Mechanoresponsive metabolism in cancer cell migration and metastasis

Abstract

Altered tissue mechanics and metabolism are defining characteristics of cancer that impact not only proliferation but also migration. While migrating through a mechanically and spatially heterogeneous microenvironment, changes in metabolism allow cells to dynamically tune energy generation and bioenergetics in response to fluctuating energy needs. Physical cues from the extracellular matrix influence mechanosignaling pathways, cell mechanics, and cytoskeletal architecture to alter presentation and function of metabolic enzymes. In cancer, altered mechanosensing and metabolic reprogramming supports metabolic plasticity and high energy production while cells migrate and metastasize. Here, we discuss the role of mechanoresponsive metabolism in regulating cell migration and supporting metastasis as well as the potential of therapeutically targeting cancer metabolism to block motility and potentially metastasis.

Introduction

Despite remarkable progress in cancer research, therapeutically targeting cancer cell migration has proved challenging, and endeavors to selectively inhibit metastasis have yielded limited clinical success (Steeg, 2016). The invasion-metastasis cascade is a complex biological process consisting of the following major events: (1) local invasion of the basement membrane and cell migration, (2) intravasation into the vasculature and/or lymphatic system, (3) survival in circulation, (4) arrest at distant organ site and extravasation, and (5) colonization at metastatic sites (Valastyan and Weinberg, 2011). This is a challenging journey and to successfully disseminate to distant organs during metastasis cancer cells are required to adapt to a dynamic tumor niche as well as the microenvironment of foreign tissue (Lu et al., 2012). To escape the primary tumor during local invasion and migration, cancer cells transverse a structurally and mechanically heterogeneous extracellular matrix (ECM) (Paul et al., 2017; van Helvert et al., 2018). The tumor-associated ECM is a critical component of the tumor microenvironment (Lu et al., 2012; Pickup et al., 2014) and altered tissue mechanics is a well-known driver of tumor progression and metastasis (Spill et al., 2016). During tumor progression, tightly controlled tissue homeostasis is lost, and dysregulation of the tumor-associated ECM alters the bi-directional mechanoreciprocity between tumor cells and the ECM to create a positive tumorigenic feedback loop (Cox and Erler, 2011; Eble and Niland, 2019; Lu et al., 2012; Zanotelli et al., 2017).

Metabolic reprogramming and altered bioenergetics are a hallmark of cancer (Hanahan and Weinberg, 2011) and growing evidence now indicate a link between mechanosignaling and metabolism during migration (DelNero et al., 2018; Papalazarou et al., 2020a; Salvi and DeMali, 2018). Migrating cells adjust their bioenergetics in response to mechanical and structural cues in their local microenvironment (Li et al., 2019b; Liu et al., 2013; Mah et al., 2018; Park et al., 2020b; Zanotelli et al., 2018; Zanotelli et al., 2019; Zhang et al., 2019). This adaptive regulation of metabolism has been proposed to be a key mechanism that cancer cells utilize to respond to fluctuating energy needs when transitioning through microenvironments with varying mechanical cues during the invasion-metastasis cascade (Epstein et al., 2017; Epstein et al., 2014). In this review, we discuss the emerging role of mechanosensitive metabolism in regulating cancer cell migration and metastasis and discuss the potential to therapeutically target metabolism to inhibit invasion.

Bioenergetic regulation of cancer invasion and cell migration

It has been suggested that cells need to decide whether to “grow or go”, where proliferation and migration are mutually exclusive to each other due to competition for shared resources (Giese et al., 1996; Hecht et al., 2015). As such, metabolic signaling for migration and proliferation can be considered to be separate and distinct. While considerable research has explored the impact of metabolic changes during cancer progression on tumor growth, our knowledge on the role of metabolism in cancer cell motility during invasion and metastasis is still emerging.

Due to both intra- and intertumoral phenotypic and metabolic heterogeneity it has become increasing important to study the bioenergetics of individual cells. The development of genetic biosensors that measure intracellular ATP:ADP ratio (Tantama et al., 2013), NADH:NAD⁺ ratio (Hung et al., 2011), glucose concentrations (Diaz-Garcia et al., 2019), glutamate metabolism (Okumoto et al., 2005), lactate flux (San Martin et al., 2013), pyruvate transport and production (San Martin et al., 2014), and other metabolic markers will support precise single-cell, real-time measurements of energetics during migration. Single-cell measurements are particularly powerful as they enable the correlation of migration behavior and metabolic activity, both of which can be highly stochastic and heterogeneous across cells even within a clonal population. Using these new tools and techniques, recent work has identified a role of bioenergetics and metabolism in regulating invasion. To meet energy requirements for migration, cells respond to the physical microenvironment and actively adapt their metabolism to support cytoskeletal activity and force generation (Cunniff et al., 2016; Kelley et al., 2019; Schuler et al., 2017; Shiraishi et al., 2015; Zanotelli et al., 2019; Zhang et al., 2019); however, more work is needed to completely understand signaling pathways used by cells to adapt their metabolism during migration. Given the complexity of metabolic traits and their maintenance, it will be important for future studies to identify short-term alterations in metabolism for cytoskeletal remodeling and the long-term response to matrix mechanics to fully understand metabolic adaptations in response to the mechanical microenvironment.

Energy metabolism and single-cell migration.

The invasion-metastasis cascade is initiated by invasion of cancer cells through the surrounding basement membrane and ECM (Valastyan and Weinberg, 2011). Cell migration and invasion is an energy-intensive process (Bursac et al., 2005; Liu et al., 2013; Meshel et al., 2005; Mizuno et al., 2007; Zanotelli et al., 2018), where cellular machinery must dynamically respond to mechanical changes in the ECM (Stevenson et al., 2012). Recent efforts have sought to quantify cellular energetics, forces, and cytoskeletal activity during motility to better understand the role of metabolism in migration (Table 1). Cells sense matrix cues through forces on integrins, which trigger changes in cytoskeleton organization and contractility during migration (Hood and Cheresh, 2002). When migrating through a matrix with high mechanical plasticity, cells extend invadopodia and apply a combination of protrusive and contractile forces to physically open spaces in the matrix and move through them (Wisdom et al., 2018). Actomyosin-based activity and actin polymerization during migration create mechanical forces and expend a significant portion of cellular energy (Bursac et al., 2005; Mizuno et al., 2007). Estimates suggest that up to approximately 50% of ATP is used to support the actin cytoskeleton (Bernstein, 2003; Daniel et al., 1986). Consequently, intracellular ATP:ADP ratio has been positively correlated with migration potential (Zanotelli et al., 2018).

Table 1.

Cellular forces related to bioenergetics during cancer cell migration.

Migration event	Measurement	Method	Value	Reference
Integrin activation	Force, unfolding of talin	Magnetic tweezers	1 pN	(Yao et al., 2016)
	Force, molecular extension	Biomembrane force probe	10 pN	(Chen et al., 2017)
Cytoskeletal remodeling	Force	Microfabricated pillar substrates	1 nN	(Gupta et al., 2015)
AMPK activation	Force	E-cadherin coated magnetic beads	10 pN	(Bays et al., 2017)
Generating strain fields during cooperative migration	Matrix strain	Cell-generated bead displacement in collagen matrix	100 pN	(Liu et al., 2013)
Force generation during collective migration	Stress in cell sheet	Fourier-transform traction force microscopy	1 kPa	(Trepat et al., 2009)
Migration	Membrane tension	Tether-pulling assay	150–450 pN/um	(Lieber et al., 2013)
	Energy consumption	Computational	4×105–1×108ATP units/s, 23 fJ	(Abraham et al., 1999; Fan et al., 2012)

To support energy requirements for cytoskeletal activity during migration, cells traffic metabolic proteins and concentrate energy production to sites with high energy needs (Cunniff et al., 2016; Kelley et al., 2019; Schuler et al., 2017; Shiraishi et al., 2015) (Fig. 1A). Due to these energy costs, cells preferentially take the path of energy minimization when traveling through a complex, physically heterogeneous microenvironment (Zanotelli et al., 2019) (Fig. 1B). Migratory cells have been suggested to favor mitochondrial respiration for increased ATP production (LeBleu et al., 2014). Cellular energy produced may be used to transport matrix metalloproteinases (MMPs) toward the leading edge and degrade the ECM in front of the invading cell and generate forces for movement (Wolf et al., 2007). To support these activities, mitochondria are trafficked for “on-site” energy production. Miro1-mediated mitochondrial trafficking to the leading edge of migrating cells is regulated by AMP-activated protein kinase (AMPK) activity and facilitates lamellipodial protrusions (Cunniff et al., 2016; Schuler et al., 2017). AMPK is a master energy metabolism regulator that allows cells to cope with stress (Hardie et al., 2012) and acts as a critical link between metabolism and mechanotransduction (Bays et al., 2017). When MMP expression is inhibited, cells adapt by localizing mitochondria to the invasive front during invasion and increasing Arp2/3-F-actin protrusive force to physically displace the matrix (Kelley et al., 2019). Other findings indicate glycolysis is associated with cytoskeletal activity and faster migration (Shiraishi et al., 2015). Glycolysis can help cells respond to fluctuating energetic demands in the membrane (Epstein et al., 2014), hence an increase in glucose uptake is often observed when the energy cost associated with migration increases rapidly (Zanotelli et al., 2019; Zhang et al., 2019). While glycolysis can synthesize ATP up to 100 times quicker than OXPHOS, the energetic yield is very low and presents a thermodynamic trade-off between rate and yield (Martinez-Outschoorn et al., 2017). It is possible that cells rapidly respond to the global energy requirement via glycolysis and precisely tune the subcellular energy requirement via mitochondria rearrangement.

Figure 1. Cellular energy metabolism in migration.

(A) Mitochondria transport toward the leading edge and localized ATP production supports the energy demand of the migration machinery, including Arp2/3 mediated actin polymerization, actomyosin force generation, and synthesis, transport, and activation of proteases. (B) Single cells migrating in collagen microtracks preferentially choose the path of least resistance to minimize energy costs. At the time of decision-making, only high-energy cells or cells that can rapidly increase their energy production can choose the more energy-costing, more resistant path. (C) During collective migration, the leader cell consumes more energy to overcome the resistance in front of it than follower cells. The original leader cell falls back once it depletes its available energy, and a new cell emerges to take over the leader role.

Energy metabolism and collective migration.

Depending on cell type and the physical microenvironment, cells can either migrate individually or collectively as multicellular groups (Friedl and Alexander, 2011). During collective migration, the invasive front is dynamic, where leader cells have higher metabolic costs for invasion compared to follower cells due to increased energy needs for ECM remodeling via high contractility and large amounts of strain energy transmitted to the ECM (Liu et al., 2013; Zhang et al., 2019). As such, contractility and metabolic signaling are related in leader cells during collective movement (Yamaguchi et al., 2015). Leader cells upregulate and localize active Rac, integrin β1, and PI3K to the leading edge of the cell, creating a feed-forward loop between Rac and PI3K and driving invasion (Yamaguchi et al., 2015). Leader cell position changes once a leader cell depletes its available energy stores and, as such, leader lifetime is directly correlated with energy consumption during invasion (Zhang et al., 2019) (Fig. 1C). When in the leader position, cells increase glucose uptake compared to follower cells, which elevates energy production likely through transient upregulation of glycolysis. Alternatively, leader cells may undergo genetic modification to permanently upregulate oxidative phosphorylation (OXPHOS), thus generating sufficient energy for migration even with lower glucose uptake than follower cells (Commander et al., 2020). However, it is not yet clear how the balance between glycolysis and OXPHOS is regulated in leader cell behavior or how the mechanical microenvironment mediates this effect.

Leader cells during collective migration resemble individually migrating cells in many aspects, as both of them need to clear the path in front of them to migrate, interact and respond directly to varying matrix densities (Zanotelli et al., 2018; Zhang et al., 2019), whereas follower cells are relatively insensitive to cues presented in the matrix (Zhang et al., 2019) likely because they are mostly surrounded by other cells instead of the matrix. While genetic variations across different tumor types and during epithelial-to-mesenchymal transition (EMT), such as the expression of E-cadherin, is a major determinant of whether cells migrate individually or collectively (Friedl and Alexander, 2011), we propose that thermodynamic energy costs of migration from the physical microenvironment as another important factor. Cancer cells are “lazy”, so they prefer the most energy cost-effective and energy minimalizing mode of migration. Consistent with this idea, with increasing matrix density and thus increasing energy cost for migration, cancer cells switch from an individual to a collective invasion mode (Haeger et al., 2014), and progressively become more cooperative by using a “drafting” mechanism to cope the increased energy cost (Zhang et al., 2019).

Altered metabolism and cancer cell migration and metastasis

Metabolism in many tumors is altered to favor glycolysis over OXPHOS, a phenomenon termed “the Warburg effect”, resulting in increased glucose uptake (Gatenby and Gillies, 2004) and upregulation of glucose transporters in almost all cancer types (Adekola et al., 2012). While energetically unfavorable, a glycolytic switch can promote survival of tumor populations over non-tumorigenic cells and promote invasive tumor growth (Gatenby and Gillies, 2004). Prostate and breast cancer cells utilize glycolysis for cytoskeletal activity and migration in vitro, where mitochondrial-derived ATP alone is insufficient to support motility (Shiraishi et al., 2015). Plasminogen activator inhibitor 1 (PAI1), which is elevated in migratory cancer cells, also promotes glycolysis in triple negative breast cancer cells in cell assays, orthotopic tumor xenografts, and lung metastases (Humphries et al., 2019). Interestingly, recent work indicates that tumors are metabolically heterogeneous and aerobic glycolysis does not predict loss of OXPHOS (Deberardinis and Chandel, 2020). The triple negative breast cancer cell line MDA-MB-468 demonstrated significantly higher levels of OXPHOS compared to MDA-MB-231 and less-metastatic breast MCF-7 cancer cells while having similar levels of glycolytic flux (Pacheco-Velazquez et al., 2018). However, under hypoxic conditions similar to tumor tissue, glycolytic pathways are promoted through hypoxia-inducible factor (HIF) signaling and all three cell lines mainly depended on glycolysis (Pacheco-Velazquez et al., 2018). The low-pH environment of solid tumors caused by elevated glycolysis can also provide a favorable environment invasion (Martinez-Zaguilan et al., 1996) and metastasis (Schlappack et al., 1991). Acid-mediated invasion has been observed in vivo, where highest tumor invasion occurred in areas with the lowest pH and no invasion was observed into peritumoral regions with approximately normal pH (Estrella et al., 2013). An acidic extracellular pH can promote migration through the activation of proteases including MMPs (Kato et al., 2007) and cysteine cathepsins (Mohamed and Sloane, 2006). Indeed, metastatic cancers often display a correlation between GLUT1 expression and MMP expression (Ito et al., 2002) and GLUT1 localizes to the tumor front at areas of localized invasion (Estrella et al., 2013). Therefore, cancers cells actively adapt metabolic machinery in response to altered environmental cues within the tumor niche to drive invasion.

While glucose is the main energy source to fuel tumor activity, the ability of cancer cells to utilize alternative energy sources has been linked with a more aggressive and migratory phenotype (Bertero et al., 2019; Pascual et al., 2017; Port et al., 2012). Increases in glutaminolysis, amnio acid and lipid metabolism, pentose phosphate pathway flux, macromolecule biosynthesis, and maintenance of the redox state also demonstrate changes in bioenergetics in cancer (Ben-Sahra and Manning, 2017). Glutamine addiction is a particularly important change in cancer metabolism that has been directly linked to mechanosignaling and motility (Wang et al., 2010). The aerobic glycolysis of cancer cells will increase glucose uptake but also divert glucose-derived pyruvate to production of lactate and away from the mitochondria (Kim and DeBerardinis, 2019), resulting in significantly elevated glutamine metabolism (Wilson et al., 2013). Glutaminase is a mitochondrial enzyme that hydrolyzes glutamine into glutamate and has traditionally been known to fuel rapid cancer cell proliferation, but emerging data reveal that it also impacts invasion and metastasis. Glutamine metabolism promotes invasion through activation of STAT3 (Yang et al., 2014) and is correlated with increased risk of metastasis and mortality in colorectal cancer (Xiang et al., 2019), drive metastasis to liver, lung, and kidney in glioblastoma (Shelton et al., 2010). Elevated activity of mitochondrial glutaminase that promotes increased metastasis in breast cancer has been linked to Rho GTPase activity (Wang et al., 2010), which plays a prominent signaling role in actin cytoskeletal rearrangements, cell polarity, migration, as well as tumorigenesis (Wilson et al., 2013). Given its importance in responding to mechanical forces, other metabolic pathways are likely also linked to Rho GTPase activity.

Nutrient flexibility and the bioavailability of other extracellular fuel sources including sugars and non-essential amino acids can also influence the invasive behavior of cancer cells. Fructose and sucrose are common dietary sugars, and many cancer cells express the fructose transporter GLUT5 (Douard and Ferraris, 2008; Port et al., 2012). Dietary fructose uptake via GLUT5 was found to influence glycan assembly and cytoskeletal organization in breast cancer cells (Monzavi-Karbassi et al., 2010) and promote sialylation via upregulating the sialyltransferase ST6Gal1 in pancreatic cancer cells (Hsieh et al., 2017), which induced an aggressive cancer phenotype. However, some sugars impact tumor invasion differently depending on cancer type. For example, uptake of monosaccharide mannose impaired tumor growth and increased sensitivity to chemotherapy in vitro and in vivo in a number of cancer types including pancreatic, colorectal, lung cancer, and osteosarcoma (Gonzalez et al., 2018). This is caused by an inhibition of glucose metabolism pathways via the accumulation of intracellular mannose-6-phosphate, and mannose sensitivity of the tumor type is dependent on phosphomannose isomerase (PMI) levels which catalyze the conversation between mannose-6-phosphase and fructose-6-phosphate (Gonzalez et al., 2018). However, when incorporated into N-glycans on the cell surface (Sharma et al., 2014), extended high-mannose glycans in cholangiocarcinoma were found to support metastasis by increasing the ability of cells to invade the surrounding matrix (Park et al., 2020a). Many non-essential amino acids, such as glutamine (Altman et al., 2016), asparagine (Knott et al., 2018), proline (Elia et al., 2017), serine (Maddocks et al., 2013), and glycine (Jain et al., 2012) can also influence metastatic progression. Notably, RNA interference screening and in vivo mouse models identified that asparagine availability, either through biosynthesis or dietary supply, strongly increase metastatic potential in breast cancer without affecting primary tumor growth in vivo (Knott et al., 2018). When supplied with elevated asparagine, breast cancer cells underwent EMT to drive invasion, whereas silence of asparagine synthetase resulted in poor invasion in vitro and smaller metastases in vivo. Asparagine induces EMT through the mechanosensitive and prototypical EMT makers, Twist1 and E-cadherin (Knott et al., 2018), but mechanosignaling of many extracellular fuel sources has not yet been examined.

Altered metabolism has been proposed to help cancer cells adapt to changing microenvironments throughout the invasion-metastasis cascade and support metabolic plasticity and flexibility (Bergers and Fendt, 2021; Epstein et al., 2017; Epstein et al., 2014). Similar to plasticity in cell adhesion, cytoskeletal dynamics, and mechanotransduction needed to adapt cell forces during migration (Friedl and Alexander, 2011), metabolic plasticity may allow cells to fuel various metabolic requirements during challenging migration events. Recent work revealed that altered cancer mechanosignaling acts to sustain high metabolic activity during metastasis when traveling across microenvironments with different stiffness. In pancreatic tumors, which have a markedly stiff stroma, YAP-driven mechanoresponsive changes support ATP production through increased mitochondrial activity and OXPHOS as well as regulating ATP recycling through the phosphocreatine-creatine kinase system (Papalazarou et al., 2020a). This increased stromal stiffness acts to drive the tricarboxylic acid (TCA) cycle and upregulate cytoplasmic creatine kinase B-type (CKB), which regenerates ATP through phosphorylation of creatine and fuels pancreatic ductal adenocarcinoma (PDAC) aggressiveness and liver metastasis (Papalazarou et al., 2020a). The phosphocreatine-creatine kinase system, or more broadly the phosphagens, may provide ATP buffering (Ellington, 2001) that rapidly responds in the event of elevated energy demand encountered during migration such as when navigating through a stiff stroma. Other work has also linked changes in metabolism to cellular force sensing and cytoskeletal architecture. PI3K-dependent Rac activation that disrupts the actin cytoskeletal to mobilize aldolase from F-actin (Hu et al., 2016) and downregulation of PFK-targeting protein TRIM21 by force-mediated stress fiber formation (Park et al., 2020b) both act to maintain high levels of energy production in response to changing mechanical cues. These findings suggest that mechanosensing may be a major driver of metabolic plasticity and maintaining high energy production during metastasis.

Metabolism mediated by extracellular cues during cell migration

Matrix cues are a significant driver of tumorigenesis and the tumor ECM has been implicated in all of the major hallmarks of cancer (Pickup et al., 2014). The tumor-derived ECM has aberrant architectural and mechanical features compared to normal ECM (Butcher et al., 2009; Pickup et al., 2014; Zanotelli et al., 2020). In solid tumors, including breast, pancreatic and colorectal cancers, tumor tissue can be tenfold stiffer than normal tissue (Kalli and Stylianopoulos, 2018; Kawano et al., 2015; Levental et al., 2009) and contain areas of dense collagen and aligned collagen fibers that allow tumor cells to migrate radially away from tumors (Provenzano et al., 2006). While many studies have examined the influence of matrix mechanics on invasion, the impact of mechanical cues on metabolism is less well understood. A growing body of work now suggests that mechanical cues can drive changes in metabolism to regulate migration and metastasis (Table 2).

Table 2.

Cell responses to altered matrix mechanics in cancer that drive changes in metabolism and migration.

Event	Force-induced event	Effect on metabolism	Effect on migration	Reference
Integrin activation	Rho/ROCK-mediated contractility with matrix density	Increased glucose uptake and glycolysis	Increased contractility; increased cellular energy state and decreased migration with matrix density	(Mah et al., 2018; Zanotelli et al., 2018)
TRIM21-mediated PFK degradation	Cell tension from increased matrix stiffness, thick actin bundles bypass stiffness response	Elevated glycolysis; actin bundles maintain high glycolysis on soft substrates	Increased contractility	(Park et al., 2020b)
NADPH oxidase assembly	Rac1b membrane localization from increased matrix stiffness	Catalyze production of reactive oxygen species	Increased EMT phenotype	(Lee et al., 2012)
AMPK activation	Force on E-cadherin stimulates LKB1 to activate AMPK	Stimulated glucose uptake and ATP production	Increased contractility, reinforced adhesion complex and cytoskeleton	(Bays et al., 2017)
Glutamine metabolism activation	Cell tension from increased matrix stiffness	Increased glutamine metabolism	Controls RhoA activity to sustain metastasis, supports CAF matrix remodeling	(Bertero et al., 2019; Haikala et al., 2018)
Creatine-phosphagen system production	YAP mechanotransduction from increased matrix stiffness	Increased ATP production from OXPHOS and controls ATP recycling	Increased invasion and metastasis	(Papalazarou et al., 2020a)

Matrix mechanics, metabolism, and cytoskeletal activity.

Recent work indicate mechanosignaling and metabolism are linked (Romani et al., 2020). Cell-matrix adhesion, cytoskeletal remodeling and turnover, and actomyosin contractility are mechanically regulated processes that are critical to migration and have been recently linked to metabolism (Bays et al., 2017; Mah et al., 2018; Park et al., 2020b). Cells sense mechanical signals from the environment through cell-cell adhesion, cell-matrix adhesions, and cytoskeletal remodeling that modulate cellular metabolic flux and energy production (Bays et al., 2017; Park et al., 2020b), and in turn, energy availability regulates adhesion and cytoskeletal dynamics and matrix remodeling (Bays et al., 2017; Papalazarou et al., 2020b). Tumorigenesis is associated with altered cytoskeletal organization, cytoskeletal proteins, and focal adhesion turnover (Zanotelli et al., 2017), all of which are important drivers of cell migration (Ridley et al., 2003; van Helvert et al., 2018).

Contractility signaling and cytoskeletal activity can also impact metabolism through mechanotransduction pathways (Bertero et al., 2019; Haikala et al., 2018; Hu et al., 2016). Metastatic cancer cells are often characterized by high contractility (Kraning-Rush et al., 2012) and prominent stress fiber formation (Stevenson et al., 2012). Regulation of glycolysis has been shown to coordinate with these actin dynamics, as actin rearrangements are energy intensive (Hu et al., 2016; Park et al., 2020b; Shiraishi et al., 2015). PI3K is known to activate AKT signaling, which increases glucose uptake and glycolysis (Elstrom et al., 2004; Fruman et al., 2017), and maybe responsible for cancer cell invasiveness (Li et al., 2017). In addition, PI3K can act in an AKT-independent manner by altering actin organization through Rac activation, which releases aldolase from F-actin to increase the catalytic activity of aldolase and promote glycolysis (Hu et al., 2016). While the precise signaling mechanism has not yet been determined, work has indicated that PI3K is an intermediate between integrin β1 and Rac (Yamaguchi et al., 2015). This suggests cytoskeletal remodeling and glycolysis could be regulated by mechanical cues through integrins. The sensitivity of cell response to mechanical forces through integrins (Elosegui-Artola et al., 2016; Roca-Cusachs et al., 2017) can allow for spatial and temporal regulation of metabolism during migration through mechanically complex microenvironments. Integrin-mediated signaling is often considered bi-directional as cues can be transmitted from inside-out and outside-in, creating a mechanism for feedback between the cell and the extracellular environment. Additionally, reciprocal feedback between mechanosignaling and non-essential amino acid flux including aspartate and glutamate within the tumor niche can further impact contractility pathways and the mechanisms by which a cell senses its mechanical microenvironment (Bertero et al., 2019). In response to increased matrix stiffness, actin remodeling and TRIM21-modulated degradation of the enzyme phosphofructokinase (PFK) has been shown to regulate glycolysis (Park et al., 2020b). PFK has a rate-limiting role in glycolysis and is normally reduced on soft substrates; however, in cancer cells, thick actin bundles can resist TRIM21-modulated PFK degradation normally observed with changes in matrix stiffness and maintain elevated glycolysis regardless of substrate stiffness (Park et al., 2020b) (Fig. 2A). Interestingly, ATP:ADP ratio and glucose uptake increase with collagen density for MDA-MB-231 cells in 3D (Zanotelli et al., 2018), suggesting that either the TRIM21/PFK pathway is cell type or environment dependent or OXPHOS increases with collagen density while glycolysis is independent of matrix density/stiffness.

Figure 2. Adaptation of cancer cell metabolism to environmental cues.

(A) Mechanotransduction through cell-matrix and cell-cell adhesions modulates energy metabolism. In soft ECMs, TRIM21s are not trapped by actin stress fiber bundles for healthy cells but not for tumor cells, which degrade PFK and decrease glycolysis. In stiff ECMs, glycolysis is promoted through adhesion force-mediated LKB1/AMPK activation, and PI3K activation, whereas OXPHOS may be activated through YAP-regulated CKB expression. (B) The metabolic symbiosis and bidirectional signaling between cancer cells and neighboring stromal cells. Mitochondria can be transferred through membrane tubes, whereas metabolic intermediates can be transferred through monocarboxylate transporters (MCTs), or extracellular vesicles. TGF-β, transforming growth factor beta; IL-6, interleukin 6; CXCL10, C-X-C motif chemokine ligand 10; CCL5, chemokine ligand 5.

Changes in tissue stiffness during tumor progression is attributed, in part, to increased crosslinking and density of the tumor-associated ECM (Lu et al., 2012), physical cues that also influence metabolism during cancer cell migration and tumor progression. Highly metastatic MDA-MB-231 breast cancer cells seeded atop collagen gels increased glycolysis via adhesion-mediated contractility with increased collagen density (Mah et al., 2018). In the same study, less invasive MCF7 and TF7D breast cancer cells exhibited no change and MCF10A non-tumorigenic breast cells shifted toward glycolysis with lower collagen density (Mah et al., 2018), suggesting tumorigenicity impacts the metabolic response to mechanical cues. During 3D migration, MDA-MB-231 intracellular ATP:ADP ratio and glucose uptake increased while velocity decreased with increasing collagen density (Zanotelli et al., 2018). Increased matrix density had similar effects on collective migration. In vitro tumor spheroids and ex vivo organoid invasion showed increased leader cell turnover to compensate for the increased energy requirements for invasion with increasing matrix density (Zhang et al., 2019). While the mechanism mediating the correlation between matrix density and ATP is not well-defined, it is likely mediated, in part, through integrin engagement. However, many cancer cells have been shown to exhibit ameboid migration with minimal integrin attachments, suggesting there are other molecules mediating the relationship between ECM density and cellular metabolism.

Cancer cells also alter metabolism and increase energy production in response to mechanical forces. Force applied to E-cadherin stimulates liver kinase B1 (LKB1) to activate AMPK and drive actomyosin contractility, glucose uptake, and ATP production to reinforce the actin cytoskeleton and resist applied mechanical force (Bays et al., 2017). The recent identification of a mechanosensitive role for E-cadherin in metabolism is particularly important because most breast cancers are invasive ductal carcinomas with E-cadherin expression (Li et al., 2003) and E-cadherin promotes metastasis in part by limiting reactive oxygen-mediated apoptosis to support survival during detachment, dissemination, and seeding (Padmanaban et al., 2019). This reveals that AMPK activation via E-cadherin may play an important role in maintaining high metabolism and surviving physiological forces while in circulation during breast cancer metastasis (Fig. 2A). However, many carcinomas show loss of E-cadherin intercellular adhesions (Birchmeier and Behrens, 1994), indicating metabolic adaptations to the physical microenvironment may be tumor specific. Further studies will need to examine metabolism and resistance to forces during metastasis across different cancers and stages of metastatic cascade.

Cell mechanics and metabolism.

The mechanically heterogeneous composition of tumors can impose additional energetic costs on invasion. Generating strain fields at the leading edge of cooperatively invading cells is energetically demanding (Liu et al., 2013; Zhang et al., 2019) and migration through tighter interstitial spaces in the matrix requires more cellular work compared to migration through less restricting spaces (Zanotelli et al., 2019). Cells tune their bioenergetics to the mechanical properties of the ECM (Li et al., 2019b; Zanotelli et al., 2018), and high energetic costs for motility through confined spaces can direct migration path during decision-making (Zanotelli et al., 2019). Compressive forces on actin networks within the cell body can further stimulate actin remodeling and drive the formation of denser and stiffer actin structures (Bieling et al., 2016). Actin network stiffening can be useful to cells migrating and pushing through tight interstitial spaces in the matrix (Wisdom et al., 2018), but it also increases energy consumption (Bieling et al., 2016). Increasing cell or matrix stiffness will additionally increase energy costs for cell-driven matrix displacement during migration through confining microtracks and add further energetic costs for motility (Zanotelli et al., 2019). Notably, cancer cells are frequently softer than healthy cells (Cross et al., 2007). This may provide a phenotypic advantage as softer cells needs less energy to deform their shape which allows them to migrate by squeezing through matrix pores in the amoeboid migration mode (Friedl and Wolf, 2010). In addition, deformable cells need less matrix remodeling thus less energy to migrate, and they could then utilize more possible migration paths when navigating the tumor stroma. However, the energetic costs associated cell deformation and matrix remodeling during cell migration have not yet been directly compared. Cancer cells have the plasticity in switching between different migration modes (Friedl and Wolf, 2010), and the associated energetic costs may be a key determinant.

Metabolism, cancer cell-stromal cell interactions, and migration.

Cancers cells can hijack stromal cells to remodel ECM fiber architecture to promote migration and alter cell metabolism to drive metastasis. Highly contractile cancer-associated fibroblasts (CAFs) can create mechanical tension (Labernadie et al., 2017) and remodel the surrounding matrix to align fibers (Attieh et al., 2017; Erdogan et al., 2017; Glentis et al., 2017). With regard to metabolism, pro-migratory collagen fiber alignment increases MDA-MB-231 cell velocity, while lowering energy utilization promoting more efficient migration (Zanotelli et al., 2018). Inhibitors of glycolysis and glucose uptake inhibit migration speed along aligned fibers (Padhi et al., 2020), suggesting these pathways are utilized during directed migration in an aligned matrix. In addition to tensile forces that may directly facilitate cancer invasion (Labernadie et al., 2017), mechanical interactions between CAFs and cancer cells can alter metabolism and increase metastasis (Kalluri and Zeisberg, 2006). Mechanical coupling between CAFs and cancer cells may activate AMPK signaling through LKB1, which increases glucose uptake and motility (Bays et al., 2017).

Cancer cells can communicate with stromal cells for the exchange of cellular components via metabolic crosstalk (Fig. 2B). Notably, the transfer of mitochondria from stromal cells, including CAFs, mesenchymal stromal cells, and endothelial cells, to cancer cells, including leukemia (Burt et al., 2019; Moschoi et al., 2016), prostate cancer (Ippolito et al., 2019), melanoma (Tan et al., 2015), breast cancer (Caicedo et al., 2015; Pasquier et al., 2013; Tan et al., 2015), ovarian cancer (Pasquier et al., 2013), may rescue functional defects and promote survival, growth, respiration, migration, and metastasis in vitro and in vivo. Although a mechanistic understanding of the role of those transferred mitochondria is not fully developed, the resulting partial or temporary reversal of the Warburg effect may play an important role in meeting energy demands necessary for metastasis (Cunniff et al., 2016). A similar metabolic crosstalk may also exist between cancer-associated macrophages and cancer cells, as cytoplasmic components transferred from the macrophages were recently reported to promote the invasion of melanoma and breast cancer cells in vitro and in vivo (Hanna et al., 2019; Roh-Johnson et al., 2017). Stromal cells, including CAFs, can mobilize energy sources stored in the cancer cells such as glycogen (Curtis et al., 2019) to fuel glycolysis and provide cancer cells with metabolic intermediates such as lactate and pyruvate (Pavlides et al., 2009), glutamine (Yang et al., 2016), non-essential amino acids (Sousa et al., 2016), and fatty acids (Romero et al., 2015) to fuel energy production and biosynthesis through the mitochondrial TCA cycle. This kind of interactive metabolic symbiosis often requires bidirectional signaling between cancer and stromal cells, including the induction of glycolysis in neighboring stromal fibroblasts (Pavlides et al., 2009), stimulation of stromal cell autophagy and metabolites secretion (Sousa et al., 2016), activation of p38α signaling in CAFs (Curtis et al., 2019), and chemokine and cytokine secretion (Curtis et al., 2019). Notably, metastatic potential of cancer cells may influence their ability in reprogramming stromal cell metabolism. In gastric cancer, highly metastatic 44As3 cells but not the weakly metastatic HSC-44PE cells induced strong glycolytic switching in the surrounding activated fibroblasts (Kogure et al., 2020).

While metabolic crosstalk has been demonstrated within tumors (Commisso et al., 2013; Davidson et al., 2017; Perera et al., 2015; Sousa et al., 2016; Yang et al., 2016), only recently has mechanotransduction been shown to play a role in this process (Bertero et al., 2019). The exchange of non-essential amino acids, such as glutamate and aspartate between CAFs and cancer cells within the tumor niche, sustains tumor cell and stromal fibroblast activation to support tumor growth and metastasis (Bertero et al., 2019). Tumor niche stiffening promotes a metabolic switch, where YAP/TAZ-dependent glutamate/aspartate crosstalk between CAFs and cancer cells promotes growth and aggressiveness, defining a mechanotransduction cascade in dysregulated tumor metabolism (Bertero et al., 2019). During this metabolic crosstalk, CAF-derived aspartate supports cancer cell proliferation and cancer cell-derived glutamate supports ECM remodeling by balancing the redox state of CAFs. Targeting altered stromal cell metabolism and mechanoresponsive metabolic crosstalk may be an alternative strategy to targeting cancer cells directly to normalize and balance the chemical landscape of the tumor microenvironment.

Hypoxia, metabolism, and migration.

Hypoxia, oncogene activation, and loss of tumor suppressors upregulate HIFs in tumor cells which in turn upregulate genes that contribute to elevated levels of glycolysis and/or reduced OXPHOS (Denko, 2008). In response to hypoxia, MDA-MB-468, MDA-MB-231, and MCF-7 cells all increase glycolysis and glycolytic flux compared to normoxia conditions via HIF1-α and activation of its target proteins including GLUT1, HK1 and HKII, and LDHA (Pacheco-Velazquez et al., 2018). Hypoxia and HIFs regulate the dynamics of the actin cytoskeleton and stress fiber assembly, mainly through Rho GTPase signaling including RhoA, Rac1, and Cdc42; however, this effect may be cell type and context dependent (Zieseniss, 2014). RhoA expression and activation were found to be elevated in many cancers including renal cell carcinoma (Turcotte et al., 2003) and breast cancer (Gilkes et al., 2014) in hypoxia, but can be either increased (Vertelov et al., 2013) or decreased (Raheja et al., 2011) in mesenchymal stromal cells. Rho GTPase-mediated contractility signaling is known to be a significant driver of migration (Ridley et al., 2003), suggesting that HIF signaling supports migration through increased cytoskeleton activity. The influence of hypoxia on contractility pathways and HIF1-α also suggests a link between increased migration and elevated glycolysis. Additionally, hypoxia may increase reactive oxygen species (ROS) level in cancer cells (Guzy et al., 2005), which can promote cancer invasion by inducing EMT and MMP expression through the mitogen activated protein kinase (MAPK) and NF-κB signaling pathways and by directly affecting cytoskeletal remodeling and contractility (Tochhawng et al., 2013). As such, targeting metabolic response to hypoxia in solid tumors may potentially impact migration and metastasis in addition to tumor growth.

Metabolic phenotype and metastasis

While cancer cells in the tumor microenvironment exhibit distinct metabolic reprogramming compared to healthy cells, changes in metabolism and metabolic preferences are heterogeneous and flexible. Within tumors, specific bioenergetic profiles have been identified for different cell behaviors. Differences in metabolism between cancer cells at the primary tumor and secondary sites as well as between different metastatic sites indicate that metabolic phenotypes are determinants of metastasis. Expression of transcriptional co-activator peroxisome proliferator-activated receptor gamma, co-activator 1 alpha (PGC1-α), which enhances OXPHOS, mitochondrial biogenesis and ATP production, has been clinically linked to the formation of distant metastases and invasion in breast cancer cells. Silencing of PGC1-α effectively suspended invasive potential of cancer cell without influencing proliferation, primary tumor growth, or EMT (LeBleu et al., 2014). To promote breast cancer metastasis, PGC1-α enhances global energy capacity and confers resistance to bioenergetic disruptors, which can facilitate metabolic adaptation (Andrzejewski et al., 2017). The identification of distinct metabolic signaling between cancer cells with a proliferative and invasive phenotype that is essential for functional metastasis is exciting, however, PGC1-α regulates numerous metabolic programs and mitochondrial functions, making it challenging to identify the impact of specific metabolic pathways on metastasis. Single-cell RNA sequencing of patient-derived-xenografts of breast cancer similarly showed micrometastases harbor a distinct transcriptomic program compared to primary tumors, where primary tumors exhibited higher levels of glycolytic enzymes while micrometastases upregulate OXPHOS pathways (Davis et al., 2020). In pancreatic cancer, tumors overexpress metabolic proteins for glucose uptake (GLUT1), glycolysis (HK2, ENO3, PKM2), and lactate production (LDHA) compared to normal tissue and further differences in metabolic profile exist between primary and secondary sites. Except for lung metastatic legions, all other metastatic sites had significantly increased GLUT1 and PKM2 expression compared to normal tissue (Chaika et al., 2012). Interestingly, dormant pancreatic cancer cells that survived oncogene ablation and responsible for tumor relapse exhibited strong reliance on OXPHOS and impaired glycolysis (Viale et al., 2014), suggesting that while increased glycolysis may promote tumor growth it is likely more important for tumor migration and invasion instead of tumor growth and colonization.

The tumor microenvironment acts as a strong determinant of metabolic heterogeneity causing differences in metabolism between metastatic sites. Organ-specific patterns of metastasis have been observed (Jin et al., 2020) and metastatic cancer cells exhibit different metabolic profiles depending on their secondary site (Andrzejewski et al., 2017; Basnet et al., 2019; Dupuy et al., 2015). Mechanical cues in tumor microenvironment can locally alter cellular metabolism and many cancers undergo a tissue-specific metabolic rewiring (Gaude and Frezza, 2016). In breast cancer, primary tumors are metabolically heterogeneous and increase both OXPHOS and glycolysis compared to normal tissue; however, liver-metastatic breast cancer cells elevate glycolysis via HIF-1α/PKD1, while bone- and lung-metastatic cells elevate OXPHOS (Dupuy et al., 2015), display high expression of PGC1-α (Andrzejewski et al., 2017), and increase oxidative stress (Basnet et al., 2019). Interestingly, brain metastasis, which is a major cause of death in patients of breast cancer, exhibit altered lipid metabolism (Jin et al., 2020). In melanoma, brain metastases exhibit increased OXPHOS as compared to both lung metastases and primary tumors, and inhibition of OXPHOS decreased brain metastasis formation (Fischer et al., 2019). However, it is unclear if a change in metabolic phenotype in the primary tumor is a determinant of metastatic site or if metabolic reprogramming occurs at the secondary site.

While metabolic adaptation to the local environment at the metastatic site is important for colonization (Schild et al., 2018), colonized cells may also actively reprogram the metastatic niche to further promote cancer progression (Psaila and Lyden, 2009), likely in a similar manner as primary tumor cells remodel the stroma and reprogram their metabolism (Mashimo et al., 2014). Some evidence indicates that cancer cells can prime the pre-metastatic niche and increasing nutrient availability prior to metastasis (Fong et al., 2015). Future work will need to investigate metabolic plasticity and flexibility during migration through complex microenvironments and the different stages of metastasis. A clear understanding of metabolic flexibility and alterations during metastasis will be critical to developing effective therapeutic strategies.

Targeting cancer cell migration through metabolism

Metabolic intervention is an emerging field of cancer treatment with several clinical inhibitors recently approved by the FDA, including ivosidenib and enasidenib, which inhibit mutated isocitrate dehydrogenase 1 (IDH1) and IDH2, respectively, and block oxidative decarboxylation of isocitrate to α-ketoglutarate during metabolism, for acute myeloid leukemia (AML) (Mullard, 2017; Norsworthy et al., 2019). While targeting cancer metabolism has traditionally been viewed as a way to target proliferation (Schulze and Harris, 2012; Tennant et al., 2010), the link between metabolism and migration and metabolic alterations in cancer provide new therapeutic targets for treating metastasis. Prior efforts to selectively inhibit cancer cell migration and metastasis have yielded limited success clinically as cell migration is a complex process (Steeg, 2006). There are many drivers of migration (growth factors and their downstream signaling); however, targeting metabolism may provide the opportunity to stall the engine that fuels migration (Li et al., 2019a), with many metabolic inhibitors having been shown to impact migration and metastasis (Table 3).

Table 3.

Treatments that targeting cancer cell metabolism and their effects on migration and metastasis.

Target	Treatment	Cancer	Result	Reference
Mitochondrial respiration	Rotenone, oligomycin, casiopeina II-gly, methoxy-TEA, celecoxib, dequalinium chloride, metformin	Breast, lung, ovarian, melanoma	Reduced cytoskeletal remodeling, migration, and metastasis	(Cerezo et al., 2013; Davis et al., 2020; Helige et al., 1993; Maddocks et al., 2013; Pacheco-Velazquez et al., 2018; Wu et al., 2012)
Mitochondrial trafficking	Cells deficient for mitochondrial RhoGTPase (miro1)	Breast	Reduced ATP:ADP at cortex; disrupted actin dynamics, lamellipodia protrusion, and membrane ruffling; slower migration	(Schuler et al., 2017)
Glycolysis	2-deoxyglucose, iodoacetate, gossypol, polyphenol phloretin, quercetin	Breast, prostate, gastric, melanoma, osteosarcoma	Reduced FAK phosphorylation; reduced cytoskeletal remodeling, migration, and invasion	(Cao et al., 2015; Pacheco-Velazquez et al., 2018; Shiraishi et al., 2015; Sottnik et al., 2011; Wu et al., 2018; Xu et al., 2018)
Pyruvate dehydrogenase kinase	Dichloroacetate	Breast	Shift from glycolysis to OXPHOS, metastasis inhibited	(Sun et al., 2010)
AMPK	Compound C	Breast	Local AMPK production; force-induced AMPK activation loop phosphorylation, glucose uptake, ATP production, and RhoA-mediated contractility blocked	(Bays et al., 2017; Cunniff et al., 2016)

Mitochondrial and glycolytic inhibitors have been shown to significantly block the migration and invasion capacity of a variety of cancer cells, including breast cancer (Pacheco-Velazquez et al., 2018; Wu et al., 2018), lung cancer (Jeon et al., 2016), ovarian cancer (Wu et al., 2012), prostate cancer (Senthilkumar et al., 2011; Shiraishi et al., 2015), gastric cancer (Xu et al., 2018), melanoma (Cao et al., 2015; Cerezo et al., 2013; Helige et al., 1993), and osteosarcoma (Sottnik et al., 2011). Many of these inhibitors have demonstrated anti-tumor and anti-metastatic effects in vivo (Dai et al., 2012; Davis et al., 2020; Helige et al., 1993; Wilson et al., 2019; Wu et al., 2012; Wu et al., 2018; Yoshinaka et al., 2006). Inhibiting pyruvate dehydrogenase kinase, which shifts metabolism from aerobic glycolysis towards OXPHOS also inhibits metastasis in vivo (Sun et al., 2010). However, other studies show that oligomycin and antimycin, which inhibit mitochondrial respiration and OXPHOS ATP production, have no effect on the migration of prostate cancer cells (Shiraishi et al., 2015) and even enhance migration of gastric (Hung et al., 2012; Hung et al., 2010) and lung cancer cells (Han et al., 2018) possibly through ROS signaling and EMT. Downregulation of mitochondrial genes is also associated with EMT signatures and worst clinical outcome across many cancer types (Gaude and Frezza, 2016). These conflicting results may be due to inter- and intra-tumor heterogeneity observed within and across cancer types (Davis et al., 2020; Gaude and Frezza, 2016). Identifying metabolic pathways critical to fueling cellular machinery for motility in specific cancers presents the opportunity for metabolic inhibitors to therapeutically target metastasis, but this may be difficult. Combination drug treatments of multiple metabolic pathways may be necessary to effectively inhibit cancer metabolism as cancer cells can shift to alternative energy sources and production pathways under energy stress.

As some metabolic pathways are mechanoresponsive, inhibitors targeting cell response to matrix mechanics have also been shown to reduce metabolism and migration. The ROCK inhibitor Y-27632 has been used to target bioenergetics and mechanosensitive metabolic adaption to changes in 2D matrix density (Mah et al., 2018). Similarly, inhibitors against ROCK, PI3K, myosin light chain kinase (MLCK), and actin polymerization all lowered intracellular ATP:ADP levels and ATP consumption in breast cancer cells during migration in 3D collagen matrices (Zanotelli et al., 2018). Understanding the metabolic pathways used to adapt to altered mechanical cues provides potential to use mechanomedicine to target abnormal ECM mechanics by disrupting the cellular response to mechanics.

Conclusions and outlook

Cancer metabolism plays an underlying role in driving cancer cell phenotype during tumor progression and promoting cancer aggressiveness. Recent studies suggest that cooperatively migrating cells dynamically switch leader position to minimize the thermodynamic costs of invasion (Liu et al., 2013; Zhang et al., 2019) and energy minimization directs cell decision-making during migration through a spatially complex matrix (Zanotelli et al., 2019). Studying cell migration from a thermodynamic perspective can provide new insights into how mechanical forces and cell mechanics impact cellular energetics during migration and direct migration path as cells preferentially migrate towards the path of energy minimization. Targeting metabolism could be used to stall the cellular engine that powers migration and limit energy-intensive migration through tight interstitial spaces. However, it remains unclear how specific metabolic pathways support cytoskeletal activity for force generation during migration and if changes in metabolism during motility are cancer-type specific. Intra- and inter-tumor metabolic heterogeneity has been observed (Davis et al., 2020; Gaude and Frezza, 2016) and metabolic reprogramming of cancer cells at metastatic sites (Dupuy et al., 2015) indicates that metabolism may be regulated by the local microenvironment. Still, targeting cancer metabolism presents an attractive strategy to limit metastasis, as the metabolic heterogeneity of tumors is likely less complex than the genetic heterogeneity of tumors (Martinez-Outschoorn et al., 2017). Hundreds to thousands of metabolites in tissues and biofluids at different stages of the invasion-metastasis cascade can be detected via the growing metabolomics technology (Beger, 2013). Extensive metabolic characterization will allow better understanding of the role of different metabolites in cancer migration and metastasis and help identify potential drug targets.