SUMMARY
GPI-anchored protein (GPI-AP) nanoclusters are generated by cortical acto-myosin activity. While our understanding of the physical principles behind this process is emerging, the molecular machinery required for the generation of these nanoclusters is unknown. Here, we show that ligand-mediated membrane receptor signaling triggers nanocluster formation. Both soluble and surface-tethered RGD ligands bind the β1-integrin receptor and activate focal adhesion and src-kinases, resulting in RhoA signaling. This cascade ultimately triggers actin-nucleation via specific formins, driving nanoclustering of both GPI-APs and a model transmembrane protein with an actin-binding domain. Integrin signaling concurrently results in talin mediated activation of vinculin. This is necessary for the coupling of the dynamic actin machinery to the inner leaflet driving GPI-AP nanoclustering. Disruption of GPI-AP nanoclustering in either GPI-anchor remodeling mutants or in cells that express vinculin mutants, provide evidence that these nanoclusters are necessary for activating cell spreading, a hallmark of integrin function.
INTRODUCTION
Sub-compartmentalization of the plasma membrane (PM) via the lateral segregation of proteins and lipids into structural and signaling platforms is likely to play pivotal roles in the spatio-temporal regulation of many signaling systems. Finely tuned signaling systems such as T-cell receptor triggering at the immunological synapse (Gaus et al., 2005), B-cell receptor activation (Gupta and DeFranco, 2007; Mattila et al., 2013), and cell-ECM adhesion (Gaus et al., 2006; Lingwood and Simons, 2010; Simons and Toomre, 2000; van Zanten and Mayor, 2015), involve the generation of membrane domains. Such membrane domains, enriched in cholesterol, sphingolipids and outer leaflet lipid-tethered glycosylphosphatidylinositol (GPI)-anchored proteins, have often been termed as membrane ‘rafts’ (Sezgin et al., 2017).
The mechanism whereby cells generate these domains remains controversial. The size, scale and statistics of membrane heterogeneities in the cell are very different than what is predicted from thermodynamically driven phase segregation observed in artificial membranes or cell-free membrane preparations such as Giant Plasma Membrane Vesicles (GUVs) (Chiantia and London, 2012; Sezgin et al., 2012a, 2012b). Many of the ‘raft’ components such as outer leaflet GPI-APs or inner leaflet Ras molecules form nanoclusters (Plowman et al., 2005; Varma and Mayor, 1998). These nanoscale clusters are hierarchically organized into larger scale optically resolvable (mesoscale) domains where a significant fraction of the lipid-anchored proteins are present as nanoclusters (Goswami et al., 2008; Tian et al., 2007; van Zanten et al., 2009).
In the ‘resting’ state, GPI-APs at the cell surface are distributed as monomers with a small fraction of nanoclusters (20-40%) that is independent of total protein expression levels (Sharma et al., 2004; van Zanten et al., 2009). Under conditions of activation such as the binding of ligand to the integrin receptor, LFA-1, in immune cells, the fraction of GPI-APs in nanoclusters increases (~ 80%), and appears in spatial proximity to LFA-1 nanoclusters, regions designated as "hot-spots". Reduction of cholesterol levels, a treatment that also prevents the formation of GPI-AP nanoclusters, drastically inhibited the ligand binding capacity of these adhesion receptors (van Zanten et al., 2009). At the same time ligand-induced or crosslinking antibody-induced clustering of GPI-APs is sufficient to drive downstream signaling responses in the cell (Harder et al., 1998; Stefanová et al., 1991; Suzuki et al., 2007). The regulation of nanoscale clustering is, thus, likely to be an important determinant in high-fidelity signal transduction processes that operate at the cell surface (Harding and Hancock, 2008; Tian et al., 2007).
Extensive studies on the organization and dynamics of GPI-APs have indicated a crucial role for a dynamic actin layer at the cortex juxtaposed to the inner leaflet of the PM in the formation of such nanoclusters at the outer leaflet (Goswami et al., 2008; Saha et al., 2015). In previous work, we had proposed that dynamic actin filaments along with myosin motors form transient remodeling contractile platforms (asters) at the inner leaflet (Gowrishankar et al., 2012). These ‘asters’ immobilize clusters of the long-acyl chain containing phosphatidylserine (PS) at the inner leaflet which interact with long-acyl chain containing GPI-APs at the outer leaflet via a transbilayer coupling interaction, thereby creating nanoclusters (Raghupathy et al., 2015). These and other observations (Köster and Mayor, 2016; Rao and Mayor, 2014) indicate that the organization at the membrane might be better understood as an active actin-membrane composite, wherein the constituents in the fluid membrane bilayer interact with the dynamic actin cortex. In this context, membrane components can be classified into three classes based on their ability to couple with and regulate this active machinery: inert, passive and active (Gowrishankar et al., 2012). Inert molecules are those that are unable to interact with the underlying actin (for example unsaturated lipids at the outer leaflet or membrane proteins that lack any linkage to actin filaments), passive molecules are those that bind (and unbind) actin filaments (such as the GPI-APs as well as transmembrane proteins that possess actin-binding motifs at their cytoplasmic tails), and active molecules which not only bind but also influence the actin cytoskeleton dynamics at the membrane and in doing so could regulate local membrane organization.
Despite the emergence of a theoretical understanding of the active mechanics behind the generation of the nanoscale assemblies and their distribution and dynamics, the molecular machinery for their formation has been missing; the nucleators of actin, triggers of myosin function, and the linkage between the actin and the PS lipid are uncharacterized. In this manuscript we uncover the molecular machinery that governs the formation of the dynamic actin-based membrane patterning system.
Here we show that integrin receptors behave as a prime example of an ‘active’ molecule that regulates actin nucleators and myosin activity necessary to build the hierarchical organization of clusters. We find that upon engagement with RGD-containing ligands, integrin receptors through their ability to activate the FAK and src kinases and the resultant RhoA activation trigger formins necessary for the generation of the dynamic actin filaments. RhoA also activates the ROCK pathway, required for myosin activation. Importantly, we also identify vinculin, a ubiquitous protein that associates with focal adhesions, as a molecule necessary for directing the generated dynamic actin filaments to the inner-leaflet lipids and thereby generating GPI-AP-nanoclusters. Furthermore, using GPI-anchor remodeling mutants as well as vinculin mutants, which fail to support nanocluster formation, we show that the nanoclusters created by this active machinery are necessary for activating cell spreading, a hallmark of integrin function.
RESULTS
Integrin activation generates nanoclusters of the outer leaflet GPI-APs
The integrin family of heterodimeric transmembrane receptors binds various extracellular ligands that activate a multitude of structural and signaling molecules (Hynes, 2002; Vicente-Manzanares et al., 2009). Integrins exhibit the hallmarks of an ‘active’ molecule that upon ligand engagement could alter the nanoscale organization of cell surface molecules in its vicinity through its ability to regulate the cortical acto-myosin network. Earlier studies of the integrin LFA1 activation in fixed immune cells have shown that upon binding to its ligand, ICAM-1, hot spots of GPI-AP clusters are formed, localized to the site of integrin activation within 30 mins of activation (van Zanten et al., 2009). This prompted us to test whether the activation of other integrins had a similar effect on the nanoclustering of GPI-APs albeit in a different cellular context. We used fluorescence emission anisotropy based microscopy to assess the extent of homo-FRET between fluorescently-tagged GPI-APs (Ghosh et al., 2012). Homo-FRET results in the lowering of anisotropy, providing a facile way to monitor nano scale clustering in intact living cells (Sharma et al., 2004; Varma and Mayor, 1998). Cells (CHO) stably expressing mEGFP or mYFP-tagged GPI (MYG-1) were deadhered and re-plated under serum-free conditions either on glass coated with 1%BSA or on glass coated with the extracellular matrix (ECM) protein fibronectin (FN) (Figure 1A). FN is capable of engaging with a specific subset of integrins (Humphries et al., 2006; Hynes, 2002) that promotes cell spreading (Figure1C), whereas the BSA surface is relatively inert to cell spreading at a similar time point (Figure 1C).
Although the amount of EGFP-GPI expressed on the cell surface is comparable (Figure1B-C; Total Intensity axis), the anisotropy (Figure 1B-C; Anisotropy axis) is much lower in cells plated on FN compared to that measured on BSA coated glass. The low (or high) anisotropy is discerned by ‘blue (or red)’ pixels in the heat map-encoded anisotropy image in Figure 1B and used throughout this manuscript.
The observed decrease in anisotropy occurs in a FN-concentration dependent manner saturating at ~10μg/ml FN solution concentration (Figure 1D). This decrease is specific for FN since when plated on 0.01% Poly-L-Lysine (that permits integrin-independent adhesion; (Schlaepfer et al., 1994) or on Laminin, Collagen-1 or Vitronectin (that engages a different subset of integrins), there is no significant reduction in the anisotropy of GFP-GPI (Figure S1A, B).
An increase in anisotropy upon photo-bleaching would indicate homo-FRET as a cause for the lowering of anisotropy (Sharma et al., 2004). Photo-bleaching YFP-GPI results in a net increase in anisotropy (Figure 1E, F) confirming our expectation. The typical profile of a linear increase is consistent with the presence of nanoclusters of YFP-GPI when cells are plated on FN (Sharma et al., 2004). The higher value of initial anisotropy and the minimal change in the YFP-GPI anisotropy value in cells plated on glass upon photo-bleaching, corroborates the low fraction of nanoclusters that are formed under this condition (Figure1E, F). Additionally, the decrease in anisotropy observed when cells are plated on FN is sensitive to the removal of cholesterol by the cholesterol-sequestering agent methyl β-cyclodextrin that disrupts nanoscale organization of GPI-APs [(Raghupathy et al., 2015); FN+mβCD; Figure 1B, C], confirming the enhancement in nanoclustering of GPI-APs on this substrate. The decrease in anisotropy occurs only for specific membrane constituents; there is a decrease in anisotropy of an exogenously incorporated fluorescent GPI analogue (NBD-GPI) (Figure 1G, I; exo-GPI) whereas an ‘inert’ fluorescent short chain-containing sphingomyelin analogue (C6-NBD-SM) incorporated into cells plated on FN does not exhibit this decrease (Figure 1H, J; exo-scSM).
The aV-class (αVβ3) and β1-class (α5β1) integrins are the primary integrins that mediate fibroblast cell spreading on FN (Humphries et al., 2006; Leiss et al., 2008).We utilized various function perturbing antibodies targeted against either the β1 or the αV class of integrins to discern which of these integrin sub-types are involved in the FN mediated generation of GPI-AP nanoclusters in human U2OS cells (Byron et al., 2009). We observed a loss in nanoclustering of GPI-APs (increase in GFP-GPI anisotropy) when U2OS cells were pre-treated with the increasing concentrations of p1-blocking antibody and subsequently plated on FN (Figure S1D-E). There was also a significant decrease in the cell spread area as a function of β1-blocking antibody concentration indicating that U2OS cells predominantly utilize the β1 integrin to spread on FN. At the highest concentration, the anisotropy values obtained were comparable to those obtained after treatment with mβCD (Figure S1D). There was no increase in GPI-AP anisotropy when U2OS cells were treated with antibodies that do not block spreading [(neutral non-function perturbing β1 antibody (K20) or Transferrin-receptor antibody (OKT9); Figure S1E-F) or αV-blocking antibody (17E6; data not shown)] and subsequently plated on FN. Additionally, U2OS cells plated on the αVβ3 ligand vitronectin (Charo et al., 1990) does not exhibit an increase in GPI-AP nanoclustering. Together, these data indicate that the enhanced nanoclustering occurs when cells are plated on FN, and this is mediated by the activated β1-class of integrins.
We next probed if an increase in affinity of the integrin for its ligand can alter the nanoclustering of GPI-APs. To test this, we plated U2OS cells on low concentrations of ligand (0.5μg/ml) in the presence of increasing amounts of Mn2+, an ion that potentiates integrin activation (Dransfield et al., 1992; Mould, 2002; Takagi et al., 2002). Prior treatment of cells with increasing amounts of Mn2+ resulted in a dose-dependent decrease in anisotropy of GPI-APs in the presence of low FN (Figure S1G-H). On high FN (10μg/ml; and higher), addition of Mn2+ did not result in a further decrease in anisotropy (Figure S1G-H) indicating no further increase in GPI-AP nanoclustering.
Taken together, these data indicate that shifting the equilibrium towards a ligand-engaged integrin, either by increasing FN density or by activation through Mn2+ promotes the generation of GPI-AP nanoclusters.
Localized nanoclustering in the vicinity of the activated integrin receptor
When plated on FN, the cells go through three major phases of behavior from a round state in suspension to a fully-flattened circular morphology (Dubin-Thaler et al., 2004). These are: Phase 0 (P0), the wetting phase mediated by the initial engagement of the integrin with its ligand; Phase 1 (P1), the rapid expansion phase where sensing of the mechanical rigidity and chemical suitability of the substrate and the establishment of a large contact area takes place (Giannone et al., 2004); and finally, Phase 2 (P2), where myosin II contractility based probing of the substrate via periodic protrusion/retraction of the cell edge and continued asymptotic spreading to maximum area is attained. These phases define critical checkpoints for progression from a suspended state to a fully spread state. This is organized in a specific spatial and temporal order involving distinct sets of protein modules in each phase (Wolfenson et al., 2015), allowing us to correlate the changes in nanoclustering of GPI-APs with these universal characteristics of a spreading cell.
We examined the effect of cell spreading (CHO) on fibronectin-coated surfaces on GPI-AP nanoclustering (Figure 2A-C; Supplementary Movie 1). At the level of a single cell, the cell surface that first comes in contact with the FN-coated area appears to be devoid of nanoclusters (red areas in Figure 2A) and starts to acquire nanoclusters (‘blue’ pixels at the cell periphery) co-incidental with the P0-P1 transition phase in cell spreading (Figure 2B and C; Pink-yellow transition zone, Figure S2B). Analysis of this behavior over a large number of cells shows that there is a sudden and consistent decrease in the steady-state anisotropy of GFP-GPI (δAnisotropy) that precedes the peak in cell expansion (δArea; that occurs in P2 phase) by ~100-200 seconds (Figure 2C).
In the P0 phase, integrin engagement and clustering is an early step in the formation of cell-ECM adhesions and is independent of force (Choi et al., 2008). To probe if the effects of integrin activation on the promotion of GPI-AP nanoclustering are force-dependent and localized to the sites of integrin activation, we employed a supported lipid bilayer (SLB) system functionalized with a mobile lipid-attached cyclic-RGD ligand (Figure S2C); Arg-Gly-Asp (RGD) ligand is the sequence motif in FN that mediates integrin-engagement (Ruoslahti, 1996). Here, the transiently immobilized fluorescently-tagged ligands serve as reporters of the ligated integrins (Yu et al., 2011, Figure S2C). This system facilitates the observation of local membrane organization during the early stages of integrin mediated cell adhesion by enabling the simultaneous tracking of the dynamics of the nascent integrin clusters and the nanoclustering of GFP-GPI in the membrane in the vicinity of this cluster. (Figure 2D; Figure S2H). Although the engaged integrin is unable to exert significant traction on the fluid bilayer which results in the inability of cells to fully spread and arrest in the P0 phase (Figure S2F), there is enhanced nanoclustering of GPI-APs on cells engaged with lipid-attached mobile RGD ligands compared to cells plated on glass (Figure S2D, E). In many cases we observe a characteristic pattern where integrin cluster formation often precedes the local decrease in anisotropy of GFP-GPI (Figure 2D; see more examples in Figure S2G, H). These results show that the activation of the FN binding integrin receptors triggers a localized change in GPI-AP nanoclustering.
The time from the initial contact until initiation of cell spreading is inversely correlated with the ligand density (Dubin-Thaler et al., 2004), suggesting that the process is triggered by the integration of chemical signals via integrin receptor engagement to its ligand. To test the possibility that the extent of GPI-AP nanoclustering is also an integral response of a chemical signaling process, we treated cells with a soluble cRGDfV peptide that has been shown to activate signaling molecules downstream of integrin receptor binding (Huveneers et al., 2008; Zhang et al., 2014). Strikingly, we also observe an increase in nanoclustering of GFP-GPI in cells plated on glass and treated with the soluble cRGDfV in a cholesterol and dose-dependent manner (Figure 2E, F) indicating that the increase in nanoclustering is triggered by a signaling response initiated by integrin-RGD-binding.
GPI-AP nanoclustering is mediated via a RhoA signaling pathway downstream of integrin activation
RhoGTPases, tyrosine kinases, and various bona fide cytoskeletal modifying proteins are involved in the steps of integrin-mediated cell spreading behavior (Vicente-Manzanares et al., 2009). To investigate the signaling pathway activated by integrins that leads to GPI-APs nanoclustering, we employed a chemical and genetic perturbation approach. Pre-treatment of cells with the src-family kinase (SFK) inhibitor, PP2 (Hanke et al., 1996) and the focal adhesion kinase (FAK) inhibitor PF 573 228 (Slack-Davis et al., 2007) results in a dramatic decrease in GPI-AP nanoclustering in cells plated on FN (Figure 3A,B). Correspondingly, FAK null fibroblasts also fail to support FN-induced GPI-AP nanoclustering (Figure S3C, D).
A downstream target of the SFK and FAK kinases during integrin mediated signaling is the Rho family GTPase member, RhoA (Cox et al., 2001; Guilluy et al., 2011; Ren et al., 1999). Increasing concentrations of the cell permeable Rho inhibitor C3 exoenzyme which specifically inhibit RhoA activity (Aktories et al., 1987; Braun et al., 1989)), also inhibited GPI-AP nanoclustering in a dose dependent manner when cells were plated on FN (Figure 3C, D). In contrast, addition of a cell-permeable RhoA activator [CN03; (Flatau et al., 1997; Schmidt et al., 1997)] induced enhanced nanoclustering of GPI-APs even when cells were plated on plain glass, a condition where there is minimal integrin activation (Figure S3E, F). Furthermore, the failure in promoting nanoclustering upon FN engagement when the cells were treated with both SFK-FAK inhibitors can be rescued by the ectopic addition of CN03 (Figure 3G, H). These results indicate that RhoA operates downstream of SFK-FAK in the molecular pathway that mediates the nanoclustering of GPI-APs.
Formin nucleators are necessary for GPI-AP nanoclustering
We next investigated the role of the actin-nucleators that are downstream targets of integrin activation in mediating the nanoclustering of GPI-APs. Pre-treatment of cells with SMIFH2, a small molecule inhibitor of the formin class of actin nucleators (Rizvi et al., 2009) led to loss of FN-mediated nanoclustering of GPI-APs (Figure 3E, F), whereas the Arp2/3 inhibitor CK666 (Nolen et al., 2009) had no effect on GPI-AP nanoclustering (Figure 3E, F). Moreover, the acute loss of GPI-AP nanoclusters observed when cells were treated with inhibitors of SFK and FAK could be rescued by treatment of cells with a formin activator (IMM01; Lash et al., 2013) which in turn is reversed by SMIFH2 treatment (Figure 3G, H), suggesting that formins are downstream of SFK/FAK-RhoA in this pathway that mediates nanoclustering of GPI-APs. In addition, treatment of cells plated on uncoated glass with the formin activator (IMM01) resulted in an increase in nanoclustering of GPI-APs even in the absence of integrin ligand engagement (Figure S3G, H). This suggests that formin activation is an important step in the integrin mediated signaling response that drives enhanced nanoclustering of GPI-APs.
To investigate the identity of the specific formin that mediates the nanoclustering of GPI-AP, we utilized specific RNAis to reduce the expression of two candidate formins, mDia1 (DIAPH1) and FHOD1 in U2OS cells. Upon reduction of the levels of mDia1 and FHOD1 between ~80 and ~40%, respectively (Figure S4A, B), there was a drastic decrease in nanoclustering of GPI-APs. The loss of nanoclustering was more significant when FHOD1 levels were reduced when compared to mDia1 (Figure S4C, D), implicating FHOD1 as one of the major formin members involved in this process. Taken together, these results indicate that actin filaments nucleated by specific formins are involved in the FN-integrin induced nanoclustering of GPI-APs.
Integrin activation triggers an acto-myosin-based clustering mechanism
To test whether changes in nanoclustering of GPI-APs induced by integrin activation are mediated by the upregulation of the cortical acto-myosin based machinery described previously (Gowrishankar et al., 2012), we monitored the organization of a model chimeric receptor composed of an extracellular reporter domain derived from folate receptor [FR], a transmembrane segment [TM] and a cytosolic domain derived from the actin binding domain of ezrin [Ez-AFBD], FRTM-Ez-AFBD (Figure 4A). This chimeric protein also forms nanoscale clusters that are dependent on its ability to bind actin and associate with a dynamic acto-myosin machinery (Gowrishankar et al., 2012). Steady-state anisotropy measurements of fluorescently labeled FRTM-Ez-AFBD expressing cells showed a decrease in anisotropy when plated on FN (Figure 4B, C), similar to that observed for the GPI-APs (Figure 1C). By contrast, cells expressing a mutated version of the FRTM-Ez-AFBD that is incapable of binding actin (FRTM-Ez-AFBD*; Figure 4D) did not display changes in the steady-state anisotropy upon engagement with FN (Figure 4E, F).
A signature of the dynamic actin-filaments at the membrane surface is the decreased diffusion of GFP-tagged-actin filament binding domain of Utrophin (GFP-Utr) as measured by fluorescence correlation spectroscopy (FCS) (Gowrishankar et al., 2012). When FCS traces were taken from regions in the periphery of the cell that were devoid of stress-fibers (Figure S4E), we detected at least two diffusing species (Figure S4F); one corresponding to the diffusion timescale of unbound GFP-Utr (0.3 ms < τ < 3 ms) and another slower component (<>10ms) that corresponds to GFP-Utr bound to actin filaments with an approximate filament length of ~200nm (Gowrishankar et al., 2012). Treatment of cells with the formin inhibitor SMIFH2, resulted in a loss of only the slow diffusing component (Figure S4F, G). This coincides with the loss of the dynamic pool of actin filaments that is likely to mediate the nanoclustering of membrane proteins (Saha et al., 2015). The nanoclustering of the chimeric transmembrane receptor (FRTM-Ez-AFBD) was also abrogated upon inhibition of formins as well as SFK/FAK on FN (Figure 4G, H).
Next we tested the role of integrin-stimulated ROCK activation in nanoclustering of GPI-APs. RhoA via ROCK can stimulate myosin light chain (MLC) phosphorylation directly or indirectly through the inhibition of MLC phosphatase. Inhibition of ROCK using the Y-27632 inhibitor (Uehata et al., 1997) results in loss of nanoclusters of GPI-APs when cells are plated on FN (Figure S4H, I) as does treatment with the MLC kinase (MLCK) inhibitor ML-7 (Figure S4H, I).
Together these data provide evidence that integrin signaling mediated by src and FAK kinases through active RhoA regulates actin polymerization via form ins. This couples integrin ligation to the generation of nanoclusters. The role of myosin activity in promoting nanoclustering indicates that signaling activates a dynamic acto-myosin machinery to promote the nanoclustering of GPI-anchored and other actin-filament binding domain (AFBD) containing proteins.
Talin and vinculin are necessary for the generation of GPI-AP nanoclusters
To further understand the mechanism of GPI-AP nanocluster generation, we tested the role of focal adhesion proteins, talin and vinculin, that form an integral part of the mechano-chemical signal transduction machinery downstream of integrin engagement (Vicente-Manzanares et al., 2009). Anisotropy measurements on vinculin knock out (vin-/-) mouse embryonic fibroblasts (MEFs) (Janostiak et al., 2014) transfected with GFP-GPI, and freshly plated on FN showed a relatively high anisotropy value, which was unaffected by treatment with mβCD (Figure 5A,B). This indicates that cholesterol-sensitive nanoclusters do not form without vinculin. To restore vinculin function, we transiently transfected full-length mCherry-tagged vinculin (Thievessen et al., 2013a) into vin-/- MEFs re-plated on FN and measured anisotropy of co-transfected GFP-GPI. Re-introduction of vinculin restored cholesterol-dependent GPI-AP nanoclustering ascertained by decreased GFP-GPI anisotropy that was sensitive to cholesterol removal (Figure 5A, B). These data indicate that the loss of nanoclustering in vin-/-cells was only due to the absence of vinculin, providing a convenient test bed to explore the role of vinculin in GPI-AP nanoclustering.
Vinculin exists in an auto-inhibited state in cells, and is activated by several interacting molecules, which bind to specific domains (Figure S5C). A well-characterized mode of activation of vinculin is through the binding of its head domain to talin, opening up the tail domains for interaction with actin and lipids (Case et al., 2015; Golji and Mofrad, 2010). We first addressed the role of talin in nanoclustering of GPI-APs, using talin1 knockout MEFs. Since loss of talin1 leads to over expression of the talin2 isoform (Zhang et al., 2008), we additionally depleted these cells of talin2 with talin2-shRNA co-expressed with GFP (Figure S5A, B). We monitored the nanoclustering of GPI-APs in these cells using a fluorescently-tagged oligomerization-defective aerolysin variant (A568-FLAER) previously characterized to report on the native distribution of endogenous GPI-APs (Raghupathy et al., 2015). A568-FLAER exhibited a higher fluorescence emission anisotropy in the talin2 shRNA expressing cells compared to the talin-1 alone deficient cells (Figure 5C, D), consistent with a loss of nanoclustering of GPI-anchored proteins in talin1-/- cells after talin2 depletion. However, this increase was less than that observed for the loss of vinculin; the partial loss of talin2 (as indicated by immunostaining for talin2) could serve as a confounding factor in these experiments (Figure S5A, B).
Expression of Vin-A50I, which is incapable of binding talin (Case et al., 2015; Figure S5C) in vin-/-MEFs, fails to support GPI-AP nanoclustering (Figure 5E, F). However, a constitutively activated vinculin, Vin-A50I-CA, which does not require talin for its activation (Case et al., 2015) restored GPI-AP nanoclustering in the vin-/- MEFs (Figure 5E, F). Together these results suggest that GPI-AP nanoclustering normally requires vinculin activation by talin.
Vinculin activation specifically links integrin signaling and GPI-AP nanoclustering
We next asked if vinculin is necessary for triggering the actomyosin-based clustering machinery downstream of integrin activation. We examined the status of nanoclustering of FRTM-Ez-AFBD in vin-/- cells (Figure 5G, H). Monitoring anisotropy of FRTM-Ez- AFBD and the FRTM-Ez-AFBD* mutant constructs indicated that the nanoclustering mechanism is unaffected in the absence of vinculin; FRTM-Ez-AFB exhibited a lower anisotropy value compared to the FRTM-Ez-AFBD* in vin-/-cells as well as in the vinculin restored cells (Figure 5G, H). Furthermore, the restoration of GPI-AP nanoclustering observed in vin-/- MEFs by the expression of vinculin is completely disrupted upon pre-treatment with the src family inhibitor, PP2, or FAK inhibitor, PF 573228 or the formin inhibitor, SMIFH2 (Figure 6A, B). GPI-AP nanoclustering is not restored by treatment of cells with the formin activator, IMM01 in the vin-/- cells. By contrast, transmembrane protein clustering is brought about upon integrin activation in these cells, indicating that vin-/- cells are not defective in generating the acto-myosin machinery responsible for clustering. Addition of an artificial linker (LactC2-Ez-AFBD) (Raghupathy et al., 2015) is able to fully restore the nanoclustering of GPI-APs, in the vin-/- cells. This indicates that vinculin activation is not necessary for the creation of the acto-mysoin machinery but is rather involved in the pathway that links actin to the inner-leaflet lipids (Figure 5I, J).
Lipid and actin binding capacity of vinculin are necessary for GPI-AP nanoclustering
Vinculin possesses a negatively charged lipid binding site in its tail domain and mutation of this site results in a loss in its ability to bind to negatively charged lipids in the membrane (Humphries et al., 2007). Importantly, expression of this Vin-Ld mutant that lacks lipid-binding capacity (Figure S5C) in vin-/- MEFs failed to restore GPI-AP nanoclustering (Figure S6A, B). To determine if the failure to bind lipids, keeps vinculin in an inactive state, we generated Vin-Ld-CA*, that is constitutively activated (Humphries et al., 2007) (Figure S5C). This mutant also failed to restore GPI-AP nanoclustering in vin-/- MEFs (Figure S6A, B). These results indicate that the lipid binding capacity of vinculin is necessary for it to catalyze GPI-AP nanocluster formation and accounts for the mechanistic differences between the nanoclustering of GPI-APs and those of transmembrane proteins with actin binding motifs.
To assess if the actin-binding capacity of vinculin is necessary to bring about GPI-AP nanoclustering, we expressed a mutant version of vinculin which has reduced capacity to bind to actin (Case et al., 2015) Vin AB1 (Figure S5C). Even though this mutant of vinculin localizes to focal adhesions, it failed to rescue nanoclustering of GPI-APs consistent with the role of the actin binding domain of vinculin in GPI-AP nanoclustering (Figure 6E, F).
Cells defective in the GPI-AP nanoclusters formation exhibit aberrant integrin function
Cells that lack vinculin have defective integrin mediated responses; they lack the P1 phase of integrin mediated spreading and exhibit aberrant FAs (Figure S7 H-I; see also (Thievessen et al., 2013)). We hypothesized that some of these defects may be a consequence of the inability of cells to build functional nanoclusters. To test this, we utilized mutant cells that are deficient in two enzymes (PGAP2 and PGAP3) required for the remodeling of the unsaturated GPI-anchor acyl chains to long, saturated chains. This defect results in the inability of PGAP2/PGAP3 mutant cells to make GPI-AP nanoclusters (Raghupathy et al., 2015), and inefficient GPI-AP incorporation into detergent-resistant membranes (Maeda et al., 2007).
When freshly plated on fibronectin, these mutant cells failed to exhibit a decrease in the anisotropy of GFP-GPI and this defect could be reversed by restoring the activities of the PGAP2 and PGAP3 enzymes (Figure S7A, B). In comparison with either wild type cells or mutant cells rescued with wild type copies of PGAP2 and PGAP3, the PGAP2/PGAP3 mutant cells lack the P1 spreading phase when plated on substrates coated with fibronectin (Figure 6A, B). These mutant cells also lack a protrusive lamellipodia and possess fewer smaller adhesions (Figure S7F-G) and exhibit bleb-based cell spreading (Supplementary Movie 2). They also do not exhibit a rapid increase in cell area when spreading on FN, characteristic of the P1 phase (Figure 7C; Supplementary Movie 3). This defect is not due to defects in integrin activation in the mutant cells, since antibodies that bind to either active or in-active conformations of the β1-integrins bind equivalently to the mutant and wild type cells (Figure S7C).
Several proteins that either create or reside within lo-like regions on the cell membrane have been implicated in the process of cells spreading and migration (Moissoglu et al., 2014; Navarro-Lérida et al., 2012). Recently we have shown that GPI-AP nanoclusters are associated with specific regions of the membrane that have an lo- like character (Saha et al, manuscript under preparation). Therefore, we tested if the lack of GPI-AP nanoclusters in the PGAP2/3 mutants could lead to a global disruption of ordered domains. Using the polarity-sensitive membrane dye Laurdan (6-lauryl-2-dimethylamino-napthalene) as a reporter of membrane order (Owen et al., 2012), we found that the PGAP2 and PGAP3 mutant cells have a lower generalized polarization (GP) value (Figure 7 D-E), which is consistent with the loss of ordered lo domains on the cell surface. The decrease in the GP value was similar to that observed when membrane cholesterol was depleted using mβCD treatment, and is fully restored in the PGAP-2/3 add-back cell line (Figure 7D-E). To further confirm that the loss of lo domains is due to a specific defect in GPI-AP nanoclustering and not due to global alterations of cholesterol or phospholipid composition of the plasma membrane, we compared the levels of filipin-labeled cholesterol and performed mass-spectrometric measurements on blebs extracted from them. We do not find any significant difference in the levels of either membrane cholesterol or phospholipid profile of the mutant cells (Figure S7 D-E; Supplementary Table S3). This suggests that the lack of GPI-AP nanoclustering specifically contributes to the loss of /o-domains at the cell surface and could account for the observable cell spreading defects.
DISCUSSION
Our results using both chemical and genetic perturbation show how GPI-AP nanoclustering is initiated via a signaling cascade triggered by β1-integrin receptors upon binding to its bonafide ligands: fibronectin (FN) or the fibronectin-derived peptide RGD (see model in Figure 7F). Ligand binding results in the activation of the src and FAK kinases. Potentially this step may involve activation of additional molecules including ILK and kindlin-kinases (Calderwood et al., 2013). Regardless, downstream of the kinases are the RhoA GTPases (Ishizaki et al., 1996; Leung et al., 1996), which directly activate formins, necessary for nanoclustering. The nucleator of branched actin filaments, Arp2/3, was not required for this process. Knockdown of both FHOD1 and mDia via RNAi-mediated depletion inhibited nanoclustering of GPI-APs, although the effect of FHOD1 depletion was more drastic. RhoA via ROCK activates FHOD1 through the phosphorylation of C-terminal serine/threonine residues in its DAD region thereby relieving its auto inhibition (Takeya et al., 2008). FHOD1 is also recruited to nascent sites of integrin ligand engagement (Changede et al., 2015; Iskratsch et al., 2013), implicating formins in effecting the nanoclustering of GPI-APs during early stages of integrin mediated signaling. Consistent with this, we found that GPI-AP nanoclusters were also formed on the supported lipid bilayer system, in the vicinity of integrin clusters. Together with previous observations that the integrin LFA1-binding to its ligand ICAM-1 results in a local concentration of GPI-AP nanoclusters, (van Zanten et al., 2009), these results show that integrin signaling generates a localized nanoclustering response.
In parallel, this mechanism requires a way to control myosin function. Indeed, RhoA activates ROCK that regulates myosin light chain kinase (MLCK), and perturbation of ROCK (via Y27632) and MLCK (via ML-7), inhibited actin-based nanoclustering of GPI-APs. However, we do not exclude the possibility of this being a myosin II independent ROCK or MLCK effect. Ectopic activation of RhoA via the agonist CNO3 was sufficient to generate the necessary machinery for creating the GPI-AP nanoclusters, independent of receptor signaling and despite the inhibition of the upstream signaling cascade. Once generated, this actin machinery was also sufficient to cluster transmembrane proteins with actin-binding capacity, exemplified by the model transmembrane protein, FRTM-Ez-AFBD. Transmembrane proteins with actin-binding motifs directly associate with the dynamic acto-myosin machinery, whereas GPI-APs at the outer leaflet, require transbilayer interactions with long acyl-chain containing lipids such as PS at the inner leaflet (Raghupathy et al., 2015). This in turn might require an entirely different mechanism to connect to the actin machinery in the cortex.
Vinculin is known to be activated downstream of integrin signaling through the activation of talin, and has several binding partners such as actin, paxillin and negatively charged lipids like PS and PIP2 (Niggli et al., 1986). The role of these two proteins in supporting GPI-AP nanoclustering was verified by the depletion of talin and vinculin in MEFs, wherein GPI-AP nanoclustering was disrupted. The observation that nanoclusters of TM-ABDs in vin-/- null cells were formed without any alteration implies that the integrin receptor recruits distinct molecular players to facilitate a link between the dynamic actin machinery and membrane lipids. Our results indicate a role for vinculin as a molecular player which may direct actin to the membrane or serve as a linker in connecting the inner leaflet to actin. The failure of the lipid and actin binding mutants to restore GPI-AP nanoclustering support the latter hypothesis. However, vinculin was not found measurably enriched at the membrane outside of focal adhesions when examined at time points where GPI-AP nanoclustering is restored in vin-/- cells (Figure S6C, D), supporting the former.
These results provide a molecular mechanism for the control of an active actin-membrane composite, wherein the fluid membrane is inextricably coupled to the cortical actin-substructure beneath (Koster and Mayor, 2016; Rao and Mayor, 2014). The functioning of this composite implies the existence of three types of membrane components: inert, passive and active. While we have previously described inert and passive components and their characteristics (Gowrishankar et al., 2012), here we provide evidence for the functioning of an active element, exemplified by the integrin receptor family.
The relevance of GPI-AP nanoclustering in membrane function has been difficult to probe because of the use of drastic perturbations such as cholesterol removal (Kwik et al., 2003) or alterations in specific phospholipid levels (Lipardi et al., 2000). The identification of a molecular mechanism behind the generation of these nanoclusters, and the key role of integrin signaling in cell spreading provides an opportune physiological context. Pertinently, many of the key components of the nanoclustering molecular machinery identified here such as src, FAK, RhoA, formin, myosin, talin and vinculin, are in the pathway of integrin-mediated signaling, and also have multiple roles in cell physiology. Therefore, the well-documented cell spreading defects and alteration in focal adhesion patterns that are exhibited by perturbations of these players, may be difficult to directly relate to nanoclustering defects. As a consequence, we explored the role of nanoclustering in membrane function by studying defects in integrin-mediated functions in the GPI-anchor remodeling mutants. These mutants lack the ability to make GPI-AP nanoclusters but they support FA-formation as well as integrin-mediated activation. However, they exhibit dramatic defects in cell spreading that are restored upon restoration of the cell’s ability to support nanocluster formation, similar to those observed in vin -/- cells. This implicates a functional role for GPI-AP nanoclustering in integrin-mediated signaling.
Why does signaling via integrin-ligation target the local construction of GPI-AP nanoclusters? An answer to this, is related to the fact that GPI-AP nanoclusters form lo nanodomains (Raghupathy et al., 2015) which in turn generate larger meso scale lo domains (Saha et al, manuscript in preparation). Here we show that the loss of GPI-AP nanoclustering results in the failure to enhance the overall lo characteristics of the membrane observed upon integrin-mediated signaling. Coupled with the observation that large cross-linked patches of GPI-APs accumulate src family kinase members at the inner leaflet (Harder et al., 1998; Stefanová et al., 1991b; Suzuki et al., 2007), these results suggest a function for the lo-like GPI-AP nanocluster rich-regions in effecting integrin signaling responses. Lipid modifications such as palmitoylation enable molecules to partition into locally generated lo micro-environments. These membrane domains are also likely to be important for the signaling activity of SFK and FAK (Seong et al., 2011). Rac1 is a palmitoylated Rho family GTPase that regulates leading edge protrusion dynamics, and its activity is restricted to lo domains (Moissoglu et al., 2014; Navarro-Lerida et al., 2012; del Pozo et al., 2004). Moreover, the GAP activity of the p190RhoGAP is also localized to potentially lo domains (Sordella et al., 2003) and its recruitment to lo-like regions has been shown to be necessary for the cell spreading process (Arthur and Burridge, 2001), as well as for the localized inhibition of the Rho GTPase and the regulation of FA size. Palmitoylation of the fyn kinase, implicated in rigidity sensing, is also required for the P1-based cell spreading process (Kostic, 2006).Thus, it is likely that the lo microenvironment created by the mechanism proposed here, could serve to localize a number of important components of the effector cascade in integrin-based activity to the leading edge where such sensing takes place.
Since the activation of the small GTPase, RhoA, is a pivotal feature downstream of many signaling receptors besides integrins, such as Cadherins, RTKs, GPCRs (Olson and Nordheim, 2010), this will likely culminate in the activation of such an acto-myosin-based mechanism as described here. Vinculin is also a downstream effector of many signaling based systems (Hazan et al., 1997), ensuring that these dynamic acto-myosin filaments also generate GPI-AP nanoclusters. The resultant membrane domains that ensue, will serve as allosteric modulators of the output of the signaling system that generates it (Harding and Hancock, 2008). This will naturally allow the cell to integrate information that is encoded primarily in the composition of its membrane bilayer. In conclusion, our results suggest a generalizable picture of how lo nanodomains may be created and deployed in the context of a number of different signaling systems.
AUTHOR CONTRIBUTIONS
J.M.K., A.A.A., and S.M designed the study. J.M.K, A.A.A set up the EA-TIRFM microscope, performed experiments, and analysed the data; C.P performed the lipid ordering experiments and filipin-labelling experiments; J.M.K also standardized the supported lipid bilayer experiments working in collaboration with M.P.S laboratory. A.A.A also performed the mass spectrometric experiments; T.S.V.Z performed and analysed the FCS measurements and analysed the data involving the cell spreading experiments. J.M.K., A.A.A., and S.M. drafted the manuscript with input from all the authors.
Supplementary Movie Legends
Supplementary Movie 1: Cell spreading dynamics of GFP-GPI expressing CHO cells on 10μg/ml FN; Imaged after every 15 seconds in 100X EA-TIRFM mode at 37°C. Left panel: Total Intensity image; Right Panel: GFP-GPI anisotropy image with the corresponding LUT bar. Scale bar 10μm. Notice that the cells initially blebs and rapidly acquire GPI-AP nanoclusters (blue pixels) before the cell extends out a prominent lamellipodia and begins spreading rapidly.
Supplementary Movie 2: 20X Phase contrast time series images of WT or PGAP2&3 double mutant CHO cells spreading on FN. Notice that the PGAP2&3 double mutants spread slowly and extend by producing blebs (and lack a prominent lamellipodia). Scale bar 50μm.
Supplementary Movie 3: Cell spreading dynamics of GFP-GPI (Cell membrane marker) expressing WT (Left Panel), PGAP2&3 mutant (Middle Panel) or Rescue (Right Panel) CHO cells spreading on 10μg/ml FN;imaged every 15 seconds in 100X TIRF at 37°C. Notice that the PGAP2&3 mutant cells exhibit defects in cell spreading due to their inability to produce a prominent lamellipodia. Scale bar 10μm.
ACKNOWLEDGMENTS
We thank Cheng-han Yu for help in standardizing the use of the supported-lipid bilayer system; Kabir Husain and Balaji for Matlab codes to analyze the anisotropy and bilayer data; Taroh Kinoshita, Yusuke Maeda, Daniel Rosel, Clare M. Waterman, Lindsey Case, Ana Pasapera for their generous gifts of various reagents (as indicated in the Supplemental Information); Max Planck-NCBS Lipid centre; Bini Ramachandran for mass spectrometry; and H. Krishnamurthy and Manoj Mathew at the Central Imaging and Flow Facility (NCBS). We thank Madan Rao, and Subhasri Ghosh for inspiration and SM lab members for their critical comments on the manuscript. J.M.K. acknowledges pre-doctoral fellowship from NCBS-TIFR. A.A.A acknowledges N-PDF fellowship from DST-SERB (Government of India). T.S.V.Z. acknowledges EMBO fellowship (ALTF 1519-2013) and NCBS fellowship. S.M. acknowledges JC Bose Fellowship from DST (Government of India), a grant from HFSP RGP0027/2012 and Wellcome Trust-DBT Alliance Margadarshi fellowship.