Introduction

Rho GTPases regulate the cytoskeletal dynamics that power morphogenetic processes at the cell- and tissue level (Etienne-Manneville and Hall, 2002). Their activity is tightly controlled in time and space by GEFs and GAPs, which respectively activate and inhibit Rho GTPases. In the GTP-loaded, active state, Rho GTPases bind downstream effectors that regulate cytoskeletal dynamics and other cellular processes. Based on classic experiments that include measuring population-averaged Rho GTPase activity levels, as well as overexpressing dominant positive/negative Rho GTPase mutants, it was proposed that Rac1 controls lamellipodial protrusion, Cdc42 regulates filopodium formation, and RhoA controls myosin-based contractility.

Recent studies in which spatio-temporal Rho GTPase activation dynamics were measured using fluorescence resonance energy transfer (FRET)-based biosensors, revealed a much higher signaling complexity. During cell migration, RhoA, Rac1 and Cdc42 are all activated in membrane protrusions at the leading edge (Itoh et al., 2002; Kraynov et al., 2000; Nalbant et al., 2004; Pertz et al., 2006). Further, the three Rho GTPases are activated in specific spatio-temporal sequences during the protrusion/retraction cycles that fluctuate on time and length scales of seconds and single micrometers (Machacek et al., 2009). A further level of complexity is that these Rho GTPase activity sequences vary depending on the cell system (Hu et al., 2022; Machacek et al., 2009; Martin et al., 2016). This indicates that complex spatio-temporal Rho GTPase signaling programs regulate processes such as cell migration. Similarly as for Ras GTPases (Schmick et al., 2015), it is now clear that these Rho GTPase signaling patterns most likely result from a precise spatio-temporal regulation of Rho GTPase signaling fluxes controlled by GEFs and GAPs (Bement and von Dassow, 2014; Fritz and Pertz, 2016; Pertz, 2010). In this signaling flux, specific GEFs will be locally activated to load Rho GTPases with GTP depending on different cellular inputs. These GTP-loaded GTPases will then subsequently diffuse in the plasma membrane by virtue of their C-terminal lipid moiety until they encounter a locally activated GAP which eventually deactivates them. The interplay of GEFs, GAPs and diffusion of Rho GTPases in the PM will ultimately lead to the formation of a Rho GTPase activity pattern. Modelling studies have suggested that sophisticated dynamic signaling behaviors can emerge from the interplay of GEFs, GAPs, and GTPases within a signaling network (Tsyganov et al., 2012). Steady-state images of Rho GTPase activity patterns do not provide information about the whole Rho GTPase flux mentioned above.

This signaling complexity is consistent with the large amount of GEFs and GAPs that are ubiquitously expressed in cells (Fusco et al., 2016, 2016; Moon and Zheng, 2003; Mosaddeghzadeh and Ahmadian, 2021; Müller et al., 2020; Rossman et al., 2005). Recently, a system-wide screen has revealed that many GEFs and GAPS localize to cytoskeletal structures as well as adhesion complexes such as focal adhesions (FAs) (Müller et al., 2020). This strongly suggests the existence of feedback mechanisms from the cytoskeleton and FAs to Rho GTPase signaling. In this study, it was shown that Rac1-specific GEFs bind to FAs at the front, while Rac1-specific GAPs bind to FAs and at the back of the lamella of fibroblasts. This precise, asymmetric positioning of GEFs and GAPs might then regulate the Rho GTPase flux that produces the wide gradient of Rac activity observed at the leading edge of fibroblasts (Itoh et al., 2002; Kraynov et al., 2000; Martin et al., 2016). Such feedback regulation might allow leading edge Rac1 activity to constantly integrate mechanosensitive inputs from FAs, allowing dynamic regulation of Rac1 activity required to fine tune cell migration. Another prominent example of such spatio-temporal feedback regulation are the excitable RhoA activity patterns observed in the cortex during cytokinesis in Xenopus Laevis (Bement and von Dassow, 2014), that have also been observed in mammalian cells (Graessl et al., 2017). Here, RhoA activity waves that control F-actin wave patterns are spatially regulated by a RhoGAP that is locally regulated by F-actin (Bement et al., 2015a). Thus, cytoskeletal feedback to Rho GTPase regulation might be crucial to generate dynamic Rho GTPase signaling patterns. Understanding such spatio-temporal feedback regulation is not accessible with classic genetic approaches in which long term perturbation of Rho GTPases, their regulators and effectors rapidly result in a new mechanochemical state of the cell, that will not be informative about an initial signaling/mechanical state of interest (Isogai and Danuser, 2018). Further, it is currently unknown which specific features of spatio-temporal Rho GTPase signaling are regulated to produce signaling patterns. For example, does negative regulation by a GAP solely control Rho GTPase signal termination, or could it also modulate, less intuitive parameters such as the rate of activation? Tackling such questions requires new tools to acutely and transiently perturb the Rho GTPase flux to understand its spatio-temporal regulation, directly in single living cells. Note that this experimental paradigm was successful at analyzing the Mitogen activated protein kinase (MAPK network) (Blum et al., 2019; Dessauges et al., 2022; Ryu et al., 2015).

A promising candidate to investigate how mechanical feedback from FAs can spatio-temporally shape Rho GTPase activity is the RhoA-specific GAP Deleted in Liver Cancer 1 (DLC1). DLC1 has been shown to both bind the plasma membrane (PM) (Erlmann et al., 2009), as well as FAs (Haining et al., 2018; Kaushik et al., 2014). At the PM, interaction with phosphatidylinositol-4,5-bisphosphate (PI(4,5)P2) can regulate GAP activity (Erlmann et al., 2009). At the FA, DLC1 interacts with the r7-r8 domains of the FA protein talin in a mechanosensitive fashion (Gingras et al., 2008; Goult et al., 2021; Haining et al., 2018; Zhao et al., 2022), as well as with Focal adhesion kinase (Li et al., 2011). When talin is in a state of low mechanical strain, it binds to DLC1, and presumably leads to local RhoA inactivation. However, upon force application, the r7-r8 domains unfold leading to dissociation of DLC1 from talin, leading to loss of negative RhoA regulation. This suggests that DLC1 is regulated at FAs by mechanosensitive interactions, which can relay information about the mechanical state of the cell to control RhoA activity.

In this work, we explore the spatio-temporal regulation of RhoA activity by DLC1. We show that knocking out DLC1 in REF52 fibroblasts leads to increased amplitude in RhoA throughout the cell. This is accompanied by an increased contractility that augments FAs and stress fibers (SFs). To explore how DLC1 contributes to Rho GTPase activity fluxes, we built a genetic circuit consisting of an optogenetic actuator to activate Rho with light, and a spectrally compatible biosensor to measure Rho activity. Transient optogenetic recruitment of a Rho GEF domain at FAs or at the PM led to an increase in the rate of activation of Rho in DLC1 KO versus WT cells at both subcellular localizations. However, we found that in both control and DLC1 null cells, RhoA activation was more efficient at FAs compared to the PM. Further, acute and local manipulation of contractility with the optogenetic actuator revealed that DLC1 dynamically associates/dissociates with FAs under acute mechanical tension/relaxation. Our results suggest a complex mechanism in which DLC1-dependent Rho regulation can occur both at the PM and at FAs, with a dependence on mechanosensitive signals on the latter.

Results

DLC1 deficient cells show increased formation of SFs and FAs

To investigate the function of DLC1 in cytoskeletal and adhesion dynamics, we knocked down (KD) DLC1 in rat REF52 fibroblasts via small interfering RNA (siRNA). We also created a knockout (KO) cell line via Crispr/Cas9, that has a frameshift deletion at position 841 of Exon 5 of the gene, resulting in an early in frame stop codon 100 bp downstream of the cut. Gene-editing was validated by sequencing the endogenous locus. We were unable to test expression levels using western blot because of the absence of commercially available antibodies specific for rat DLC1. We then also rescued the DLC1 KO cells with an mCherry-labeled DLC1 construct using stable transfection with a piggyBac transposase system. REF52 fibroblasts were seeded at subconfluent density onto fibronectin-coated glass bottom wells and allowed to spread for one hour before fixation, leading to an isotropic spreading state in control, wild-type (WT) cells. To assess the impact of DLC1 deficiency on the formation of cytoskeletal and adhesion structures, we stained for F-actin (phalloidin) and paxillin (Figure 1A) or phospho-myosin light chain (pMLC) (Figure 1B). Consistent with an expected increase in RhoA activity, we observed a phenotype of increased contractility as documented by increased lamella size, SF content, number of FAs, and increased pMLC. This increased contractility led to the loss of the isotropic spreading observed in WT cells. Rescue with the DLC1 construct reverted the cells to a less contractile phenotype that displayed a lamella of identical size as WT cells, enabling the characteristic isotropic spreading observed in WT cells. Quantification of these images revealed both an increase in lamella size (Figure 1C) and FA area (Figure 1D) in DLC1 KD and KO cells versus WT and rescue cells (Figure 1C). As previously documented in fibroblasts (Kaushik et al., 2014), these results indicate that DLC1 feeds into the regulation of contractility, SF and FA formation in fibroblasts. The less penetrant phenotype observed in KO versus KD suggests that the KO cells might be able to adapt to the long-term absence of DLC1, while KD cells had less than 48 hours to adjust to the new mechanical state. These experiments also validate the DLC1 KO cells and DLC1 KO rescued with mCherry-DLC1 that will be used in subsequent experiments.

Figure 1.

DLC1 deficient cells display altered spreading dynamics, lack of polarization and efficient migration

We timelapsed REF52 WT and DLC1 KO cells for multiple hours in presence of platelet derived growth factor (PDGF), a potent stimulator of fibroblast motility (Martin et al., 2014), and evaluated their ability to migrate. We found that DLC1 KO cells displayed a reduction in both their speed of migration and their ability to migrate directionally (Figure S1A-C). We then investigated cytoskeletal dynamics during spreading using REF52 fibroblasts stably expressing the F-actin marker LifeAct (Riedl et al., 2008). Imaging early spreading (e.g. 20 minutes after plating) revealed no clear differences in edge dynamics in WT versus DLC1 KO cells (Figure 2A). However, kymograph analysis clearly showed that the lamella that characteristically starts directly at the leading edge of the highly contractile REF52 cells (Martin et al., 2016, 2014), was wider in DLC1 KO versus WT cells. Later, 1 hour after plating, when the cells are still spreading, lamellipodial protrusion/retraction cycles of wider amplitude were observed in DLC1 KO versus WT cells (Figure 2B, Movie S1), with again a large increase in SF content (Figure 2B, right panel). After initial spreading, REF52 fibroblasts break their symmetry and eventually display short episodes of polarized cell migration. Later, contractility builds up leading cells with spindle-like morphology. This process occurs on timescales of multiple hours. To capture these behaviors, we investigated cytoskeletal dynamics over the next 10 hours after reseeding at a slower timescale that precludes observation of lamellipodial dynamics. While WT cells efficiently broke their symmetry, and could display short stretches of polarized migration, DLC1 KO cells displayed robust stress fibers that hampered symmetry breaking and polarization (Figure 2C, Movie S2). DLC1 KO cells therefore rapidly adopted a spindle-like shape. Quantification of the adoption of such a spindle shape phenotype showed that WT cells will assume the spindle-like shape much later than DLC1 KO cells and often fail to do so entirely within the 10 hours of imaging (Figure 2D). Further highlighting a role for DLC1 in RhoA-mediated control of contractility, we found that DLC1 overexpression leads to strong reduction in the number of SFs (Figure S2A). Together, these results show that DLC1 loss of function leads to increased contractility at the onset of spreading, which later impedes polarization and directional cell migration.

Figure 2.

DLC1 deficient cells display an increase in global RhoA activity, without any marked differences in RhoA activity pattern

To explore how DLC1 spatio-temporally controls RhoA activity, we used RhoA2G (Fritz et al., 2013), a fluorescence resonance energy transfer (FRET)-based biosensor that reports on RhoA activity together with the Lifeact F-actin marker during spreading. As previously observed (Martin et al., 2014), a RhoA activity band that correlates with the lamellar contractile actomyosin network of the cell can be observed at the cell edge of WT REF52 fibroblast (Figure 3A, Movie S3). In DLC1 KO cells, we observed global elevation of RhoA activity throughout the cell. Quantification of both the average activity per cell and the local edge RhoA activity revealed increased RhoA activity (Figure 3B). Kymograph analysis revealed that the lamellar RhoA activity band remained of constant size during spreading (Figure 3A). Examining cells approximately 10 hours after spreading when they had fully spread, and WT cells had adopted a more contractile state, the DLC1 KO cells still exhibited a global increase in RhoA activity (Figure 3C,D). This increase was however less strong than in spreading cells, most likely because of a higher state of contractility in WT well-spread versus spreading cells. We also evaluated the effect of overexpression of an mCherry-tagged DLC1 on RhoA activity (Figure S2B). We observed that mild DLC1 expression, that remains mostly bound to FAs and to the PM and does not compromise cell morphology, leads to a decrease in global RhoA activity. Strong DLC1 expression that leads to a large cytosolic pool of DLC1 further diminishes global RhoA activity but then compromises cell morphology. Together, these results suggest that DLC1 signaling at the PM and at FAs controls global levels of RhoA activity within cells, rather than controlling the spatio-temporal RhoA activity pool observed at the lamella.

Figure 3.

Design of an optogenetic actuator - Rho biosensor to interrogate the Rho GTPase flux

While FRET measurements as shown above provide insight about a steady-state RhoA activity pattern, they cannot probe in detail how a GEF/GAP/GTPase signaling network might affect the Rho GTPase flux which most likely emerges from the regulation of multiple GEFs and GAPs. For example, GAP-mediated negative regulation might control Rho GTPase signal duration, but might also be involved in adjusting the rate of activation. We reasoned that transient, acute optogenetic perturbations of the Rho GTPase flux, followed by measuring Rho activity dynamics might provide new insights about its regulation. To manipulate Rho activity in single living cells, we re-engineered an iLID-based optogenetic actuator based on a DH/PH domain of the GEF LARG (Oakes et al., 2017) (Figure 4A). We used a high affinity nano SspB domain for efficient light-dependent recruitment of the LARG GEF domain to the iLID anchor. To better focus optogenetic activation, we also fused the iLID anchor to a stargazin membrane anchor that slows down its diffusion (Natwick and Collins, 2021). Both the stargazin-iLID anchor and the SspB-LARG domains were separated by a P2A self-cleaving peptide, allowing for equimolar expression of both units from a single operon. We refer to this construct as optoLARG. To interrogate the Rho GTPase flux, we engineered a REF52 line that stably expressed the optogenetic construct as well as a rhotekin-based G protein binding domain (rGBD) that reports on Rho activation (Mahlandt et al., 2021). The rGBD probe is labelled with a tandem dimeric Tomato (tdTomato) fluorophore that is spectrally orthogonal to optogenetic activation. We used a truncated CMV promoter to warrant low expression level of the rGBD probe to avoid dominant negative effects of the construct through inhibition of Rho signaling. We also expressed a spectrally orthogonal far-red Lifeact-miRFP construct that reports on F-actin dynamics. Because rGBD cannot distinguish between the activity of the RhoA, RhoB and RhoC isoforms, we refer to any rGBD measurement as Rho activity. This is in marked contrast with RhoA2G, the FRET biosensor we used above that is specific for RhoA (Fritz et al., 2013), but is not spectrally compatible with optoLARG. Note that both biosensors are able to measure active Rho GTPase pools during robust morphogenetic events such as tail retraction (Fritz et al., 2013; Mahlandt et al., 2021), while rGBD misses leading edge Rho GTPase activity pools widely observed using FRET probes (Machacek et al., 2009; Martin et al., 2016; Pertz et al., 2006).

Figure 4.

We then used a digital micromirror device (DMD) to spatially shine repetitive pulses of blue light delivered at different frequencies on 2 distinct regions of interest (ROIs) at the edge of fibroblasts. When light pulses were delivered at high frequency, we observed robust Rho activation at the edge, which correlated with assembly of contractile F-actin structures and potent edge retraction (Figure 4B, kymograph shown in 4C, Movie S4). Removal of the blue light pulses led to a subsequent decrease in Rho activity, and reversion of the edge to a protrusive state. Further, stimulating a ROI at the opposite edge of the cell with a lower light pulse frequency led to a lower Rho activation, and less robust edge retraction. These results indicate that depending on the strength of the light input, optoLARG can locally, reversibly, and quantitatively control Rho activity in single living cells on timescales of seconds, allowing us to acutely manipulate the Rho GTPase flux.

Optogenetic interrogation of the Rho GTPase flux in WT and DLC1 KO cells

Because DLC1 might both regulate the Rho GTPase flux both at the PM (Erlmann et al., 2009) and at FAs (Haining et al., 2018), we performed a series of experiments using optoLARG to transiently activate RhoA with the expectation of observing different perturbations in Rho activity fluxes in control versus DLC1 KO cells. For that purpose, we produced WT and DLC1 KO REF52 lines expressing optoLARG, the tdTomato-rGBD biosensor and miRFP-paxillin as a readout for focal adhesions. We first sought to locally activate RhoA by illuminating large ROIs at edges and evaluate any difference in Rho activity patterns using the spectrally compatible rGBD biosensor in WT versus DLC1 KO cells. We however were not able to identify a threshold optogenetic light input that would induce different RhoA activity patterns and edge dynamics in WT versus the more contractile DLC1 KO cells. We observed that the same light input could induce edge retraction in both WT and DLC1 KO cells in a non-predictable manner. This might be due to the spatial heterogeneity of the mechanical states within cells, as well as small differences in optoLARG and rGBD expression levels in the stable cell lines. To address this caveat, we reasoned that we should activate Rho using optoLARG with a more subtle modality by illuminating only small ROIs and measuring small local fluctuations of Rho activity using rGBD without inducing large changes in mechanical states. Further, to dissect potential regulation of DLC1 GAP activity by PI(4,5)P2 at the PM (Erlmann et al., 2009) versus mechanosensitive interactions at the FA (Haining et al., 2018), we decided to make measurements on FAs but use ROIs outside of FAs (non-FAs ROIs) to probe the PM. For that purpose, we built an image analysis pipeline to first segment a cell in a field of view and to then segment FAs based on their fluorescent intensity, size and localization at the cell periphery using the paxillin-miRFP signal (Figure 5A). We then used these FA segmentations to illuminate small ROIs within FAs or in neighboring non-FA ROIs with one pulse of blue light using the DMD and evaluated rGBD fluorescence fluctuations at this location for a couple of minutes until Rho activity returned to baseline (Figure 5B). This optogenetic stimulation had no visible effect on edge retraction, or FA morphodynamics, as observed when stimulating larger ROIs. Quality control of our automated segmentation revealed high versus low paxillin signals in FA versus non-FA ROIs (Figure 5C). This experiment was performed on multiple cells, allowing to average out any experimental noise due to small differences in expression levels of optoLARG or rGBD. These data were then averaged for FA and non-FA ROIs. We observed that this pulsed optoLARG light input led to a local transient of Rho activity for approximately 2.5 minutes in WT cells when FAs or the PM were stimulated (Figure 5D). The identical optoLARG input led to an increased rate of Rho activation leading to higher amplitude when a FA versus a non-FA ROI was stimulated. Surprisingly, in DLC1 KO cells, we observed increased Rho activation rate that also led to augmented Rho activity amplitude in comparison with WT cells in both FA and non-FA ROIs. This was also accompanied with faster deactivation kinetics of RhoA. These data suggest that DLC1 limits the rate of Rho activation rather than signal duration to regulate RhoA activity in spread cells. This can occur both at the PM and at FAs. Further, FAs seem to provide a subcellular environment that favors GEF-mediated activation of Rho activity. The finding that DLC1 KO also leads to faster deactivation suggests the existence of a negative feedback that points to complex modalities of spatio-temporal regulation that most likely involves additional GEFs/GAPs.

Figure 5.

DLC1 differently interact with FAs depending on their mechanical states

Our finding that we did not observe striking patterns of RhoA activity in the vicinity of FAs in spread cells challenges the idea that mechanical inputs regulate Rho activity in this specific regime of mechanical forces. We speculated that the DLC1/talin system in FAs is only mechanosensitive to stronger mechanical inputs. To evaluate FA-DLC1 interactions, we imaged fibroblasts rescued with mCherry-DLC1 and miRFP-paxillin as a marker for FAs. We observed that DLC1 binds to FAs at the onset of FA assembly, and that the local DLC1 density augmented with the paxillin density (Figure S3A-C). During FA disassembly, DLC1 density diminished concomitantly with decreasing paxillin density until the FA was fully disassembled (Figure S3D-F).

To characterize force-dependent interactions of FAs with DLC1 in regimes of strong mechanical perturbations, we engineered our mCherry-DLC1 rescue cells to stably express the optoLARG system, as well as the miRFP-paxillin construct. We then used the DMD to activate Rho-mediated contractility locally and transiently in small ROIs containing FAs with a standardized light input. The transient light pulse induced either of the two FA behaviors: light-mediated FA reinforcement as evidenced by a local increase in paxillin density, was followed by 1. FA relaxation (and decrease in paxillin density) upon light removal (Figure 6A-C, quantified in Figure 6D, Movie S5), or 2. FA rupture (Figure S4A-C). This was accompanied by a striking simultaneous decrease in DLC1 intensity at the FA during FA reinforcement, and a reassociation of DLC1 during FA relaxation upon light input removal. If the light input led to FA rupture, then the DLC1 signal gradually decreased along FA reinforcement until the FA ultimately ruptured in an all-or-nothing manner. FAs in the same cell that were not subjected to the light input displayed dynamics as in normal cells, as shown in Figure 6 (Figure S4D-F). These results indicate that DLC1 dynamically dissociates and associates during acute FA reinforcement and relaxation in response to a strong mechanical input.

Figure 6.

Discussion

A key question in Rho GTPase biology is how multiple GEFs and GAPs control spatio-temporal Rho GTPase signaling patterns to regulate cytoskeletal dynamics that power morphogenetic processes such as cell migration. The recent finding that many GEFs and GAPs themselves bind cytoskeletal and adhesion structures (Müller et al., 2020) strongly suggests direct feedback from the cytoskeleton to Rho GTPase signaling. Self-organizational properties of cytoskeletal and adhesions structures might therefore contribute to spatio-temporal control of Rho GTPase signaling. Because DLC1 has been proposed to bind to FAs in a mechanosensitive manner (Haining et al., 2018), we decided to study it as a prototypical example of an adhesion feedback to Rho GTPase signaling. Using novel technologies that allow for acute spatio-temporal perturbations, we provide some new insights about how Rho GTPases are spatio-temporally controlled by a RhoGAP that integrates mechanosensitive inputs.

We found that in spreading fibroblasts, DLC1 KO globally augments contractility, as illustrated by a robust increase in SFs and FAs (Figure 1). Evaluation of F-actin dynamics immediately after spreading, when global contractility levels are still low, reveal that DLC1 KO fibroblasts rapidly assemble a wider contractile lamella than WT cells, suggesting aberrant increase of myosin-based contractility already at an early stage of spreading (Figure 2A). One hour after cell plating, when cells are well spread but still retain an isotropic shape, we observed a global increase in contractility in DLC1 KO versus WT cells (Figure 3B). On a timescale of multiple hours after spreading, when WT cells break symmetry, display polarized cell migration episodes, and finally adopt a contractile state leading to a spindle morphology, we find that DLC1 KO cells transition directly to the contractile phenotype without being able to polarize (Figure 2C,D). Thus, DLC1 loss of function leads to a global increase of aberrant contractility.

We had previously shown that REF52 fibroblasts display a wide band of RhoA activity that correlates with the lamellar contractile myosin network during spreading (Martin et al., 2016, 2014). We found that DLC1 KO leads to a global increase in RhoA activity throughout the cell, without however markedly modifying the spatial band of RhoA activity (Figure 3A,B). Similar results were observed when cells were allowed to further polarize and assemble a robust actomyosin cytoskeleton (Figure 3C,D). Further, we observed that mild DLC1 expression, that can still mostly be titrated by FAs and the PM without spilling out into the cytosol, is already able to strongly downregulate global RhoA activity levels (Figure S2B). We propose that DLC1 regulates global levels of RhoA activity at steady state, rather than spatially patterned RhoA activity pools as observed in protrusive edges and filopodia (Fritz et al., 2013; Machacek et al., 2009; Pertz et al., 2006), in the lamella (Martin et al., 2016, 2014) or at the cytokinetic furrow (Basant and Glotzer, 2018; Bement et al., 2015b). Note that the classic genetic perturbation paradigm cannot be used to study more subtle functions of DLC1 at FAs because long-term loss of DLC1 function results in cells being locked in a state of excessive contractility.

How does DLC1 contribute to the Rho GTPase flux in cells? To answer this question, we used optoLARG to interrogate the Rho GTPase flux at the PM and at FAs, 2 sites at which DLC1 can control Rho (Erlmann et al., 2009; Haining et al., 2018). We 1st found that a transient optoLARG light input of identical amplitude at FAs led to increased Rho activity amplitude compared to stimulation in non-FA PM regions (schematized in Figure 6E). This strongly suggest that the FA is a subcellular region which is highly permissive for GEF-mediated activation of Rho, which could involve a number of mechanisms including Src regulation of Rho release from RhoGDI (DerMardirossian et al., 2006). 2nd, surprisingly, we observed that DLC1 KO cells displayed increased rates of Rho activation kinetics in both FA and non-FA regions in response to a transient optoLARG light input, also leading to increased Rho activity amplitude. This strongly suggests that DLC1 regulates the rate of Rho activation rather than signal duration/termination. Given the promiscuous expression of GEFs and GAPs within a cell type (Moon and Zheng, 2003; Rossman et al., 2005), other Rho GAP(s) might specifically control signal termination, which might by example be important to specifiy the lamellar pattern of RhoA activity in our REF52 cell system. The role of DLC1 in controlling the rate of Rho activation remains to be understood. One can however speculate that DLC1 provides negative feedback to control the rate of Rho activation. Such feedback structures are prevalent in different GTPase signaling networks (Tsyganov et al., 2012; Wu and Lew, 2013), and can endow the latter with properties such as robustness to perturbations. 3rd, we observed that in addition to the increased rate of Rho activation in response to the optoLARG input, DLC1 KO cells also display faster adaptation back to baseline Rho activity compared to control cells. Such non-linear behavior again suggests the existence of additional feedback regulation that might involve other GEFs/GAPs. An important feature of our optogenetic approach in which we perturb cells minimally without eliciting cytoskeletal responses is that we might be able to capture subtle dynamic processes occurring at steady state. Strong optoLARG input that leads to robust actomyosin contractility and edge retraction, as seen in Figure 4, might activate numerous mechanical feedback not present at steady-state, and thus not provide access to cellular regulation processes occurring in an unperturbed, normal state.

Together, our data strongly suggests that in spread fibroblasts at steady state, in absence of any acute mechanical input, DLC1 can exert its GAP activity both at the PM and at FAs to control the rate of Rho activation. In these conditions, DLC1 association with FAs follows FA assembly and disassembly (Figure 6), strongly suggesting that mechanosensitive interactions with talin are not relevant in this context (Haining et al., 2018). Note that RhoA localization is homogeneously distributed at the PM, and any RhoA pool being regulated by DLC1 at the FAs will immediately diffuse in the PM, explaining the absence of local pools of RhoA activity at FAs. This global regulation of RhoA activity levels by DLC1 in fibroblasts is in marked contrast with neuronal growth cones, in which DLC1 regulates a highly focused RhoA activity pattern that localizes to the tip of F-actin bundles in filopodia, most probably to regulate formin-dependent actin polymerization (Fusco et al., 2016). In this context, DLC1 loss of function leads to an enlargement of the RhoA activity pattern at filopodia, which results in longer and stabler filopodia, ultimately inducing more processive growth cone motility and neurite outgrowth. This shows that a specific RhoGAP, such as DLC1, can function in different ways in distinct cellular systems and mechanical contexts.

In the presence of acute mechanical perturbations, we observe distinct DLC1 dynamics at FAs that most likely involve documented mechanosensitive interactions with talin at FAs (Haining et al., 2018). Using optoLARG, we were able to control actomyosin contractility transiently and locally in a small ROI leading to FA reinforcement followed by relaxation (Figure 6). We found that in response to such a mechanical input, DLC1 rapidly dissociates from FAs that exhibit robust reinforcement (Figure 6, schematized in Figure 6F). The release of DLC1 in response to strong FA reinforcement correlates well with in vitro measurements that have shown that DLC1 dissociates from stretched talin due to unfolding of the r7-r8 domain (Haining et al., 2018). Our data also shows that FA relaxation can lead to rebinding of DLC1 to FAs (Figure 6), suggesting that the r7-r8 talin domain might refold in these conditions. Our results illustrate how optoLARG enables spatio-temporal, reversible control of contractility that might help unravel mechanosensitive interactions that are not accessible with classically used global chemical perturbations such as nocodazole-mediated microtubule depolymerization (Krendel et al., 2002), or treatment with lysophosphatidic acid (Pertz et al., 2006) leading to global RhoA activation and massive stimulation of contractility.

We propose that DLC1 provides two levels of regulation of Rho. In absence of any strong mechanical input, active DLC1 at FAs and at the PM throughout the cell seems to contribute to the regulation of steady-state global Rho activity in the cells by controlling its rate of activation. The significance of this mode of regulation is yet to be understood but most likely involves simultaneous regulation with additional Rho GEFs and GAPs. Distinct Rho GAPs might spatially control signal termination that shape the characteristic band of RhoA activity that coincides with the lamella. In response to strong mechanical stimuli leading to FA reinforcement and talin stretching, rapid unbinding of DLC1 from FAs might locally decrease local RhoGAP activity providing positive feedback that could rapidly increase the local rate of Rho activation, possibly strengthening contractility that might lead to FA disassembly as we have observed. Conversely, when the mechanical input vanishes, recruitment of DLC1 to relaxing FAs might rapidly switch off RhoA activity to return to a steady state of low contractility. We speculate that these non-linear signaling behaviors might allow to amplify local mechanical inputs to regulate all-or-nothing cytoskeletal processes such as tail retraction during cell migration. Such tail retraction events are characterized by short bursts of highly localized RhoA activity that are followed by rapid relaxation once the tail has been retracted (Martin et al., 2014; Pertz et al., 2006). Here, SF pulling on FAs at the back of a cell might allow local release of DLC1, rapid increase of local RhoA activity, leading to strong local contraction and tail retraction. Experimentally testing this hypothesis in the context of tail retraction is technically challenging because long term DLC1 perturbation leads to global contractility that precludes cell polarization, directional migration and thus tail retraction episodes. This illustrates an important limitation of the classic genetic perturbation toolkit that has been used to study Rho GTPase signaling.

Our results provide new insights into the complex network circuitry that regulates spatio-temporal Rho GTPase signaling and illustrate how we must tackle this problem. By allowing for precise and reversible perturbations, signaling optogenetics provide the opportunity to dissect the spatio-temporal Rho GTPase fluxes that control patterns of Rho GTPase activity at adequate time/length scales. In the future, a full understanding of Rho GTPase signaling fluxes and formation of signaling patterns will require more systematic approaches to consider the multiple GEFs/GAPs that are involved in shaping cell morphogenesis.

Materials and methods

Cell culture

Rat Embryo Fibroblast 52 (REF52) were cultured in Dubelcco’s Modified Eagle Medium with 4.5 g/l glucose, 10% 4mM L-Glutamine, and 100 U/ml penicillin/streptomycin. Cells were grown at 37 °C and 5% CO2.

SiRNA knockdown

4µl lipofectamine RNAi max was mixed with a total of 500µl of OPTIMEM medium and a pool of SiRNAs against the gene of interest at a final concentration of 60 nM (SiTools Biotech). The mix was added to a 6 well plate containing 100000 cells in a 2 ml medium volume medium.

CrispR/Cas9 Knockout generation

Two guide sequences targeting DNA within exon 5 of the DLC1 gene were selected using the CRISPOR tool for predicted high-specificity protospacer adjacent motif target sites in the rat genome ³². Two complementary oligos each containing the DLC1 guide sequence and Bbsl ligation adapters were synthesized (Microsynth) (sense sequence: 5’-CACCGAACCGAGAGAGCTACCCGG-3’, antisense sequence: 5’-AAACCCGGGTAGCTCTCTCGGTTC-3’). The guide sequences were annealed and ligated into a pSpCas9(BB)-2A-GFP vector. REF52 cells were grown in six-well plates to 60% confluency and transfected with 1µg of the vector together with 1 µl of Lipofectamine 3000 and 5 µl of P3000 solution (ThermoFisher Scientific). Two days post transfection cells were detached, suspended in PBS +1% FBS and sorted into 96 well-plates using fluorescence-activated cell sorting. After expansion, individual clones were detached and suspended in the lysis buffer (10mM TRIS, 50mM KCL, 2.5mM MgCl2, 0.45% Tween-20, 0.05% Gelatine, 0.12 mg/ml of proteinase K). The cell solution was frozen at -80 °C and subsequently kept at 63 °C for one hour. Proteinase K was inactivated by heating the solution to 95 °C for 15 minutes. PCR primers were designed to amplify a 500bp region around the CRISPR cut site (forward primer: 5’-AAGGAGTGTGTCTAACTCCACGCAGACCAG-3’, reverse primer: 5’-CTCCTTAGGACTGTCGCTGCTGTTTTCTCT-3’). Genomic sequences were amplified by PCR and sequenced by Sanger sequencing.

DNA constructs

A “pB3.0” piggybac vector was created by adapting the pPBbsr2 (Matasci et al., 2011) to make it smaller in size. pPBbsr2 was digested with AscI and PacI, yielding 3922 bp and 2864 bp fragments. Full size pPBbsr2 was used as template for a PCR with forward primer 5’-TTAGCATTAATTAAGCGGCCGCGTTGCTGGCGTTTTTCC-3’ and reverse primer 5’-GTGCCTTTACAACTTATGAGTAACCCCGCGCGGACGATT-3’, yielding a 1671bp insert that was ligated back into the receiving 3922bp part of the digested pPBbsr2 vector, creating “pB3.0-BLAST” (Blasticidine resistance). To introduce different resistance cassettes, we performed an overlap extension PCR with pB3.0-BLAST as template for PCR1 (primers A+B) and PCR2 (primers C+D) with the following primers: primer A: 5’-AAGGATGCCCAGAAGGTACCCCATTGTATGGGATCTGATCTGGG-3’, primer B: 5’-GGAAACTTTTTGTGCTATTATGGTGGCCATTCAGCTCTACGTAGCTACT-3’, primer C 5’-CCTTTGAAAAACACGATAATACCACCGGTAAGTCGAGATGCATCGATGA-3’, primer D: 5’-CTCCGCCTTTCTTGGACGTCGGGTTCGAACCGCATTAGT-3’. Primers A and D, the product of PCR1 and PCR2 were used for fusing the two fragments by PCR. The resulting PCR product was ligated into the pB3.0-BLAST vector, digested with Kpn1 and Pst1, by fusion cloning, yielding a pB3.0-noAB vector with a single Age1 site to introduce antibiotic resistance cassettes. To produce pB3.0-HYGRO and pB3.0-PURO, PCRs were performed on template vectors pHygro-PB and pPuro-PB, respectively. For pB3.0-HYGRO forward primer 5’-AACACGATAATACCACCATGAAAAAGCCTGAACTCACCGC-3’ and reverse primer 5’-GCAGGCTCCCGTTTCCTTATCGGCCATTCAGCTCTA-3’ were used. For pB3.0-PURO forward primer 5’-AACACGATAATACCACCATGACCGAGTACAAGCCCACG-3’ and reverse primer 5’-GCGTTCGGGCCACGGACTGGCCATTCAGCTCTA-3’ were used. The respective inserts were inserted via fusion cloning in the pB3.0-noAB vector ³³, digested with AgeI.

Construction of the pB3.0-optoLARG-mVenus-SspB-p2A-stargazin-mtq2-iLID optoLARG system was as follows. The cDNA for optoLARG-mVenus-SspB(nano)-p2A-iLID-CAAX was synthesized by custom gene synthesis (Genewiz, Azenta life sciences, US). The sequence contained the catalytical DH-PH domain of human RhoGEF12 (LARG, aa 766-1138), connected via a linker (GSGSGSGS) to full-length mVenus fused to SSPB(nano). This is followed by a p2A sequence (GSGATNFSLLKQAGDVEENPGP), the iLID module (Guntas et al., 2015) and a CAAX box (KRAS). The construct was cloned into pB3.0-BLAST containing a CAG promotor, yielding pB3.0-optoLARG-mVenus-SspB-p2A-iLID-CAAX. After publication of an improved plasma membrane anchor for the iLID-module (Natwick and Collins, 2021), we changed the part after the p2A sequence. The pB3.0-optoLARG-mVenus-SspB-p2A-iLID-CAAX vector was digested with BsrGI and AflII, and a custom gene synthesized stretch of cDNA containing SspB(nano)-p2A, full-length stargazin, mTurqoise2 and iLID was inserted. This yielded pB3.0-optoLARG-mVenus-SspB(nano)-p2A-stargazin-mtq2-iLID (exact sequences of the gene synthesis products are available on request).

An eGFP DLC1 plasmid containing the 1091aa long isoform originating from mouse (kindly provided by Monilola A. Olayioye) was used to produce mCherry-DLC1. The DLC1 gene was cut out of this plasmid with BamHI and, N-terminally fused to FKBP12-mCherry and inserted into pB3.0-BLAST containing a CAG-promotor, yielding pB3.0-FKBP12-mCherry-DLC1.

Spreading Assay

Glass bottom well plates (Celvis) were coated with 5 µg/ml of human plasma fibronectin purified protein (Merck) for one hour at room temperature. REF52 cells were seeded at a density of 7000 cells per well (using 24 well plates) in FluoroBrite DMEM medium containing 1% FBS, 0.1% BSA, 4mM L-Glutamine and 100 U/ml penicillin/streptomycin. Imaging was done with a Nikon Eclipse Ti-E inverted microscope with an automatic stage. Temperature was kept at 37°C with a temperature control system, humidity (100%) and CO2 (∼5%) with a gas mixer (Life Imaging Services). Focus drift was prevented by the equipped Perfect Focus System (Nikon). The microscope was controlled with Metamorph software (Universal Imaging).

Optogenetic Stimulation experiments

Glass bottom well plates (Celvis) were coated with 5 µg/ml of human plasma fibronectin purified protein (Merck) for one hour at room temperature. REF52 cells were seeded at 6000 cells/well into 12 well plates and allowed to attach for 12 hours. For both RhoA activity and mCherry-DLC1/miRFP-paxillin optogenetic stimulation experiments cells were incubated in FluoroBrite DMEM medium overnight. The microscope was controlled with NIS-Elements (Nikon) for observing DLC1 dynamics and with open-source micromanager software (Edelstein et al., 2014) for edge stimulation. Optogenetic stimulation was localized to specific regions using an Andor mosaic 3 DMD. For mCherry-DLC1/miRFP-paxillin experiments, ROIs were user-defined. For RhoA activity experiments, cells in different fields of view were selected by a user, cells were then segmented with a custom pixel-classifier based on a pre-trained VGG16 convolutional neural network, which was trained on a field of view of the experimental data. ROIs for stimulation masks were then automatically defined at the top and bottom of the cell for consecutive illumination with different light pulse regimes. Fields of views were imaged one after the other on multiple cells. For data extraction the stimulation mask was expanded by gaussian smoothing and then thresholded to cover a wider region, reducing artifacts caused by membrane dynamics. Image intensities were calculated for the overlap of this mask with the segmented cells and the full segmented cells respectively.

Immuno-histochemistry

One hour after seeding, cells were fixed for 10 minutes with 0.2% paraformaldehyde. A 0.1% Triton X solution was used for permeabilization, after which cells were incubated overnight at 4C with the blocking buffer, containing 5% Bovine-Serum-Albumin and 0.05% Tween20 in PBS. After each step of the following protocol cells were washed three times with Phosphate-Buffered-Saline (PBS). Cells were incubated overnight with the primary antibody (1:250 dilution for paxillin (Abcam ab32084), 1:50 for pMLC (Cell Signaling Antibody #3671). Secondary antibody incubation (1:1000) lasted one hour. Phalloidin incubation (1:200 Phalloidin-Atto647N) was performed for 20 minutes and DAPI incubation for 10 hours. Cells were imaged immediately after the last step.

Image analysis

Lamella size, focal adhesion areas (Figure 1) and RhoA activity ratio measurements (Figure 3) were computed with CellProfiler in conjunction with a pixel classifier trained with Ilastik (Berg et al., 2019). Spreading dynamics (Figure 2D) were manually analyzed. FRET ratio images were computed and analyzed with custom python scripts. DLC1 and paxillin dynamics (Figure S3D) were calculated with manually drawn ROIs in Fiji (Schindelin et al., 2012).

Acknowledgements

This work was supported by the Swiss National Science Foundation IZSAZ3_173462 Argentinian-Swiss Joint Research Programme and by the Sinergia CRSII5_183550 grants to Olivier Pertz. We are grateful to the Microscopy Imaging Center of the University of Bern for support (https://www.mic.unibe.ch). We thank Kazuhiro Aoki (NIBB, Japan) for providing the pPBbsr2 vector, David L. Hacker (EPFL, Switzerland), for the pPuro-PB and pHygro-PB plasmids, Judith Trüb (UniBE, Switzerland) for the cloning of pB3.0-BLAST, Monilola A. Olayioye (University of Stuttgart, Germany) for the eGFP DLC1 plasmid, Dean E. Natwick and Sean R. Collins (University of California Davis) for sharing the Stargazin-iLID constructs. We are grateful to Miguel Vicente-Manzanares, Bernhard Wehrle-Haller and Daniel Riveline for constructive comments on the manuscript

Author contributions

O.P. supervised the research. O.P. and M.H. wrote the paper. M.H. designed and performed experiments. L.H. designed and performed optogenetic experiments. J.v.U. built the optoLARG/rGBD circuit. M.D. designed data analysis tools.

Additional Information

The authors declare no competing interests.

Supplementary Figure legends

Figure S1:

Figure S2.

Figure S3.

Figure S4.

Supplementary Movie legends

Movie S1: F-actin dynamics of WT and DLC1 KO REF52 cells during late spreading (relevant to Figure 2B).

Representative ibw contrast movie of WT and DLC1 KO cells expressing Lifeact-mCherry during late spreading. Time: Minutes:seconds. Scale bar = 20 mm.

Movie S2: F-actin dynamics and edge dynamics of WT and DLC1 KO REF52 cells during acquisition of a polarized cell migration phenotype over a period of 15 hours (relevant to Figure 2C).

Representative ibw contrast movie of WT and DLC1 KO cells expressing Lifeact-mCherry for 15. Edge dynamics were color-coded with respect to time. Time: Minutes:seconds. Scale bar = 20 mm.

Movie S3: F-actin and RhoA activity dynamics WT and DLC1 KO REF52 cells during late spreading (relevant to Figure 3A).

Representative movie of WT and DLC1 KO cells expressing Lifeact-mCherry and the RhoA2G FRET biosensor during late spreading.

Left panels: F-actin channel in ibw contrast. Right panels: RhoA activity emission ratio color coded according to color scale.

Time: Minutes:seconds. Scale bar = 20 mm.

Movie S4: Characterisation of the optoLARG/rGBD genetic circuit (relati to Figure 5)

REF52 cells stably expressing the optoLARG/rGBD circuit as well as Lifeact-miRFP are locally stimulated with blue light of identical intensity, first in a ROI at the cell top with a high stimulation frequency, and then with a ROI at the cell bottom with a lower stimulation frequency. The stimulation frequencies are indicated by the appearance/disappearance frequency of the ROIs.

Left panel: rGBD signal, Right panel: F-actin lifeact-miRFP. Note the more robust edge contraction, as well as increased RhoA activity in response to high versus low light frequency stimulation.

Time: Minutes:seconds. Scale bar = 10 mm.

Movie S5: DLC1 dynamics at FAs during an optoLARG-induced contractility pulse (relevant to Figure 6A-C, cell denoted by pink arrow)

REF52 KO cells rescued with mCherry-DLC1 and expressing miRFP-paxillin, and the optoLARG/rGBD circuit were imaged before, during and after a transient optogenetic stimulation in the whole field of view shown in the movie.

Time: Minutes:seconds. Scale bar = 5 mm.