Forcing a growth factor response – tissue-stiffness modulation of integrin signaling and crosstalk with growth factor receptors

Research throughout the 90s established that integrin crosstalk with growth factor receptors stimulates robust growth factor signaling. These insights were derived chiefly from comparing adherent versus suspension cell cultures. Considering the new understanding that mechanosensory inputs tune adhesion signaling, it is now timely to revisit this crosstalk in different mechanical environments. Here, we present a brief historical perspective on integrin signaling against the backdrop of the mechanically diverse extracellular microenvironment, then review the evidence supporting the mechanical regulation of integrin crosstalk with growth factor signaling.We discuss early studies revealing distinct signaling consequences for integrin occupancy (binding to matrix) and aggregation (binding to immobile ligand). We consider how the mechanical environments encountered in vivo intersect with this diverse signaling, focusing on receptor endocytosis. We discuss the implications ofmechanically tuned integrin signaling for growth factor signaling, using the epidermal growth factor receptor (EGFR) as an illustrative example. We discuss how the use of rigid tissue culture plastic for cancer drug screening may select agents that lack efficacy in the soft in vivo tissue environment. Tuning of integrin signaling via externalmechanical forces in vivo and subsequent effects on growth factor signaling thus has implications for normal cellular physiology and anti-cancer therapies.


Introduction
Heterodimeric transmembrane integrins are major regulators of cellular interaction with the extracellular matrix (ECM), controlling essential cellular functions by signaling bi-directionally between the cell and the ECM. Alongside work in the 90s showing that integrin ligation and subsequent clustering instigates a hierarchy of signaling events (Miyamoto et al., 1995b), a picture of force-mediated activation of integrin transduced signals has emerged (for examples, see Sawada et al., 2006;Bradbury et al., 2017;Janostiak et al., 2014). The demonstration of the pro-tumorigenic effects of the stiff collagenous matrix that characterizes many solid tumors (Butcher et al., 2009) has led to a heightened focus on the effects of a rigid tissue. However, this does not address the entire biomechanical lifecycle of the (cancer) cell; while many tumors have a rigid ECM, the extent to which this is applicable to all solid tumor types has not been comprehensively quantified. Moreover, invasive cancer cells transmigrate tissues that may be softer than the primary tumor. Considerably less is known about the effect of the entire range of tissue forces exerted on cells in vivo, yet this is likely to have important consequences for cell function. This Review first provides a brief historical perspective on integrin signaling. Next, the evidence supporting mechanical regulation of integrin crosstalk with growth factor signaling is reviewed. Finally, potential mechanisms of mechanical modulation of crosstalk are considered, with the epidermal growth factor receptor (EGFR) as an illustrative example. This Review departs from the many excellent reviews emphasizing the role of a stiff tissue environment by instead exploring how the range of tissue stiffnesses encountered by cells in the body may tune adhesion signaling and modulate the crosstalk between integrins and growth factors, with the attendant consequences for anti-cancer therapies.

Tuning in
Far from being a simple mechanism whereby ligand-binding activates downstream signaling, integrin signaling constitutes a hierarchical recruitment of signaling molecules that is modulated by the states of ligand occupancy, receptor aggregation and the transmission of extracellular and intracellular forces (Kechagia et al., 2019;Humphries et al., 2019;Geiger and Yamada, 2011). Thus cells "tune in" to their environment via their integrins, adjusting downstream signaling according to combined inputs.

Occupancy and aggregation
Miyamoto and colleagues demonstrated that receptor occupancy versus aggregation, alone and in combination, each generate diverse downstream integrin-signaling responses (Miyamoto et al., 1995a,b). To briefly summarize their findings: receptor aggregation with antiintegrin antibody-coupled beads triggers recruitment of 'first-stage' molecules, including a tensin family member and focal adhesion kinase (FAK, also known as PTK2). Ligand occupancy by binding to cognate ligands stimulates so-called 'outside-in' signaling, which causes the accumulation of cytoskeletal proteins, such as vinculin, talin and α-actinin proteins. 'Inside-out' signaling following tyrosine phosphorylation of cytoplasmic molecules, results in the focal accumulation of signaling molecules at the receptors, activating mitogen-activated protein kinase (MAPK) extracellular related kinase ERK1 and ERK2 (ERK1/2, also known as MAPK3 and MAPK1, respectively) and Jun N-terminal kinase (JNK) (also known as stressactivated protein kinase, SAPK) signaling. Maximal integrin signaling depends on the combination of ligand occupancy (outside-in signaling), tyrosine phosphorylation (inside-out signaling) and integrin aggregation (Miyamoto et al., 1995a,b;Meyer et al., 2000). Together, these actions facilitate accumulation of the cytoskeletal proteins that are required for association of bundles of filamentous (F)-actin at the cytoplasmic tail of the integrins, setting up the structures required to mediate force sensing through the focal adhesions (see section on 'Probing tissue forces' below). In the presence of aggregated integrins but the absence of integrin activation, extracellular forces fail to stimulate cyclic AMP production, indicating that integrin occupancy and aggregation is required in order for cells to transmit force signaling (Meyer et al., 2000). In the artificially rigid environment of commonly used ECM-coupled plastic or glass culture dishes, conditions thus favor maximal integrin signaling (Pelham and Wang, 1997). The understanding of integrin-mediated signaling that has been described through the lens of cells grown on such dishes may therefore obscure the full spectrum of the in vivo response.

Probing tissue forces
Organs and tissues in the body each have unique and characteristic mechanical features, which can be described by the elastic modulus, a factor describing the ratio between the force exerted and the resulting deformation. Mechanically static or compliant tissues, such as brain or lung, are characterized by a low elastic modulus (≤10 kPa), whereas tissues exposed to a high mechanical loading, such as bone, have elastic moduli orders of magnitudes greater (∼10 GPa); the elastic moduli of plasticware and glassware far exceeds any in vivo modulus (Cox and Erler, 2011). Cells sense the mechanical features of the external tissue by exerting intracellularderived forces on the matrix that are generated by myosin-motor mediated contraction of bundles of F-actin (stress fibers) (Kechagia et al., 2019). The stress fibers are coupled via adaptor proteins to the cytoplasmic tail of ECM-bound integrins and 'pull' on ECM fibers on the external cell surface through the integrins . The resultant macromolecular structure consisting of ECM-bound transmembrane integrins, actomyosin filaments and associated cytoplasmic signaling molecules is termed a focal adhesion (Burridge et al., 1988). Integrins diffuse laterally across the plasma membrane and cluster into focal adhesions in response to increased mechanical load, thereby dispersing force and reducing the tension through individual receptors (Shemesh et al., 2005;Cao et al., 2015). In rigid environments, aggregations of ligand-bound integrins generate large focal adhesions (Riveline et al., 2001) (Fig. 1A) where integrins are simultaneously pulled by the ECM and internal cytoskeleton with equal magnitude and in opposing directions (Kechagia et al., 2019). The magnitude of the force required to deform the ECM by the cell is therefore related to the ECM stiffness and, in this manner, cells can probe the mechanical forces in their local tissue. Most natural ECMs and tissues have both elastic and viscous qualities; viscous materials strain linearly over time, while elastic materials return to their original form once the stress is released (Cantini et al., 2020). Most of the current understanding of integrin activity in response to the forces of the underlying tissue has focused on the elastic features (here referred to as the tissue stiffness). For purely viscous substrates, it has been reported that focal adhesions and FAK phosphorylation increase in proportion to increased viscosity (Bennett et al., 2018), and recent studies suggest that the impact of viscosity may be greater for soft substrates than for stiff substrates (Gong et al., 2018). However, the interrelationship between focal adhesions, viscous materials and viscoelastic (i.e. having both viscous and elastic properties) tissue is currently less well understood.
Tuning in to modulate signaling A renewed vigor in the area of mechanosensing (sensing mechanical cues in the environment) and mechanotransduction (converting mechanical stimuli into biochemical signals) was sparked by the demonstration that substrate rigidity can independently cue stem cell fate through directing the differential expression of key lineagespecific markers (Engler et al., 2006). Cells 'tune in' to the In the presence of extracellular ligands on highly rigid surfaces (plastic and/or glass), integrins aggregate in focal adhesions, with bidirectional forces transmitted through the integrins. This leads to high-level phosphorylation (P) of downstream target molecules, culminating in the activation of downstream signals regulating proliferation, survival and migration. (B) Under conditions of cell suspension, integrins do not aggregate, and there is no integrin-mediated stimulation of phosphorylation. For most normal (non-cancerous) cells, this leads to cell death; in contrast, cancer cells are anchorage-independent and thus can continue to grow in suspension. (C) The anchorage-independent growth that characterizes cancer cells suggests that the rigidity sensing machinery is altered in cancer cells; however, cancer cells still respond to matrix signals via the integrins. The status of integrin signaling in response to different substrate stiffness (indicated by wedge with gradient of color, reflecting increased substrate rigidity) is not well understood (indicated by ?). As integrin engagement increases and focal adhesions grow in response to increased external force, potentially there are accompanying changes in integrinmediated phosphorylation signaling cascades. biomechanical features of the extracellular environment using varied and interconnected mechanisms. So far, these have been found to include the mechanosensitive signaling YAP/TAZ pathway (Dupont, 2016), protein activation through force-mediated conformational changes to protein structure, for example the focal adhesion adaptor protein p130Cas (also known as BCAR1) (Sawada et al., 2006;Bradbury et al., 2017), force-dependent hierarchies of kinase activation (Prager-Khoutorsky et al., 2011) and changes to protein trafficking (Du et al., 2011;Yeh et al., 2017). A common mechanism in many of these mechanosensitive pathways is the uncovering of cryptic binding sites when molecules are placed under tension (Vogel and Sheetz, 2006). For example, p130Cas contains an unstructured substrate-binding domain that is stretched under force, facilitating access by Src tyrosine kinase (Sawada et al., 2006). Src then phosphorylates multiple tyrosine residues within the substratebinding domain, thereby amplifying p130Cas-mediated signals (Sawada et al., 2006;Bradbury et al., 2017). While it has long been known that focal-adhesion-associated molecules, such as FAK, are rapidly dephosphorylated following cell detachment (Schlaepfer et al., 1994) (Fig. 1B), there have been limited studies of the effect a soft substrate has on their phosphorylation state. Since cancer cells avoid the detachment-induced apoptosis that characterizes normal cells, their rigidity sensing machinery is clearly altered from normal (Khwaja et al., 1997;Wolfenson et al., 2016), yet, integrin signals received through matrix adhesion still regulate cancer cell behavior (Levental et al., 2009). Conceivably, different levels of force (in the presence of ECM ligand) may correspond with different phosphorylation events (Fig. 1C). Indeed, recent work showing that tyrosine phosphorylation of the Cas-family substrate-binding domain is reduced, but not absent, on soft substrates (Bradbury et al., 2017), suggests the need to reconsider the activation of adhesion signaling from an on-off (for attached and detached, respectively) mechanism, to a more nuanced vision, in which cells 'tune in' to the stiffness of the surrounding tissue and adjust downstream signaling accordingly.

Force-regulated receptor endocytosis
Endocytic trafficking of the integrins to and from the plasma membrane precisely controls their signaling, providing both spatial and temporal regulation of downstream signaling Moreno-Layseca et al., 2019). There are intriguing hints in the literature that the mechanical features of the extracellular environment may influence the endocytosis of integrins. Yu and colleagues demonstrated that a low-force microenvironment is a prerequisite for recruitment of the clathrin-mediated endocytosis (CME) adaptor protein Dab2 to active β3 integrins; loss of traction force on ligand-bound β3 integrin thereby stimulates clathrinmediated integrin endocytosis (Yu et al., 2015). Similarly, when cells were plated on a soft substrate matching the elastic modulus of the in vivo brain environment, β1 integrin endocytosis was increased (Yeh et al., 2017;Du et al., 2011). Increased endocytosis resulted in loss of β1 integrin from the cell surface and formation of a complex on the endocytic vesicles between β1 integrin and the neuronal regulator bone morphogenic protein receptor 1A (BMPR1A). This complex prevents BMPR1A-mediated phosphorylation of its downstream target, the Smad proteins (Du et al., 2011). In turn, unphosphorylated Smad proteins are free to robustly promote neuronal induction (Chambers et al., 2009). Thus, on soft substrates, increased β1 integrin endocytosis stimulates neuronal differentiation through dampening an otherwise inhibitory pathway. Mimicking the changes that occur when cells are plated on soft substrates by stimulating biochemical relaxation of acto-myosin cellular tension drives integrins into flat clathrin lattices (Zuidema et al., 2018). Although it is unclear whether these structures are the result of frustrated endocytosis or if they play an active role in adhesion signaling (Sochacki and Taraska, 2019), this is another example of how underlying tissue stiffness feeds into the presentation of integrins at the cell surface. These examples illustrate how cellular interpretation of altered tissue rigidities modulates integrin signaling to generate force-specific signaling outcomes.

It takes two to signal
In addition to probing the external environment and transducing biochemical and mechanical signals, integrins also play an important role in modulating growth factor receptor signaling through crosstalk. In the context of mechanical tuning of integrin signaling, this suggests a new mechanism for modulating growth factor signaling.
In common with integrins, receptor tyrosine kinases (RTKs), such as insulin receptor, platelet-derived growth factor receptor (PDGFR), vascular endothelial growth factor receptors (VEGFRs), EGFR and hepatocyte growth factor receptor (Met), are continually internalized from the cell surface. Depending on ligand interaction they are either recycled on endosomes back to the plasma membrane or targeted for degradation by the lysosome. Integrins physically interact with the RTKs in the plasma membrane (Miyamoto et al., 1996;Falcioni et al., 1997;Schneller et al., 1997;Moro et al., 1998;Soldi et al., 1999;Borges et al., 2000;Sieg et al., 2000;Mariotti et al., 2001;Trusolino et al., 2001), and physically constrain and direct their response to soluble growth factors (Yamada and Even-Ram, 2002;Giancotti and Tarone, 2003;Asthagiri et al., 2000;Barrow-McGee et al., 2016). Furthermore, integrins and RTKs have been shown to undergo endocytosis to the same endosomes, where integrins may support RTK signaling (Barrow-McGee et al., 2016). RhoA-mediated regulation of this process (Mai et al., 2014) highlights a link between acto-myosin contractility and endocytosis. Moreover, the same oncogenic pathways, such as those downstream of mutant p53, may drive recycling of both RTKs and integrins (Muller et al., 2009(Muller et al., , 2013. Several elegant studies have demonstrated that outside-in-mediated integrin signaling shares a large overlap with RTK signaling pathways (Sieg et al., 2000;Ivaska and Heino, 2011). Here, key shared molecules, such as FAK and Src, represent an important point of convergence between the signaling pathways. Given the overlap in downstream signaling and their membrane proximity, it is unsurprising that crosstalk exists between integrins and RTKs and growth factor receptors.
The intimate relationship between integrin-mediated adhesion and changes in growth factor-directed behavior has long been appreciated. In fact, signals derived through ECM-integrin binding can promote ligand-independent RTK activation (Sieg et al., 2000;Miyamoto et al., 1996;Sundberg and Rubin, 1996;Wang et al., 1996;DeMali et al., 1999;Mitra et al., 2011). Integrin-stimulated EGFR auto-activation is sufficient to induce partial activation of EGFR signaling to induce ERK and phosphoinositide 3-kinase (PI3K) pathways, which promote a degree of cell survival even in the absence of growth factors (Moro et al., 1998). After binding to the ECM, integrins cluster and form a complex with, and activate, Src, which then induces EGFR phosphorylation. However, the consequent signaling is not identical, as the specific phosphorylation of residue Y1148 on EGFR that is seen following EGF-ligand stimulated EGFR activation does not occur in the integrin-Src-EGFR pathway (Moro et al., 2002).
Crosstalk between integrins and growth factor signaling highlights the potential for upstream ECM rigidity to impact integrin-mediated regulation of RTK signaling and influence downstream cellular responses. Indeed, stiff substrates induce osteogenic commitment of stem cells regardless of the presence of growth factors, whereas culture on soft substrates requires additional biochemical cues to induce the same cell fate (Banks et al., 2014). Among a subset of tyrosine kinases that are activated by distinct matrix rigidities are a number of RTKs, namely, colony stimulating factor 1 receptor (CSF-1R), neurotrophic receptor tyrosine kinase 3 (NTRK3), Met, receptor tyrosine kinase-like orphan receptor 2 (ROR2) and fibroblast growth factor receptor 4 (FGFR4) (Prager-Khoutorsky et al., 2011). Based on their findings, the authors proposed that the kinase response to tissue rigidity serves as a molecular checkpoint to jointly control cell contractility and focaladhesion-mediated mechanosensing (Prager-Khoutorsky et al., 2011). Moreover, RTKs control the contractile pinching that is necessary for integrins to detect force (Yang et al., 2016). This was demonstrated by the report that depletion of the RTKs AXL and ROR2 from fibroblasts grown on micropillars coupled to substrata of various rigidities altered rigidity sensing, and increased the magnitude or duration of local micropillar contraction ('pinching') in the absence of receptor ligands (Yang et al., 2016).
Crosstalk between RTKs and integrins influences signaling and trafficking of both receptor types (Ivaska and Heino, 2011). Integrins promote the activation of RTKs (Wang et al., 1996;Mitra et al., 2011) and vice versa (Liang and Chen, 2001;Chiu et al., 2002;Trusolino et al., 1998). Ligand stimulation of the PDGFR activates β1 integrin, resulting in activation of the actin-nucleating endocytic protein neuronal Wiskott-Aldrich syndrome protein (N-WASP, also known as WASL), increased localized endocytosis of PDGF and consequent downstream signaling (King et al., 2011). Under conditions of cell suspension, co-internalized β1 integrin and Met sustains Met-dependent ERK1/2 activation in the autophagy-related endomembranes, promoting the malignant characteristics of survival and invasion (Barrow-McGee et al., 2016). An open question is whether a soft tissue environmentas opposed to cells in suspensionmight similarly stimulate this pro-malignant β1 integrin-Met signaling. EGF stimulation causes β1 integrin to be lost from from cell-cell adhesions (Mukoyama et al., 2007), while EGFR inhibition on stiff surfaces where integrin signaling is active alone is sufficient to reduce contractility-mediated spreading in fibroblasts (Saxena et al., 2017). These examples demonstrate the intimate crosstalk between the functioning of growth factor receptors and integrin signaling. To date, the role of substrate viscosity in mediating this crosstalk is unclear; however, it is notable that FAK phosphorylation correlates with the extent of substrate viscosity (Bennett et al., 2018). Collectively, the data highlight integrin-mediated rigidity-dependent RTK activation, and suggest a role for biomechanical tuning of integrin and growth factor crosstalk, as discussed below using EGFR signaling as an illustrative example.
EGFRa case study EGFR is a highly studied receptor and a major regulator of cell proliferation and survival. It has become an important target for anticancer therapies because of its oncogenic role in a number of cancer types (Tomas et al., 2014). There has been intensive effort in the development of EGFR inhibitors, constituting either monoclonal antibodies that target the extracellular domain or tyrosine kinase inhibitors that target the intracellular kinase domain . However, resistance to EGFR inhibition is a significant problem (Camp et al., 2005). Mechanisms of resistance described to date include mutation of the receptor that prevents action of tyrosine kinase inhibitors, bypass mechanisms, such as compensatory upregulation of alternative proliferative signaling pathways, resistance as a result of epithelial-mesenchymal transition (EMT) (Gibbons and Byers, 2014), and aberrant EGFR internalization or a switch of internalization routes (Wheeler et al., 2008;Menard et al., 2018;Chew et al., 2020). Resistance to the anti-EGFR monoclonal antibody Cetuximab and the potent EGFR kinase inhibitor Osimertinib has been shown to be due failure of EGFR internalization (Wheeler et al., 2008;Menard et al., 2018). Despite the fact that crosstalk with integrins is known to generate robust EGFR signaling Morello et al., 2011;Ricono et al., 2009;Sieg et al., 2000;Umesh et al., 2014;Moro et al., 2002;Moro et al., 1998), little attention has been paid to date to the role that integrin signaling plays in the response to anti-EGFR therapies. The recent demonstration that pharmacological inhibition of EGFR endocytosis can sensitize cancer cells to treatment with EGFRtargeting monoclonal antibodies (Chew et al., 2020), emphasizes the importance of understanding the regulation of the EGFR life-cycle for the development of successful cancer therapies. In the following sections, EGFR is used as an illustrative example to examine the concept of a mechanosensitive integrin-mediated control of growth factor signaling and the implications for response, or lack thereof, to molecularly targeted therapies, such as the EGFR inhibitors.
The 'EGFR odyssey' EGFR signaling is tightly controlled by endocytic recycling and degradation, leading to spatiotemporally distinct signaling outcomes (Bakker et al., 2017). EGFR bound to its cognate ligand is packaged into endocytic vesicles followed by transport to lysosomes for proteolytic degradation, or recycling to the membrane surface in ligand-free form where it is then ready to recommence signaling (Bakker et al., 2017). The binding of soluble cognate ligands to inactive EGFR monomers on the cell surface stimulates receptor autoactivation, initiating intracellular signaling that predominantly targets cell growth such as the well-documented pro-proliferative Ras-Raf-MEK-ERK pathway (Weinstein-Oppenheimer et al., 2000). Strikingly, the fate of internalized EGFR can be directed by the concentration of ligand in the external environment (Sigismund et al., 2005;Caldieri et al., 2017). Under conditions of high EGF concentration, the receptor undergoes rapid endocytosis accompanied by ubiquitylation, tagging it for lysosomal degradation; at low EGF concentrations, receptor trafficking takes the form of a slower CME (Sigismund et al., 2005;Caldieri et al., 2017). Generally, minimally activated EGFR moves from early endosomes into recycling endosomes that are recycled back to the membrane surface (Menard et al., 2018). The intracellular EGFR tail remains phosphorylated and active throughout the process of endocytosis, regardless of the mode of internalization, and thus EGFR continues to signal even when internalized (Menard et al., 2018).

Forcing the odyssey
The potent crosstalk between integrin and EGFR signaling suggests that the interpretation of tissue rigidity through integrins may affect the regulation of EGFR signaling, and unpicking this may reveal tactics to increase the potency of EGFR-targeted therapies. Clustering of EGFR, Src and integrins (Plopper et al., 1995;Moro et al., 2002;Schwartz and Ginsberg, 2002;Ricono et al., 2009) in a stiff tissue environment sets the scene for enhanced EGFR signaling. Presumably, the converse may also be true, in that loss of EGFR-Src-integrin clustering on soft substrates may mute EGFR signaling (Fig. 2).
Answers to the role of tissue stiffness in the response to EGFRtargeted therapies may be found in studies on related transmembrane receptors, such as VEGFR. Vandetanib (ZD6474) is a commonly used anti-angiogenic agent that inhibits both VEGFR and EGFR (Ciardiello et al., 2003). Wu and colleagues demonstrated that in stiff environments that favor increased integrin clustering, human umbilical vein endothelial cells (HUVECs) display a rigidityregulated response to Vandetanib, with increased sensitivity on soft substrates changing to a resistance phenotype on rigid surfaces (Wu et al., 2015). Increased matrix stiffness have been shown to encourage increased internalization of VEGFR into early endosomal vesicles (LaValley et al., 2017) and this may account for the rigidity response. Potentiated VEGFR internalization can be halted by treatment with Y27632, an inhibitor of Rho-kinase (ROCK)-mediated contractility, which potentially may rescue  Fig. 2. Model for how mechanosensing might tune EGFR signaling via integrin crosstalk. (A) On a rigid substrate, large focal adhesions form with bidirectional force sensing through the integrins (black arrows). This stimulates high-level activation of the integrin-associated kinases FAK and Src, which in turn crosstalk with EGFR, inducing ligand-independent activation of EGFR signaling. When ECM-adherent cells are stimulated with EGF ligand, maximal signaling downstream of the receptor is induced, resulting in phosphorylation (P) and activation of MAPKs, which are important EGFR targets. Under conditions of maximal EGFR signaling, EGFR is targeted to the lysosome via clathrin-independent endocytosis. (B) Conversely, on a soft matrix, limited forces are exerted on the integrins and they undergo clathrin-mediated endocytosis (CME). At the same time, reduced integrin engagement is proposed to result in reduced signaling through the EGFR; this alters the pathway of EGFR internalization, favoring its CME and trafficking back to the membrane via the recycling endosome.
Vandetanib action in stiff tissues (LaValley et al., 2017). β1 integrin inhibition disrupts EGF binding, leading to decreased EGFR activity and correspondingly, sensitivity to EGFR inhibitors such as Gefitinib can be restored (Morello et al., 2011). Since β1 integrin expression is reduced (Yeh et al., 2017) and integrin adhesion lowered, with cells rounding and losing actin stress fibers on soft substrates (Engler et al., 2004), this suggests a mechanism for how a soft tissue environment may reduce sensitivity to EGFR inhibition whereby loss or suppression of the integrin activation changes the activity of the EGFR and hence response to EGFR inhibition. These examples reveal how integrin mechanosensing could alter the cellular response to growth factor inhibitors (Fig. 2). We propose that loss of integrin crosstalk on soft substrates may dampen EGFR signaling, thereby driving the 'less active' receptor into a rapid recycling pathway (Fig. 2).
In addition to forces from the surrounding tissue, tension of the plasma membrane also has an important part to play in the regulations of endocytosis, and serves to maintain membrane homeostasis (Thottacherry et al., 2018). Under conditions of low membrane tension, endocytosis is induced, thereby removing membrane from the cell surface and generating a corresponding increase in membrane tension in order to maintain membrane tension homeostasis (Diz-Muñoz et al., 2013;Barbieri et al., 2016;Sitarska and Diz-Muñoz, 2020). Since adhesion and cytoskeletal forces regulate both tension of the cortical actin skeleton (required to sense external force through the integrins) and plasma membrane tension, and thus there likely are feedback loops between the two systems (Thottacherry et al., 2018), it has been difficult to delineate the contribution of plasma membrane tension versus adhesion-mediated cortical tension (Le Roux et al., 2019). With the development of new tools for measuring the tension changes in these two systems independently (Sitarska and Diz-Muñoz, 2020), the contribution of each to receptor internalization and the potential consequence for drug resistance will become clearer.
Therapy for the entire tissue rigidity spectrum Tissue elastic moduli in vivo are heterogeneous, both within and between tissues, as well as temporally. Therefore, features of the relevant in vivo biomechanical milieu over the time course of tissue development or during disease pathogenesis are important considerations for elucidating relevant cell biology. The promalignant effects of a stiff ECM, which characterizes many solid tumors (Paszek et al., 2005), has led to increased efforts by investigators to consider mechanosignaling in cancer biology studies, including exploiting this effect by various means with a view to improve therapeutic response (for example, see Jiang et al., 2016;Zhu et al., 2017). While the research focus has consequently shifted to the impact of rigid tissue microenvironments specifically, invasive cancer cells are exposed to a mechanically heterogenous environment when they move from a highly rigid primary tumor tissue environment (Levental et al., 2009) through and to softer tissues (Box 1). Similar to the example shown for breast cancer (see Box 1), highly invasive brain cancer cells migrate away from the primary tumor bulk into the soft, compliant healthy brain tissue (Gritsenko et al., 2012). Thus, we need to broaden our focus and consider the diverse tissue environments that cancer cells are exposed to throughout the entire disease trajectory in order to design or improve treatments.
Although anchorage-independent growth is a characteristic feature of cancer cells, they still respond to signals from the ECM (Levental et al., 2009), displaying mechanosensitive changes in cell migration, proliferation and differentiation. There appears to be a spectrum of mechanosensitivity, with distinct cellular mechanophenotypes (rigidity sensitive and insensitive) displayed in response to varying environmental rigidities (Ulrich et al., 2009;Tilghman et al., 2010;Grundy et al., 2016). For example, analysis of primary patientderived glioblastoma brain cancer cells revealed that while some cells displayed a monotonic increase in cell migration speed relative to increasing substrate stiffness (rigidity sensitive), other cells move with the same speed regardless of the substrate stiffness (rigidity insensitive) (Grundy et al., 2016). Tilghman and colleagues analyzed the proliferation rates relative to substrate stiffness in cells from 14 different cancers, including breast, lung, prostate, pancreatic, melanoma and prostate (Tilghman et al., 2010). They demonstrated that cells in which growth was slowed on soft substrates, displayed a correspondingly reduced ability to form tumors in the lungs of recipient mice. Cancer cells with a proliferation mechanophenotype, where changes in rates of proliferation change depending on substrate Box 1. Elastic moduli of the diverse tissue environments encountered by metastasizing breast cancer cells.
During metastatic progression, breast cancer cells leave the primary tumor, invade through surrounding tissue to reach the blood and lymph systems where they travel around the body followed by exit from the circulation and the establishment of secondary tumors (Thiery, 2002). Breast cancers frequently metastasize to the brain, lung, bone and liver (Lorusso and Rüegg, 2012) (see figure). Each of the tissues that the invading cancer cell traverses have distinct rigidities. Primary tumors are characterized by a stiff, collagen-dense matrix that can be detected by physical palpation of the tumor (Paszek et al., 2005). In contrast, the surrounding healthy breast tissue has been reported to be very soft (Paszek et al., 2005), similar to the environment in the lung and some areas of the brain (Cox and Erler, 2011). As the metastasizing cancer cells transit through the lymph nodes, they are exposed to a stiffer environment (∼10 kPa; Arda et al., 2011) than the healthy brain tissue. Finally, bone presents a very stiff environment (Butcher et al., 2009), although the bone marrow is considerably softer, and thus the mechanical influence here will depend on where the secondary tumor takes up residence. The metastasis of all cancer cells is characterized by these transitions through local and distant tissues having alternative characteristics of elasticity and/or stiffness. Consideration of the cancer cell response to therapy in these mechanically distinct environments may help to achieve complete eradication of all cancer cells within the body.
The compliance, are likely to exhibit altered response to anti-proliferative agents, depending on the rigidity of their surrounding environment (Tilghman et al., 2010;Rehfeldt et al., 2007;Feng et al., 2013;Wu et al., 2015). The mechanical features of the substrate also regulate cellular differentiation programs (Engler et al., 2006), and for some cancer cells, culture in rigid plastic dishes pushes cells to undergo EMT (Tilghman et al., 2010;Rice et al., 2017), which is associated with resistance to a range of anti-cancer drugs (Hanahan and Weinberg, 2011). Current standard drug-screening practices, which favor volume over comprehensive candidate drug property profiling, focus on high-throughput assays of cells grown on plasticware. However, assaying drug potency in such rigid, plastic environments may speed up rates of proliferation, switch cells to a mesenchymal phenotype and alter crosstalk between integrin and growth factor receptors, all of which can potentially skew results towards the identification of candidate drugs that possess a potency against cancer cells in vitro (plastic), which is not replicated in the in vivo physiological mechanical environment. In contrast to the links between a rigid matrix and drug resistance, examples are emerging revealing that a soft matrix can also increase drug resistance in some situations. For example, despite distinct differences in their mode of action, higher concentrations of both the DNA-damaging agent cisplatin and the anti-microtubule agent paclitaxel, are required to induce cell death in MCF-7 breast cancer cells when they were cultured on a soft substrate; this has been attributed to slower cell cycle progression in these cells on the soft substrate (Feng et al., 2013). Similarly, mesenchymal stem cells are more resistant to the anti-proliferative effects of mitomycin C when cultured on soft polyacrylamide hydrogels compared to when cultured on more rigid hydrogels (Rehfeldt et al., 2007). Therefore, although there is a clear relationship between matrix rigidity and drug response, the specific effect depends on the cell and tissue context.

Conclusions and perspectives
As discussed in this Review, tissue mechanics are likely to play an important role in the response to EGFR inhibitors owing to crosstalk with mechanosensing integrins, and this may in part account for some of the disappointing results with the clinical use of EGFR inhibitors Záhonero and Sanchez-Gómez, 2014). In the body, transmigrating and metastatic tumor cells are potentially exposed to a broad spectrum of extracellular forces. Given that treatment-resistant cancer cells lead to disease relapse, it is critical that the design of novel therapies considers all potential mechanisms of therapy resistance, including the role of the mechanoenvironment. In order to achieve this outcome, it will be critical to map integrin and growth factor signaling pathways in cancer cells grown under different conditions of matrix rigidity. This will not only shed light on potential new resistance mechanisms but may also reveal previously unappreciated mechanosensitive targets, facilitating the delivery of treatments that are tailored to match the in vivo conditions.