Tumour-directed microenvironment remodelling at a glance

In tissue biology, the terms stroma or microenvironment have been used to distinguish supporting tissue from the parenchyma, a historical term for the functional component of the organ. However, we now know that, far from being non-functional, the microenvironment is complex and dynamic, being composed of supporting cells including fibroblasts, immune cells, endothelial cells and adipocytes, and the extracellular matrix (ECM), a meshwork of structural, fibrillar proteins and non-structural matricellular proteins that regulate interactions between the ECM and cells (Xu et al., 2009). The microenvironment is critical for the normal functioning of tissues and organs. Broadly speaking, the cellular constituents of the microenvironment transmit nutrients and oxygen, tackle infection, produce the ECM and initiate biochemical cues, while the ECM maintains structural integrity, transmits biophysical cues and focuses biochemical signals (Theocharis et al., 2019).

The normal tissue microenvironment can suppress the formation of cancers via anti-tumour immune responses elicited by tumour cell-specific neoantigens (Jiang et al., 2019), by providing a hostile environment for the survival of delaminated cells (Kaukonen et al., 2016) and arguably also by preventing the basal extrusion of epithelial cells exhibiting oncogene activation (Hendley et al., 2016). These functions are underpinned by direct cell–cell and cell–ECM physical interactions, as well as a complex paracrine signalling programme between the microenvironment and epithelium, driven, at least in part, by a wound healing response (Huet et al., 2019).

Tumour cells can co-opt all aspects of normal interaction with the microenvironment in such a way that, in a tumour, a distinct microenvironment is established that differs from the normal tissue microenvironment in the structure and composition of the ECM and in the phenotypes and functions of its cellular components. In general, a tumour is considered to contain tumour cells and the tumour microenvironment (TME). The TME encompasses all other tumour constituents that are not cancer cells and may also include normal cells (Montagner et al., 2020).

Recruitment and polarisation of various inherently normal stromal cell types (Jackson et al., 2020) via a tumour-specific suite of signalling molecules allows cancers to build a diverse population of supporting stromal cells. Interactions of cancer cells with stromal cells are likely to be spatially regulated, resulting in local microenvironments. For example, interactions of tumour cells with blood vessel endothelia will result in a TME that is distinct from that formed by interactions with fibroblasts or resident immune cells at tumour–stroma interfaces (Jackson et al., 2020). It is worth noting that one of the key differences between the normal microenvironment and the TME is the complete metabolic reprogramming of the stroma, reflecting the changed metabolic status of the cancer cells themselves (Faubert et al., 2020), which underlies the distinct functions of tumour-associated stroma. The pro-tumorigenic feedback from these reprogrammed stromal cells is both direct (secreted paracrine signals) and indirect (remodelling of the ECM by these stromal cells), and is discussed further in following sections.

Here, we will briefly discuss how a tumour modifies its microenvironment to a permissive form to facilitate disease progression by modifying both its biochemical and biophysical properties. In the first part of this Cell Science at Glance article, we will discuss how the stromal cells of the microenvironment are reprogrammed, and in the second part, how the ECM is remodelled. While we focus here on tumour-promoting changes in the microenvironment, some elements of the microenvironment have been demonstrated to suppress tumour growth (Ozdemir et al., 2014). However, whether tumour-suppressive components of the tumour microenvironment can be effectively distinguished from the normal microenvironment, which, as we have discussed above suppresses tumour formation, remains unclear.

Fibroblasts are a diverse cell population, including myofibroblasts, mesenchymal cells, tissue-resident fibroblasts and vasculature-derived fibroblasts (see poster). For instance, it is proposed that breast cancers exhibit distinct cancer-associated fibroblast (CAF) populations with potentially distinct functions, as inferred from their gene expression profiles (Bartoschek et al., 2018; Valdes-Mora et al., 2020 preprint), that vary by cancer subtype (Ali et al., 2020). Although the origins of CAFs appear to be largely context and tumour dependent, it is clear that their presence in the tumour stroma is associated with a poor prognosis (Sahai et al., 2020).

Many tumour-derived factors have been reported to reprogram fibroblasts to an activated or educated, tumour-promoting CAF form, with transforming growth factor β (TGFβ) in particular being well canvassed (Calon et al., 2014). Other factors shown to induce CAF reprogramming include epidermal growth factor (EGF), platelet-derived growth factors (PDGFs) α and β, basic fibroblast growth factor (bFGF), interleukin (IL)-6 and IL-1β (Santi et al., 2018), cysteine-rich EGF-like domains 2 (CRELD2) (Boyle et al., 2020), tumour necrosis factor (TNF) (Vennin et al., 2019), activin A (Cangkrama et al., 2020) and thrombospondin 2 (THBS2) (Del Pozo Martin et al., 2015). Hypoxic conditions within tumours can potentiate fibroblast reprogramming. For instance, NADPH oxidase 4 (NOX4) released by tumours exhibiting the Warburg effect (Arcucci et al., 2016) can prolong TGFβ signalling to reprogram fibroblasts (Jain et al., 2013).

Once activated, CAFs can secrete several cancer-promoting factors, including hepatocyte growth factor (HGF), EGF, connective tissue growth factor (CTGF) and insulin-like growth factor (IGF) (Santi et al., 2018). Many cytokines and chemokines are also released, acting on cancer cells as well as recruiting other pro-tumorigenic cells to the tumour milieu, including macrophages, so-called myeloid-derived suppressor cells, and regulatory T cells (Tregs) (Santi et al., 2018). An emerging concept is that reprogramming of CAFs can induce senescence and the senescence-associated secretory phenotype (SASP), which is driven by the p38 MAPK, nuclear factor (NF)-κB and mammalian target of rapamycin (mTOR) signalling pathways, and is tumour promoting (Faget et al., 2019; Krtolica et al., 2001). The specific functions of CAFs in remodelling the ECM are discussed below.

Tumour cells can subvert, or activate, normal adipocytes in the stroma or adjacent fat depots into cancer-associated adipocytes (CAAs) by means of metabolic reprogramming, as well as through appropriate CAA metabolites to promote tumour growth and progression (see poster). This reprogramming is facilitated by elevated aerobic glycolysis in tumours (Wu et al., 2019a) but may also be mediated by exosomal micro (mi)RNAs delivered to adipocytes by cancer cells to regulate MAPK, insulin-receptor substrate (IRS) and AMPK signalling (Wu et al., 2019b). The CAA phenotype is characterised by lipid depletion and higher expression of differentiation markers, matrix metalloproteases (MMPs) and ILs compared to normal adipocytes (Dirat et al., 2011). CAA-derived metabolites (ketones, fatty acids, pyruvate and lactate) directly promote cancer cell invasion and metastasis (Dirat et al., 2011; Wang et al., 2017). CAAs also display differential production of chemokines (high CCL2 and CCL5), cytokines [high IL-1β, IL-6, TNF and vascular endothelial growth factor (VEGF)] and adipokines (low adiponectin and high leptin), which not only act on cancer cells but also on surrounding stroma (Vansaun, 2013; Wu et al., 2019a). For example, IL-6, leptin and TNF from CAAs promote angiogenesis, while IL-10 can stabilise Tregs, PD-L1 (also known as CD274) restrains cytotoxic T lymphocytes (CTLs), and leptin can polarise macrophages to an M2 phenotype (Deng et al., 2016; Guo et al., 2012; Wu et al., 2019a). Furthermore, CAAs can secrete matrix-remodelling enzymes (including TNF, osteopontin and MMP9) (Ribeiro et al., 2012). It has also been postulated that one origin of CAFs may be from trans-differentiation of CAAs (Jotzu et al., 2010), although the body of evidence supporting this is currently small.

It is now well-established that cancer cells secrete pro-angiogenic factors that promote the growth of blood vessels, a process termed angiogenesis, to supply the tumour with nutrients and oxygen. These factors include chemoattractants that recruit endothelial cells (Lugano et al., 2020) and well-known pro-angiogenic growth factors such as VEGF, delta ligand-like 4 (DLL4), angiopoietin 2 (ANGPT2) and FGF that are also produced by stromal cells under the influence of the cancer (De Palma et al., 2017). Under the influence of these factors, a new network of blood vessels is set up to service the tumour (see poster). Intriguingly, cancer cells themselves have been demonstrated to take the place of endothelial cells to generate a network of vessels that are contiguous with the blood vasculature and capable of supplying the tumour with nutrients and oxygen (Folberg et al., 2000). This process, termed vasculogenic mimicry, is associated with progressive cancers (Lugano et al., 2020).

The tumour vasculature has an additional role in some tumours, providing a highly restricted microenvironment termed the perivascular niche (PVN) (Ghajar et al., 2013), which has been reported to harbour so-called cancer stem cells and provide an environment that protects these cells against chemotherapy (Carlson et al., 2019). In breast cancers, it is thought that the PVN is established and maintained via a combination of secreted factors, such as thrombospondin 1, and ECM proteins, such as periostin (Ghajar et al., 2013), although it is not yet completely clear precisely what, if any, mechanistic role is played by tumour cells in the establishment of the PVN.

It is worth noting that stromal cell signalling significantly contributes to angiogenesis. These indirect interactions are highlighted in the relevant sections below.

Tumour-associated macrophages (TAMs), which can constitute up to 50% of the solid tumour mass, arise from recruitment of monocytes to the tumour site and the co-opting of tissue-resident macrophages involved in immune surveillance (Solinas et al., 2009). Although their inherent function is tumour suppression owing to their pro-inflammatory role in fighting infection, many studies have reported immune-suppressive, tumour-promoting roles for TAMs (Mantovani et al., 2017), with an emerging role for Rho-associated kinase (ROCK) signalling within cancer cells in orchestrating these and other immune changes (Georgouli et al., 2019; Orgaz et al., 2020).

The functions of TAMs are often associated with a phenotype switch, traditionally classified into binary M1 (pro-inflammatory, anti-tumour) and M2 (anti-inflammatory, pro-tumour) polarisation states (see Box 1), although it has now become clear that TAMs exhibit context-dependent plasticity along this continuum (Aras and Zaidi, 2017; DeNardo and Ruffell, 2019). Several cancer cell-derived factors have been reported to recruit or polarise TAMs to M2-like in the TME (see poster). These include chemokines (CCL2, CCL5, CCL7, CCL8, CCL9, CCL18, CCL20 and CXCL12) and cytokines, including TGFβ, IL-4, IL-6, IL-10, IL-13, colony-stimulating factor-1 (CSF-1) and VEGF (Balkwill et al., 2005; Boyle et al., 2015; Lin et al., 2019).

Tumour-promoting TAMs secrete IL-1β, IL-8, IL-10, TNF and TGFβ to induce epithelial-to-mesenchymal transition (EMT) in cancer cells (Aras and Zaidi, 2017; Lin et al., 2019), and chininase-3-like protein 1 (CHI3L1), CSF-1 and EGF to promote invasion, migration and metastasis (Chen et al., 2017; Goswami et al., 2005; Wyckoff et al., 2004; Zeng et al., 2019). TAMs also promote tumour progression indirectly via the ECM (discussed in detail below) and by dampening anti-tumour immunity by secreting immunosuppressive cytokines (IL-10 and TGFβ), prostaglandins and metabolites, and inhibiting immune-checkpoint proteins (Mantovani et al., 2017). Furthermore, signalling of TAMs to CAFs through CCL2, CCL5, IL-6 and TGFβ can aid CAF reprogramming to a pro-tumour form (DeNardo and Ruffell, 2019).

TAMs are able to promote the growth of blood vessels through secretion of pro-angiogenic factors and Wnts (DeNardo and Ruffell, 2019). Interestingly, hypoxia in progressing tumours contributes to the pro-angiogenic function of M2-like TAMs, increasing their production of VEGFA, and GLUT1, GLUT3 and iNOS (also known as SLC2A1, SLC2A3 and NOS2, respectively) (Van Overmeire et al., 2014).

Tumour-associated neutrophils (TANs) contribute to the immune infiltrate in multiple cancers (Gentles et al., 2015; Shaul and Fridlender, 2019), and patients with advanced stage cancer exhibit neutrophilia (Schmidt et al., 2005). Like TAMs, TANs may exist on a continuum of N1 to N2 phenotypes (see Box 1) (Masucci et al., 2019). TANs can be recruited to many tumour types expressing chemokines that engage CXCR2, which is highly expressed by neutrophils (Jamieson et al., 2012), and the N2 TAN phenotype can be induced by cancer cell-derived TGFβ, TNF and granulocyte-macrophage colony stimulating factor (GM-CSF; also known as CSF2) (Fridlender et al., 2009) (see poster).

TANs support tumour progression by promoting cancer cell proliferation, motility, migration and invasion (Coffelt et al., 2015; Shaul and Fridlender, 2019; Wculek and Malanchi, 2015), either through proteins such as oncostatin M (Queen et al., 2005), or inflammation that is mediated by factors, including prostaglandin E2 (PGE2), IL-6, IL-8, TNF, CXCL1 and COX-2 (Masucci et al., 2019). Furthermore, TANs can enhance tumour-promoting genomic instability in cancer cells by releasing ROS and NO into the microenvironment (Sandhu et al., 2000), as well as stimulating angiogenesis (Kuang et al., 2011) and, notably, modulating immune cell function. Neutrophils suppress the anti-tumour responses of T cells via activation of the PD-L1–PD-1-mediated immune checkpoint (PD-1 is also known as PDCD1) (He et al., 2015), whilst also recruiting tumour-promoting macrophages and Tregs to the TME by secreting CCL2 and CCL17 (Mishalian et al., 2014; Zhou et al., 2016).

Myeloid-derived suppressor cells (MDSCs) (see Box 1) are recruited by cancers and stimulated to proliferate by tumour- and stroma-derived factors, including PGE2, macrophage colony stimulating factor (M-CSF), G-CSF (CSF3), GM-CSF, TNF and interleukins (see poster). MDSCs secrete inflammatory mediators, facilitate immune evasion and secrete cancer-promoting chemokines and cytokines such as CXCL12, which promote tumour cell migration (De Sanctis et al., 2016; Obermajer et al., 2011; Tu et al., 2008). MDSCs also contribute to immune evasion of cancer cells by inhibiting T cell surface receptors, thereby skewing their population to suppressive phenotypes; for example, CD4⁺ T cells to a Treg instead of helper T cell (Th) phenotype, and macrophages to an M2 phenotype (see Box 1). They also deplete essential metabolites in the TME required for T cell anti-tumour function and generate hypoxic conditions that promote angiogenesis (De Sanctis et al., 2016). MDSCs have also been shown to act directly on cancer cells. For instance, ovarian tumour cells exposed to MDSC-conditioned medium had increased properties of stemness and were significantly more metastatic (Cui et al., 2013), and crosstalk between breast cancer cells and MDSCs mediated by IL-6 signalling has been shown to potentiate tumour cell aggressiveness and metastasis (Oh et al., 2013).

Other granulocytic cells are also recruited to the TME, including mast cells, eosinophils and basophils (see poster). Infiltration of granulocytes can be indicative of both good and poor prognosis in cancer (Davis and Rothenberg, 2014; Rigoni et al., 2018; Wei et al., 2018). Upon granulocyte degranulation, the release of pro-inflammatory factors can support tumour growth by increasing serum levels of inflammatory cytokines and chemokines (such as IL-8, CCL4 and CCL5) that recruit tumour-promoting macrophages, neutrophils and MDSCs; this suppresses the anti-tumour immune responses through release of IL-10 and TGFβ, and promotes tumour angiogenesis (Khazaie et al., 2011; Rigoni et al., 2018; Schmielau and Finn, 2001; Trellakis et al., 2011). Granulocytes may also directly support cancer progression (Strouch et al., 2010).

Tumour-infiltrating lymphocytes (TILs) play both tumour-suppressive and tumour-promoting roles in solid tumours, with Th cells and CTLs exhibiting largely anti-tumour activities, whereas regulatory Tregs promote tumour formation. Innate lymphoid natural killer (NK) cells can also shape tumour immune responses.

Th cells (see Box 1), which are activated by antigen-presenting B cells, macrophages or dendritic cells of the TME, support anti-tumour immune responses by activating effector CTLs and recruiting innate immune cells (see poster). Th cells activate CTLs by cell–cell contact (mediated by co-stimulatory molecules such as CD27, CD134 and class II MHC), as well as in a paracrine manner (e.g. IL-2 secretion) (Knutson and Disis, 2005). However, differentiation into IL-17^hi Th17 cells (see Box 1) can be activated by cancer-cell-derived TGFβ, TNF and interleukins (Maniati et al., 2010). Th17 cells release pro-inflammatory cytokines that can support angiogenesis, recruit other inflammatory cells and upregulate pro-tumour transcription factors (Maniati et al., 2010).

Upon recognition of tumour antigens, CD8⁺ CTLs (see Box 1), recruited by chemokines, initiate cytotoxic killing of cancer cells. In general, a high CTL count signifies a good prognosis. Cancer cells, CAFs and M2 TAMs can induce PD-1 expression on CTLs to induce their exhaustion, and M2 TAMs can also secrete a battery of immune-suppressive cytokines that exhaust CTLs to suppress tumour immunity (Zhang et al., 2020). Accordingly, inducing checkpoint blockade by neutralising CTL-associated antigen 4 (CTLA4) and/or PD-1 is clinically beneficial (Gubin et al., 2014). CTLs can also become exhausted due to continuous stimulation by tumour antigens. Exhausted CTLs (see Box 1) have dysregulated production of anti-tumour cytokines [IL-2, interferon (IFN)-γ and TNF]. Because of their potent anti-tumour capacity, CTLs are being harnessed for clinical immunotherapies (Neeve et al., 2019). Chimeric antigen-receptor T (CAR-T) cells have shown promise in clinical trials against several malignancies (Gubin et al., 2014). However, cancers can co-opt immune checkpoint pathways to suppress CAR-T cell function and support tumour growth (Gubin et al., 2014).

Tregs, both the naturally occurring (nTreg) and induced (iTreg) populations (see Box 1), which are recruited from the thymus or tumour periphery, respectively, by cytokines and chemokines (for example TGFβ and CCL22) produced by the growing tumour (Curiel et al., 2004; Liu et al., 2007), are often associated with poor prognosis in cancers due to their immune regulatory roles (Ma et al., 2014). Tregs contribute to tumour growth by facilitating immune evasion, characterised by production of immunosuppressive cytokines, cell–cell inhibition of T cells, inhibition of antigen-presenting cells, and killing of anti-tumour T cells (Gallimore and Simon, 2008; Gondek et al., 2005). Treg-derived TGFβ can promote Th17 differentiation, while Treg-derived IL-23 can block CTL infiltration into the tumour (Gallimore and Simon, 2008).

High activity of NK cells (see Box 1) is associated with a reduced cancer risk (Imai et al., 2000). These cells exhibit cytotoxic killing of tumour cells and secrete cytokines such as IFN-γ that support tumour-suppressing immune cell populations (see poster). As such, NK cells may be harnessed for immunotherapy, with many approaches proposed (Vivier et al., 2012). However, tumours may modulate NK cells in their environment. Cancer-cell-secreted PGE2 can decrease the expression of NK-produced granzyme B, perforin and IFN-γ (Hodge et al., 2014), and NK cell activation can be dampened by a host of TME factors, including Treg-derived TGFβ (Gomes et al., 2014).

The ECM consists of a meshwork of fibrillar proteins (including collagens I and III, fibronectin and proteoglycans) that act primarily as a scaffold for tissues and organs (see poster). In addition to providing structural support, the ECM has key roles in transducing mechanical signals to the parenchyma and acts as a reservoir for chemokines, cytokines, growth factors and other signalling molecules via its ability to regulate their local concentrations (Poltavets et al., 2018).

The molecular structure of its components enables the ECM to provide structural integrity to tissues and transduce mechanical forces. Collagen, which forms the bulk of the structure, can be crosslinked either enzymatically or non-enzymatically, which enhances its mechanical rigidity, although the latter exhibits slower kinetics than the former (Avery and Bailey, 2006). Lysyl oxidase (LOX) and the LOX family of amine oxidases mediate collagen cross-linking enzymatically through oxidative deamination of lysine residues (Yamauchi et al., 2018).

The mechanical properties of a tissue greatly influence its function. Stiffness is the extent to which a material can resist deformation in response to applied force and is denoted by Young's modulus (Guimarães et al., 2020). Tissues exhibit widely varying Young's moduli (ranging from 0.5–1 kPa for fat to 15,000–20,000 kPa for bone) underpinned by the composition and structure of their ECM (Handorf et al., 2015). However, normal tissue mechanics is altered in cancer. For instance, stage IV colorectal tumours can have moduli ranging from 5.58 (near normal) to a high stiffness of 68 kPa (Bauer et al., 2020). ECM stiffness can be enhanced by increased production of ECM components or enhanced remodelling and cross-linking in both normal tissue physiology (e.g. during wound healing) and tumorigenesis (Poltavets et al., 2018). LOX is frequently elevated in cancers, which often results in collagen cross-linking and increased tumour ECM stiffness. Hypoxia also contributes to collagen cross-linking and thereby to increased tumour ECM stiffness (Makris et al., 2014). Mechano-reciprocity is the ability of cells to moderate intracellular tension to accommodate these changes in tumour ECM stiffness, and is mediated by mechanotransduction signalling pathways that transduce information about mechanical changes within cells and the microenvironment (Boyle and Samuel, 2016).

High stiffness of the tumour ECM has direct consequences for tumour progression, invasion and metastasis. In breast tumours and organoid cultures (Levental et al., 2009), and in skin cancers (Samuel et al., 2011), enhanced tumour ECM stiffness facilitates the formation of focal adhesions and integrin signalling, promoting cancer invasion. Elevated tumour ECM stiffness also promotes EMT, a key process in tumour progression, in pancreatic (Rice et al., 2017) and breast cancer, where mechanotransduction releases the EMT-regulating transcription factor TWIST1 from its cytoplasmic binding partner G3BP2, whereupon it translocates to the nucleus (Fattet et al., 2020; Wei et al., 2015). Consistent with the establishment of a positive-feedback loop, a stiff tumour ECM potentiates further ECM production and stiffening by enhancing mechanotransduction in CAFs (Bayer et al., 2019; Zhang et al., 2016).

The tumour ECM engages mechanotransduction signalling through the Hippo, Wnt, phosphoinositide 3-kinases (PI3K)–Akt, TGFβ and Rho-ROCK pathways, amongst others (see poster). This results in changes in cell physiology, including that of the actin cytoskeleton, which regulates cell rigidity (Huang et al., 2019; Ibbetson et al., 2013). Integrins in focal adhesions can sense changes in extracellular force, leading to the activation of FAK (also known as PTK2), Src family kinases and RhoA (Harburger and Calderwood, 2009). High matrix stiffness also increases YAP/TAZ-mediated regulation of transcription and further integrin expression (Dupont et al., 2011), potentially establishing a positive feedback loop. Elevated YAP and TAZ (also known as YAP1 and WWTR1, respectively) levels are associated with increased cancer stemness and metastatic potential (Zanconato et al., 2016). Such associations may also explain the link between tumour ECM stiffness and aggressive tumour phenotypes.

Stiffening of the tumour ECM during cancer progression can also enhance mechanical forces upon tumour cells, such as compression (pushing against ECM) and tension (stretching at cell-ECM interfaces); these stresses can potentiate tumour-promoting mechanotransduction signalling (Boyle et al., 2018).

Tumours can modify their ECM through the secretion of ECM-modifying enzymes, or cause CAFs to produce tumour ECM as discussed above. Activated CAFs secrete TGFβ along with HIF1-α to induce LOX, leading to collagen cross-linking (Levental et al., 2009). In colorectal cancer, enhanced tumour ECM stiffness feeds back on CAFs, causing them to secrete activin A, which promotes metastatic potential of tumour cells (Bauer et al., 2020).

Mechanical stress under normal physiological conditions has been demonstrated to influence macrophage functions and inflammatory cytokine production (Maruyama et al., 2019), although the molecular mechanisms underlying these processes remain to be elucidated. However, it is well-known that macrophages produce MMPs, such as MMP1, MMP2, MMP9 and MMP12, that can remodel ECM (Condeelis and Pollard, 2006). The capacity to harness macrophages therapeutically to remodel ECM has been demonstrated as a proof-of-principle using an engineered macrophage population in an allograft breast cancer model (Zhang et al., 2019). In this study, tumour progression was impaired by upregulating MMPs that degraded ECM in infused chimeric antigen receptor-expressing macrophages (Zhang et al., 2019).

Stromal dendritic cells are also affected by mechanical stimuli as they travel through different microenvironments. Dendritic cells cultured on matrices of high stiffness (50 kPa) exhibited low adhesion and reduced viability (Mennens et al., 2017). As dendritic cells play an important role in immune surveillance and tumour antigen presentation, their impaired functions on stiff substrates may reduce the ability of T cells to eradicate tumour cells.

In addition to enhancing signalling through mitogenic mechanotransduction pathways in tumour cells, mechanical forces also influence lymphatic and blood vessels. Solid stress from rapidly growing cancers deforms blood and lymphatic vessels to induce hypoxia, as has been observed in murine liver tumours, oesophageal carcinoma and salivary duct carcinoma (Stylianopoulos et al., 2013).

The preceding discussion provides a glimpse into a dynamic network of biochemical and biomechanical signals that underpin cancer-microenvironment inter-relationships. A clear understanding of these is essential to targeting them therapeutically (see Box 2).

In conclusion, the TME and, more specifically, the suite of mechanisms that tumour cells adopt to modify the microenvironment, remains a relatively under-exploited resource for novel approaches to target cancer progression. Nevertheless, the key pathways that differentiate tumour–microenvironment interactions from parenchyma–microenvironment interactions are becoming clearer. Adopting a nuanced approach to exploiting this information will determine whether microenvironment-targeted therapies will fully realise their tantalising promise.