Joining forces: crosstalk between biochemical signalling and physical forces orchestrates cellular polarity and dynamics

Abstract

Dynamic processes like cell migration and morphogenesis emerge from the self-organized interaction between signalling and cytoskeletal rearrangements. How are these molecular to sub-cellular scale processes integrated to enable cell-wide responses? A growing body of recent studies suggest that forces generated by cytoskeletal dynamics and motor activity at the cellular or tissue scale can organize processes ranging from cell movement, polarity and division to the coordination of responses across fields of cells. To do so, forces not only act mechanically but also engage with biochemical signalling. Here, we review recent advances in our understanding of this dynamic crosstalk between biochemical signalling, self-organized cortical actomyosin dynamics and physical forces with a special focus on the role of membrane tension in integrating cellular motility.

This article is part of the theme issue ‘Self-organization in cell biology’.

1. Introduction

Cells have evolved remarkable abilities to sense, integrate and respond to the multitude of external and internal stimuli they encounter during migration [1,2]. In many cases, genetic, in vitro and biochemical approaches have identified the triggers (ligands) and the sensors (receptors) that initiate cellular responses to directional cues as well as the terminal effectors of cortical actomyosin activity that power cell motility [3]. Importantly, cytoskeletal rearrangements and the consequent mechanical forces are not only downstream outputs of biochemical cascades; they also feedback into biochemical signalling networks [4,5] to enable cells to process and integrate information into coordinated cellular decisions.

The dynamic cross-talk between signals and forces plays out over a large range of spatial and temporal scales (figure 1). In many cases, regulatory biochemical signalling originating at molecular scales with relatively fast sub-second kinetics [6] organizes the cortical actin and myosin activity into periodic waves or pulses. These activities occur at mesoscopic scales (approx. 0.1–1 μm) over timescales ranging from tens of seconds to several minutes [7]. Both the signalling events [8,9] and the patterned activity of actomyosin show several signatures of self-organization [10]. Protrusive or contractile dynamics of the cortical cytoskeleton generate forces that ultimately integrate whole-cell behaviours like cell shape changes and movements [11]. The relations between these three processes/phases are not strictly linear due to complex feedbacks arising both within and between each phase (arrows in figure 1).

Figure 1.

Spatio-temporal scales in dynamic feedback between signalling and cellular dynamics. Fast biochemical signalling at the molecular scale underlies all sub-cellular actomyosin dynamics and rearrangement. These mesoscale dynamics of cortical cytoskeleton give rise to forces and global tension that integrate several cellular scale behaviours. The dimensions of each module (box) qualitatively reflect their associated spatio-temporal scales. Interestingly, several regulatory interactions and feedbacks (arrows) exist within and between each module. Understanding the nature of these complex feedbacks will unravel how these modules work together to self-organize cellular dynamics. (Online version in colour.)

In this review, we highlight recent advances in our understanding in the feedback between signalling, cytoskeletal dynamics and physical forces (figure 1). We discuss how biochemical signalling self-organizes cortical actomyosin patterns and activity at sub-cellular scales that then act as generators of intracellular forces. Shifting gear to the scale of whole cells, we review the contribution of physical forces, particularly membrane tension, in integrating global cellular behaviours. Finally, we discuss how cells control their membrane tension.

2. Of waves and pulses: self-organized cytoskeletal dynamics at sub-cellular scales

Cells must organize protrusive actin dynamics or contractile myosin stresses to control cell shape and movement. High-resolution imaging approaches have revealed numerous instances of non-equilibrium and highly dynamic actomyosin patterns like waves, flashes, pulses and flows, but many key questions remain unanswered. How do cells convert molecular-scale biochemical signals into these distinct mesoscopic self-organized spatio-temporal patterns of activity? And what roles do the waves, pulses and flows play in orchestrating cell behaviours?

(a). Actin waves

Actin-based oscillatory waves and pulses have been identified in several cell types and contexts ranging from Dictyostelium [12–16] to leukocytes [5], fibroblasts [17], keratocytes [18,19], neurons [20] and oocytes [21]. Theoretical modelling as well as experimental observations suggests that these oscillatory activities are an outcome of feedback interactions between biochemical signalling and the cortical structures they create. These waves are thought to arise from excitable circuits containing fast positive feedback (via an activator) and delayed negative feedback (due to an inhibitor) [5,15,22–25]. In chemoattractant-stimulated neutrophils, the WASP family veroprolin homologue (WAVE) regulatory complex (WRC) forms oscillatory outward propagating waves (figure 2a–c), which exhibit several features of excitability including an all-or-none response, destructive interference of wavefronts and refractory zones behind the fronts of activation [5]. The self-organization of WRC waves is driven by local positive feedback, likely arising from strong cooperativity in Rac-GTP and phosphatidylinositol (3,4,5)-trisphosphate (PIP3)-dependent recruitment and activation of WRC and self-enrichment of WRC complexes through oligomerization on the plasma membrane [5,28,29]. Activated WRC recruits actin-related protein 2/3 complex (Arp2/3) to nucleate new actin filaments, and the resulting actin polymerization powers the retrograde flow that rips WRC off of the plasma membrane [5,30]. The actin filaments hence exert delayed negative feedback to physically remove the WRC complexes away from their site of activity and recycle the components for the next cycle of WRC wave generation [5,31]. Excitable WRC dynamics allows neutrophils to coordinate chemotactic migration by amplifying shallow external gradients into a steep internal gradient of leading edge activity. These waves also enable a complex cell-level behaviour, barrier avoidance, because they need forward motion to survive (to outrun the local negative feedback). When a migrating cell encounters an immovable barrier, the sites of actin polymerization along the membrane at the barrier are extinguished because they cannot sustain forward motion [5]. Active WRC can generate flat (lamellar) protrusions even when unconstrained by two-dimensional surfaces, and these protrusions are important for changes in cell direction [32]. These WRC-dependent structures enable the guidance and navigation of leucocytes in complex three-dimensional environments [33].

Figure 2.

Self-organized sub-cellular patterns of actomyosin dynamics. (a) Self-organized waves of WAVE regulatory complex (WRC) in neutrophils organize polarity and motility. Neutrophil-like HL-60 cells stimulated with chemoattractant uniformly activate the WRC, which asymmetrically disappears and is polarized to one end of the cell (arrow) and starts generating WRC waves (adapted from [5] with permission). (b) Schematic shows the arrangement of WRC waves and F-actin showing the basis for their self-organization. (c) Kymograph showing several sets of waves with gaps indicating zone of activator (WRC, bright) and inhibitor (F-actin, gap region). (d) Starfish blastomere shows the presence of waves of Rho activity (green) and actin (copper) (adapted from [21] with permission). (e) Fast positive feedback from RhoGEF Ect2 drives Rho-driven F-actin formation possibly via formins that exercise delayed negative feedback via possible recruitment of RhoGAPs. (f) Actomyosin pulses are generated by cycles of actomyosin turnover and contractility leading to advective flows and generating contractile forces. (g) Schematic showing the core signalling network underlying the two proposed mechanisms generating pulses; the Rho-pacemaker model [26] and self-organization model [27].

Actin waves were first observed and have been extensively studied in Dictyostelium [25,34] and they share similar logic and molecular machinery for their generation and propagation to mammalian cells. The excitable Ras/PIP3 positive feedback and processive filamentous actin (F-actin) waves in Dictyostelium cells are thought to arise from coupling between this signal transduction network and the cytoskeletal oscillatory network [24,35]. These cells also use the G protein subunit, Gβ, to tune the degree of coupling between actin oscillators to coordinate internal and external cues for efficient polarization and motility [16]. Adherent migratory cells like keratocytes and fibroblasts also show protrusive and ventral actin waves, respectively, generated by spatial fluctuations of actin polymerization via shuttling of vasodilator stimulated phosphoprotein (VASP) from the leading edge to integrin-based focal adhesion [17,19], potentially enabling cells to coordinate between protrusion dynamics, local sensing and adhesion.

At anaphase onset, Xenopus and starfish oocytes and embryos also exhibit oscillatory cortical waves of Rho-GTPase activity and actin polymerization that coordinate cytokinesis [21]. Triggered by cyclin-dependent kinase 1 (Cdk1) and RhoGEF (guanine nucleotide exchange factor) Ect2, Rho–GTPase activity (activator) generates fresh rounds of polymerized F-actin that subsequently antagonize Rho activity (inhibitor), resulting in excitable spatio-temporally anticorrelated waves of both Rho activity and F-actin accumulation (figure 2d,e). Interestingly, in both systems, the Rho and actin waves propagate at similar velocities (approx. 0.2 µm s⁻¹) with wavelength of approximately 20 µm and period of approximately 80 s [21]. An order of magnitude slower and larger than other propagating actin/WRC waves observed in cellular cortex, the Rho-actin waves in oocyte cortex indicate that excitable cortical activity can be scaled from mesoscopic sub-cellular scale to larger scales during early development. The Rho-actin waves and WRC waves share similar logic underlying their excitability—local positive feedback and actin-based inhibition of the nucleation-promoting factor. The fact that these two systems use different nucleators (Arp2/3 in WRC instead of formins in Rho-actin waves), yet are both susceptible to actin-based inhibition, suggests a general feature of these actin-based excitable waves.

Finally, neuronal axons exhibit lateral F-actin waves that propagate via directed assembly and disassembly of actin cables mechanically anchored to the cell membrane via clutch molecules like L1-CAM. These waves, which are proposed to facilitate intracellular transport, provide a distinct mechanism to generate waves outside the excitable activator–inhibitor framework [20,36].

(b). Actomyosin pulses and flows

Actomyosin pulses and flows refer to a broad class of cortical biomechanical phenomenon driven by the active turnover and contractility of non-muscle myosin II on actin networks (figure 2f). They were first described in Caenorhabditis elegans zygotes where actomyosin pulses accompany global cortical flows to set-up the anterior–posterior axis [37]. Pulsatile actomyosin dynamics have now been documented in several adherent single cells [38], during a wide range of morphogenetic events from Drosophila embryos [39–41], and recently in mouse embryos [42]. While our understanding of the mechanistic basis of their regulation is still evolving, pulsing is thought to achieve two broad goals: first, it imparts fast minute-timescale contractile stresses and forces [11,43] and, second, pulses drive advective (transient local enrichment) flows of cortical actomyosin and associated regulators [44,45].

How are the actomyosin pulses and flows generated? Two related mechanisms (figure 2g) for actomyosin pulses and flows have been proposed based on recent findings from studies in C. elegans embryo and Drosophila epithelia. First, the ‘pacemaker’ hypothesis suggests myosin pulses could be driven by cyclical activity of the Rho-GTPase. Rho activates Rho-Kinase (ROCK), which in turn phosphorylates the myosin regulatory light chain and at the same time inhibits myosin phosphatase, thereby stimulating myosin activity [46]. Moreover, Rho-GTPase is under regulatory control of Rho-GEF (activator) and Rho-GAP (GTPase activating protein) (inhibitor), and recent results indicate that Rho-GTPase exhibits autocatalytic activation upon pulse initiation [47]. Finally, the pulses can enrich the inhibitor (like a Rho-GAP) via advection to regulate Rho activity. Hence cyclical regulation of Rho-GTPase via its upstream activator/inhibitors can generate pulsatile myosin activity. Recent results support the presence of oscillatory Rho activity in C. elegans embryo [26,47], and Rho-GEF2 and active Rho spatially localize to and often precede myosin II accumulation in pulses [48].

Alternatively, myosin pulses can arise as an emergent behaviour of a bio-mechanical self-organization process [27]. These pulses are thought to be generated by active turnover of myosin II via phospho-cycling of the regulatory light chain (bio-chemical component) and the advection due to the motor's contractile activity (mechanical component). Within this framework, pulses arise at an intermediate regime of myosin II activity and dissociation, allowing advective enrichment of myosin II and its upstream regulators like Rho1 and ROCK, which can further stimulate pulsing (positive feedback). However, the rapid build-up of stabilized actomyosin network and possible accumulation of the phosphatase [49] lowers further advection and triggers dissociation of myosin II filament (delayed negative feedback). Moreover, myosin II contractility can intrinsically disassemble actin filaments [50,51] and contribute to the overall turnover of the actomyosin pulse. Although this mechanism of pulsing does not require the upstream pacemaker activity of Rho, it is possible that the advective flows can potentially also recruit Rho regulators and modulate its activity.

Several recent studies have now shown that actomyosin pulses exert contractile forces to drive deformation of the cell shape via apical constriction [39] and junction remodelling [40,52], which underlie several key morphogenetic processes in Drosophila [11]. While pulsing has been studied at the level of single cells in epithelia, it is possible that synchronous patterns of pulsing can give rise to tissue-scale forces to coordinate tissue-scale dynamics and shape changes [41,53]. Recent studies also highlight the mechanisms underlying large-scale flows. In the case of C. elegans one-cell embryo, the actomyosin flows are triggered by cortical instabilities caused by the entry of sperm at the posterior end [37] leading to anisotropies in actomyosin tension [54] and providing a physical mechanism for large-scale flows. The gradient of contractility pulls the cortex towards the anterior end and deposits the anterior partitioning defective (PAR) complex, allowing components of the posterior PAR complex to localize to the posterior end. Biochemical negative feedback between components of posterior and anterior complex [44] and inhibitory interaction between PAR-2 of posterior complex and myosin further prevents myosin accumulation, thereby stabilizing the pattern [37].

3. Cells generate and use forces to organize whole-cell behaviours

The dynamic actin structures discussed in the last section also generate or impart forces over the scale of whole cells or across tissues. In this section, we focus on how physical forces arising from membrane tension integrate cellular processes ranging from cell shape [55] to vesicular trafficking [56], cell spreading [57], polarity [58] and motility [59]. The apparent plasma membrane tension is a combination of the in-plane tension of the bilayer and the energy stored in membrane-cortex attachments [60]. Membrane tension exerts a force along the cell boundary that needs to be overcome by any processes involving membrane deformations [61]. Moreover, the tension of the membrane also depends on the amount of excess membrane locked up in local membrane folds or caveolar pits; an increase in membrane tension leads to disassembly of caveolae [62]. Vesicular trafficking can change the effective cell surface area and alter membrane tension by increasing exocytosis during high membrane tension and stimulating endocytosis in response to low membrane tension [57,63–65]. During motility, cells move forward by polymerizing new actin filaments at the leading edge [66]. The increase in filamentous actin density results in protrusive forces that iron out the membrane folds [63], resulting in an increase of membrane tension (figure 3a). We will use the case of neutrophil polarity to highlight the feedback between membrane tension and actin assembly. Neutrophils undergo a rapid transition from an unpolarized resting phase (figure 3b, left) to global activation of new actin growth during chemoattractant stimulation (figure 3b, middle). The protrusive forces from the actin growth lead to a nearly fourfold increase in membrane tension, which is thought to equilibrate rapidly across the cell [58]. Initially, sites of activation of polarity factors (like WRC) are generated throughout the cell, but these nascent protrusions compete via an increase in membrane tension to enable a single ‘winner-take-all’ leading lamellipod [58]. Hence membrane tension acts as a global inhibitor of actin assembly and facilitates the competition between fronts to constrain actin growth in a single robust leading edge (figure 3b, right).

Figure 3.

Membrane tension as global mechanical integrator in polarity and motility. (a) Feedback between actin polymerization and membrane tension. Cell membrane with low F-actin density and/or excess membrane folds has low in-plane tension (left). Increase in F-actin polymerization at the membrane generates outward protrusive forces (vertical arrows) leading to increase in membrane tension (horizontal double-headed arrow, right) until it becomes limiting to growth of new filaments and effectively stalls the network. (b) Membrane tension coordinate leading edge activity and polarity. Resting neutrophils (left) have low cortical actin density and hence low membrane tension allowing uniform increase in actin density and new protrusions (middle) shortly upon chemoattractant stimulation. Consequent increase in tension can mechanically and/or biochemically inhibit secondary fronts to converge on a single polarized leading edge. (c) Feedback between membrane tension and actin growth can be both mechanical (direct) and biochemical (indirect) [58,67]. mTORC2, mechanistic target of rapamycin Complex 2; PLD2, phospholipase D2. (Online version in colour.)

Similar mechanisms also drive polarization and motility of fish keratocytes [68,69] and nematode sperm [70], where membrane tension streamlines actin polymerization in the direction of movement and acts to limit and disassemble filaments outside the moving front. Cells have a remarkable ability to resist changes in membrane tension through homeostatic mechanisms. For instance, fish keratocytes respond to acute increase of cell surface area by exogenous lipid vesicle fusion (which would be expected to decrease membrane tension) with a burst of actin polymerization that brings them back to their pre-fusion level of membrane tension and migration speed [68]. Cells can also compensate for transient increases in membrane tension. During cell spreading [63] or frustrated phagocytosis [56], actin polymerization drives the depletion of membrane reserves and hence transient increase in tension, leading to exocytosis of intracellular vesicles to decrease membrane tension and enable further membrane spreading. In subsequent sections, we discuss how membrane tension can engage with biochemical signalling and the different cellular strategies of regulating tension.

4. Membrane tension and its dialogue with biochemical signalling

How do cells sense and respond to changes in membrane tension? Membrane tension can stall actin networks directly by increasing the mechanical load on the existing network past the regime in which new actin monomers can be added [66]. This allows membrane tension to regulate the assembly of actin networks directly by purely mechanical means [61]. Indeed, recent in vitro reconstitution studies [71] and live imaging of actin assembly in fish keratocyte lamellipodia [72] suggest that dendritic actin networks can intrinsically adapt to increases in mechanical load by increasing the density and altering the geometry of newly assembled actin filaments.

However, it is becoming clear that changes in membrane tension can also act indirectly by triggering mechanosensitive biochemical signalling [67]. Since cells operate at or near constant membrane tension regimes, a combination of direct and indirect tension response allows homeostatic control over the tension set-point. One conserved node for tension regulation and cell homeostasis is the target of rapamycin Complex 2 (TORC2) pathway. In yeast, membrane stretch unfolds membrane invaginations like eisosomes to release proteins that activate TORC2 [73]. TORC2 stimulates sphingolipid biosynthesis in response to membrane stretch as a compensatory mechanism to restore membrane composition homeostasis [73]. In Dictyostelium, mechanical stimuli lead to activation all the major chemotactic signalling hubs including TORC2 [74]. In neutrophils, membrane tension changes trigger phospholipase D2 (PLD2)-dependent activation of mTORC2 [67]. This signalling axis exerts an overall negative feedback on actin polymerization (figure 3c). Cells lacking PLD2 and Rictor, a key component of mTORC2 complex, show profound defects in leading edge size, global motility and elevated membrane tension upon chemoattractant stimulation [67,75].

Acute changes in membrane tension lead to disappearance of leading signals like WAVE complex and Rac [56,58,76]. While both physical and regulatory interaction between mTORC2 and Rac has been reported [77,78], the underlying mechanisms connecting mTORC2 activity to inhibition of leading edge signals are yet to emerge. Recently, membrane tension increase was also shown to inhibit FBP17 (Bin-Amphiphysin-Rvs domain protein)-dependent N-WASP/WASP activation, thereby reducing overall actin assembly [79], highlighting the involvement of BAR domain proteins in membrane tension actin assembly feedback. However, whether BAR domain proteins play dual roles as tension sensors and signalling scaffolds has not yet been established.

5. Cellular strategies for regulating membrane tension

The previous section introduced several membrane tension homeostats, but it is also important to consider how they interact to form cohesive programs that cells can use under different physiologically relevant conditions. Cells can stretch or relax the membrane—altering the apparent plasma membrane (PM) tension—in two main ways (figure 4a–c): (i) by modulating actin assembly to change the aspect ratio of cells without changing the volume and (ii) by altering osmotics to force water movement across the membrane, which changes cell volume [80]. Both of these responses rely on the ‘excess’ PM surface area of cells, which can be released by unfolding exvaginations/invaginations [57]. Cells can change the former by varying the rate of actin protrusion at the leading edge [68] and the latter by modifying hydrostatic pressure [57].

Figure 4.

Cellular strategies for membrane tension regulation. Conditions leading to in-plane membrane tension perturbation and responses by the cell homeostatic machinery. Membrane tension is affected by changes in actin polymerization at the leading edge (a), in the balance of osmolytes across the membrane (b) and in external compression (c) of the cell (e.g. neutrophil squeezing through an epithelial layer). Cells can restore their ‘target’ membrane tension by using their homeostatic machinery to control actin polymerization at the leading edge (d) and to activate volume regulatory machinery to force water across the membrane (and changing the cell volume) (e). (Online version in colour.)

The contribution of cytoskeletal forces to membrane tension is well established. As mentioned previously, stalling the lamellipodia in moving keratocytes leads to a large drop in PM tension, and increasing keratocyte PM surface area has only minor effects on PM tension and motility because the lamellipodial actin network expands until PM tension is close to the original set-point [68]. However, the idea that such cytoskeletal tension homeostats may be complemented by osmotic regulators to buffer membrane tension during migration is attracting renewed interest [81]. Indeed, work in metastatic breast cancer cells shows that migration in confined spaces is strongly reliant on osmotic regulators, while actin polymerization is dispensable [82]. This ‘osmotic engine’ relies on directed water permeation—net inflow of water at the front of the cells and net outflow of water at the back—to push the cell forward during migration in confined environments. One of the key regulators from that study, the ubiquitous Na⁺/H⁺ antiporter, NHE-1, has long been known to localize to lamellipodia in migrating cells [83] and is required for leucocyte chemotaxis and for the increase in volume observed during migration [84]. Blocking activity of NHE-1 greatly decreases the rounding pressure exerted by HeLa cells during mitosis [85], suggesting that cells can use this regulator to increase osmotic pressure and PM tension.

Other channels, for example, Piezo1, can both read out and change membrane tension. Piezo1, a widely expressed mammalian mechanosensitive channel, responds directly to increases in in-plane membrane tension [86] and activates secondary ion channels to decrease cell volume and (presumably) membrane tension in red blood cells (RBCs) [87]. The ability of RBCs to decrease their volumes in response to squeezing has been suggested to improve their ability to pass through constrictions in the vasculature. In cases where the primary contributor to PM stretch is actin polymerization, negative feedback on polymerization through PLD2-mTORC2 [67] likely dominates (figure 4d). But when the membrane is stretched due to osmotics (like during neutrophil migration towards sites of freshwater ingress in zebrafish tail wounding experiments [88]) or confinement, cells could respond primarily with osmotic regulators to restore tension (figure 4e). However, this idea has not been formally tested and needs to be, both in terms of the relevant regulators in each context and the direct measurement of membrane tension and cell volume under these different conditions.

Most analyses of cell response to membrane stretch have been done in cases where cytoskeletal or osmotic-based feedback is likely to dominate. For example, for cells in unconstrained environments such 2D migration, cytoskeleton-based changes in cell shape are likely the dominant source of membrane tension, and feedback to actin polymerization is likely to be the primary point of control of cell polarity and movement [58,68]. The other extreme is when membrane tension is predominantly affected by large changes in external osmolarity [88] or under one-dimensional confinement [82], where osmotics dominate. But which homeostat(s) are used for conditions where both cytoskeletal-based membrane stretch and osmotics are likely to be relevant, such as cell squeezing in complex three-dimensional environments? More generally, following membrane stretch, how do cells decide whether to compensate by decreasing actin assembly or changing ionic flow to pump out water? Though these compensations both result in a decrease in tension, they could have dramatically different effects on cell shape and migration. Is the choice between homeostatic control mechanisms dependent on different timescales or magnitude of membrane stretch, and are these homeostats solely reading out changes in membrane tension or do they rely on other physical properties such as cell shape or chemical cues such as intracellular ion concentration?

6. Conclusion: unresolved questions and future challenges

There are several open questions regarding the interplay between forces and signalling in coordinating cellular dynamics. Future work will require newer interdisciplinary approaches and tools to address these questions. We highlight some of these questions and approaches.

• — What are the molecular mechanisms underlying actin waves or actomyosin pulses? While several of the core molecular players involved in migration are known, our knowledge of regulatory components is still emerging. Unbiased genetic screens (greatly facilitated by high-throughput methods such as CRISPR) could be used to reveal novel players in these pathways.

• — What are the minimal components essential for waving or pulsing? Can in vitro reconstitution of native or synthetic systems capture the rich dynamics seen in cells and tissues?

• — How do cells measure and interpret membrane tension changes during their motility? What is molecular nature of the sensors and underlying homeostats maintaining tension set points?

• — While there are few force biosensors for measuring contractile forces and tension [89], it would be important to develop similar sensors to monitor protrusive forces and local membrane tension in a non-invasive fashion that gives continuous information about these forces over the entire cell surface. Newer approaches are also necessary to accurately measure membrane surface area to complement the emerging tools for measuring volume.

• — What are the relative contributions of membrane tension and cortical tension during various migratory modes—mesenchymal versus amoeboid or in two-dimensional versus three-dimensional environment [90]?

• — Existing tools for manipulating membrane tension such as osmotic changes, aspiration and stretchable surfaces also change other cellular properties. We need a precise and titratable method for manipulating membrane tension under physiologically relevant mechanical conditions.

These are exciting times for studying cellular processes at the interface of biochemical signalling and forces, and future work will continue to shape our understanding of how both these players act together to self-organize cellular dynamics.

Data accessibility

This article has no additional data.

Authors' contributions

S.S. and O.D.W. contributed to the conception; all three authors drafted the text. S.S. and T.L.N. illustrated the figures with inputs from O.D.W.

Competing interests

We declare no competing interests.

Funding

This work was supported by NIH GM118167. T.L.N. acknowledges support from NSF Graduate Research Fellowship Program grant no. 1650113.