but not former condition (sup fig 7). Unfortunately paxillin is not a marker for integrin clustering, but rather integrin signaling. For clustering of b1integrins, 9EG7 staining should be used. Enhanced paxillin recruitment is a late event (observed only at 100s), does it correlate with adhesion maturation over time? How is the adhesion curve evolving over the time range visualized for paxillin recruitment (e.g. 5 min), comparing a b1-integrin only from a VN-mediated b1integrin-dependent adhesion?
3. There is no pure VN control for the mixture of VN with ConA: Normally cell spreading is analyzed on glass coated with purified VN. Mixing VN with ConA as adhesive support used on the cantilever, could create a signaling effect that is not induced with a pure VN coating. What happens to cells when they are plated on a ConA/Vn mixture compared to a VN/BSA mixture or pure VN coated surface. If ConA is recruiting a co-signaling receptor to the engaged aV-integrins on VN, this could induce an "integrin-independent" signaling cross-talk. In addition, labeling the figure with VN is misleading VN/ConA would be more appropriate. How do VN-only cantilever work for this assay. 4. How can the effect of the inhibitors be separated from the integrin signaling occurring at the cantilever versus the integrin signaling at the adhesion site? One solution is demonstrated by Pinon et al., in which the substrates molecules VN and FN are physically separated, but contacted by normal or mutant forms of the integrin receptors. Alternatively the authors could use signaling molecules, such as growth factors, to create an integrin-like effect in the target cell, or remove b1integrin inhibitors, or swap cytoplasmic domains, etc...

Minor issues:
-Please explain which blocking antibody was used for beta1, is it changing the conformation of the integrin.
-Line 249: does this cross-talk require cantilever pulling and force transmission on the aVintegrins? Is it possible to visualize the surface of the cantilever for integrin signaling? Does varying the pulling speed affect the integrin signaling at the level of the cantilever.
- Figure 5: I am not aware of a bent a5b1 structure (green), similar to the one's described for a2bb3 and avb3. In fact, it is likely that avb3 and a5b1 are regulated entirely different by cytoplasmic adapters, which would affect their ability to cross-talk.
-Check legend to sup figure 5: a and b appear to be exchanged.

Point-by-point response to the comments of reviewer #1
Reviewer #1: Understanding how cells adhere to extracellular matrix proteins is a very important, yet not fully understood topic in current cell biology. In particular, we know little about whether α5β1 and αV-class fibronectin-binding integrins function individually and/or cooperate, and how they strengthen adhesion. In this groundbreaking work, state of the art atomic force microscopy (AFM) is used to demonstrate, for the first time, a dual role of the two integrin classes upon interaction with fibronectin. In addition, combining AFM with total internal reflection fluorescence (TIRF) microscopy reveals that protein crosstalk triggers the clustering of α5β1 integrins and recruitment of adhesome proteins. These findings are of great biological significance and may represent a generic mechanism leading to the assembly of focal adhesions.
The work is novel and important, the methods state of the art, data very solid with appropriate controls and statistics, and the conclusions very well-supported by the results. No doubt that the study will appeal to a very broad audience. Overall, I strongly support publication after some minor revisions.
Authors: Thank you for your encouraging and constructive comments. Below, we describe point-by-point how we addressed each specific comment of the reviewer.

Reviewer #1:
The maximum adhesion force is the key parameter considered here, which makes sense. Yet, the authors may want to clarify whether changes in the shape of the force profiles were observed from on condition to another, i.e. did all the curves look qualitatively like supplem fig 1a? Could it be interesting to report the work of adhesion to take into account the whole curve?
Authors: The reviewer asks to clarify whether we have observed changes in the shape of the force profiles from one condition to another and whether all force curves look qualitatively similar to the example shown in Supplementary Fig. 1a. To evaluate this issue, we averaged force-distance curves 1 for all the three integrin reconstituted fibroblast lines as they were de-adhered from FNIII7-10 after the contact times ranging from 5 -120 s (Fig. R1). The averaged force curves differed in their adhesion force, however, no striking differences in the force profiles from one condition to another were observed. This comparison shows that the force-distance curves looked similar to that represented in Supplementary Fig. 1a. Figure R1. Irrespective of α5β1 or/and αV-class integrin expression, force-distance curves, recorded upon detaching fibroblasts adhering to FNIII7-10-coated substrates, show similar shape. Shown are averages of force-distance curves (n ≥ 30 for each condition) that were recorded upon detaching single fibroblasts from FNIII7-10-coated supports to which they adhered 5, 20, 50 or 120 s. The fibroblasts were either reconstituted with αV-class integrins (pKO-αV, yellow), or with α5β1 (pKO-β1, green), or with α5β1 and αV-class integrins (pKO-αV/β1, blue).
The reviewer further questions whether it would be interesting to report the work of cell adhesion. We agree that the work would be an interesting parameter to describe cell adhesion. However, the work of cell adhesion is substantially influenced by other physical properties/parameters of the cell, including elasticity, deformation or dissipation. Upon de-adhering from the substrate, the rather soft fibroblast is considerably stretched over the distance of several tens of micrometers (Fig. R1). Thus, as the 'work of cell adhesion' describes the adhesive region (blue) of the force-distance curves (Fig. R2), it substantially characterizes the mechanical aspect of stretching and deforming the cell instead of the integrin-mediated adhesive bonds formed to the substrate. The example of rupturing integrins at the tip of cell membrane tethers highlights this effect (Fig. R2). Although, such a membrane tether quantifies the rupture of a single integrin bond, it can be pulled by several micrometers from the cell membrane 2,3 . Thereby, the force needed to extract the tether from the cell membrane remains constant whereas the tether length depends largely on the lifetime of receptorligand bond tethering the tip of the tether to the substrate. Consequently, the 'adhesion work' required to detach a cell from a support can largely depend on other parameters than the cell adhesion and thus can be difficult to interpret 3 . For such reasons, the measured work is not the direct measure of adhesion work/energy of the cell or of individual cell adhesion molecules. Since so far, the work of cell adhesion and cell deformation cannot be distinguished properly in SCFS measurements, we preferred to stay on the conservative side by interpreting and analyzing only the maximum cell adhesion force. This maximum cell adhesion force is a direct measure of adhesion and is easy to interpret, as fewer physical properties of the cell influence it. Figure R2. Detailed description of a force-distance curve recorded while detaching a fibroblast from a substrate. In our SCFS assay, a fibroblast attached to a functionalized AFM cantilever is (i) approached to a substrate until reaching a preset contact force. (ii) For a given contact time, the fibroblast is kept in contact with the FN-coated substrate to initiate adhesion. (iii) Thereafter, the fibroblast is separated from the substrate by retracting the AFM cantilever for up to 100 µm until (iv) the fibroblast is fully detached from the substrate. During this experimental approach and retraction cycle, the force acting on the cantilever is recorded. The approach force-distance curve (red) records the cantilever-bound cell contacting the substrate. The retraction force-distance curve (black) highlights the de-adhesion process and the adhesive characteristics of the fibroblast for the substrate. The blue shaded area estimates the work needed to de-adhere the cell from the substrate. This work is a composition of cell adhesion, cell deformation and extraction of the cell membrane (tethers) 5 . The adhesion force of the fibroblast, measured by maximum downward force deflecting the cantilever, is the maximum force needed to initiate the detachment of the fibroblast from the substrate. Thenceforth, single unbinding events of receptor-ligand bonds are observed. Rupture events are recorded when a cytoskeleton-linked adhesion receptor-ligand bond fail. Tether events are recorded when a membrane tether is pulled away from the cell membrane with one or multiple adhesion receptors at its tip 4,5 . In this case, the cytoskeleton linkage of the adhesion receptor was either too weak to withstand the mechanical load or non-existent. In contrast to the rupture force describing the force required to stress and separate a receptor-ligand bond, the force required to pull a tether from the cell membrane is constant and depends on the cell membrane properties (fluidity, tension, etc) 4,5 . However, the length of the tether, which can be extracted several tens of µm from the membrane describes the lifetime of the receptor-ligand bond stressed at constant force.

Reviewer #1: Also, little is said about the adhesion probability. Could it be an interesting parameter to look at when changing conditions?
Authors: The reviewer suggested to elaborate more on the adhesion probability. This has now been done in the revised manuscript (see revised Results, section 'Differential contributions of α5β1 and αV-class integrins'). Furthermore, the reviewer asks whether it would be interesting to look at adhesion probability at different conditions. In fact studying how integrins change adhesion probability under different conditions would be an exciting work on its own. However, addressing this issue would go beyond the scope of this manuscript as these measurements are time consuming and address a different scientific question. We thus decided to shortly elaborate on this possible follow up study as a brief outlook in the discussion of our revised paper (see revised Discussion).

Reviewer #1:
The clustering section is an important one. The TIRF data should be better presented and discussed for a broad audience, not necessarily expert in the technique. In particular, how can we proof clustering from the data, please explain in detail.

Authors:
The reviewer asked for a better presentation of the TIRF data and to discuss this data for a broader audience. Particularly, we were asked to explain how we can prove clustering from the data. We have improved the presentation of the TIRF data ( Fig. R3, now included as revised Fig. 4) and explained it for the broader audience (see revised Results, section 'Engagement of αV-class integrins clusters α5β1 integrins'). The clustering from the TIRF data can be proven from the enhanced paxillin signal observed at the adhesion site. Paxillin has been shown to be an active marker for adhesion initiation in the hierarchical model for the formation of adhesions (from nascent adhesions to mature focal adhesions) and multiple integrin-trafficking pathways 6 . Therefore, we used paxillin as the marker for integrin clustering. This information has now been included into the revised manuscript (see revised Results, section 'Engagement of αV-class integrins clusters α5β1 integrins'). (a) Time series of TIRF images of GFP-labeled paxillin expressed in pKO-αV/β1 (blue), pKO-αV (yellow) and pKO-β1 (green) fibroblasts adhering to FNIII7-10-coated substrates. To record the images, single fibroblasts were attached to ConA-(non-stimulated) or VN-coated (stimulated) cantilevers, incubated for 7-10 min and then approached to the FNIII7-10-coated substrate. Paxillin-GFP-intensity was detected by TIRF microscopy after 5 s and then after every 20 s for up to 500 s contact time with the substrate. To stimulate fibroblasts by VN, cantilevers were coated using 5 µg ml -1 VN diluted in ConA. Scale bars, 10 µm. (b) Paxillin-GFP-intensity over contact time. The data was taken from TIRF images such as shown here. Dots show mean fluorescence intensities of ten fibroblasts at a given time point with s.e.m.
Reviewer #1: Do we have an idea of the size of the clusters? The authors may want to discuss the limitations of TIRF and the potential, for future research, of higher resolution techniques (superresolution, single-molecule AFM imaging) to analyze nanoscale clusters.

Authors:
The reviewer asked if we have an idea about the size of the adhesion clusters from the TIRF data. In our SCFS-TIRF assay, fibroblasts were brought in contact with the FNIII7-10 for up to 500 s. As stated by the reviewer, pertaining to the innate resolution limitation of TIRF, we could not use the TIRF images to determine the sizes of individual clusters below the lateral resolution limit of ≈500 nm. Because of this resolution limit, we can hardly define a cluster size since many clusters overlap. Therefore, we have characterized the intensity of paxillin-GFP instead of cluster size (Fig. R3). However, to address the lower limit of the cluster size, super resolution microscopy would be better suited 7,8 . In our revised manuscript, we now briefly discuss this issue and the potential for future research to analyze nanoscale clusters (see revised Discussion).

NCOMMS-16-14815-T αV-class integrins exert dual roles on α5β1 integrins to strengthen adhesion to fibronectin
Point-by-point response to the comments of reviewer #2 Authors: Thank you for your encouraging and constructive comments. Below, we describe point-by-point how we addressed each specific comments of the reviewer.

Reviewer #2:
My main concern is with figures 4 and s7, and their interpretation. First, the authors should show the time course of the evolution of fluorescence, and not only one time point (which I assume corresponds to the end of the experiment).

Authors:
The reviewer asked to show the time course of the evolution of fluorescence. Thank you for your suggestion. We have improved the presentation of the TIRF data (Fig. R3, now included as Fig. 4) to show the time course of the evolution of fluorescence and explained the data in detail (see revised Results, section 'Engagement of αV-class integrins clusters α5β1 integrins'). The improved presentation of the data shows that within the initial ≈ 40 s of adhesion, fibroblasts adhering to FNIII7-10 steeply increase the fluorescent intensity of GFP-labeled paxillin clusters and thereafter do not show any further steep increase for up to 500 s. However, within this first ≈ 40 s of contact time, VN-stimulated pKO-αV/β1 fibroblasts increased the fluorescence of GFP-paxillin clusters faster and to higher values compared to non-stimulated pKO-αV/β1 fibroblasts. Likewise, the TIRF signal, obtained for pKO-αV and pKO-β1 fibroblasts adhering to FNIII7-10, increased within the initial ≈ 40 s of contact time. Moreover, the TIRF signal for pKO-αV and pKO-β1 fibroblasts adhering to the FNIII7-10 was not affected upon VN-stimulation and was less strong than that of VN-stimulated pKO-αV/β1 fibroblasts. In previous sections of results, we show that αV-class integrins, engaged by VN, signal to enhance adhesion of pKO-αV/β1 fibroblasts to FN, which is primarily mediated by β1 integrins. Therefore, our results suggest that in fibroblasts, VN-bound αV-class integrins signal to cluster β1 integrins resulting in adhesion strengthening. (a) Time series of TIRF images of GFP-labeled paxillin expressed in pKO-αV/β1 (blue), pKO-αV (yellow) and pKO-β1 (green) fibroblasts adhering to FNIII7-10-coated substrates. To record the images, single fibroblasts were attached to ConA-(non-stimulated) or VN-coated (stimulated) cantilevers, incubated for 7-10 min and then approached to the FNIII7-10-coated substrate. Paxillin-GFP-intensity was detected by TIRF microscopy after 5 s and then after every 20 s for up to 500 s contact time with the substrate. To stimulate fibroblasts by VN, cantilevers were coated using 5 µg ml -1 VN diluted in ConA. Scale bars, 10 µm. (b) Paxillin-GFP-intensity over contact time. The data was taken from TIRF images such as shown here. Dots show mean fluorescence intensities of ten fibroblasts at a given time point with s.e.m.

Authors:
The reviewer queried about the interpretation of the TIRF data and was concerned how does VN-binding of αV-class integrins induce α5β1 clustering. We apologize for not being clear enough in explaining our results.
Prior to the TIRF experiments, we observed that αV-class integrins outcompete α5β1 integrins for binding FNII7-10-coated supports. Interestingly, when αV-class integrins are sequestered to the cantilever using VN, we observe an increased adhesion to FNIII7-10. We determined the integrin localization on the cantilever and substrate sides of the fibroblast. Our study revealed that fibroblasts upon attachment to VN-coated cantilevers recruited αV-class integrins to the cantilever and stimulated the adhesion to FNII7-10-coated supports at the opposing surface of the fibroblast. This enhanced adhesion to FNII7-10 was predominantly mediated by α5β1 integrins. To answer the intriguing question, whether this enhanced fibroblast adhesion to FNII7-10 is due to the elimination of competition or an active crosstalk between αV-class and α5β1 integrins, we performed different experiments: First, we used cilengitide (CiL) to successfully sequester αV-class integrins to the cantilever; importantly, unlike VN-bound αV-class integrins, CiL-bound αV-class integrins could not elicit a comparable signaling response (Fig. 2a). Moreover, FNIII7-10 adhesion of CiL-bound fibroblasts was higher compared to ConA-attached fibroblasts ( Supplementary Fig. 2e) and lower compared to fibroblasts attached to a VN-coated cantilever. These results indicate that only eliminating the competition between αVclass and α5β1 integrins for binding FN is not sufficient to induce enhanced cell adhesion. It also requires substantial signaling.
Second, we altered the amount of sequestered αV-class integrins on the cantilever by coating them with different concentrations of VN ( Supplementary Fig. 4c). Fibroblasts bound to cantilevers coated with 5 µg ml -1 VN (diluted in ConA) showed substantial adhesion to VN-coated substrates indicating that not all αV-class integrins were sequestered to the cantilever (Fig. 2c and Supplementary Fig. 4c). Under this condition, adhesion of fibroblasts to FNIII7-10 was significantly higher than those bound to cantilevers coated with the highest VN concentration, which sequestered all αV-class integrins to the cantilever. Therefore, adhesion of fibroblasts was also enhanced when the competition between αV-class and α5β1 integrins was not fully eliminated.
Third, the TIRF experiments show that α5β1 integrins in pKO-β1 fibroblasts (in the absence of αV-class integrins) do not assemble significant paxillin clusters within the first 40 s (Fig. R3, now included as Fig. 4). This lack of assembled clusters becomes particularly evident compared to VN-stimulated pKO-αV/β1 fibroblasts. Thus, we conclude that the formation of paxillin-rich α5β1 integrin clusters in pKO-αV/β1 fibroblasts attached to VN-coated cantilevers is due to a crosstalk between αV-class and α5β1 integrins.
Altogether, our results suggest that this enhanced adhesion of VN-stimulated pKO-αV/β1 fibroblasts did not result merely from elimination of competition but also by the crosstalk originating from αV-class integrins, which induce α5β1 integrins to establish additional adhesion sites distant from those formed by αV-class integrins. In our revised manuscript, we have paid particular attention to describe these experimental findings more clearly and how we can deduce from these findings our conclusions (see revised Results and Discussion).

Reviewer #2:
To address this, the author should repeat the experiment after labelling fluorescently not paxillin, but avb3/a5b1 integrins. I know that this would likely alter the respective concentrations of integrins, but it would still allow to observe where and to what extent the different integrins localize as a function of the coating both of the substrate and of the cantilever. Carrying out not only TIRF but also epifluorescence imaging would be useful to see what is happening at the cantilever/cell interface.

Authors:
The reviewer asked to repeat the TIRF experiment with fluorescently labeled αVβ3 and α5β1 integrins. The experiment was suggested to observe where and to what extent the different integrins localize as a function of the coating both of the substrate and of the cantilever.
Initially, we sought to visualize integrins directly using labeled primary antibodies. However, due to the low signal-to-noise ratio in these experiments, we could not characterize integrin clustering. Furthermore, unfortunately we could not obtain fluorescently labeled integrins in our fibroblasts lines either, as many of fluorescently labeled integrins were retained in the endoplasmic reticulum and only a small number was transported to the plasma membrane. Additionally, we could not exclude the possibility that labeling integrins with a fluorescent protein would not interfere with the function of the integrin, especially signaling. Thus, since this manuscript is about an integrin crosstalk, we chose to prevent direct integrin labeling and used a focal adhesion protein instead to study integrin clustering. Paxillin has been shown to be an active marker for adhesion initiation in the hierarchical model for the formation of adhesions and multiple integrin-trafficking pathways 6 . Therefore, we chose to employ paxillin to visualize integrin localization as paxillin shows both integrin clustering and signaling (either as signal sender or receiver).
Nevertheless, we could visualize integrin clusters in fibroblasts adhered to both the cantilever and the substrate. To this end, we coated glass coverslips, which have similar chemical and physical properties as the silicon-nitride cantilever, and stained for integrins, phospho-tyrosine and actin ( Fig. 2a and Supplementary Fig. 4b). Using this approach, we succeeded to mimic the cantilever condition and the substrate condition on glass, which provided insights on where and to what extent the different integrins localize as a function of the coating both of the substrate and of the cantilever. Our confocal studies revealed that both phospho-tyrosine-and β3 integrin-clusters were formed in VN-bound pKO-αV/β1 fibroblasts. This result suggests that αVβ3 integrins are engaged in signaling when pKO-αV/β1 fibroblasts are attached to VN-coated cantilevers (Fig. 2a). Importantly, neither integrin clusters nor phospho-tyrosine accumulations were observed in pKO-αV/β1 fibroblasts attached to ConA. However, visualizing cell/cantilever and cell/substrate interface simultaneously in one experiment is technically very challenging and so far not working for us.

Reviewer #2: Relatedly, the abstract states that "once engaged, av class integrins activate a5b1 integrins to establish additional adhesion sites to fibronectin, dislocated from those formed by av class integrins". This is not direct evidence shown in this work, but rather a proposed mechanism based on previous literature. The way it is written, it seems as if those were results presented here. This should be corrected.
Authors: The reviewer commented that the statement "once engaged, αV-class integrins activate α5β1 integrins to establish additional adhesion sites to fibronectin, dislocated from those formed by αV-class integrins" has not been directly shown in this study but rather a proposed mechanism based on previous literature. We apologize for not being more explicit in explaining the results, which led to this conclusion. We have now revised our manuscript to clearly explain how our study provides direct evidence that ligand-bound (VN-stimulated) αV-class integrins signal to α5β1 integrins to bind FN. These evidences include: Firstly, the fibroblasts are bound to VN-coated cantilevers for 7-10 min before they are brought in contact with FN-coated supports. Hence, αV-class integrins engage to the VN-coated cantilever prior to the engagement of α5β1 integrins to the FN-coated support. Secondly, we observe the crosstalk across the fibroblasts with αV-class integrins bound to VN-or FN-coated cantilevers and signaling to cluster α5β1 integrins at the opposite FN-coated support. Hence, we can conclude that "once engaged, αVclass integrins signal α5β1 integrins to establish additional adhesion sites to fibronectin (FN), dislocated from those formed by αV-class integrins". To avoid confusion, we have revised our abstract replacing term 'activate' by 'signal' and have further revised the manuscript text to better describe our experimental findings and how we deduce our conclusions (see revised Results and Discussion). (fig. 2). I agree with the authors that the different binding properties (functional binding in VN, but mere inhibition in CiL) may play a role. However, the main role of VN and CiL in this case is to sequester integrins away from the substrate, and in principle one would expect that potential differences in binding properties would apply to the site of binding (the cantilever) and not so much to the substrate, which is the one that determines the adhesion measurements. Why would the lack of signaling (in the case of CiL) reduce the effect of integrin sequestration to the cantilever with respect to VN?

Reviewer #2: It is unclear why the effects of coating tips with VN or CiL are so different
Authors: The reviewer queried why the effects of coating cantilevers (please be aware that our cantilevers do not have tips) with VN or CiL are so different as the main role of VN and CiL is to sequester integrins away from the substrate and hence should not affect the adhesion measurements.
The reviewer is correct. Indeed one role of VN and CiL is to sequester αV-class integrins away from the substrate and potential differences in binding properties should apply to the site of binding. However, the functional state of αV-class integrins upon sequestration is very different. αV-class integrins are capable of signaling upon binding of fibroblasts to VN-coated cantilever, while they do not or signal relatively less when bound to CiL (Fig. 2a and Supplementary Fig. 2e). This functional state of αV-class integrins dictates the FN-adhesion of pKO-αV/β1 fibroblasts, which provided the evidence for the integrin crosstalk. The differential adhesion of CiL-bound versus VN-bound fibroblasts to FNIII7-10 is a key finding for the crosstalk between αV-class and α5β1 integrins. We conclude from this result that signaling capable -active αV-class integrins regulate the binding of α5β1 integrin to FNIII7-10. This finding is further supported by the TIRF experiments and the VN-dilution experiments.
The reviewer further questioned why would the lack of signaling (in the case of CiL) reduce the effect of integrin sequestration to the cantilever with respect to VN. As stated above, CiL-bound αV-class integrins bind but elicit relatively less signaling response compared to VN-bound αV-class integrins (Fig. 2a). We observe α5β1 integrins establish enhanced adhesion to fibronectin only when signaled from VN-bound αV-class integrins and not from CiL-bound αV-class integrins. We have revised our manuscript to describe these issues more clearly and thus to avoid confusion of the reader (see revised Results, section 'αV-class integrins stimulate fibroblast adhesion to FN' and Discussion).

Reviewer #2:
The results of figure 3 are very interesting but somewhat confusing. For instance, it is very intriguing that some contractility inhibitors (the Y compound or blebbistatin) inhibit the effect of VN cantilever coating, but some others (ML-7) do not. In this respect, carrying out fluorescence experiments such as those in figure 4, by tagging 12 fluorescently avb3 and a5b1 integrins, after the different inhibitions would be very useful. This would allow to see whether the different inhibitions alter the differential recruitment of the two integrin types at the substrate versus the cantilever, for instance. I realize that carrying out such experiments for all the conditions tested would represent an enormous amount of work, but doing it for the Y compound and ML7 for instance would add very valuable information on the differential effect of the two drugs.

Authors:
The reviewer questioned about differential effects of contractility inhibitors like Y27632, blebbistatin and ML-7 on VN-stimulated fibroblasts adhesion to FN. We conclude from these results that not all integrin-mediated signaling molecules are involved in the early crosstalk between αV-class and α5β1 integrins. Based on our observation and existing literature, we speculate how an integrin-mediated signaling molecule would be involved in this crosstalk. With respect to this specific question, we would like to elaborate on the role of the targets of the inhibitors (Y27632, blebbistatin or ML-7) contributing to the crosstalk. Myosin-II activity is regulated by myosin light chain (MLC) phosphorylation (inhibited by blebbistatin), which is either directly positively regulated by MLC kinase (MLCK; inhibited by ML-7) or by RhoA kinase (ROCK; inhibited by Y27632) 9 . We observed that the crosstalk does not require MLCK activity. Therefore, we speculate that during the crosstalk, stimulated adhesion via α5β1 integrins is governed by myosin-II induced tension regulated by RhoA/ROCK activity.
The reviewer further suggested for fluorescence experiments such as those in our SCFS-TIRF studies with different inhibitions using fluorescently tagged αVβ3 and α5β1 integrins. Indeed, it would be interesting to observe how different inhibitions cause the differential recruitment of the two-integrin types at the substrate and at the cantilever. However, as stated above we could not use fluorescently labeled integrins as they were retained in the ER. Therefore, we chose an alternative approach, wherein we mimicked both cantilever and substrate conditions on glass and observed for localization of integrins upon drug administration by fixing the fibroblasts after 10 min (SCFS condition) and after 90 min (as a control) (Fig. R4). Upon depletion/inhibition of integrin-associated proteins, pKO-αV/β1 fibroblasts showed varied phenotypes on FNIII7-10 and VN. Each perturbation had a striking effect on the cell-spreading pattern and hence focal adhesions (clusters) were disrupted in almost all conditions. Concerning the reviewer's specific question on the contractility inhibitors (Y27632, blebbistatin and ML-7), pertaining to the differences in the phenotypes upon inhibitions, we cannot say much about how the integrins were localized. However, unlike pKO-αV/β1 fibroblasts treated with Y27632 and blebbistatin, in ML-7 treated pKO-αV/β1 fibroblasts adhering to substrate-coated glass for 10 min, we observed both integrin classes at similar localization but at less concentration compared to the unperturbed state. This result suggests that in pKO-αV/β1 fibroblasts adhesion initiation is not severely affected upon ML-7 treatment, which could provide reasoning behind the differences in the effect of the contractility inhibitors on fibroblasts adhesion to FN. Figure R4. Disruption of the integrin-mediated signaling machinery affects the integrin assembly at the cell-substrate interface. pKO-αV/β1, talin KO, kindlin KO, ILK KO fibroblasts and pKO-αV/β1 fibroblasts treated with specific inhibitors against integrin-mediated signaling molecules were allowed to adhere to FNIII7-10-and VN-(high, 50 µg ml -1 ) coated glass surfaces for 10 min (substrate and cantilever conditions) or for 90 min (control). The adhered fibroblasts were then fixed to restrict integrin localization. Thereafter, fixed pKO-αV/β1 fibroblasts were stained for αVβ3 integrins (green), actin (red) and α5β1 integrins (pink) using β3 integrin specific antibodies, phalloidin and β1 integrin specific antibodies (Methods), respectively. Immunostaining of αV-class and β1 integrins and actin in pKO-αV/β1 fibroblasts adhering to FNIII7-10-coated substrates are used as a positive control. Scale bars, 10 µm. Despite of multiple attempts, kindlin KO fibroblasts did not adhere strongly to FNIII7-10 within the first 10 min of attachment and were washed off during the staining protocol; therefore, the panel is blank.
Reviewer #2: Given the rich previous literature on the topic of a5b1 versus avb3 integrins, a short discussion of how this work fits in with or reinterprets previous data would be useful. This is already done for some publications, but adding for instance Balcioglu et al. , jcs 2015, or some of the work by the spatz/cavalcanti groups would be useful.

Authors:
The reviewer suggested to add a short discussion of how the previous 15 literature on the topic of α5β1 versus αVβ3 integrins relates to our work. Thank you for your suggestion. We have extended the discussion in our revised manuscript by mentioning the suggested references.
Reviewer #2: An effect of differential on/off rates between alphav and a5b1 integrins was already described previously (Elosegui-Artola et al., nat. mater. 2014). This previous work does not affect the novelty of this submission since it involved different measurements and a different av integrin, but it should still be mentioned.

Authors:
We have done this.

Point-by-point response to the comments of reviewer #3
Reviewer #3: This manuscript describes a detailed analysis using single-cell force spectroscopy, of the differential interaction of aV-containing integrins, typically binding extracellular ligands such as vitroectin (VN), osteopontin and weaker binding to fibronectin (FN), with the classical FN-receptor a5b1

. The experiments are very well controlled and use genetically engineered cells and extracellular ligands, to exclude any contamination with the other type of integrin receptors, or contaminations of ligand preparations. The result is an impressive study showing that av-integrins have the capacity to inhibit the recruitment and adhesion reinforcement of a5b1-integrins on the cell-binding fragment of FN. Although this rather surprising observation is not entirely new (Pinon et al, 2014, not cited!), new elements are added to explain the cross-talk between the aV and a5b1 integrin receptors. Notably, the authors show that the inhibition by aV-integrins on a5b1-mediated FN binding can be prevented by physically separating the ligands (not surprising), but that this separation is creating a positive intracellular feedback, involving classical integrin-signaling pathways that enables efficient b1-reinforcement after previous binding and signaling of aV-integrins on VN.
Although the work is interesting for specialist in the domain of integrins, the study is not really revealing the mechanistic basis for the observed competition and cross-talk between aV-and b1-integrins (while aV binds weaker to FN, it has a faster on rate than b1-integrins). Therefore, I suggest a couple of issues that should be evaluated: Authors: Thank you for your encouraging and constructive comments. Below, we describe point-by-point how we addressed each specific comments of the reviewer.
Reviewer #3: Unfortunately a key question remains unanswered. What is the qualitative difference, between aV integrin versus a5b1 integrin binding, in respect to adapter recruitment, signaling, and differential adhesion to VN versus FN. Maybe avb3, originally termed the VN-receptor, is never intended to bind to FN, but competes with endogenous a5b1 because there is no better extracellular binding partner, implying that the observed difference is due to the extracellular ligand binding domain, and faster on-rate of aVintegrins. However an alternative explanation is that the observed competition between aV and a5b1 integrins is not due to the extracellular domain, but linked to differences at the level of the cytoplasmic domain. Is not talin a better binder for b3 than b1 (Anthis et al., 2010)? For example Pinon et al, (2014) has shown that avb3 can cluster on FN after 1 hour of spreading, preventing a5b1-integrin recruitment, very similar to what is shown here in this manuscript. However, when b3 is absent or mutated in its talin-binding motif, a5b1-is readily engaged on FN to induce adhesion and spreading. Thus, despite a normal extracellular domain, such a aVb3-integrin with a cytoplasmic talin-binding mutation, can no longer compete with a5b1 integrins. This can lead to two conclusions: (1) talin is required for aVb3-activation and binding to FN, or (2) b1-integrin is normally prevented from rapid binding because it has a lower affinity for talin, or because its cytoplasmic domain is interacting with inhibitory adapter proteins. It would therefore be important to test how the deletion of b1-integrin specific inhibitory adapters, such as ICAP-1 or filamin would affect the recruitment dynamic or the here called crosstalk between the analyzed integrins. A plausible explanation, consistent with the observed effects, could involve the signaling-dependent inactivation (e.g. by phosphorylation) of b1-specific inhibitors and thus rapid a5b1 activation and spreading as for example seen in ICAP-1 deleted fibroblasts (Millon-Fremillion).

Authors:
The reviewer queried if the competition between αV-class and β1 integrins could be due to their differences in the binding affinities/abilities for talin. This could indeed be the case as previous work suggested that αV-class and β1 integrins compete for the cytoplasmic talin pool leading to negative, trans-dominant effects 10 . Intrigued by reviewer's query, we characterized the binding of talin to integrin β1-and β3-tails by a pull down assay (Fig. R5a). This pull down observed equivalent binding of talin to both the β1-and β3-tail of integrins. We have now included this data and discussed it accordingly in our revised manuscript (see revised Discussion and Supplementary Fig. 8).
The reviewer further suggested to perform competition studies in the absence of β1 integrin specific inhibitory adapters, such as ICAP-1 or filamin, to see if this competition is due to inactivation of β1 integrins. This was a fantastic suggestion. We quantified the adhesion of ICAP-1 deficient mouse embryonic fibroblasts (ICAP-1 KO MEFs) 11 and of wild-type MEFs, as a control, to FNIII7-10 (Fig. R5b). As hypothesized by the reviewer, the adhesion of ICAP-1 KO MEF to FN was indeed higher than that of wild-type MEFs. This result suggests that during adhesion initiation, since ICAP-1 is bound to the cytoplasmic domain of β1 integrins, talin is unable to bind and activate β1 integrins. Therefore, talin readily binds to αV-class integrins resulting in higher binding rates of αV-class integrins to FN and hence αV-class integrins outcompete β1 integrins. We have now included this result and discussed it accordingly in our revised manuscript (see revised Discussion and Supplementary Fig. 8).  figure  1a (second column), the reinforcement of binding on FN is seen for cells expressing only b1-integrins. However, this type of binding is never compared with FN binding in av-b1 cells when bound to a ConA/VN cantilever. Is binding significantly enhanced in the latter condition? If so, this could explain the enhanced recruitment of paxillin for the latter but not former condition (sup fig 7). Authors: Thank you for your suggestion. We had compared the adhesion force of pKO-β1 and VN-stimulated pKO-αV/β1 fibroblasts in the result section 'Engagement of αV-class integrins stimulate fibroblast adhesion to , now revised to 'αV-class integrins stimulate fibroblast adhesion to FN' of our manuscript but maybe we were not explicit enough to point out the difference. Therefore, we now compare the adhesion force of pKO-β1 and pKO-αV/β1 fibroblasts to FNIII7-10 at the suggested conditions directly in one figure (Fig. R6). The comparison shows that the adhesion of VN-stimulated pKO-αV/β1 fibroblasts is enhanced compared to ConA-bound pKO-β1

Reviewer #3: Reinforcement of binding in ConA/b1-cells versus VN/av-b1-cells. In
fibroblasts. In addition, we have now included the comparison into the revised manuscript (see revised Discussion and Supplementary Fig. 5). Reviewer #3: Unfortunately paxillin is not a marker for integrin clustering, but rather integrin signaling. For clustering of b1-integrins, 9EG7 staining should be used ?
Authors: Initially, we used antibodies to stain integrins for visualizing integrin clustering. We tried different antibodies including the 9EG7 antibody. Unfortunately, the signal-to-noise ratio in these experiments was not satisfying in fibroblasts adhering for only 10 min to the substrate (SCFS conditions). Therefore, we could not image integrin clusters at a satisfying quality. Further, we encountered problems of bleaching of the fluorophores when we used antibodies as we had exposure for every 20 s for up to 10 min. Paxillin has been shown to accumulate during the formation of adhesions (from nascent adhesions to mature focal adhesions). Therefore, we have chosen GFP-tagged paxillin as the marker for adhesion initiation via active integrins and integrin clustering.

Reviewer #3: Enhanced paxillin recruitment is a late event (observed only at 100s), does it correlate with adhesion maturation over time? How is the adhesion curve evolving over the time range visualized for paxillin recruitment (e.g. 5 min), comparing a b1integrin only from a VN-mediated b1-integrin-dependent adhesion?
Authors: The reviewer suggested to show the adhesion maturation over time and to compare adhesion maturation of β1 integrins in pKO-β1 and VN-stimulated pKO-αV/β1 fibroblasts. Thank you for your suggestion. To address this issue, we have revised Fig. 4 (below shown as Fig. R3) and the Results section 'Engagement of αV-class integrins clusters α5β1 integrin'. The revised figure now shows that within the initial ≈ 40 s of adhesion, fibroblasts adhering to FNIII7-10 steeply increase the fluorescent intensity of GFP-labeled paxillin and thereafter do not show any substantial increase in the TIRF signal for up to 500 s. However, within this first ≈ 40 s of contact time, VN-stimulated pKO-αV/β1 fibroblasts recruited paxillin much quicker to integrin clusters and thus the fluorescence intensity increased much quicker and to higher values compared to non-stimulated pKO-αV/β1 fibroblasts. Likewise, the TIRF signal obtained for adhesion of pKO-αV and pKO-β1 fibroblasts to FNIII7-10 increased within the initial ≈ 40 s of contact time. Moreover, the TIRF signal for pKO-αV and pKO-β1 fibroblasts adhering to the FNIII7-10 was not affected by VN-stimulation and was less than that of VN-stimulated pKO-αV/β1 fibroblasts. Specifically, paxillin recruitment in pKO-β1 fibroblasts was strikingly lower compared to all other conditions including VN-stimulated pKO-αV/β1 fibroblasts as they bound FNIII7-10. Thus, the adhesion maturation of β1 integrins in pKO-β1 fibroblasts is different from the adhesion maturation of β1 integrins in VN-stimulated pKO-αV/β1 fibroblasts. Unfortunately, since after 500 s of contact time, the fibroblasts established substantial adhesion to FNIII7-10, the fibroblast detached from the cantilever due to their stronger adhesion to the substrate. Therefore, we could not measure adhesion forces to FNIII7-10. However, to answer the reviewer query, in a few successful SCFS attempts, we could observe an adhesion force ≥ 25nN, for VN-stimulated pKO-αV/β1 fibroblasts, as they adhered FNIII7-10 for 500 s. (a) Time series of TIRF images of GFP-labeled paxillin expressed in pKO-αV/β1 (blue), pKO-αV (yellow) and pKO-β1 (green) fibroblasts adhering to FNIII7-10-coated substrates. To record the images, single fibroblasts were attached to ConA-(non-stimulated) or VN-coated (stimulated) cantilevers, incubated for 7-10 min and then approached to the FNIII7-10-coated substrate. Paxillin-GFP-intensity was detected by TIRF microscopy after 5 s and then after every 20 s for up to 500 s contact time with the substrate. To stimulate fibroblasts by VN, cantilevers were coated using 5 µg ml -1 VN diluted in ConA. Scale bars, 10 µm. (b) Paxillin-GFP-intensity over contact time. The data was taken from TIRF images such as shown here. Dots show mean fluorescence intensities of ten fibroblasts at a given time point with s.e.m.

Reviewer #3:
There is no pure VN control for the mixture of VN with ConA: Normally cell spreading is analyzed on glass coated with purified VN. Mixing VN with ConA as adhesive support used on the cantilever could create a signaling effect that is not induced with a pure VN coating. What happens to cells when they are plated on a ConA/Vn mixture compared to a VN/BSA mixture or pure VN coated surface. If ConA is recruiting a cosignaling receptor to the engaged aV-integrins on VN, this could induce an "integrinindependent" signaling cross-talk.

Authors:
The reviewer is concerned if VN diluted in ConA would elicit a specific signaling effect on fibroblasts different from that of pure VN and therefore suggested to compare the status of fibroblasts on VN diluted in ConA and pure VN. Thank you for your suggestion. We sought to use the VN/ConA mixture to functionalize cantilevers as fibroblasts bound to cantilevers coated by VN only at a concentration of 5 µg ml -1 repeatedly dropped off the cantilever during SCFS measurements. This dropping off the 22 cantilever was pertaining to the higher binding strength of VN-stimulated fibroblast to FNIII7-10. Therefore, to aid SCFS measurements, we used ConA to further dilute VN for cantilever functionalization, which assisted the binding of the fibroblast to the cantilever irrespective. In our revised manuscript, we describe this issue more clearly (see revised Results, section 'αV-class integrins stimulate fibroblast adhesion to FN').
To further characterize what happens to fibroblasts when plated on a VN/ConA mixture compared to fibroblasts plated on a VN/BSA mixture or pure VN-coated surface, we visualized the localization of integrin for each of the cantilever conditions mimicked on a glass substrate (Fig. R8). Fibroblasts were plated on pure 50 µg ml -1 VN (labeled as VN HIGH ), 5 µg ml -1 VN diluted in ConA (labeled as VN), cilengitide (CiL), ConA and FNIII7-10 as a control. The degree of clustering of αV-class integrins decreased, as VN was further diluted (10 times) in ConA compared to pure VN (VN HIGH ). Therefore, we could exclude the possibility of having ConA as a co-signaling receptor. Moreover, there was no specific αV-class integrin clustering on pure ConA. Thus, these results confirm that ConA-binding does not initiate integrin signaling as reported previously by others 12 .
We have now included the controls into the revised manuscript and discuss them accordingly (see revised Results and Fig. 2a). Authors: The reviewer suggested to change the label of figure from VN to VN/ConA as it is misleading. Since we excluded a role of ConA as co-signaling receptor with VN and explicitly mentioned in the Results, Legends, and Methods that VN refers to 5 µg ml -1 VN diluted in ConA, we prefer to remain with the old labeling.
Furthermore, reviewer asked us about the effect of pure VN in our crosstalk assay. We did perform crosstalk experiments with 50 µg ml -1 VN only (labeled as VN HIGH ). Fibroblasts attached to VN HIGH functionalized cantilevers exhibited enhanced adhesion to FNIII7-10 compared to fibroblasts attached to ConA (Supplementary Fig. 4c).

Authors:
The reviewer asked how the effect of the inhibitors could be separated from the integrin signaling occurring at the cantilever versus the integrin signaling at the adhesion site. We agree with the reviewer. As mentioned in the discussion, we could not determine whether the inhibitors affected the origin of signaling or reception or response to the signal and thus would like to address this issue in our future work with mutant forms of the integrins. Moreover, an appropriate answer to this question would be beyond the scope of this work, as we think that it is already quite complex. Physical separation of FN and VN on one substrate would not be possible in the SCFS setup, as the fibroblasts need to adhere to VN for longer time than to FN for crosstalk studies. Thus, we need to attach fibroblasts to VN-coated cantilevers. So far, we wanted to sum up our findings as a crosstalk between α5β1 and αV-class integrins wherein αV-class integrins signal to cluster α5β1 integrins. Furthermore, the reviewer made multiple suggestions including to remove β1-integrin inhibitors to study the effect of the inhibitors in the crosstalk. Thank you for your suggestions. It is interesting yet quite time consuming to study effect of all the inhibitors, used in this work, in the absence of a β1-integrin inhibitor such as ICAP-1. However, driven by curiosity, we characterized the crosstalk using ICAP-1 deficient mouse embryonic fibroblasts (ICAP-1 KO MEFs) and wild-type MEFs (as a control). VN-stimulated ICAP-1 KO MEFs adhered much stronger to FNIII7-10 compared to VN-stimulated wild-type MEFs and ConA-bound ICAP-1 KO MEFs (Fig. R9). This result suggests that constitutively active β1 integrins 11 bind FN even much more stronger in response to signaling originating from engaged αV-class integrins. We have now included this result in the revised manuscript and discussed it accordingly (see Discussion and Supplementary Fig. 9). Reviewer #3: Minor issues: -Please explain which blocking antibody was used for beta1, is it changing the conformation of the integrin.

Authors:
We used function-blocking α5β1 integrin antibody, BMC5 clone at recommended concentration of 10 µg ml -1 to block α5β1 integrin binding.
Reviewer #3: Minor issues: -Line 249: does this cross-talk require cantilever pulling and force transmission on the aV-integrins?
Authors: The reviewer queried if this crosstalk depends upon cantilever pulling and force transmission on the αV-class integrins. For our study, we used constant pulling speed of 5 µm s -1 for all experiments and therefore, we cannot elucidate if this crosstalk depends on the pulling and force transmission of αV-class integrins. However, inspired by the reviewer's question, we initiated pulling speed and mechanical load dependent SCFS experiments to investigate to which extent these parameters modulate the crosstalk. As these experiments are very time consuming, the experimental work will increase by several factors and thus we would like to address this complex and data intensive topic in a separate study.

Reviewer #3:
Is it possible to visualize the surface of the cantilever for integrin signaling? Does varying the pulling speed affect the integrin signaling at the level of the cantilever.

Authors:
The reviewer asked if we could visualize the surface of cantilever for integrin signaling. We closely monitor cell morphology on the cantilever using wide-field microscopy during SCFS experiments. However, it is very challenging to monitor integrin-mediated signaling in living cells during SCFS experiments because of following reasons: Firstly, SCFS experiment has an innate technical disadvantage when it comes to simultaneous high resolution imaging. Since cantilever is moved during the SCFS experimental cycle, the focal plane at the cell-cantilever interface is constantly changing. Therefore, imaging this surface with the required resolution is technically very challenging and impossible with the conventional microscopy. In addition, the cantilever is intrinsically tilted by 10 degrees, which tilts the focal plane making standard confocal microscopy technically impossible. Secondly, to visualize integrin-based signaling, we would need to use an active integrin marker like paxillin as we used in TIRF studies. Since paxillin is diffusing across the cell, it would not be trivial to identify its confinement on the surface of cantilever using standard fluorescence microscopy. Therefore, to circumvent these problems, we had to combine SCFS with TIRF to study integrin localization on the surface of fibroblast in contact with the substrate.
The reviewer further asked whether integrin signaling depends on the pulling speed of cantilever. The regulation of adhesion by mechanical load (conferred upon by different pulling speeds) is an intriguing question. We would, however, expect that a possible mechanosensitive response would be based on different signaling pathways, crosstalks and thus regulate cell adhesion differently. But, since we performed all the experiments, in this study, with the constant pulling speed of 5 µm s -1 , we cannot confirm if varying the pulling speed of the cantilever would affect the integrin signaling.
Reviewer #3: Minor issues: - Figure 5: I am not aware of a bent a5b1 structure (green), similar to the one's described for a2bb3 and avb3. In fact, it is likely that avb3 and a5b1 are regulated entirely different by cytoplasmic adapters, which would affect their ability to cross-talk.