Discrete spatial organization of TGFβ receptors couples receptor multimerization and signaling to cellular tension

Cell surface receptors are central to the cell's ability to generate coordinated responses to the multitude of biochemical and physical cues in the microenvironment. However, the mechanisms by which receptors enable this concerted cellular response remain unclear. To investigate the effect of cellular tension on cell surface receptors, we combined novel high-resolution imaging and single particle tracking with established biochemical assays to examine TGFβ signaling. We find that TGFβ receptors are discretely organized to segregated spatial domains at the cell surface. Integrin-rich focal adhesions organize TβRII around TβRI, limiting the integration of TβRII while sequestering TβRI at these sites. Disruption of cellular tension leads to a collapse of this spatial organization and drives formation of heteromeric TβRI/TβRII complexes and Smad activation. This work details a novel mechanism by which cellular tension regulates TGFβ receptor organization, multimerization, and function, providing new insight into the mechanisms that integrate biochemical and physical cues. DOI: http://dx.doi.org/10.7554/eLife.09300.001


Introduction
The diversity and specificity of cellular responses rely on the precise integration of biochemical and physical cues from the microenvironment. Cells generate a coordinated response through interactions among signaling pathways -from ligands and receptors to intracellular effectors. Receptors are a particularly versatile locus of control since they undergo regulated microdomain clustering, internalization and homo/hetero-meric multimerization. Because these mechanisms affect ligand binding, enzymatic activity, and effector recruitment, receptors play a crucial role in defining signal intensity, duration, location, and quality (Bethani et al., 2010;Di Guglielmo et al., 2003;Groves and Kuriyan, 2010;Salaita et al., 2010). However, many questions remain about the mechanisms by which receptors participate in the concerted cellular response to a multitude of concurrent cues.
The TGFb signaling pathway exemplifies the importance of regulated receptor multimerization. TGFb signals through a heterotetrameric complex of transmembrane receptor kinases. Once the TGFb ligand is activated from its latent form, it binds directly to a dimer of type II receptors (TbRII) (Annes, 2003;Munger et al., 1999;Wipff et al., 2007;Munger and Sheppard, 2011). The ligandbound TbRII complex recruits and phosphorylates two type I receptors (TbRI) -either Alk5 or Alk1 (Wrana et al., 2008). TbRI, in turn, phosphorylates and activates canonical Smad proteins and multiple non-canonical effectors, such as RhoA, TAK1 and Akt (Massague, 1998;Feng and Derynck, 2005). Specifically, recruitment of Alk5 to the TbRII complex stimulates phosphorylation of Smad2/3, whereas Alk1 recruitment drives activation of Smad1/5/8 (Lin et al., 2008). The inappropriate shift of TbRII multimerization partner from Alk5 to Alk1 underlies disease processes ranging from vascular disorders to osteoarthritis (Blaney Davidson et al., 2009;Goumans, 2002). Not only do TGFb receptors associate with one another, but also with a number of other receptor families, notably integrins (Scaffidi et al., 2004;Garamszegi et al., 2010). Garamszegi et al. revealed a physical interaction between integrin a2b1 and TGFb receptors involved in collagen-induced Smad phosphorylation (Garamszegi et al., 2010). TGFb receptor interactions alter ligand specificity and effector selection, offering a regulatory mechanism to calibrate TGFb signaling based on the cellular microenvironment.
Integrins, another class of multimeric receptors, are central to cellular mechanotransduction. Upon integrin binding to the extracellular matrix, the formation of focal adhesions stimulates actomyosin contractility to generate cellular tension (DuFort et al., 2011;Giancotti, 1999;Ingber, 1997). Through this Rho/ROCK-dependent mechanism, cells establish tensional homeostasis with the physical features of the extracellular environment (DuFort et al., 2011). Cellular tension can amplify, alter, or suppress cellular responses to growth factor signaling (Allen et al., 2012; eLife digest Cells constantly encounter diverse physical and biological signals in their surroundings. Information contained in these signals is transmitted from the cell surface to the interior to trigger coordinated changes in the cell's behavior. Physical signals include the forces generated by cells pulling on one another or on their surroundings. These pulling forces calibrate the cell's response to biological signals through mechanisms that remain unclear. The cell surface contains many different proteins that are specialized to sense these signals and guide the cell's response. In animals, these membrane proteins include the receptors that detect a small signaling protein known as TGFb. TGFb first binds to one of these receptors (called TbRII). Next another receptor (called TbRI) is recruited to the complex. Once this complex is formed, the TGFb receptors activate a complicated signaling pathway that controls how cells grow and divide. Previous work has shown that the TGFb pathway can also sense and respond to mechanical forces. But it remains poorly understood how pulling forces (or tension) impact TGFb receptors at the cell surface.
Rys, DuFort et al. have now used cutting-edge microscopy and biochemical techniques to analyze individual TbRI and TbRII receptors and observe how they respond to mechanical forces in real-time. This revealed that TbRI and TbRII exist in discrete regions on the cell surface. Rys, DuFort et al. observed that TbRI is enriched at assemblies of molecules called focal adhesions. Focal adhesions are the sites on cell surfaces that allow cells to adhere to one another and to the molecular scaffolding in their surroundings. Unlike TbRI, TbRII was often excluded from these sites and more commonly appeared to 'bounce' around the edges of individual focal adhesions. Therefore, focal adhesions limit the interactions between TbRI and TbRII, by sequestering one away from the other.
Rys, DuFort et al. next treated cells with a chemical that disrupts tension, and saw that the physical separation between TbRI and TbRII collapsed, which permitted these two receptors to interact and form a working signaling complex. Further work is needed to understand how physical control of TGFb receptor interactions helps cells coordinate their tasks in response to the myriad biological and physical signals in their surroundings. McBeath et al., 2004;Wang et al., 2012). The functional state of many intracellular effectors, including b-catenin, YAP/TAZ, and MAPK, is modulated by cellular tension (Samuel et al., 2011;Dupont et al., 2011;Wang et al., 1998). In the case of TGFb signaling, we and others have identified several mechanosensitive responses (Allen et al., 2012;Wang et al., 2012;Leight et al., 2012). The activation of latent TGFb ligand, as well as the phosphorylation, nuclear translocation and transactivation of Smads is regulated by cellular tension in a Rho/ROCK-dependent manner (Allen et al., 2012;Wipff and Hinz, 2008). However, the mechanisms by which changes in cellular tension modulate effector activity remain unclear.
The effect of cellular tension on the multimerization of receptors other than integrins is largely unexplored. In spite of the established tension-sensitive regulation of downstream signaling effectors, the effect of physical cues on growth factor receptor interactions is unknown. This gap in understanding is partly due to the fact that until recently, studies of cell surface receptor colocalization and physical interactions have mostly utilized biochemical, biophysical, or fluorescence imaging approaches. While invaluable, these approaches are limited by their inability to discriminate spatially discrete molecular interactions that occur in specific cellular domains. Novel super-resolution imaging approaches provide the capability to visualize receptor responses to biochemical and physical cues at the single molecule level with spatial and temporal specificity (Coelho et al., 2013;Manley et al., 2008;Rossier et al., 2012;Calebiro et al., 2013;Xia et al., 2013). To elucidate mechanisms by which physical cues regulate growth factor signaling, we utilize high-resolution imaging, single particle tracking, mass spectrometry and biochemical assays to test the hypothesis that cellular tension regulates TGFb receptor multimerization. We find that cellular tension controls the spatial organization, multimerization and activity of a discrete population of TGFb receptors at integrin-rich focal adhesions, suggesting a novel mechanism by which physical cues calibrate the activity of the TGFb signaling pathway.

Discrete localization of TbRI and TbRII to segregated spatial domains
To investigate the spatiotemporal control of TGFb receptors, we evaluated the localization of endogenous and fluorescently tagged TbRII and TbRI in ATDC5 chondroprogenitor cells and NIH3T3 fibroblasts. Immunofluorescence of TbRII in both wildtype and transfected ATDC5 cells yielded similar results, revealing specific punctate staining that did not provide structural information ( Figure 1A, Figure 1-figure supplement 1). Proceeding with fluorescently tagged TbRII allowed for visualization of fine structural features in static and dynamic conditions. Spinning disc confocal microscopy of TbRII-mEmerald allowed visualization of its spatial organization, revealing shadowed regions where TbRII expression is completely absent (indicated by arrows, Figure 1B-D). Total internal reflection fluorescence (TIRF) microscopy improves visualization of transmembrane proteins by examining a thin section of the sample at the adherent cell surface. Switching from widefield microscopy ( Figure 1E) to TIRF on the same cell vividly revealed segregated domains of TbRII ( Figure 1F) and TbRI ( Figure 1G,H). The sequestration of TbRII from TbRI was present with either the canonical (Alk5) or non-canonical (Alk1) type I TGFb receptors ( Figure 1G,H). Indeed, when co-expressed in the same cell, TbRII is enriched at the boundary of discrete TbRI domains, demonstrating a novel spatial segregation of these signaling partners ( Figure 1I-L).

Single molecule trajectories reveal specific regulation of TGFb receptor dynamics
Since dynamic recruitment of TbRI to TGFb-bound TbRII complexes stimulates downstream effectors, we sought to determine if spatial segregation of TGFb receptors affects receptor mobility. Single-particle tracking photoactivated localization microscopy (sptPALM) resolves the dynamics of individual molecules in live single cells. Using sptPALM, we captured thousands of trajectories of individual TbRI (Alk5) and TbRII proteins labeled with photoswitchable mEos2 (Figure 2A Figure 2C) allowed us to visualize single molecule track behavior and describe molecular environments within individual cells. For both TbRI and TbRII, individual receptors showed a range of mobility, resulting in groups of immobile, confined, or freely diffusive receptors (representative tracks, Figure 2D).
Mobility of each group of TbRI did not differ significantly from TbRII ( Figure 2E), but the diffusion coefficient of TbRI was slightly higher ( Figure 2F), perhaps because of its lower molecular weight (TbRI/Alk5 56 kDa vs. TbRII 65 kDa). Relative to whole cell TGFb receptor dynamics, TbRI and TbRII are significantly less mobile in cellular domains enriched with clusters of spatially organized receptors ( Figure 2F). Thus, this spatially organized population of TGFb receptors is slower and more confined, possibly due to interactions with other proteins.  Figure 1-figure supplement 1) demonstrates punctate staining. Imaging of mEmerald-labeled TbRII (B) reveals TbRII-absent domains in ATDC5 (B,C) and NIH3T3 (D) cells expressing mEmerald-TbRII. Switching from widefield (E) to TIRF mode imaging (F) on the same cell unveils a specific spatial organization of TbRII, which is discrete from that of TbRI (Alk5 and Alk1) (G,H). ATDC5 cells co-expressing mEmerald-TbRII and mCherry-TbRI (Alk1) reveal that TbRII surrounds specific domains of TbRI (I-L). Quantitative profile plot of expression intensity demonstrates separate and distinct localization patterns of TbRI and TbRII (L). DOI: 10.7554/eLife.09300.003 The following figure supplement is available for figure 1:

Focal adhesions organize TbRII around a segregated pool of TbRI
The distinct localization of TGFb receptors could result from physical interactions with any number of known TGFb receptor-associated proteins. Among these, integrins bind to TbRI and TbRII and functionally interact with the TGFb pathway at multiple levels (Wrana et al., 2008;Scaffidi et al., 2004). The primary integrins in chondrocytes are integrins a2 and aV, which bind collagen and vitronectin/ fibronectin (Loeser, 2000). Both integrins interact with the TGFb pathway (Scaffidi et al., 2004;Garamszegi et al., 2010). TIRF imaging of mCherry-labelled integrin a2 revealed the presence of focal adhesions at these TbR-rich sites. Specifically, TbRII is absent from sites of adhesions and forms a peripheral ring surrounding integrin a2, resulting in distinct patterns of spatial localization ( Figure 3A). Interestingly, this spatial organization is absent in cells grown on poly-l-lysine-coated substrates that facilitate integrin-independent cell adhesion ( Figure 3-figure supplement 1). Therefore, TbRII organization at sites of adhesion is dependent upon integrin activity. Profile plots of intensity and a custom analysis (Figure 3Ai,Bi,Ci) were utilized to quantify colocalization between  . Focal adhesions sequester TbRI from TbRII. TIRF mode imaging and a custom colocalization analysis were used to evaluate localization of TbRII (A), Alk5 (B), or Alk1 (C) with integrin a2 in ATDC5 cells. TbRII surrounds integrin a2 (A), whereas both subtypes of TbRI, Alk5 (B) and Alk1 (C), are included within integrin-rich focal adhesions, as reflected by profile plots and the slope values of the regression lines (Ai,Bi,Ci). Quantification of colocalization reveals that Alk5 and Alk1 are significantly more colocalized with integrin a2 relative to TbRII (**p < 0.001, mean ± SD, D, Figure 3source data 1). This organization is also present in ATDC5 cells when the fluorescent labels for TbRII and integrin a2 have been switched (E), in osteosarcoma Saos-2 cells (F), or in epithelial MCF10A cells (G), when labeling focal adhesions with integrin aV (G), and when TbRII is expressed and imaged alone (H). TbRII spatial organization is unaffected by addition of TGFb, indicated by red outlines in the same cellular region following 15 min of TGFb treatment (I). See Source code 2. DOI: 10.7554/eLife.09300.007 The following source data and figure supplement are available for figure 3: Source data 1. Colocalization Index DOI: 10.7554/eLife.09300.008 This analysis reveals that integrin a2 colocalizes significantly more with Alk5 and Alk1 than with TbRII ( Figure 3D). The specific localization of TbRII near focal adhesions is apparent in cells of both mesenchymal (ATDC5, Figure 3A-C,E; Saos-2, Figure 3F) and epithelial (MCF10A, Figure 3G) origin and is observed whether integrin a2 or integrin aV is tagged with a fluorescent protein ( Figure 3). Furthermore, this observation still holds if the fluorescent labels for TbRII and integrin a2 are switched, as well as if TbRII is expressed and imaged alone ( Figure 3E,H). The overall spatial organization of TGFb receptors at sites of adhesion is not affected upon stimulation with exogenous TGFb ( Figure 3I), suggesting that this spatiotemporal organization is regulated through mechanisms independent from TGFb ligand addition. Given the critical role of integrins in mechanotransduction and the known sensitivity of TGFb signaling to cellular tension (Allen et al., 2012), the unique pattern of TGFb receptor and integrin localization could prime TGFb receptors for regulation by elements of the mechanotransduction pathway.

Focal adhesions immobilize TbRI and limit the integration of TbRII
To investigate the effect of focal adhesions on TbRI (Alk5) and TbRII dynamics, we used sptPALM to visualize TGFb receptor trajectories near or within these vinculin-rich domains ( Figure 4A,B). SptPALM shows, both qualitatively and quantitatively, that TbRI is preferentially enriched and TbRII is preferentially excluded at sites of adhesion ( Figure 4A-C). Analysis of individual TbRII trajectories shows that TbRII 'bounces' around the edges of individual focal adhesions ( Figure 4E) but is rarely incorporated within the focal adhesion, as is common for TbRI ( Figure 4D,i-ii). To determine if focal adhesions shifted the fractions of freely diffusive, confined, or immobile receptors, TbRI and TbRII . Quantification of these regions shows that TbRI is preferentially enriched inside adhesions relative to outside, and that TbRII is preferentially excluded at these same sites (*p < 0.01, mean ± SD, C). Representative single molecule trajectories show sequestration of TbRI in focal adhesions (D, i-ii) and free diffusion outside adhesions (D, iii-iv), whereas TbRII bounces around the edges of focal adhesions in a freely diffusive (E, i-ii) or confined (E, iii-iv) manner. Analyzing TbR trajectories at focal adhesions based on diffusion (Red: Immobile, Green: Confined, Blue: Freely Diffusive) shows a higher density of tracks inside adhesions for TbRI (F) compared to TbRII (G), and demonstrates a higher fraction of immobile TbRI tracks inside relative to outside adhesions (H). The diffusion coefficient of TbRI trajectories decreases inside adhesions (mean ± SD, I). See Source code 1 and trajectories near sites of adhesion were mapped based on receptor mobility. Trajectory maps reveal that TbRII mobility is confined near focal adhesions, which sequester and immobilize TbRI ( Figure 4F,G). Indeed, a higher fraction of immobilized TbRI is present inside adhesions relative to outside ( Figure 4H). Accordingly, the diffusion coefficient for TbRI decreases for tracks inside adhesions compared to those outside, demonstrating that this spatial organization specifically limits TbRI mobility ( Figure 4I). The differential localization and dynamics of TbRI and TbRII in adhesion-rich domains, relative to one another and to the whole cell TGFb receptor population, indicates that this spatial control has functional implications for TGFb signaling and for mechanotransduction.

TGFb receptors form complexes with integrin aV and the actin-binding protein cofilin
To determine whether these changes in receptor mobility at sites of adhesion are due to direct or indirect physical interactions with other proteins, we performed mass spectrometry and co-immunoprecipitation experiments. Mass spectrometric analysis of proteins that precipitate with Flag-tagged TbRI (Alk5) and TbRII revealed hundreds of proteins, several of which were specifically enriched compared with precipitates of untransfected (mock) cells. The analysis identified proteins already known to interact with TbRs, such as PRMT5 and PRMT1 (Xu et al., 2013). TbRs also precipitated several adhesion-related proteins, including integrin aV and endogenous cofilin, as shown in the annotated spectra ( Figure 5A,B). The peptide counts (graph insets) indicate that integrin aV associates with both TbRI and TbRII, and that cofilin preferentially associates with TbRII ( Figure 5A,B). Cofilin is an actin-binding protein that severs ADP-actin filaments at the leading edge of migratory cells (Pollard and Borisy, 2003). Previous reports implicate cofilin as a target of TGFb-activated RhoA, which promotes actin reorganization through ROCK, LIMK and cofilin (Vardouli et al., 2005;Lamouille et al., 2014). However, this is the first report, to our knowledge, of a complex between TGFb receptors and cofilin. To confirm these mass spectrometry findings, we performed co-immunoprecipitation on cells expressing Flag-tagged TbRI/II and tagged integrin aV or cofilin ( Figure 5C,D). Consistent with the mass spectrometry peptide counts, integrin aV forms a complex with both TbRI and TbRII, whereas cofilin primarily interacts with TbRII. Although the novel finding of a complex formation, either through direct or indirect interactions, between TbRII and cofilin remains to be further explored, it suggests a potential mechanism underlying the discrete spatial organization of TbRII at focal adhesions.

Cellular tension regulates TGFb receptor organization at focal adhesions
Integrins transmit changes in the physical microenvironment across the plasma membrane to modulate cellular tension and signaling. The presence of a focal adhesion-associated TGFb-receptor population suggests a novel mechanism by which cellular tension may regulate TGFb signaling. To test the hypothesis that TGFb receptor organization at focal adhesions is sensitive to cellular tension, we Colocalization quantification (Ai,Bi,Ci) demonstrates that TbRII is significantly more colocalized with integrin a2 post-treatment (Y27632, blebbistatin) relative to pre-treatment (**p < 0.001, mean ± SD, E, Figure 6-source data 1). Disruption of tension with Y27632 enhances integrin aV association with TbRI but reduces its association with TbRII (F). See Source code 2. DOI: 10.7554/eLife.09300.013 The following source data is available for figure 6: Source data 1. Colocalization Index (vehicle and treatment) DOI: 10.7554/eLife.09300.014 treated ATDC5 cells with the ROCK inhibitor Y27632 or the myosin II inhibitor blebbistatin. Within 15 min of adding Y27632 ( Figure 6A,B) or blebbistatin ( Figure 6C,D), the peripheral ring of TbRII completely collapses. The segregation of TbRII from TbRI and integrin a2 at sites of adhesion is dynamically released, such that TbRII (Video 1) converges and colocalizes with integrin a2 (Video 2, Video 3). Quantitative analysis demonstrates that TbRII is significantly more colocalized with integrin a2 after addition of Y27632 and blebbistatin ( Figure 6E).
To assess the effect of cellular tension on physical associations among TbRI, TbRII and integrins, we performed co-immunoprecipitation experiments. We find that cellular tension not only regulates the spatial organization of integrins and TGFb receptors, but also affects their physical associations with each other; though these interactions may be direct or indirect. Specifically, while disruption of tension with the ROCK inhibitor enhanced integrin aV association with TbRI, it almost completely blocked the association between integrin aV and TbRII ( Figure 6F).

Tension-sensitive regulation of TGFb receptor heteromerization and signaling
Since a reduction in cellular tension drives colocalization of TbRI and TbRII, we sought to determine if this change in spatial organization had functional consequences for TGFb signaling. We first evaluated the effect of reduced cellular tension on TbRI/TbRII heteromerization using co-immunoprecipitation. Release of this discrete spatial segregation of TGFb receptors at focal adhesions allows the receptor subunits to interact such that ROCK-inhibition stimulates formation of heteromeric TbRI/ TbRII complexes ( Figure 7A). To examine the effect of manipulating cellular tension under physiological conditions, we cultured cells on polydimethylsiloxane (PDMS) substrates of varying stiffness. A reduction in cellular tension through culture on compliant substrates significantly drives TbRI/TbRII complex formation ( Figure 7B). Therefore, a reduction in cellular tension, due to pharmacologic ROCK inhibition or changes to the stiffness of the microenvironment, drives formation of a multimeric TbRI/TbRII complex that is required for the activation of downstream TGFb effectors.
To determine the effect of tension-sensitive TbR localization and heteromerization on downstream TGFb effectors, we evaluated the phosphorylation of Smad3. Culturing cells on compliant substrates leads to significantly increased endogenous Smad3 phosphorylation ( Figure 7C). Interestingly, the effect of TGFb on Smad3 phosphorylation is substrate-dependent, such that TGFb induces Smad3 phosphorylation on 0.5 kPa substrates but not on 16 kPa substrates ( Figure 7C). This is consistent with the established non-linear response of TGFb signaling and other cellular behaviors to cellular tension (Allen et al., 2012;Rape et al., 2015). Thus the spatial organization of TbRI and Video 1. Disruption of cellular tension leads to dynamic disassembly of TbRII spatial organization at sites of adhesion ( Figure 6). TbRII-mEmerald spatial organization collapses within 15 min of adding ROCK inhibitor Y27632 in ATDC5 cells (45 min, 7 fps). DOI: 10.7554/eLife.09300.015 Video 2. Disruption of cellular tension leads to dynamic disassembly of TbRII spatial organization at sites of adhesion ( Figure 6). Integrin a2-mCherry adhesions disassemble within 15 min of adding ROCK inhibitor Y27632 in ATDC5 cells (45 min, 7 fps). DOI: 10.7554/eLife.09300.016 TbRII by integrins at focal adhesions affords tension-sensitive control of TbRI and TbRII multimerization and activation of Smad3, providing a mechanosensitive mechanism by which cells calibrate their response to TGFb.

Discussion
Here we show that cellular tension regulates TGFb receptor spatial organization and interactions at focal adhesions, providing a novel mechanism for the cellular integration of signaling by physical and biochemical cues. We observe a novel spatiotemporal regulation of the TGFb pathway such that TbRII is segregated from TbRI and integrins at sites of adhesions. Single particle tracking reveals the dynamics of individual TGFb receptor molecules, and identifies populations of TGFb receptors with distinct behaviors and mobility near and far from sites of focal adhesions. The confined population of TGFb receptors at focal adhesions has lower mobility than the freely diffusive receptor population far from sites of adhesion. TGFb receptors associate with several adhesion-related proteins, including the actin-binding protein cofilin, which preferentially associates with TbRII relative to TbRI. This novel spatial organization of TbRI and TbRII at sites of adhesion provides mechanosensitive control of TGFb receptor multimerization and function independently of TGFb ligand stimulation. Overall, this reveals the potential of two differentially regulated populations of TGFb receptors -one that is TGFb-sensitive and one that is tension-sensitive -a finding that may contribute to the context-dependent signaling outcomes of this pathway.
This tension-dependent mechanism for the regulation of TGFb receptors has a number of interesting functional implications. At the level of the TGFb ligand, integrins activate TGFb from its latent form through cellular tension generated by actomyosin contraction (Wipff et al., 2007;Munger and Sheppard, 2011;Wells and Discher, 2008;Giacomini et al., 2012). The observed recruitment of TGFb receptors to focal adhesions would enrich their access to this reservoir of integrin-activated TGFb. At the receptor level, focal adhesions may sequester TbRI from TbRII to limit their activity in the presence of ligand. The extent to which this sequestration is cofilin-dependent requires further investigation. This sequestration of TbRI may contribute to its slow internalization, relative to TbRII, following TGFb stimulation (Vizan et al., 2013;Ma et al., 2007). Alternatively, focal adhesions may create structured TbRI and TbRII boundaries that prime a robust response when cells encounter the correct combination of physical and biochemical cues. We demonstrate tension-sensitive regulation of endogenous downstream Smad3 phosphorylation by cellular tension and TGFb. In addition, chondrocytes grown in TGFb on 0.5 MPa substrates induce differentiation markers far beyond levels induced by either cue alone (Allen et al., 2012). We and others have reported that the effect of substrate stiffness or cellular tension/Rho/ROCK activity on downstream TGFb signaling is synergistic and nonlinear (Allen et al., 2012;Wang et al., 2012;Leight et al., 2012). Therefore, it is possible that lower cell tension in one cell type may have a differential effect on Smad phosphorylation, nuclear localization, and transactivation than in another cell type. It would be interesting to examine this effect utilizing a substrate system that provides independent and continuous gradients of ligand density and substrate stiffness (Rape et al., 2015). The mechanisms responsible for such synergy have been unclear, but this newly described regulation of TGFb receptor multimerization and downstream signaling may couple the mechanosensitive activity of the TGFb pathway to physical cues. Fully understanding the functional implications of this spatially-distinct TGFb receptor population will require the development of new imaging tools, such as those that can dynamically visualize TGFb effector activity locally at focal adhesions.
Video 3. Disruption of cellular tension leads to dynamic disassembly of TbRII spatial organization at sites of adhesion ( Figure 6). Composite of TbRII and integrin a2 (Video 3) demonstrate a tension-sensitive collapse of this discrete spatial organization at sites of adhesion and a reorganization at the cell periphery. DOI: 10.7554/eLife.09300.017 The current study of TGFb receptors opens the possibility that tension-sensitive receptor multimerization may underlie mechanosensitive signaling by other pathways. Cellular tension impacts the activation, translocation, and function of intracellular effectors including small GTP-ases, kinases and transcriptional regulators such as Smads and YAP/TAZ (Allen et al., 2012;Wang et al., 2012;Dupont et al., 2011;Leight et al., 2012). However, known mechanisms are insufficient to explain the ability of physical cues to modulate cell-type specific responses to BMP, EGF, and other growth factors (Wang et al., 2012;Paszek et al., 2005). Several receptor families share features with TGFb receptors that may contribute to their mechanosensitivity, such as their association with integrin-rich focal adhesions and their potential for the formation of stable receptor clusters by geometric constraints (Bethani et al., 2010;Salaita et al., 2010;Hartman and Groves, 2011). Previous studies have established important physical and functional links between focal adhesion components and growth factor receptors. EGF receptor binds actin and colocalizes with integrin a2b1 (Alam et al., 2007), while the receptor CD44 interacts with several components of the focal adhesion complex, such that hyaluronan-bound CD44 activates c-Src and Rac1 (Turley et al., 2002). Aside from TGFb, shown herein, the extent to which these physical associations contribute to mechanosensitive control Figure 7. Disruption of tension-sensitive TbR segregation increases TbRI/TbRII multimerization and phosphorylation of Smad3. ROCK inhibition releases the discrete spatial organization of TbRs at focal adhesions and drives the formation of heteromeric TbRI/TbRII complexes within 15 min of Y27632 exposure (A), as shown by Flag co-immunoprecipitation (IP) and immunoblotting (IB). Likewise, manipulation of cellular tension through culturing cells on collagen II-coated glass or 0.5 kPa PDMS substrates increases co-immunoprecipitation of TbRI with Flag-tagged TbRII (p < 0.05, B). In cells grown on collagen II-coated compliant (0.5 kPa, p < 0.05) or stiff (16 kPa) PDMS substrates, endogenous Smad3 phosphorylation is increased (C). The effect of TGFb on Smad3 phosphorylation is substrate-dependent, such that maximal TGFb-inducibility is observed on 0.5 kPa substrates (p < 0.05), consistent with a tension-sensitive calibration of TbR localization and activity (C). See Figure 7 -source data. DOI: 10.7554/eLife.09300.018 The following source data is available for figure 7: Source data 1. Western Quantitative Analysis DOI: 10.7554/eLife.09300.019 of receptor multimerization or downstream signaling remains to be determined. Nonetheless, others have postulated receptor multimerization as a mechanism for mechanocoupling of TGFb, ephrin, and T cell receptor signaling (Hartman and Groves, 2011;Hynes, 2009). In each case, the solid state presentation of the ligand is thought to play a critical role in structuring multimeric receptor clusters. In T cell receptor and ephrin signaling, the solid state is provided by ligands on the neighboring cell, which create geometric constraints that mechanically trap receptors to induce clustering (Bethani et al., 2010;Salaita et al., 2010;Hartman and Groves, 2011). For growth factors like TGFb, BMP, and EGF, the ECM serves as the solid state (Hynes, 2009). ECM proteins such as collagen II bind both TGFb and integrin a2b1 (Zhu, 1999), imposing geometric constraints that may structure receptor clusters. Therefore, growth factor receptor multimerization at focal adhesions, controlled by receptor interactions with integrins and with solid state growth factors, provide focal adhesions with the capability to integrate signaling between physical and biochemical cues.
Understanding the mechanosensitive regulation of TGFb signaling has significant biological implications. We find that focal adhesions segregate TbRI from TbRII in both epithelial and mesenchymal cell lineages, opening the possibility that this is a general cellular mechanism for the control of TGFb signaling. It will be interesting to determine if TGFb receptor multimerization at focal adhesions responds to physical cues that aberrantly promote TGFb-induced epithelial-mesenchymal transition (EMT) in cancer or the loss of chondrocyte homeostasis in osteoarthritis. On stiff substrates, TGFb preferentially activates PI3K to induce EMT instead of apoptosis (Leight et al., 2012). In osteoarthritis, the material properties of the cartilage ECM deteriorate as chondrocytes inappropriately shift the balance from canonical (Alk5/Smad2/3) to non-canonical (Alk1/Smad1/5/8) TbRI signaling (Blaney Davidson et al., 2009). In each case, the extent to which changing the physical environment alters TGFb effector selection through differential TGFb receptor multimerization remains to be determined. Applied physical cues, such as compression or shear flow, also regulate TGFb signaling in cartilage, vasculature, and other tissues (Li et al., 2010;Sakai et al., 1998;Streuli, 1993). Whether similar mechanisms operate in response to exogenous physical cues remains to be elucidated.
In conclusion, we utilized novel high-resolution imaging and single particle tracking microscopy coupled with biochemical assays to explore the spatial organization of TGFb signaling at the receptor level. At focal adhesions, TbRII is uniquely segregated from its TbRI counterpart. Cellular tension modulates the spatial organization, multimerization, and downstream signaling of TGFb receptors at sites of adhesion, suggesting the existence of a functionally distinct subpopulation of TGFb receptors. Overall, this finding provides a new mechanism by which cellular tension and physical cues exert control of growth factor signaling at the cellular membrane.

Plasmids
The plasmids pRK5 TGFb type I receptor Flag and pRK5 TGFb type II receptor Flag were gifts from Rik Derynck (Addgene plasmids 14,831 (Feng and Derynck, 1996), 31719). The plasmid pRK5 TGFb type I receptor Myc was also a gift from Rik Derynck. All fluorescent protein expression vectors are available in the Michael Davidson Fluorescent Protein Collection on Addgene. All fluorescent protein expression vectors were constructed using C1 or N1 (Clontech-style) cloning vectors and initially characterized using the advanced EGFP variant mEmerald to verify proper localization of the fusions. To construct the N-terminal (with respect to the fluorescent protein) human integrin alpha2 (NM_002203.3) fusions, the following primers were used to amplify the integrin alpha 2, and create the 18-amino acid linker (GSAGGSGVPRARDPPVAT): XhoI forward: CTC CGT CTC GAG ACC GCC ATG GGG CCA GAA CGG ACA GGG GCC KpnI reverse: CGG AAC GGT ACC CCG CTT CCG CCT GCG CTG CCG CTA CTG AGC TCT GTG GTC TCA TCA ATC TCA TCT GGA TTT TTG GTC Following digestion and gel purification, the PCR product was ligated into a similarly digested mEmerald-N1 cloning vector to produce mEmerald-Integrin alpha2-N-18. The resulting fusion, along with mCherry-N1 cloning vector, was sequentially digested with AgeI and NotI to yield mCherry-Integrin alpha2-N-18. To generate the N-terminal human integrin alpha V (NM_002210.4) fusions, the following primers were used to amplify the protein and create the 25-amino acid linker (PGSRAQASNSAVDGTAGPGSPPVAT): AgeI forward : CCC GGG ATC CAC CGG TCG CCA CCA TGG CTT TTC CGC CGC GGC GAC GGC  TGC GCC TCG GTC  HindIII reverse: AAT TGA AGC TTG AGC TCG AGA TCC CGG AAG TTT CTG AGT TTC CTT CAC  CAT TTT CAT GAG GTT GAA GCT GCT CCC TTT CTT GTT CTT CTT GAG The PCR product was digested, gel purified, and ligated into a similarly treated mEmerald-N1 or mCherry-N1 cloning vector to yield the mEmerald-Integrin alphaV-25 or mCherry-Integrin-alphaV-25 fusions. To construct the N-terminal tagged human vinculin (NM_003373.3) plasmids, the following primers were used to PCR-amplify and create a 21-amino acid linker (SGGSGILQSTVPRARDPPVAT): NheI forward: GTC AGA TCC GCT AGC ACC GCC ACC ATG CCA GTG TTT CAT ACG CGC ACG ATC GAG AGC EcoR1 reverse: CGA CTG CAG AAT TCC GCT GCC ACC GGA CTG GTA CCA GGG AGT CTT TCT AAC CCA GCG CAG The PCR product was digested and ligated into a similarly cut mEmerald-N1 or mCherry-N1 cloning vector to yield mEmerald-Vinculin-N-21 or mCherry-Vinculin-N-21 expression vectors. To construct the C-terminal human TbR2 (NM_001024847.2) fusion plasmids, the following primers were used to amplify the TbR2 and generate an 18-amino acid linker (SGLRSRESGSGGSSGSGS): XhoI forward: GAC GAG CTC GAG AGA GTG GCT CTG GTG GGT CGA GTG GAA GTG GCA GCG GTC GGG GGC TGC TCA GGG GCC TG BamHI reverse: CGT CTA GGA TCC CTA TTT GGT AGT GTT TAG GGA GCC GTC TTC AGG AAT CTT CTC C Following digestion and gel purification, the PCR product was ligated into a similarly digested mEmerald-C1 cloning vector, to produce mEmerald-TbRII-C-18. The fusion, along with mCherry-C1 and mEos2-C1 cloning vectors, was sequentially digested with AgeI and BamHI and ligated to yield mCherry-TbRII-C-18 and mEos2-TbRII-C-18. To generate the N-terminal human TbRII plasmids and create an 18-amino acid linker (SSGGASAASGSADPPVAT), the following primers were used: NheI forward: CGA TCC GCT AGC GCC ACC ATG GGT CGG GGG CTG CTC AGG GGC BamHI reverse: CCT GTA CGG ATC CGC GCT ACC ACT GGC TGC GCT TGC TCC ACC GCT GCT TTT GGT AGT GTT TAG GGA GCC GTC TTC AGG AAT CTT CTC C The PCR fragment was digested, gel purified, and ligated with a similarly treated mEmerald-N1 cloning vector to produce mEmerald-TbRII-N-18. The resulting fusion, along with mCherry-N1 and mEos2-N1, was double digested with BamHI and NotI to yield mCherry-TbRII-N-18 and mEos2-TbRII-N-18 respectively. To construct the N-terminal human Alk1 (NM_000020.2) expression vectors, the following primers were used to amplify the plasmid, and create a 13-amino acid linker (GSAGGSGDPPVAT): EcoRI forward: GCG TTG AAT TCA CCG CCA TGA CCT TGG GCT CCC CCA GGA AAG GCC BamHI reverse: CGG AAC GGA TCC CCG CTT CCG CCT GCG CTG CCT TGA ATC ACT TTA GGC TTC TCT GGA CTG TTG CTA ATT TTT TGT AGT GTC TTC TTG ATC Following amplification, the PCR fragment was digested, purified, and ligated to a similarly treated mEmerald-N1 cloning vector to yield mEmerald-Alk1-N-13. Upon sequence verification, the resulting fusion, along with mCherry-N1, was digested with BamH1 and NotI and ligated to yield mCherry-Alk1-N-13. To generate the C-terminal human Alk5 (NM_004612.2) expression vectors, the following primers were used to amplify the plasmid, and create an 18-amino acid linker (SGLRSGSSAGSASGGSGS): BglII forward: GAC TCG AGA TCT GGC TCC AGC GCA GGC AGC GCA TCC GGC GGA AGC GGA AGC GAG GCG GCG GTC GCT GCT CCG CGT C HindIII reverse: CGG TCA AAG CTT TTA CAT TTT GAT GCC TTC CTG TTG ACT GAG TTG CGA TAA TGT TTT CTT AAT CCG C Following amplification, the PCR fragment was digested, purified, and ligated to a similarly treated mEmerald-C1 cloning vector to yield mEmerald-Alk5-C-18. The resulting fusion, along with mCherry-C1 and mEos2-C1, was double digested with BglII and NheI to yield mCherry-Alk5-C-18 and mEos2-Alk5-C-18. To construct the N-terminally labeled Alk5 fusions, the following primers were used to amplify the Alk5 and generate a 13-amino acid linker (GSGGAGGGGPVAT): BglII forward: GTC TGT AGA TCT GCC ACC ATG GAG GCG GCG GTC GCT GCT CCG Affinity purification and reversed-phase liquid chromatographyelectrospray tandem mass spectrometry (LC-MS/MS) Cells expressing TbRII-Flag or TbRI-Flag and integrin aV-mCherry in 10 cm cell culture dishes were lysed as above and affinity-purified with M2-Flag magnetic beads (Sigma-Aldrich), followed by onbead trypsin digestion (Kean et al., 2012) and mass spectrometry approaches to study associated proteins (N=3). Peptides recovered were analyzed by reversed-phase liquid chromatography-electrospray tandem mass spectrometry (LC-MS/MS) as described (Duong et al., 2015). Briefly, peptides were separated by nano-flow chromatography in a C18 column, and the eluate was coupled to a hybrid linear ion trap-Orbitrap mass spectrometer (LTQ-OrbitrapVelos, Thermo Fisher Scientific, Waltham, MA) equipped with a nanoelectrospray ion source. Following LC-MS/MS analysis, peak lists generated from spectra were searched against the human subset of the SwissProt database using in-house ProteinProspector (Clauser et al., 1999). For analysis, peptide counts of each protein were normalized by the total protein content in the sample and the molecular weight of the respective protein. This provided an abundance index for each protein that served as a comparison between pulldowns. The ratio between abundance indices for TbR pulldowns to untransfected control (mock) pulldowns was used to screen candidate proteins.
Image acquisition and analysis ATDC5, NIH3T3, and MCF10A cells were transiently transfected with fluorescently labeled expression plasmids and plated on collagen II, fibronectin or poly-l-lysine-coated glass-bottom imaging wells. Cells were imaged 24 hr after transfection, and treated with Y27632, blebbistatin, or TGFb as indicated. Confocal images were obtained on a motorized Yokogawa CSU-X1 spinning disk confocal unit on an inverted microscope system (Ti-E Perfect Focus System, Nikon, Tokyo, Japan), with either a 100X/NA 1.49 oil-immersion objective (CFI Apo TIRF, Nikon) or a 40X/NA 1.15 water-immersion objective (CFI Apo LWD, Nikon), on a front illuminated CMOS camera (Zyla sCMOS, Andor, Belfast, United Kingdom). For TIRF and sptPALM, imaging was performed on a motorized objective-type TIRF inverted microscope system (Ti-E Perfect Focus System, Nikon) with activation and excitation lasers of 405 nm, 488 nm, and 561 nm, and an electron-multiplying charged-coupled device camera (QuantEM 512, Photometrics, Tuscon, AZ), a 100X/NA 1.49 oil-immersion objective (CFI Apo TIRF, Nikon), a stage top incubator (Okolab, Burlingame, CA), and controlled by NIS-Elements software (Nikon). Cells expressing mEos2-tagged constructs were simultaneously activated with a 405 nm laser and excited with a 561 nm laser. Laser intensities were adjusted to maintain a constant sparse population of activated molecules that were spaced well enough for accurate localization and tracking. Prior to each sptPALM imaging sequence and photoconversion of mEos2, the mEmerald signal from mEmerald fusions of vinculin was imaged to localize focal adhesions. NIS-Elements software (Nikon) was used for the acquisition of images at 10 fps. Individual receptors were localized and tracked using a previously described algorithm (Sbalzarini and Koumoutsakos, 2005) written in MosaicSuite for ImageJ and available at (www.mosaic.mpi-cbg.de). All images were processed using ImageJ with a 0.6 gaussian blur filter to remove noise. Images shown are representative of multiple cells (N!5) for at least three independent experiments for each condition.

Colocalization quantification
TIRF mode imaging was used to obtain intensity profiles of two distinct molecules over adhesionrich regions of interest (for example, regions shown in Figure 3A-C). The similarity of the two profiles was quantified to provide a measure of colocalization, specifically by comparing pixel intensities (8-bit grayscale) at each point across the two profiles. For each pixel, an ordered pair containing the intensities at that particular coordinate from both images was plotted. Values closer to the line y=x refer to coordinates that have very similar intensities in both profiles. Values further from y=x are coordinates that have a mismatch in intensities. By reflecting all points in the top half of this graph across y=x, a distribution of points is created between y=x and the x-axis, but the distance of individual points from y=x is preserved. The magnitude of the slope of the regression line through these points can be used as a quantitative metric of colocalization. The greater this slope, the higher the degree of colocalization. Plots are representative of multiple cells (N!3) and multiple regions of interest (N=5). ANOVA followed by Bonferroni correction was used to evaluate statistical significance.

Single-molecule tracking
Each sptPALM imaging sequence generates tens of thousands of molecule trajectories per cell (N=6 cells for each TbR). From these, only trajectories lasting between 0.5 seconds and 2 seconds (5 to 20 frames at 10 fps) were selected for analysis. Tracks that were not confined to either inside or outside focal adhesions were not considered in the quantitative analysis. For each individual track, a series of parameters were calculated to quantify receptor dynamics. These parameters include mean squared displacement (MSD), diffusion coefficient (D), and radius of confinement (r conf ). MSD was computed as per Equation 1 (Rossier et al., 2012): Where x i and y i are the coordinates of the molecule at time i Ã Dt and N is the number of frames for which the trajectory persisted. The radius of confinement (r conf ) of a track is defined to be the magnitude of the radius of the smallest circle that encloses all points in that track. D is defined as one-fourth of the slope of the regression line fitted to the first four values of the MSD as per Equation 2.
MSDðtÞ ¼ 4Dt: (2) Using these variables, trajectories were pooled into three fractions: immobile, confined, and freely diffusive. Immobile molecules were defined as being restricted to a radius of confinement equal to one pixel (r conf < 0.166 mm). Confined molecules were defined as non-immobile tracks with a diffusion coefficient D of less than 0.2 mm 2 /s, and the remaining tracks were considered freely diffusive. Custom routines written for Python were used for track quantification, analysis, and visualization (source code). To account for variability in these large data sets consisting of tens of thousands of tracks, we report mean and standard error of the mean (SEM). ANOVA followed by Bonferroni correction and student's t test were used to evaluate statistical significance.
To quantify the diffusive behavior of TbRI and TbRII around focal adhesions, we calculated an enrichment ratio to compare track densities in several (N!5) focal adhesion-rich regions (where vinculin covered more than 25% of total area) across at least three cells. The enrichment ratio was defined as the ratio of the track density (tracks/mm 2 ) inside adhesions to the track density outside adhesions within a given area. Student's t test was used to evaluate statistical significance.

Statistical analysis
For colocalization quantification (Figure 3 and Figure 6) and enrichment ratio ( Figure 4C), we report mean and standard deviation (SD). There are three circumstances in which it was more statistically appropriate to report the standard error of the mean (SEM). Specifically, to account for variability in large sptPALM data sets (consisting of tens of thousands of tracks), we report mean and SEM ( Figure 2E,F and Figure 4I). Significance was calculated with ANOVA followed by Bonferroni correction and student's t test, with significance defined as p<0.01. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Additional information
Author contributions JPR, CCDuF, DAM, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article; MAB, MWD, Drafting or revising the article, Contributed unpublished essential data or reagents; JAOP, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article; SC, Acquisition of data, Analysis and interpretation of data; ALB, Conception and design, Analysis and interpretation of data; TNA, Conception and design, Analysis and interpretation of data, Drafting or revising the article