The Immune Microenvironment and Cancer Metastasis

Abstract

The dynamic interplay between neoplastic cells and the immune microenvironment regulates every step of the metastatic process. Immune cells contribute to invasion by secreting a cornucopia of inflammatory factors that promote epithelial-to-mesenchymal transition and remodeling of the stroma. Cancer cells then intravasate to the circulatory system assisted by macrophages and use several pathways to avoid recognition by cytotoxtic lymphocytes and phagocytes. Circulating tumor cells that manage to adhere to the vasculature and encounter premetastic niches are able to use the associated myeloid cells to extravasate into ectopic organs and establish a dormant microscopic colony. If successful at avoiding repetitive immune attack, dormant cells can subsequently grow into overt, clinically detectable metastatic lesions, which ultimately account to most cancer-related deaths. Understanding how disseminated tumor cells evade and corrupt the immune system during the final stages of metastasis will be pivotal in developing new therapeutic modalities that combat metastasis.

Metastasis is a multistep process characterized by the spread of malignant cells to distant organs. This is evolutionary driven by genetic and/or epigenetic alteration within cancer cells but is also dependent on an intricate interaction with stromal cells at a local and systemic level (Fig. 1). Tumor cells must overcome several hurdles to establish macroscopic metastasis, collectively referred to as the invasion–metastatic cascade (Valastyan and Weinberg 2011). During this process, carcinoma cells deviate from their primary growth site (local invasion and intravasation), disseminate systemically via circulation through the lymphatic or vascular system (survival, arrest at distant site, extravasation), and survive and adjust to a new microenvironment (colonization and outgrowth) (Lambert et al. 2017). For cancer cells to migrate through tissue, they acquire several abilities that act in concert to promote invasion, including cytoskeletal reorganization, secretion of proteases, and altered adhesion receptor-ligand interactions (Kessenbrock et al. 2010; Quail and Joyce 2013). In addition, cancer cells can undergo a reversible program termed epithelial to mesenchymal transition (EMT) through which loss of epithelial polarization and intercellular adhesion allows for motility and invasiveness (Thiery et al. 2009). Neoplastic cells leaving the primary tumor and intravasating into circulation must then survive severe physical shear stress, attack by the immune system, and oxidative stress (Labelle et al. 2011; Le Gal et al. 2015). Once cells arrest at a distant site or are physically trapped in minute capillaries, they can proliferate within the intravascular space or extravasate by modifying the endothelium (Al-Mehdi et al. 2000; Deneve et al. 2013; Strilic and Offermanns 2017). Only a small fraction of cells that disseminate from the primary tumor will manage to invade and adapt to their new surrounding (Nagrath et al. 2007). Thus, the ability of neoplastic cells to co-opt immune processes to aid in these steps and evade immune recognition are necessary for successful metastasis (Sosa et al. 2014; Massagué and Obenauf 2016).

Figure 1.

Establishment of the tumor immune microenvironment: Environmental insults or spontaneous genetic alterations in epithelial cells can lead to immune surveillance (1) or the growth of premalignant cells that evade clearance by innate and adaptive immune cells (2). In addition to driving cell proliferation, certain oncogenic alterations are associated with the release of cytokines that recruit additional immune cells to aid neoplastic progression by providing mitogenic factors or producing reactive oxygen species that result in the accumulation of additional genetic mutations (3). Uncontrolled cellular growth at this stage causes environmental changes such as reduced oxygen levels (hypoxia), extracellular matrix (ECM) turnover, and stromal disruption (4). These signals of tissue damage result in a pathological healing response that includes the creation of a disordered and permeable vascular network (angiogenic switch), as well as morphological and transcriptional changes in epithelial cells that reflect a mesenchymal phenotype (EMT). Malignant cells are thus able to migrate toward and intravasate into the vascular lumen, becoming circulating tumor cells.

ESTABLISHMENT OF THE TUMOR IMMUNE MICROENVIRONMENT

The tumor immune microenvironment (TIME) promotes every aspect of carcinogenesis, including initiation, survival, growth, metastasis, and immune evasion. Depending on the tissue of origin, the TIME shows a unique immune repertoire, with different proportions and functional states of lymphocytes, granulocytes, monocytes/macrophages, and dendritic cells (DCs) (Gentles et al. 2015). Although several of these immune populations have the potential to eradicate malignant cells, numerous factors in the tumor act to blunt this activity or redirect the cells toward promoting tumorigenesis. In addition to relative proportions, the locations of immune cells are associated with disease progression. For example, sites of focal myoepithelial cell layer dysregulation in ductal carcinoma in situ (DCIS) display an enrichment of leukocytes, suggesting a role in driving or sustaining local invasion (Gil Del Alcazar et al. 2017), whereas in late stage disease the exclusion of T lymphocytes from tumor beds promotes resistance to immune checkpoint blockade (Jayaprakash et al. 2018; Jerby-Arnon et al. 2018). The degree to which parallel programs are used to establish the metastatic immune microenvironment (MIME) are largely unknown, with the exception of myeloid cell recruitment to establish the premetastatic niche (Liu and Cao 2016).

Tumor Intrinsic Signaling

In addition to the type of cancer, immune composition correlates with the underlying genetic alterations driving carcinogenesis (Rooney et al. 2015), as oncogenic signals in transformed cells are associated with the release of chemokines that attract immune cells to the tumor (Wellenstein and de Visser 2018). As one of the most frequently mutated genes in cancer, loss of Trp53 activates the NF-κB pathway, stimulates cytokines release from cancer cells, which through paracrine interactions modify the immune landscape of cancer (Meylan et al. 2009; Kastenhuber and Lowe 2017). Similarly, the activation of the transcription factor MYC endows cancer cells with the ability to overproduce potent proinflammatory cytokines that facilitate tumor angiogenesis and recruitment of protumoral mast cells, macrophages, and neutrophils (Soucek et al. 2007), whereas activation of the oncogenic EGFR family of receptor tyrosine kinases results in secretion of CSF-1, a chemotactic and survival factor for macrophages. Interestingly, oncogenic Kras^G12D drives unique programs in different cancer types, although each cumulates in the recruitment of an immunosuppressive myeloid population. This includes expression of GM-CSF and CXCL3 in pancreatic cancer and colon carcinoma, respectively (Pylayeva-Gupta et al. 2012; Liao et al. 2019), as well as CCL9 expression in lung cancer in the context of Myc activation (Kortlever et al. 2017). Oncogenic Kras and Myc also cooperate to induce interleukin (IL)-23-dependent exclusion of lymphocytes (Kortlever et al. 2017). Collectively, these data establish an important role for tumor genotype in driving the immune repertoire in cancer.

There are also several studies showing that cells undergoing EMT modify the TIME through chemokine secretion (Dongre and Weinberg 2019). The EMT master regulator SNAIL promotes expression of CCL2 and CCL5, leading to macrophage recruitment (Hsu et al. 2014), and expression of CXCR2 ligands leading to recruitment of suppressive myeloid cells (Taki et al. 2018). Other mediators of epithelial mesenchymal transition, such as SRC-1, promote macrophage recruitment via expression of CSF-1 (Qin et al. 2009; Wang et al. 2009), and mesenchymal tumors display prominent macrophage infiltration (Dongre et al. 2017). The ability of EMT to promote an immunosuppressive environment is likely further augmented by high production of transforming growth factor (TGF)-β, which can lead to the exclusion of lymphocytes from tumor beds (Mariathasan et al. 2018; Tauriello et al. 2018) and regulate the activation state of most immune cells. This has been shown in one study with overexpression of SNAIL, which leads to production of TGF-β and thrombospondin-1, and increased conversion of regulatory T cells (T_regs) (Kudo-Saito et al. 2009). Given the critical role of EMT in the metastatic process it will be important to determine if these pathways are also involved in establishing the MIME.

Homeostatic Imbalance

Solid tumors show large regions with low oxygen levels (hypoxia) and nutrient insufficiency (Gatenby and Gillies 2004). Hypoxia has the ability to modulate immune functions directly by activating HIF-1α-dependent transcriptional changes within the immune cells, or indirectly by rewiring tumor cell secreted metabolites and cytokine repertoire (Cairns et al. 2007). For example, hypoxia was shown to trigger HIF-1α-dependent immunosuppressive activity in macrophages (Doedens et al. 2010). Through indirect effects on the tumor and the stroma, hypoxia also causes the overproduction of monocyte recruitment factors including CCL2, CCL5, CXC-chemokine ligand 12 (CXCL12), colony-stimulating factor (CSF-1), and vascular endothelial growth factor (VEGF) (Murdoch et al. 2004). As an indirect consequence of hypoxia, tumor cells up-regulate glycolysis and secrete lactic acid, which induces phenotypic changes and immunosuppressive function in macrophages (Colegio et al. 2014). Lactate-activated macrophages in turn enhance the glycolytic phenotype of breast cancer cells through the release of extracellular vesicles containing a HIF-1α-stabilizing long noncoding RNA (Chen et al. 2019). The acidic microenvironment of tumors also directly regulates macrophage phenotype, as seen following the genetic ablation of the macrophage pH sensor, Gpr132, or the buffering of tumor secreted acids, which reduces the spontaneous lung metastasis of breast cancer cells and prostate cancer invasiveness, respectively (Chen et al. 2017a; El-Kenawi et al. 2019). Finally, areas of hypoxia frequently coincide with regions of cell death, leading to recognition of apoptotic cells by phagocytes and/or the release of damage-associated molecular patterns (DAMPs) during secondary necrosis. Phosphatidylserine recognition receptors thus promote myeloid-dependent immune suppression in the TIME (Cook et al. 2013; Crittenden et al. 2016; Ubil et al. 2018); whereas several DAMPs have been shown to activate DCs and elicit a cytotoxic T-cell response (Kroemer et al. 2013).

Immune-Stroma Crosstalk

In addition to regulating the immune compartment directly, tumor cells drive phenotypic changes in the nonimmune stromal cells, including fibroblasts, endothelial cells, and pericytes. This “reactive” stroma possesses a unique gene/protein expression profile that differs from the signature of the corresponding normal tissue (Finak et al. 2008; Nonn et al. 2009) and is created, in part, by the influence of inflammatory mediators and cytokines. For example, TGF-β cooperates with SDF-1 (CXCL12) to generate myofibroblasts with enhanced tumor-promoting capability through autocrine signaling loops (Kojima et al. 2010). Cancer-associated fibroblasts can suppress a T-cell response through a myriad of pathways, including direct suppression via PD-L2 and FAS ligand expression (Lakins et al. 2018), expression of cytokines (Kato et al. 2018), and recruitment of immunosuppressive myeloid cells (Kumar et al. 2017). Cancer cells can even create gap junctions with astrocytes to transfer the second messenger cGAMP to astrocytes, activating the STING pathway and production of type I interferons (Chen et al. 2016). It should also be noted that changes in the physical properties of the stroma further modulate the TIME (DeNardo and Ruffell 2019). For example, higher levels of collagen (Wesley et al. 1998; Stahl et al. 2013; Pickup et al. 2014) and collagen cross-linking (Van Goethem et al. 2010; McWhorter et al. 2015) enhance the protumoral phenotype of macrophages, and inhibiting the hyperactivation of focal adhesion kinase (FAK) is sufficient to alter macrophage polarization and increase response to immune checkpoint blockade (Jiang et al. 2016).

There is also an elaborate interplay between immune cells in the tumor that controls the phenotype and functional role of the cells. Most well established is the ability of macrophages, monocytes, neutrophils, and other myeloid populations to suppress the cytotoxic function of lymphocytes, through mechanisms that have been reviewed elsewhere (Veglia et al. 2018; DeNardo and Ruffell 2019). As will be discussed, myeloid cells are a critical component of the immunosuppressive TIME and MIME, and depleting these cells can reduce metastasis by unleashing T and natural killer (NK) cell cytotoxicity. Similarly, T_regs promote pulmonary metastasis by blocking NK cell function (Olkhanud et al. 2009), with T_reg conversion occurring through the activity of B_regs (Olkhanud et al. 2011). T and B lymphocytes can also promote the protumorigenic properties of macrophages (Gunderson and Coussens 2013), including IL-4-producing CD4⁺ T cells enhancing the ability of macrophages to foster invasion (DeNardo et al. 2009).

INVASION AND INTRAVASATION

Cancer cells must acquire phenotypic changes that endow them with ability to invade surrounding stroma as a precursor to intravasation. These invasive characteristics include alterations in expression of adhesion molecules and expression of proteases to degrade the extracellular matrix (ECM). Each of these characteristics is fuelled by the presence of immune cells and can potentially cross-regulate each other. For example, EMT leads to the recruitment of macrophages, which can secrete and activate TGF-β, further supporting the induction of EMT (Dongre and Weinberg 2019). Similar feedback loops are required for intravasation, with tumor cells stimulating the chemotactic and angiogenic properties of macrophages (Lewis et al. 2016). Interestingly, recent studies have shown that cells egressing the primary site early during cancer progression are more metastasis-efficient than cells leaving the tumor at later stages (Hüsemann et al. 2008; Ghajar and Bissell 2016; Harper et al. 2016; Hosseini et al. 2016). The metastatic competency of early disseminated cells—which would harbor fewer mutations—reiterates that microenvironmental factors contributes to the acquisition of metastatic traits (Hendry et al. 2017; Linde et al. 2018). In support of this, although the molecular profiles of invasive breast cancer and preinvasive was indistinguishable, aggressive lesions displayed high infiltration of macrophages and T_regs (Abba et al. 2015; Nelson et al. 2018).

Epithelial-to-Mesenchymal Transition

An established TIME provides several cytokines and chemokines with the potential to activate latent EMT programming in epithelial cells, causing decreased E-cadherin expression, loss of cell–cell junction and, morphological changes that enable cells to invade and migrate to distant sites (Dongre and Weinberg 2019). This can be mediated through cooperation with the TGF-β signaling pathway, one of the primary induces of EMT (Lu et al. 2014; Su et al. 2014; Chockley and Keshamouni 2016), although independent mechanisms have also been described. For example, in a model of gastric cancer, macrophage production of CXCL1 and CXCL5 increase Snail transcription through activation of a CXCR2/STAT3 feed-forward loop (Zhou et al. 2019), whereas in a model of Her2⁺ breast cancer Tie2⁺ macrophages initiate EMT through expression of Wnt (Linde et al. 2018). This association in breast cancer drives dissemination in the absence of a palpable lesion (Linde et al. 2018), and thus may also explain the impact of macrophage ablation on the invasive phenotype in a model of Trp53 null breast cancer (Carron et al. 2017). Intriguingly, macrophage expression of IL-35 at the site of metastasis has recently been shown to promote mesenchymal-to-epithelial transition (Lee et al. 2018), highlighting the critical role of macrophages at multiple stages of the metastatic process.

Immature monocytes and neutrophils have also been shown to contribute to EMT through the production of TGF-β, epidermal growth factor (EGF), and hepatocyte growth factor (HGF), and these cells are found around the invasive edge of tumors in which markers of EMT are most prevalent (Toh et al. 2011; Sangaletti et al. 2016; Ouzounova et al. 2017; Wang et al. 2019), with neutrophil levels at the invasive margin of gastric cancer being an independent, negative predictor of disease-free survival (Li et al. 2019). Neutrophils have also been shown to contribute to EMT by creating abnormal vasculature in lung tumors, resulting in hypoxia and induction of a Snail-dependent EMT program (Faget et al. 2017). Given an analogous role for macrophages in breast cancer (Stockmann et al. 2008), it may be that macrophages promote EMT through the same mechanism.

Although it has not been extensively evaluated, there are also a few reports suggesting that T cells can activate EMT. In vitro, coculturing premalignant cells with CD4⁺ T cells decreased E-cadherin expression and elevated expression of mesenchymal proteins (Goebel et al. 2015; Chen et al. 2017b). Whether this is mediated by TGF-β production by T_regs is not clear (Li et al. 2007; Worthington et al. 2012). One study also observed an in vivo role for CD8⁺ T cells involving selection of Her2/neu escape variants (Santisteban et al. 2009). However, this may be an artifact of this particular model system, and a role for T cells in vessel normalization (Motz et al. 2014; Schmittnaegel et al. 2017; Tian et al. 2017) suggests these cells could actually reduce hypoxia and EMT in other settings.

Extracellular Matrix Remodeling

The ECM is a dynamic, acellular, three-dimensional structure that serves as scaffold, on-demand reservoir for growth factors, and ligand for cellular adhesion. The ECM undergoes a continuous process of remodeling that is tightly regulated by the MMP and cathepsin family of proteases (Bonnans et al. 2014). Not surprisingly, ECM remodeling is highly abnormal in tumors, with increased deposition of key components such as collagen and fibronectin, increased degradation and turnover, and an abnormal structure marked by higher density and tensile force (Northey et al. 2017). Navigating the ECM is thus critical for cancer cells to invade tissue and gain access to the vasculature.

Studies documenting a role for immune cells in ECM deposition are limited, but production of SPARC (secreted protein, acidic rich in cysteine) by macrophages leads to increases in collagen and fibronectin, along with higher levels of metastasis (Sangaletti et al. 2008). Inflammatory monocytes have also been shown to elicit FXIIIA-mediated fibrin cross-linking, which provides a scaffold for local tumor cell invasion (Porrello et al. 2018). Conversely, production of interferon (IFN)-γ by NK cells restricts metastasis by inducing expression of fibronectin (Glasner et al. 2018), highlighting the complex association of ECM location and structure with the invasive process. A number of stromal proteases are also known to promote local tumor invasion, including urokinase/plasminogen activator (uPA), MMP-9 and cathepsin B, S, or Z (Almholt et al. 2005; Vasiljeva et al. 2006; Yang et al. 2008; Gocheva et al. 2010; Yan et al. 2010; Bekes et al. 2011; Akkari et al. 2014). Most of these proteases are expressed primarily by tumor macrophages, with the exception of MMP-9, which is also expressed by neutrophils and mast cells (Coussens et al. 2000; Bekes et al. 2011). However, it is currently unclear whether these proteases act by clearing out the ECM to permit invasion, release factors that promote an invasive phenotype (e.g., active TGF-β), alter the structure of the ECM to facilitate the creation of “tracks,” or act through some combination thereof (Condeelis and Segall 2003).

Intravasation

Angiogenesis is a hallmark of cancer, and angiogenic factors such as VEGF, fibroblast growth factor (FGF), and placental-derived growth factor (PlGF) are produced by immune cells in the tumor, especially macrophages (Rivera and Bergers 2015), which are critical for the angiogenic switch during tumor development (Lin et al. 2006). Macrophage production of these factors does not necessarily regulate the extent of angiogenesis in late stage tumors, as VEGF overexpression can bypass a requirement for macrophages (Lin et al. 2007). However, macrophages retain their importance as regulators of vascular structure and permeability (Stockmann et al. 2008). Thus ablating macrophages production of VEGF increases pericyte coverage, improves tissue perfusion, and increases tumor growth (Stockmann et al. 2008), while simultaneously increasing susceptibility to chemotherapy (Stockmann et al. 2008; Hughes et al. 2015) and reducing intravsation and metastasis (Harney et al. 2015). This phenotype is driven largely by a population of Tie2⁺CXCR4⁺ macrophages that reside in the perivascular space around vessels (De Palma et al. 2005; Welford et al. 2011), allowing for local production of VEGF to increase vessel permeability and ease transendothelial migration (Harney et al. 2015). Targeting these cells by blocking the ligand for Tie2, angiopoietin (Ang)-2, thus also results in reduced metastasis (Mazzieri et al. 2011). Although not necessarily related to the metastatic cascade, it is worth noting that vessel normalization promotes T-cell infiltration, which then maintains vessel normalization by producing IFN-γ (Motz et al. 2014; Schmittnaegel et al. 2017; Tian et al. 2017).

Macrophages are also responsible for recruiting tumor cells to these sites of vascular permeability through a well-described paracrine loop involving epidermal growth factor (EGF). As mentioned, tumor cells are an important source of CSF-1, and this induces the production of EGF in macrophages, resulting in tumor cell chemotaxis toward the vasculature (Wyckoff et al. 2004, 2007; Wang et al. 2009; Ishihara et al. 2013). This process is not mediated by Tie2⁺ perivascular macrophages, but rather the recently recruited population of monocyte-derived macrophages that forms the dominant population within tumors (Arwert et al. 2018). After entering the tumor parenchyma, these macrophages up-regulate expression of CXCR4 and are slowly recruited back to the vasculature via CXCL12-expressing fibroblasts, creating an EGF chemotactic gradient and pathway of altered ECM that directs cancer cell migration to the vasculature (Arwert et al. 2018). Finally, endothelial cell interactions promote differentiation into the Tie2⁺CXCR4⁺ perivascular macrophage population that facilitates transendothelial migration. Thus, interfering with monocyte recruitment (CCR2-CCL2), survival and differentiation (CSF1R-CSF1), migration to the vasculature (CXCR4-CXCL12), or maturation to perivascular cells (Tie2-Ang2) functions to severely limit the ability of tumor cells to intravasate into the blood stream (Qian and Pollard 2010; Kitamura et al. 2015). As with vascular normalization, the ability of macrophages to promote tumor cell chemotaxis is controlled by infiltrating T cells, with IL-4-expressing CD4⁺ T cells augmenting EGF expression by macrophages (DeNardo et al. 2009).

SURVIVAL AND EXTRAVASATION

Most primary tumors release millions of cells into the blood stream, but only a small number of metastatic lesions usually develop, indicating the inefficiency of tumor cell dissemination (Nagrath et al. 2007). In particular, circulating tumor cells (CTCs) have to resist physical shear stress caused by blood flow, death by anoikis, and avoid recognition by cytotoxic lymphocytes and phagocytic macrophages (Fig. 2) (Mohme et al. 2017). With this limited capacity to survive within the vasculature, the capacity of CTCs to adhere to the vasculature and migrate into the extravascular space is also central to their metastatic potential (Reymond et al. 2013), processes that are linked to the recruitment of neutrophils and monocytes/macrophages.

Figure 2.

Immune evasion and assistance during cancer cell dissemination. During circulation within the intravascular space, major histocompatibility complex (MHC) I expression protects by circulating tumor cells (CTCs) from natural killer (NK) cell recognition (1), whereas down-regulation of MHC I and up-regulation of NK cell activating ligands can lead to cytotoxicity (2). Conversely, recognition can be antagonized via up-regulation of NK cell inhibitory ligands (3) or physical shielding using platelet/fibrin coagulates (4). To extravasate into the tissue, CTCs interact with endothelial cells via selectins, cadherins, integrins, CD44, junctional adhesion molecules (5). Neutrophils are able to assist this process via ICAM-1 and VCAM-1 interactions, as well as through the release of NETs that trap CTCs and hide the cells from lympohcytes (6). Recruitment of CCR2⁺Ly6C⁺ monocytes (7) results in local production of VEGF, thereby increasing vascular permeability to facilitate CTC extravasation (8).

Avoiding Cytotoxic Cells

Increased NK cell cytotoxicity or enhanced expression of NK cell activation receptors are associated with good prognosis for patients at risk of metastatic disease (López-Soto et al. 2017). For example, the infiltration of NK cells expressing high levels of the activating receptors NCR1/NKp46 (natural cytotoxicity triggering receptor 1) and NKG2D (killer cell lectin-like receptor K1) was positively associated with recurrence-free survival in prostate cancer following surgical resection (Pasero et al. 2016). Several studies have shown that NKp46 expression in particular is critical for NK cell recognition of metastatic cells (Halfteck et al. 2009; Lakshmikanth et al. 2009; Elboim et al. 2010; Glasner et al. 2012) and loss of NK cells increases metastasis (Smyth et al. 1999; Bidwell et al. 2012). Although it is difficult to distinguish between a role for NK cells within the primary tumor, circulation, or metastatic sites, at least some experimental evidence exists for the direct killing of CTCs (Palumbo et al. 2005; Hanna et al. 2015), and poor infiltration by NK cells in the absence of treatment likely limits their role in most primary tumors (Böttcher et al. 2018).

Not surprisingly, tumors use multiple mechanisms to avoid recognition by NK cells (Box 1). First and foremost, down-regulated protein or surface expression of NK activating ligands, especially MHC I polypeptide-related sequence A (MICA), and MICB, can reduce NK cell recognition and killing of tumor cells (Nausch and Cerwenka 2008; Wang et al. 2014). Conversely, tumor cells can secrete soluble versions of these same molecules to suppress NK cell immunosurveillance (Schlecker et al. 2014; Zhang et al. 2015). This occurs within the primary tumor, but is sufficient to drive a systemic reduction in the cognate receptors on NK cells (Pasero et al. 2016), and high levels of these soluble ligands in the circulation correlate with disease progression (Paschen et al. 2009; Yamaguchi et al. 2012). That said, a high-affinity NKG2D ligand (MULT1) has also been shown to stimulate NK cell-mediated killing of tumor cells after shedding (Deng et al. 2015), so there may be a cost to this approach in some instances. Second, although HLA down-regulation is a common mechanism of evading T-cell recognition (Campoli and Ferrone 2008), HLA expression is never completely lost, likely because of the selective pressure of NK cells. Instead, HLA loss of heterozygosity is commonly observed as an alternative to loss of neoantigen expression, allowing tumors to limit detection by both of the cytotoxic lymphocyte populations simultaneously (McGranahan et al. 2017; Rosenthal et al. 2019).

BOX 1. NONCLASSICAL MHC SIGNALING.

Histocompatibility antigen class I G (HLA-G) is a human leukocyte antigen that belongs to the “nonclassical” HLA-class Ib molecules (Choo 2007). HLA-Ib molecules have minimal polymorphisms compared with the HLA-Ia molecules and are involved in immune modulation rather than antigen presentation (Kievits et al. 1987; Braud et al. 1997). HLA-G has been described to contribute to immune evasion in cancer (Lin and Yan 2015) and expression correlates with poor patient survival in melanoma, colorectal, hepatocellular, breast, glioma, ovarian, and lung cancers (Cai et al. 2009; de Kruijf et al. 2010; He et al. 2010; Guo et al. 2015). Moreover, soluble HLA-G is secreted by the primary tumor were identified and negatively correlate with patient outcome (Contini et al. 2003; Zheng et al. 2014). HLA-G can bind to a variety of receptors, such as CD8, leukocyte immunoglobulin-like receptor subfamily B member 1 (LIR-1), and killer cell immunoglobulin-like receptor (KIR), all of which protect cancer cells from CD8 T cell- or NK cell-induced killing (Wiendl et al. 2002; Kochan et al. 2013; Loumagne et al. 2014; Lin and Yan 2015). HLA-G polymorphisms also impact graft-versus-tumor responses in renal cell carcinoma (Crocchiolo et al. 2018) and correlate with outcome in epithelial ovarian carcinoma (Schwich et al. 2019).

There is also a complex interaction between tumor cells, platelets, and NK cells that regulates cytoxicity in the blood. Tumor cells become coated with platelets on entering the circulation, and this can promote survival, EMT, and adhesion to the endothelium (Gay and Felding-Habermann 2011; Labelle et al. 2011). This also has the effect of physically masking CTCs from NK cells (Camerer et al. 2004; Palumbo et al. 2005) and/or transferring platelet-derived membrane vesicles containing MHC I molecules that act to “normalize” the tumor cells (Placke et al. 2012). Additionally, platelet-derived TGF-β has been shown to induce down-regulation of NKG2D to prevent NK cell reactivity (Kopp et al. 2009). Cumulatively, platelets and their associated factors promote metastatic spread, a phenotype that is dependent on the presence of NK cells in most models (Camerer et al. 2004; Palumbo et al. 2005; Palumbo 2008; Turpin et al. 2014; Adams et al. 2015).

Avoiding Phagocytosis

CD47 is a surface protein belonging to the immunoglobulin superfamily that regulates a broad range of cellular functions, including adhesion, migration, proliferation, apoptosis and—most relevant for metastasis—recognition by phagocytes via signal regulatory protein α (SIRPα) (Reinhold et al. 1995; Jiang et al. 1999). SIRPα is expressed on monocytes, macrophages, and CD11b⁺ classical dendritic cells (cDCs), and on binding to CD47 the phagocytic activity of these cells is blocked, thereby acting as a “don't eat me” signal (Jaiswal et al. 2009). CD47 is ubiquitously expressed on mouse and human myeloid leukemias, thereby allowing them to evade clearance and populate organs with an abundance of macrophages (Jaiswal et al. 2009) and hinting that the same process may allow for CTCs to colonize ectopic organs.

Certainly, CD47 is overexpressed in many solid tumors (Liu et al. 2017), and it can protect cancer stem cells from elimination during conventional antitumor therapies, which in turn increases the probability of recurrence (Soltanian and Matin 2011). Gene expression profiling of primary tumors and CTCs from patients with colorectal cancers also found pronounced expression of CD47 (Steinert et al. 2014), whereas the presence of CD47⁺MET⁺CTCs in breast cancer patients was associated with increased metastasis and decreased survival (Baccelli et al. 2013, 2014). Taken together, the evidence favors CD47 expression reducing clearance by monocytes/macrophages. The question remains whether this clearance is relevant within the intravascular or extravascular space. It seems unlikely that monocytes and CTCs are interacting while they are both in circulation. However, nonclassical Ly6C⁻ monocytes actively patrol the endothelial lumen (Auffray et al. 2007; Carlin et al. 2013) and rapidly take up cellular debris (Headley et al. 2016), which induces chemokine expression and recruitment of NK cells to limit metastasis (Hanna et al. 2015). Whether these patrolling monocytes also phagocytose live cells is unclear, and it may be that phagocytosis in the extravascular space or tissue parenchyma plays a more critical role in restricting metastasis.

Adhesion to the Endothelium

Circulating neutrophils correlate with poor prognosis (Teramukai et al. 2009; Lee et al. 2012; Gondo et al. 2013), and although this may partially reflect altered hematopoiesis and the establishment of an immunosuppressive microenvironment (Coffelt et al. 2015), it has also become increasingly clear that circulating neutrophils promote the conversion of CTCs to disseminated tumor cells (DTCs) through two distinct mechanisms. The first involves binding between integrins expressed by neutrophils (i.e., CD11b/CD18, VLA-4) and immunoglobulin superfamily ligands expressed by endothelial cells and CTCs (ICAM-1, VCAM-1). This allows neutrophils to form a cellular bridge for CTCs to adhere to the endothelium (Box 2) (Huh et al. 2010; Spicer et al. 2012). These interactions also induce “outside-in” signaling that promotes the proliferative capacity of CTCs, enhancing their metastatic potential (Szczerba et al. 2019). As the majority of CTCs are associated with neutrophils—at least in breast cancer patients—this appears to be an important mechanism underlying metastatic spread (Szczerba et al. 2019).

BOX 2. TUMOR–ENDOTHELIAL CELL INTERACTIONS.

CTCs becoming trapped in the capillaries is thought to be one of the major mechanisms of arrest in the circulation, but the process of adhesion to the endothelial layer and migration into the tissue is an active process (Chambers et al. 2002; Wirz et al. 2008; Strilic and Offermanns 2017). The vascular endothelium differs structurally depending on the organ, and as a result, leukocyte extravasation in lung and liver occurs in the microvasculature, whereas it occurs in the postcapillary venules in the skin, muscle, and mesentery (Aird 2007; Strell and Entschladen 2008). However, although the processes of adhesion and transmigration (diapedesis) are at least partially shared between leukocytes and tumor cells (Miles et al. 2008), the molecules described for tumor cell transmigration vary and are often tumor-specific (Bendas and Borsig 2012; Strilic and Offermanns 2017). However, it is still not entirely clear the degree to which all of these molecules are involved in tumor cell adhesion/transmigration in vivo, or whether different tumor subclones may use unique strategies to colonize different tissues (Caswell and Swanton 2017).

The other major mechanism by which neutrophils promote metastasis is through the release of neutrophil extracellular traps (NETs) containing their nuclear DNA (Park et al. 2016). Originally described to trap CTCs during injury or infection (Tohme et al. 2016), it has also been shown that release of NETs can be induced by the presence of CTCs in the lungs, leading to an increase in micrometastasis (Cools-Lartigue et al. 2013; Najmeh et al. 2017). Beyond simply sequestering CTCs, NETs inhibit NK cell-mediated cytotoxicity, and the release of factors such as IL-1B, MMP-9, and HMGB1 promotes extravasation (Spiegel et al. 2016; Tohme et al. 2016). Importantly, the formation of NETs in postoperative patients was linked to a reduction in survival, highlighting the potential clinical relevance of this mechanism (Tohme et al. 2016). Given the prevalence of CTC-neutrophil clusters in circulation (Szczerba et al. 2019) and the ability of tumor cells to induce the release of NETs¹⁰⁹, it will be interesting to determine if this alters the dynamics of CTC retention and/or allows CTCs to create their own protective shell.

Extravasation

Ly6C⁺CCR2^HI monocytes are recruited to the site of CTC arrest via a classical multistep adhesion cascade involving selectins, chemokines, and integrin ligands (Qian et al. 2011; Ferjančič et al. 2013; Häuselmann et al. 2016). In addition to CCL2 emanating from the tissue, cancer cell production of CCL2 is necessary for this process (Qian et al. 2011; Häuselmann et al. 2016), suggesting that it initiates the inflammatory cascade. Reminiscent of the role of macrophages in tumor cell intravasation, monocytes recruited to the endothelial lumen via CCL2 secrete VEGF and increase endothelial permeability, thereby enhancing transmigration of tumor cells (Qian et al. 2011). Interfering with the CCR2/CCL2 chemokine axis thereby reduces liver and lung metastasis in multiple murine model systems (Qian et al. 2011; Sanford et al. 2013), provided that the mice are left on treatment for the duration of the study (Bonapace et al. 2014). Platelet-bound CTCs can also recruit monocytes to promote metastasis, but the degree to which this process involves CCL2 and/or VEGF is unknown (Gil-Bernabe et al. 2012).

ECTOPIC GROWTH

Outgrowth of metastatic colonies represents the final stage of disease and is responsible for the majority of cancer-related deaths, and unfortunately, there is evidence that most patients have DTCs at the time of diagnosis (Friberg and Nystrom 2015; Massagué and Obenauf 2016). These DTCs are largely in a dormant state and thus resistant to conventional cytotoxic chemotherapies, resulting in metastatic recurrence for many patients long after the primary tumor has been removed. Addressing the issue of DTCs, as well as the treatment of overt metastasis, are therefore the two most critical areas of cancer research (Sosa et al. 2014; Massagué and Obenauf 2016). Thankfully, it is becoming clear that modulating the immune response has the potential to address both of these issues, with the approval of immune checkpoint blockade agents for metastatic disease being the most significant advancement.

Formation of the Premetastatic Niche

The premetastatic niche was originally described as VEGFR1-expressing myeloid cells accumulating at sites of distant metastasis before the arrival of tumor cells (Kaplan et al. 2005), and is usually defined by the recruitment of macrophages and neutrophils (Fig. 3) (Liu and Cao 2016). The formation of the niche is driven by a range of tumor and stroma-derived soluble factors (e.g., G-CSF, VEGF, CCL2, TNF-α, TGF-β, S100A8/A9) in addition to the release of tumor-derived extracellular vesicles (usually containing microRNAs [mRNAs]) (Liu and Cao 2016). Regardless of the molecules identified or the mechanism proposed, each appears to act primarily by inducing the recruitment of myeloid cells. For example, S100A8/A9 promotes the release of serum amyloid A3, which activates an inflammatory program in macrophages via toll-like receptor (TLR)4 (Hiratsuka et al. 2008). These cells then promote the ability of DTCs to become established by generating an immunosuppressive and/or permissive microenvironment, as we will discuss below. It is unclear why so many divergent approaches are used to achieve a similar goal, but presumably there are subtleties in the activation state of the recruited myeloid cells, as well as alternations in the functions of other stromal cells that play a complementary and supportive role in establishing the niche (McAllister and Weinberg 2014; Peinado et al. 2017).

Figure 3.

Establishment of the metastatic immune microenvironment: The primary tumor can actively prime organs by forming a metastasis-permissive microenvironment before the arrival of disseminated tumor cells (DTCs) (1). This premetastatic niche includes recruited monocytes, macrophages, and neutrophils, as well as tumor-secreted factors such as cytokines, chemokines, growth factors, and proteases. Enhanced vascular leakiness, reorganization of the stroma and extracellular matrix, and the establishment of an immunosuppressive environment are hallmarks of the metastatic microenvironment (2). These allow DTCs to evade natural killer (NK) cell immunity and colonize the ectopic organ as single cells or micrometastatic lesions. Micrometastatic lesions often enter a period of dormancy, as they are ill-adapted to the new microenvironment, encounter growth suppressive factors such as TGF-β, and are subjected to NK cell-mediated immune surveillance. Outgrowth involves tumor-intrinsic changes alongside the induction of angiogenesis and the creation of a mature immunosuppressive microenvironment to evade recognition by cytotoxic lymphocytes (3).

Escaping Dormancy

There is only limited evidence that immune cells can promote the survival and growth of DTCs (Qian and Pollard 2010), with macrophages shown to provide survival signals via binding to VCAM-1 (Chen et al. 2011) and neutrophils enhancing tumorigenicity via leukotriene production (Wculek and Malanchi 2015). However, macrophages in primary tumors promote therapeutic resistance by producing factors such as TNF-α and IL-6 (Ruffell and Coussens 2015), which activate signaling pathways known to promote the proliferation of dormant cells (Massagué and Obenauf 2016). Thus, it is likely that these and other pathways are important at the metastatic site, and that our limited knowledge in this area reflects experimental difficulties evaluating the early stages of colonization. This includes the use of murine cancer cell lines that do not display dormancy or extended periods of latency, and evaluation of metastasis primarily to the lung even though there is evidence of a tissue-specific role for critical molecules such as TGF-β (Massagué and Obenauf 2016). Indeed, using a model of metastatic dormancy, it was recently shown that NETs promote metastasis via release of neutrophil elastase and MMP9, resulting in cleavage of the ECM component laminin (Albrengues et al. 2018). Rather than activating myeloid cells via TLRs, as has been described for versican in the premetastatic niche (Kim et al. 2009), cleaved laminin triggered tumor cell proliferation via integrin activation and FAK signaling (Albrengues et al. 2018).

Other studies that have addressed these experimental issues have noted a clear role for lymphocytes in maintaining tumor dormancy. CD8⁺ T-cell-depletion dramatically increased the rate of metastasis in a model of spontaneous uveal melanoma (Eyles et al. 2010) and following the surgical resection of implanted fibrosarcoma (Romero et al. 2014). Interestingly, although CD8⁺ T-cell-depletion resulted in a 100% rate of metastasis, NK cell-depletion led to a rate of 87%, and CD4⁺ T-cell-depletion resulted in a rate of 23% (Romero et al. 2014). This may reflect a complex interplay between these cell types in maintaining dormancy, or simply the production of a common set of cytokines such as IFN-γ and TNF-α. Certainly, the incidence of primary tumors is suppressed by these cytokines (Müller-Hermelink et al. 2008), either through cytostatic/cytotoxic effects on carcinoma cells or via suppression of angiogenic pathways (Müller-Hermelink et al. 2008; Andreu et al. 2010). Critically, a functional immune system prevents disease recurrence in humans, as seen from the unfortunate development of metastasis following kidney transplantations from donors with a previous history of melanoma or lung cancer (MacKie et al. 2003; Xiao et al. 2013).

Evading Immunity

Leaving the immunosuppressive microenvironment of the primary tumor exposes metastatic cells to recognition by cytotoxic lymphocytes, as seen from the ability of NK cells to protect against experimental and spontaneous metastasis but not primary tumor growth (Smyth et al. 1999, 2001; Takeda et al. 2001; Olkhanud et al. 2009; Milsom et al. 2013; Paolino et al. 2014). Thus, it is not surprising that successful metastasis is linked to the formation of a myeloid-rich environment (de Mingo Pulido and Ruffell 2016). What remains largely unexplored is the degree to which immunosuppressive mechanisms used by metastatic cells recapitulate those found in the primary tumor, represent the unique biology of the ectopic organ, and/or reflect an incipient microenvironment. For example, in a model of lobular breast cancer, primary tumors drive neutrophil expansion and an immunosuppressive phenotype via initiation of an IL-1β/IL-17/G-CSF cascade, which acts to blunt CD8⁺ T-cell-dependent rejection of early metastatic lesions (Coffelt et al. 2015). Interfering with any component of this pathway reversed this immune suppression, but had no impact on any measured parameter within primary tumors. This discrepancy may relate to the size of the lesion, as the complexity of an advanced microenvironment would make it resistant to perturbations in a single pathway and late stage metastatic lesions were not impacted by neutrophil depletion (Coffelt et al. 2015). However, it has also been described that macrophages in primary tumors indirectly suppress a CD8⁺ T-cell-response through production of IL-10, whereas their ability to limit T-cell activity in metastatic tumors occurs through a different mechanism of action (Ruffell et al. 2014), such as the production of reactive oxygen species or expression of CTLA4 ligands (Kitamura et al. 2018). Conversely, the clinical success of immune checkpoint blockade in metastatic and primary tumors, including in relatively immune privileged sites, highlights the importance of tumor immunogenicity (Hodi et al. 2014; Topalian et al. 2014; Overman et al. 2018). Thus, intrinsic, extrinsic, and temporal factors appear to regulate the vulnerability of metastatic lesions to immune surveillance, and understanding these factors should assist in the development of immunotherapies for patients with metastatic disease.

SUMMARY

The mechanisms underlying early steps of the metastatic process have been relatively well studied compared with colonization and immune evasion, despite early dissemination of tumor cells making these later steps more clinically relevant. The critical role for T cell- and NK cell-mediated surveillance in restraining the growth of metastatic lesions represents a unique therapeutic opportunity as agents that promote the activity of these cells become available. In addition to approved antagonist antibodies against CTLA-4 and PD-1/PD-L1, agents blocking LAG-3, TIGIT, and TIM-3 are in early phase clinical trials (Anderson et al. 2016), whereas several new targets have been identified, including those that unleash NK cell activity (André et al. 2018). Combining these agents with those that block the inhibitory function of myeloid cells, or activate their immunostimulatory potential, have already shown potential in murine models of primary tumors (DeNardo and Ruffell 2019).

Beyond the treatment of overt metastatic lesions, the role of lymphocytes in maintaining dormancy suggests the possibility of developing immunotherapies in an adjuvant setting. Although long-term treatment is difficult and expensive, the potential benefit of preventing recurrence would be immense, and the use of a minimal effective dose could improve the viability of this approach. Alternatively, understanding how the immune system promotes escape from dormancy could allow for the development of therapies targeting systemic inflammation. Recent technological developments including CRISPR/Cas9 screens, single cell genomics, and mass cytometry will allow progress to be made in this area, and dissecting how life events such as inflammatory diseases, infections, and injury relate to recurrence-free survival should allow findings in murine models to be correlated with patient data.