Metastasis organotropism: redefining the congenial soil

Abstract

Metastasis is the most devastating stage of cancer progression and causes the majority of cancer-related deaths. Clinical observations suggest that most cancers metastasize to specific organs, a process known as “organotropism”. Elucidating the underlying mechanisms may help identify targets and treatment strategies to benefit patients. This review summarizes recent findings on tumor intrinsic properties and their interaction with unique features of host organs, which together determine organ-specific metastatic behaviors. Emerging insights related to the roles of metabolic changes, the immune landscapes of target organs and variation in epithelialmesenchymal transitions open avenues for future studies of metastasis organotropism.

Zhang ETOC

Clinical observations suggest that most cancers metastasize to specific organs, a process known as “organotropism”. This review from Gao et al. highlights recent findings on tumor intrinsic properties and their interaction with unique cellular, architectural, metabolic and immune features of host organs, which together determine organ-specific metastatic behaviors.

Introduction

Metastasis is a process by which cancer cells are dispersed from their primary site of tumorigenesis and disseminated to a different part of the body. It remains the major cause of morbidity and mortality in cancer patients (Chaffer and Weinberg, 2011). Before colonization at secondary sites, cancer cells generally undergo a complicated cascade, including invasion to surrounding tissue, intravasation to the blood vessels, survival in the circulation, and extravasation to colonize and thrive at distant sites (Obenauf and Massagué, 2015).

Metastasis follows a non-random distribution among distant organs, known as “organotropism” or “organ-specific metastasis”. Different cancer types and subtypes display distinct organotropisms. For example, prostate cancer preferably relapses in bone while uveal melanoma typically colonizes in liver (Nguyen et al., 2009). Breast cancer can metastasize to different sites, including bone, lung, liver, and brain. However, the luminal subtype has a higher propensity to metastasize to the bone, whereas metastases of triple-negative breast cancer (TNBC) prefer visceral organs (Chen et al., 2018; Wu et al., 2017). Accumulating evidence suggests that organotropism is regulated by multiple factors, including the circulation pattern, tumor-intrinsic factors, organ-specific niches, and the interaction between tumor cells and the host microenvironment (ME). In this review, we summarize recently emerging concepts and mechanisms underlying organ-specific metastasis.

General metastasis mechanisms

Metastatic tumors largely rely on the same driver mutations found in primary tumors (Zehir et al., 2017), suggesting that the hallmark functions for tumor maintenance and progression remain critical in metastases. Regardless of their final destination, the early steps in the cancer metastasis cascade are generally similar among different cancers. For example, the epithelialmesenchymal transition (EMT) program is thought to play a central role in the departure of cancer cells from primary tumors (Lambert et al., 2017). EMT refers to the loss of epithelial features and acquisition of mesenchymal properties, which is of paramount importance for preparing carcinoma cells to invade the surrounding parenchyma and intravasate to enter the bloodstream. Many EMT-inducing transcription factors (EMT-TFs), including Snail, Slug, Twist, and Zeb1 coordinately regulate this critical process. EMT has been shown to participate in almost all the aspects of tumor dissemination, although some recent studies revealed more complicated dynamics between EMT and metastasis (reviewed in Lambert et al., 2017; Yeung and Yang, 2017). After departure from the primary tumor, cancer cells traveling in the circulation are designated as circulating tumor cells (CTCs). CTCs can disseminate either as single cells or clusters. In order to survive in the bloodstream, CTCs enlist platelets and leukocytes, particularly neutrophils, to evade immunosurveillance (reviewed in Lambert et al., 2017; Riggi et al., 2018). Upon arrival at secondary sites, cancer cells may remain dormant to facilitate adjustment to the new niche environment. Disseminated cancer cells (DTCs) are thought to retain stem cell properties which may be required to re-initiate tumor growth in distant organs (reviewed in Lambert et al., 2017; Oskarsson et al., 2014).

Mechanisms underlying organ-tropism of metastasis

Classic “seed and soil” mechanisms

Stephen Paget stipulated that both cancer cell-intrinsic properties (“seed”) and the congenial ME (“soil”) are essential for metastasis formation (Paget, 1889). Research in the past a few decades has greatly enhanced our understanding of the molecular and cellular nature of both “seed” and “soil”. In this section, we will review some mechanisms of organ-specific metastasis focusing on the organ-specific ME.

Bone tropism

Cancers disseminate to bone with different frequencies (Table 1). Breast and prostate cancers are the principal cancers that metastasize to bone (Budczies et al., 2015; DiSibio and French, 2008). Bone and bone marrow comprise unique cell types including osteoblasts, osteocytes, and osteoclasts. Osteoclasts resorb the bone matrix, while osteoblasts refill osteolytic cavities with new bone deposition to either mature into lining cells or to become embedded in the bone matrix to form osteocytes. DTCs can hijack osteoblasts activity and promote osteoclastogenesis which leads to increased bone resorption (Thomas et al., 1999). This process releases numerous factors from the bone matrix, including calcium, collagens, glycoproteins, hyaluronans, proteoglycans, growth factors, proteinases and cytokines (Casimiro et al., 2009) which boost the proliferation of tumor cells, thereby creating a vicious cycle between osteoblasts, osteoclasts and tumor (Figure 1) (Celià-Terrassa and Kang, 2018; Guise, 2002). Therapeutic interventions targeting the vicious cycle, or more specifically the activation of osteoclasts, exhibited clinical benefit in treating bone metastases. The approved drugs include bisphosphonates (Fulfaro et al., 1998), which induce osteoclast apoptosis, and denosumab (Ford et al., 2013), which is a RANKL antibody preventing osteoclast maturation. These drugs can significantly strengthen bones and delay tumor progression, although their effects on overall survival remain questionable (Croucher et al., 2016). Therefore, more effective treatments are still urgently needed. In recent years, multiple lines of studies further substantiated this paradigm and also began to reveal early-stage events prior to the onset of the vicious cycle (Eyob et al., 2013; Korpal et al., 2009; Lu et al., 2011; Ross et al., 2017; Sethi et al., 2011; Waning et al., 2015; Zhang et al., 2019; Zheng et al., 2017). These studies provided additional therapeutic targets.

Table 1:

Incidence of metastasis to different organs at autopsy (%)

Cancer╲Organ	Bone	Lung	Liver	Brain	Peritoneum	Reference
Breast	71	71	62	22	NA	(Lee, 1985)
Prostate	90.1	45.7	25	1.6	7	(Bubendorf et al., 2000)
Lung	34	-	21	39	NA	(Riihimäki et al., 2014)
Melanoma	48.6	71.3	58.3	54.6	42.6	(Patel et al., 2004)
Pancreas	25	55	62	NA	NA	(Kamisawa et al., 1995)
Renal	44.5	74	34.5	NA	NA	(Johnsen and Hellsten, 1997)
Thyroid	13	78	20	18	13	(Besic and Gazic, 2013)
Gastric	12	15	48	3	32	(Riihimäki et al., 2016)
Colon	8	32	70	5	21	(Riihimaki et al., 2016)
Liver	8	44	-	1	9	(Lee and Geer, 1987)
Ovarian	11.2	33.9	47.9	3	83.6	(Rose et al., 1989)

Figure 1. Bone-specific metastasis.

The bone microenvironment (ME) secretome generated by osteoblasts, osteoclasts or other cells may promote bone metastasis, while tumor cells can produce factors such as LOX to induce pre-metastatic niche formation. Interactions between tumor cells and osteoblasts through adherens junctions (via E-cadherin/N-cadherin, JAG1/Notch) and gap junctions (via CX43) also facilitate bone metastasis. Colonizing tumor cells express osteoblast-specific markers such as ALP and RUNX2 to escape immunosurveillance. In addition, tumor cells secrete factors promoting bone turnover to induce osteolysis, which in turn produces factors to stimulate tumor growth, creating a “vicious cycle”.

Molecular pathways involved in bone metastasis

Multiple mechanisms have been proposed in bone metastasis (Supplemental table 1) and involve both tumor intrinsic and extrinsic factors. Some mechanisms (e.g., CCXL12/CXCR4-mediated chemotaxis) are common among different cancer types (e.g., Supplemental table 1; Taichman et al., 2002, Shiozawa et al., 2011; Zhang et al., 2009 and 2013), whereas many other mechanisms appear to be cancer type- or even subtype-specific. The expression of estrogen receptor (ER) defines the largest subtype of breast cancer (ER+), which exhibited a stronger bone-tropism as compared to ER- subtype (Kennecke et al., 2010). ER+ breast cancer appears to adopt different mechanisms for bone colonization (Supplemental table 1; ER+ oriented) as compared to ER- (Supplemental table 1: TNBC and HER2+ oriented). Androgen receptor (AR) is a major driver of prostate cancer, and it has long been thought that development of castration resistance (i.e. independence of AR signaling) is associated with bone metastasis. However, recent systematic studies identified three clusters of prostate cancer (PCS1, PCS2, and PCS3) with PCS2 displaying higher androgen receptor (AR) signaling and bone metastasis, indicating an unexpected role of AR in bone metastasis (Thysell et al., 2017; You et al., 2016). Interestingly, a recent study found that TMPRSS2-ERG gene fusions can promote osteoblastic bone metastasis in prostate cancer (Delliaux et al., 2018), indicating that specific mutations drive bone metastasis. In lung cancer, several studies suggest the requirement of epithelial markers such as CD24, discoidin domain receptor-1 (DDR1), and melanoma cell adhesion molecule (MCAM) for bone metastasis (Supplemental table 1; lung cancer). Limited studies for other cancers including melanoma, myeloma, bladder cancer have been reported for bone metastasis (Supplemental table 1).

Metastatic niches in bone

The most studied bone niches include the hematopoietic stem cell (HSC), osteoblastic, vascular endothelial, and neural niches (Calvi et al., 2003; Ding and Morrison, 2013; Katayama et al., 2006). These niches are essential for normal bone development and maintenance. It is increasingly apparent that various niches sustain different stages of cancer metastasis (Celià-Terrassa and Kang, 2018; Ren et al., 2015). For example, the high vascularization of the bone may contribute to cancer progression and dissemination. Indeed, several studies indicated that the perivascular niche maintains metastatic dormancy through cancer-endothelium interactions (Ghajar et al., 2013; Price et al., 2016). The osteogenic niche, on the other hand, was found to promote metastasis progression (Wang et al., 2015). In prostate cancer, the tumor-promoting function of osteoblasts was associated with the HSC niche, suggesting a direct competition between cancer cells and HSCs (Shiozawa et al., 2011). These findings are intriguing, but much remain to be learned. Both endothelial cells and osteogenic cells are highly heterogeneous in bone and bone marrow (Bussard et al., 2008; Yu and Scadden, 2016), and exhibit interactions that are temporally and spatially dynamic. For instance, the type H endothelium (CD31^high/Endomucin^high sinusoidal vessels enriched in growth plates of long bones) maintains perivascular osteoprogenitors, coupling osteogenesis and angiogenesis (Kusumbe et al., 2014). A recent study, employing continuous micro-endoscopic multiphoton imaging over several months, provides evidence for an exceptionally dynamic vasculature in bone marrow (Reismann et al., 2017). These results suggest potential overlap of or inter-conversion between different niches, which may affect the cellular fates of bone metastatic seeds. Precise mapping of heterogeneous cancer, endothelial and osteogenic cells will be required to delineate the co-evolution of different niches and the consequent impact on metastasis progression.

Osteomimicry

The selection of bone ME may drive the development of “osteomimcry”, i.e., metastatic cancer cells may evolve to resemble bone cells. Both breast and prostate cancers can express osteoblast-specific markers including alkaline phosphatase (ALP) and Runt-related transcription factor 2 (Runx2), as well as other factors involved in bone turnover including osteoprotegerin, PTH-related peptide (PTHrP), RANKL, and macrophage colony-stimulating factor (M-CSF) (Rucci, 2010). Similarly, tumors can also express bone matrix proteins such as osteocalcin, sialoprotein, osteopontin, and osteonectin to mimic osteoblast activity (Huang et al., 2005; Rucci, 2010), enabling cancer cells to directly foster osteoclast maturation without osteoblasts. This may be essential in advanced osteolytic metastasis with a decreased osteogenic population. This process can be altered by microRNAs (miRNAs) such as miR218, which regulate the expression of osteomimetic genes in breast cancers with metastatic properties and high Wnt signaling (Hassan et al., 2012). Osteomimicry can be mediated by osteomimetic genes (Knerr et al., 2004). For instance, endothelin-1 was shown to be a bone-induced factor that drives osteomimicry in breast cancer (Bendinelli et al., 2014). More molecular and cellular mechanisms are reviewed in-depth elsewhere (Rucci, 2010).

Seed pre-selection and pre-metastatic niche

Bone-tropism may already arise in primary tumors. Previous studies suggest that cancer-associated fibroblasts contribute to creating a cytokine environment resembling the bone marrow, thereby selecting cancer cells that are more fit to colonize the bone ME even before dissemination. This is termed “seed pre-selection” and may explain why gene expression profiles of primary tumors can be used to predict bone metastasis (Zhang et al., 2009, 2013). Even before their arrival, tumor cells can induce the formation of a “pre-metastatic niche,” a supportive ME in distant organs that is conducive to their survival, attachment, invasion, immune evasion, and outgrowth (Peinado et al., 2017). Hypoxic tumors, for example, can recruit bone marrow cells to pre-metastatic sites through secretion of lysyl oxidase (LOX) (Erler et al., 2009). Other factors, including miRNAs, can also promote pre-metastasis niche formation. For example, cancer-derived miR25–3p exosomes were found to promote pre-metastatic niche formation through recruitment of hematopoietic progenitor cells (HPCs) and induction of vessel permeability and angiogenesis (Zeng et al., 2018).

Liver tropism

Liver is one of the favored distant metastatic sites for solid tumors such as breast cancer, lung cancer and gastrointestinal cancers (Table 1). It receives a dual blood supply from the hepatic portal vein and hepatic arteries and has a much lower sinusoid blood pressure gradient (Kumar et al., 2008; MacPhee et al., 1995). This unique architectural feature allows CTC access and facilitates their attachment to the sinusoidal endothelium for seeding. For example, the blood circulation of the colon and proximal rectum is drained through the hepatic portal system, while the blood of the distal rectum goes to the lung. This vascular organization correlates with the fact that colorectal cancer prefers liver metastasis with lung as the second favored metastatic site (Riihimaki et al., 2016). Indeed, a greater number of colorectal CTCs are trapped in the liver than the peripheral blood (Denève et al., 2013). Furthermore, the endothelial layer of the liver sinusoid is fenestrated, which may be more permissive to extravasation, as compared to the well-organized endothelial wall and basement membrane in other organs (Figure 2) (Nguyen et al., 2009).

Figure 2. Liver-specific metastasis.

Hepatocytes directly interact with tumor cell enriched in claudin-2 to promote liver metastasis. They also promote the formation of a pro-metastatic niche though secretion of serum amyloid A1 and A2 (SAA). Hepatic stellate cells can produce PDGF, HGF, and TGFβ to induce liver metastasis. Tumor cells secrete exosomes, which are taken up by Kupffer cells. Integrin αvβ5 enriched exosomes stimulate Kupffer cells to produce pro-inflammatory S100A8, whereas MIF enriched exosomes trigger Kupffer cells to secrete TGFβ, which activates hepatic stellate cells, inducing liver-specific metastasis. LSECtin generated by sinusoidal endothelial cells also facilitates metastasis by inhibiting the T cell immune response.

Genes and pathways specifically implicated in homing and colonization to liver

Systematic studies have identified tumor intrinsic factors favoring liver organotropism. By transcriptional profiling of breast cancer metastases, Kimbung et al. identified a 17-gene liver metastasis–selective signature (Kimbung et al., 2016). Of note, the majority of these genes are ECM genes involved in cadherin and integrin signaling pathways. In addition, citrullination of the ECM by colorectal cancer cells-derived peptidylarginine deiminase 4 (PAD4) is essential for the growth of liver metastasis, consistent with the finding that inhibition of PAD4 altered EMT markers and diminished metastasis (Supplemental table 2). Furthermore, the direct binding and interaction between cancer cells and hepatocytes also play a role in liver tropism (Mook et al., 2003; Tabaries et al., 2012). For instance, claudin-2 is prevalent in breast cancer liver metastases but not in bone or lung metastases, and is essential for cancer cell-hepatocyte interactions and liver metastasis (Supplemental table 2).

Pre- and pro-metastatic niches in liver

Liver metastasis depends on the formation of both pre- and pro-metastatic niches. Various circulating factors from cancer cells, particularly in the form of exosomes, can help establish the pre-metastatic niche. For example, exosomes from pancreatic ductal adenocarcinoma (PDAC) cells, enriched in macrophage migration inhibitory factor (MIF) can activate Kupffer cells, inducing secretion of TGFβ. TGFβ triggers hepatic stellate cells to produce fibronectin, which promotes recruitment of bone marrow-derived macrophages that induce liver-specific metastasis (Costa-Silva et al., 2015). Exosome proteomics of several tumor models has identified different exosomes that establish the pre-metastatic niches in different organs. Specifically, Kupffer cells in the liver absorbed PDAC-secreted exosomes expressing integrin αvβ5 and released pro-inflammatory S100A8, resulting in liver tropism. Targeting integrin αvβ5 inhibited liver metastasis (Hoshino et al., 2015).

In addition, liver resident cells can promote the formation of pro-metastatic niches, which support the outgrowth of DTCs. For instance, hepatocytes coordinate myeloid cell accumulation and fibrosis within the liver to direct the formation of a pro-metastatic niche through IL-6-STAT3-serum amyloid A1 and A2 (SAA) signaling (Lee et al., 2019). Hepatic stellate cells, upon activation, secreted growth factors and cytokines such as PDGF, HGF, TGFβ, to promote ECM degradation, which stimulates angiogenesis and inhibits the immune response, establishing a permissive ME for tumor cells (Van Den Eynden et al., 2013; Kang et al., 2011).

Lung tropism

Lung is another frequent metastatic site in cancers such as breast, melanoma and thyroid (Table 1). The physiology of the lung makes it ideal for colonization and metastasis. The broad surface area and numerous capillaries provide opportunities for cancer cells to adhere, extravasate and colonize. At the same time, the endothelial layer in the lung has tight junctions between endothelial cells and an intact basement membrane, thus representing a more restrictive barrier for extravasation as compared to bone and liver (Figure 3). One strategy adopted by tumor cells to traverse the endothelial wall and basement membrane is to induce the formation of discrete foci of vascular hyperpermeability by increasing focal adhesion kinase (FAK)/ E-selectin and MMP9 expression in lung endothelial cells (Hiratsuka et al., 2002, 2011).

Figure 3. Lung-specific metastasis.

Tumor-derived factors such as SPARC, VCAM1, and ANGPTL4, as well as EVs have been shown to be involved into tumor cell extravasation to the lung parenchyma. Tumor cells can activate TLR3 signaling in alveolar type II cells, which in turn recruit neutrophils and promote lung metastasis. Chemokines enriched in the lung such as CXCL12 and CCL21 recruit CXCR4 and CCR7 positive tumor cells. Alveolar macrophages can secrete the pro-inflammatory mediator Leukotriene B4 to suppress the T cell response and facilitate metastasis. Fibroblasts secrete CSTB to induce tumor cell survival. Also, fibronectin-enriched fibroblasts recruit VEGFR1 and integrin α4β1 positive hematopoietic progenitor cells to facilitate metastasis. Tumor cells also produce tenascin C to initiate metastasis and GALNT14 to overcome dormancy signals from fibroblasts. TSP-1 secreted from endothelial cells inhibits tumor cells self-renewal, while CX3CL1 expression leads to recruitment of CX3CR1-positive patrolling monocytes, preventing metastasis.

Tumor intrinsic factors for lung metastasis

Many cell-intrinsic factors associated with lung metastasis are present in primary tumors, suggesting, as discussed above for other metastatic sites, that lung tropism may be preselected. Transcriptomic profiling of metastatic and non-metastatic breast cancer cells generated a list of 54 genes implicated in lung metastasis. The list includes secreted protein acidic and cysteine-rich (SPARC), vascular cell adhesion molecule 1 (VCAM1), angiopoietin-like 4 (ANGPTL4), ID1 and Tenascin C (TNC) (Minn et al., 2005). Many of these tumor-intrinsic factors disrupt vascular endothelial cell-cell junctions, increased the permeability of lung capillaries, and facilitated the trans-endothelial passage of tumor cells (Supplemental table 3). For instance, melanoma-derived SPARC promoted lung metastasis through inducing vascular permeability and extravasation in an endothelial VCAM1 dependent manner (Tichet et al., 2015). TGFβ in breast tumors induced ANGPTL4 to facilitate extravasation (Padua et al., 2008). Other factors may be involved in cell proliferation and survival (Supplemental table 3). For example, ID1 expression is selectively enriched in breast cancer lung metastasis and mediates lung colonization and sustained cell proliferation (Gupta et al., 2007). TNC, an ECM protein of stem cell niches, is produced by breast cancer cells that infiltrate the lung to increase stem cell-related signaling such as Notch and WNT pathways to initiate metastasis (Oskarsson et al., 2011).

Extracellular vesicles (EVs) and pre-metastatic niche in the lung

Similar to liver metastasis, cancer cells also actively secreted EVs, including exosomes, to build the pre-metastatic niche in the lung (Supplemental table 3). Liu et al. found that both lung cancer and melanoma-derived exosomal RNAs activate toll like receptor 3 (TLR3) signaling in alveolar type II cells to induce chemokine secretion and recruit neutrophils to build up the premetastatic niche (Liu et al., 2016). Melanomas can also produce EVs to downregulate interferon alpha and beta receptor subunit 1 (IFNAR1) and IFN-inducible cholesterol 25-hydroxylase (CH25H) in normal cells to promote the formation of a pre-metastatic niche enriched with CD11b+ myeloid clusters and fibronectin deposits (Ortiz et al., 2019). In addition, breast cancer cells secreted integrins α6β4- and α6β1-positive exosomes, which in turn enhanced pro-inflammatory S100A4 expression in lung-resident fibroblasts to establish a pre-metastatic niche and promote lung metastasis (Hoshino et al., 2015). Furthermore, Keklikoglou et al. found that chemotherapy-elicited breast cancer EVs were enriched in annexin A6 (ANXA6), a Ca2+-dependent protein that promoted NF-κB-dependent endothelial cell activation, CCL2 induction and Ly6C+CCR2+ monocyte expansion in the pulmonary pre-metastatic niche to facilitate the establishment of lung metastasis (Keklikoglou et al., 2019).

Pro-metastatic and metastasis-suppressive niches

Lung resident cells can establish a pro-metastatic niche for many types of cancer, e.g. they secrete abundant chemokines such as CXCL12 and CCL21 that direct breast cancer and melanoma cells that highly express CXCR4 and CCR7 to the lung (Müller et al., 2001). Fibroblasts also contribute to pro-metastatic niche formation in the lung. Liu et al. showed that fibroblast-secreted cathepsin B (CSTB) activated stearoyl-CoA desaturase 1 (SCD1), a critical modulator of cell proliferation, through the ANXA2 and PI3K/Akt/mTOR pathway, and promoted metastatic colonization of melanoma cells. Besides modulating tumor cells, lung fibroblasts also generated fibronectin to recruit VEGFR1 and integrin α4β1-positive bone marrow-derived HPCs to terminal bronchioles and bronchiolar veins, providing a permissive niche for incoming tumor cells (Kaplan et al., 2005).

In contrast to the pro-metastatic niche, several recent studies suggest that the lung also has metastasis-suppressive niches which inhibit cancer cell proliferation (Altorki et al., 2019). For example, a perivascular niche expressing thrombospondin-1 (TSP-1) induced sustained breast cancer cell quiescence, and this inhibitory effect was lost in sprouting neovasculature where active TGF-β1 and periostin were upregulated (Ghajar et al., 2013). The inhibitory effect of TSP-1 from perivascular niche has also been identified in bone, suggesting a shared metastasissuppressive mechanism among different organs (Ghajar et al., 2013). Lung fibroblast-derived bone morphogenetic proteins (BMPs) had an inhibitory effect on cancer stem cell self-renewal. Tumor cells, however, can secret polypeptide N-acetyl-galactosaminyltransferase 14 (GALNT14) to overcome this effect (Song et al., 2016). Furthermore, CX3C-chemokine receptor 1 (CX3CR1)+ monocytes, which were attracted by lung endothelial cells-derived CX3-chemokine ligand 1 (CX3CL1), in turn, recruited and activated natural killer (NK) cells to prevent lung metastasis (Hanna et al., 2015).

Brain tropism

The majority of brain metastases come from lung cancer, breast cancer and melanoma (Table 1). The brain is protected by the blood-brain barrier (BBB), which distinguishes it from other organs. BBB is a continuous, nonfenestrated endothelium stitched together by tight junctions and supported by a basement membrane, astrocytes, and pericytes (Figure 4) (Weidle et al., 2016). These features protect the brain from being invaded by cancer cells. However, some tumors cells produce cathepsin S to proteolyze the junctional adhesion molecule (JAM)-B and facilitate the transmigration through the BBB (Sevenich et al., 2014). They may also release miR181c-enriched EVs to target PDPK1/cofilin modulated actin dynamics to subsequently destroy the BBB (Tominaga et al., 2015). Once the extravasation of cancer cells is accomplished, the BBB may invert its roles from a barrier against cancer cells to a barrier against therapies. Indeed, the BBB protects brain metastatic tumor cells from many chemotherapeutic drugs (Valiente et al., 2018).

Figure 4. Brain-specific metastasis.

Tumor cells colonizing the brain need to produce cathepsin S, miR181c-enriched EVs and HPSE to overcome the defense provided by the blood-brain barrier. They also generate anti-plasminogen activator serpins to inhibit plasmin production from astrocytes to initiate metastasis. Astrocytes secrete many factors such as IL6, TGFβ, and IGF-1 that induce the growth of brain metastases. They also secrete miR19a-enriched exosomes, which inhibit PTEN expression and facilitate metastasis. Furthermore, gap junction characterized by CX43 between astrocytes and tumor cells and induced by PCDH7 stimulate brain metastasis. Exosomes enriched with miR503 induce M2 polarization of microglia which can promote metastasis through WNT signaling. Of note, bone and liver contain a fenestrated endothelium in contrast to the smooth lining of the lung endothelium and blood-brain barrier. Therefore, tumor cells develop different strategies to colonize at these different sites.

Brain-tropic genes

Brain metastasis-related genes and signaling pathways have been identified by different groups. A study of brain-seeking breast cancer cells in clinical samples identified 17 genes associated with brain metastasis. Among them, PTGS2 and the EGFR ligand HBEGF, are commonly expressed in both pulmonary and cerebral metastases, whereas the α2,6-sialyltransferase ST6GALNAC5 is a brain metastasis-specific mediator (Bos et al., 2009). Similarly, a number of studies suggest brain metastatic cancer cells are capable of producing other factors to break through the BBB (Supplemental table 4). For example, heparanase, a potent proangiogenic enzyme, is overexpressed in breast cancer brain metastases and mediates transendothelial migration. Inhibition of HPSE by miR1258 suppressed brain metastasis (Zhang et al., 2011).

Special contribution of astrocytes to brain metastasis

The brain neural niche is primarily composed of neurons and glial cells (astrocytes, oligodendrocytes, microglia) (Termini et al., 2014). Among them, astrocytes are the best-understood in brain metastasis. Astrocytes support neurons by secreting growth factors and cytokines and this ability can be hijacked by tumor cells to facilitate metastasis (Valiente et al., 2018). For example, astrocyte-derived IL6, TGFβ, and IGF-1 increase tumor cell proliferation (Seike et al., 2011; Sierra et al., 1997). Also, miR19a-enriched exosomes, derived from astrocytes and taken up by tumor cells, reduced the expression of PTEN, a major tumor suppressor. Astrocyte-specific depletion of miR19a, or blockade of astrocyte exosome secretion, rescued PTEN loss and suppressed brain metastasis. Furthermore, loss of PTEN and tumor cell-induced CCL2 expression recruited pro-metastatic myeloid cells (Zhang et al., 2015).

The interaction between cancer cells and astrocytes can promote tumor survival and protection from chemotherapy (Kim et al., 2011). Chen et al. found that cancer cell-derived protocadherin 7 (PCDH7) promoted the assembly of carcinoma-astrocyte gap junctions composed of CX43. These gap junctions transferred cGAMP to astrocytes to activate the STING pathway and release inflammatory cytokines that support tumor growth and chemoresistance (Chen et al., 2016). Furthermore, plasmin from reactive astrocytes induces cancer cell death in a FasL-dependent manner and cleaves cancer cell-derived L1 cell adhesion molecule (L1CAM), a molecular required for vascular co-option and metastatic outgrowth. Both lung and breast cancer cells can produce serpins that inhibit astrocytes-derived plasmin in order to initiate metastasis in the brain (Valiente et al., 2014).

General principles underlying organotropism

Despite the uniqueness of each distant organ, certain general principles underlying organotropism emerge. First, tropism may develop before dissemination, either through “seed pre-selection” or formation of a pre-metastatic niche. These mechanisms may explain why metastatic tropism often correlates with specific gene expression patterns in primary tumors. Second, specific chemotactic and adhesive factors may facilitate retention of DTCs in specific organ. This process may resemble immune cell homing to different peripheral organs. Third, distinct vascular structures in target organs establish specific requirements for cancer cell extravasation. For instance, the architecture of blood barriers in different organs may select for metastatic seeds with different capacity for breaking down endothelial junctions. Fourth, unique resident cells, together with the secretome and ECM generated by these cells, determine the initial fate of cancer cells upon their arrival. It should be noted that even in the same tissue, different MEs may provide dramatically distinct milieus. Therefore, precise mapping of various niches in different organs is of critical importance, especially for understanding the early-stage colonization process. Fifth, the ability of cancer cells to hijack resident cells and remodel the ME determines if overt metastases can be established. During these initial interactions, cancer cells and the cancer-entrained microenvironmental cells may form vicious cycles that are difficult to terminate. Finally, all of the above processes involve specific interactions between “the seeds” (cancer cells) and “the soil” (ME), which can be dynamic and continuously evolving throughout the colonization process.

Passive dissemination

Passive dissemination represents an alternative mechanism of metastatic spread, in contrast to hematogenous or lymphatic metastasis which requires the active steps of intravasation and extravasation. Passive dissemination includes the intra-peritoneal dissemination of ovarian cancer and the direct invasion to adjacent organs by gastrointestinal cancers, including gastric, colorectal and pancreatic cancers (Mikuła-Pietrasik et al., 2018). For example, ovarian tumor cells can disseminate to other organs in the peritoneal cavity through the passive movement of ascitic fluid (Lengyel, 2010; Mitra, 2016). Single or clusters of ovarian cancer cells detached from the tumor mass are carried by the peritoneal fluid and preferentially land on the abdominal peritoneum or omentum (Lengyel, 2010; Mitra, 2016). However, this process of intraperitoneal spread is not completely passive. EMT may still be required for the initial step of intraperitoneal dissemination (Lengyel, 2010). The ovarian tumor cells floating in ascites and in metastatic sites show reduced E-cadherin expression compared to cells in the primary tumors (Veatch et al., 1994). Down-regulation of E-cadherin decreases intercellular adhesion and promotes the invasion of epithelial cells, which facilitates the detachment of ovarian cancer cells (Ellerbroek et al., 1999; Kalluri and Weinberg, 2009). The mesothelium is traditionally considered a passive barrier preventing the intraperitoneal spread of cancer cells (Davidowitz et al., 2014; Iwanicki et al., 2011). However, recent studies suggest that mesothelial cells may play an active role in promoting the progression of intraperitoneal metastasis (Kenny et al., 2014; Mikuła-Pietrasik et al., 2014, 2016a, 2016b). Specifically, TGFβ1 from ovarian cancer cells stimulates the secretion of fibronectin through the TGFβ receptor/RAC1/SMAD-dependent signaling pathway in mesothelial cells (Kenny et al., 2014). The increased deposition of fibronectin promotes the adhesion, invasion, proliferation, and metastasis of ovarian cancer cells (Kenny et al., 2014). Attachment of ovarian cancer cells to mesothelial cells also stimulates the secretion of IL-6 and IL-8, and thereby promotes the proliferation of ovarian cancer cells (Mikuła-Pietrasik et al., 2014). The frequency of intraperitoneal spread positively correlates with the age of ovarian cancer patients (Mikuła-Pietrasik et al., 2016a). Furthermore, aged mice are more susceptible to developing metastases in an ovarian allograft model (Loughran et al., 2018). Senescent peritoneal mesothelial cells may create a metastasis-favorable niche for ovarian cancer cells by releasing pro-cancerous factors and via direct cell-cell contact, which can be blocked by neutralization of p38 MAPK (Mikuła-Pietrasik et al., 2016a). In addition, the senescent mesothelial cells are reported to promote the neoangiogenesis by stimulating the production of pro-angiogenic factors such as CXCL1 and VEGF by ovarian cancer cells (Mikuła-Pietrasik et al., 2016b).

Emerging organotropism mechanisms

Recent advances, which still await full validation, suggest additional organotropic mechanisms (Figure 5).

Figure 5.

Emerging facets of the “seed and soil” hypothesis. Green: organotropic cancer cells possess distinct metabolic features. Bone metastases utilize nutrients released during bone remodeling, such as serine, glycine, glucose, and glycerol. Liver metastases are highly glycolytic and consume local glucose. Lung metastases develop antioxidant strategies for survival in the pro-oxidant lung environment. Brain metastases exploit brain metabolites such as acetate, glutamine, amino acids and glutamine due to limited glucose resources. Blue: EMT regulates cancer organotropism. Liver metastases of pancreatic cancers require at least one copy of P120CTN, a stabilizer for membranous E-cadherin. Pancreatic cancer cells harboring homozygous P120CTN mutations can only metastasize to lung due to loss of E-cadherin. EMT may also contribute to organotropism through regulating metabolism and the immune microenvironment (ME). Red: different immune MEs regulates organ-specific metastasis. Bone metastasis is facilitated by osteoclasts, myeloid-derived suppressor cells (MDSCs), Tregs, and other bone-resident cells. Kupffer cells, CD11b/Gr1mid myeloid cells, metastasis-associated macarophages (MAMs), and neutrophils contribute to liver metastasis. Lung metastasis is regulated by alveolar macrophages, monocytes, MAMs, neutrophils, Tregs, and other cells. Microglia, the tissue-resident macrophages of the brain, promote brain metastasis.

Metabolic features of the microenvironment

Significant metabolic changes accompany tumor progression and metastasis and raise the possibility of therapeutic targeting of metabolic pathways (Elia et al., 2018; Teoh and Lunt, 2018). However, it is often difficult to differentiate between pathways promoting metastasis and those simply supporting tumor growth. Tumor growth is usually characterized by rapid cell division, whereas metastasis involves migration, invasion, and survival, and these processes often exhibit divergent metabolic activities relative to the primary tumor. For example, in breast cancer, the secretion of miR122 downregulates the level of pyruvate kinase isozyme M2 (PKM2), which in turn increases glucose availability in tumor cells but decreases glucose consumption of the ME niche. Interestingly, while overexpression of miR122 promotes metastasis to brain and lung, primary tumor growth is reduced (Fong et al., 2015). Similarly, while the fatty acid receptor CD36 drives metastasis in oral squamous cell carcinoma, melanoma and breast cancer, the effect of CD36 on primary tumor growth is limited, suggesting a unique role of lipid metabolism in metastasis initiation (Pascual et al., 2017). Metabolic differences between primary tumors and metastases are further illustrated by differential regulation of mitochondrial metabolism. The downregulation of mitochondrial genes correlates with poor clinical outcomes across several cancer types in primary tumors (Gaude and Frezza, 2016). However, mitochondrial metabolism is paradoxically upregulated in metastatic breast cancer (Kim et al., 2014). In this section, we will focus on organ-specific metabolic reprogramming of metastasis. More comprehensive insights into metabolic phenotypes in the general metastatic cascade have been reviewed previously (Elia et al., 2018; Luo et al., 2017; Pascual et al., 2018).

Only a very small minority of DTCs ultimately initiate the outgrowth of an overt metastasis (Obenauf and Massagué, 2015). A growing body of evidence suggests that metabolic interactions with the tumor ME indeed play an important role in metastasis propensity and progression. Upon colonization, survival of cancer cells usually requires adaptation to the metabolism of local tissues with respect to energy, nutrient and oxygen availability (Schild et al., 2018). Consequently, cancer cells undergo a tissue-specific metabolic rewiring for colorization, survival, and outgrowth (Gaude and Frezza, 2016).

Bone

Most overt bone metastases are osteolytic in breast cancer, but osteoblastic in prostate cancer. Upon the onset of bone resorption, metastatic metabolism may be rewired by nutrients released from the bone matrix, including serine, glycine, glucose, and glycerol (Shi et al., 2014). In contrast, metabolic reprogramming of osteoblastic metastasis is still poorly understood. Considering the recent finding that early-stage bone colonization of breast cancer often occurs in the osteogenic niche (Wang et al., 2015, 2018), an environment that is more “osteoblastic”, it is possible that the metabolic activities of DTCs need to undergo a drastic change when transiting from microscopic osteogenic to macroscopic osteolytic metastases. And this change may represent a bottleneck in the escape from metastatic dormancy, an as yet to be understood step in breast cancer metastasis.

Liver

The liver plays a central role in all metabolic processes in the body, especially for the equilibrium of glucose and fatty acid synthesis. Similar to the hypoxic glycolytic profile of local hepatic cells, liver metastases are also characterized by high glycolytic activity and reduction in mitochondrial metabolism (Dupuy et al., 2015). Additionally, Vriens et al. discovered an unconventional fatty acid desaturation pathway, involving sapienate biosynthesis in liver carcinomas, which further increased hepatic metabolic plasticity in cancer cells (Vriens et al., 2019). Furthermore, in hepatic metastasis of colorectal cancer, an enhanced fructose metabolism was observed via the upregulation of enzyme aldolase B (ALDOB), hence providing extra fuel for metastatic outgrowth (Bu et al., 2018). Thus, as the metabolic center of the entire organism, the liver appears to provide a unique milieu enabling or forcing cancer cells to assume specific metabolic activities for their colonization.

Lung

Lung is the organ of respiration, meaning that local tissues are exposed to high oxygen levels. Accordingly, cancer cells metastasized to lung often develop antioxidant strategies to overcome oxidative damage and stress. For example, by upregulating peroxisome proliferator-activated receptor gamma, coactivator 1 alpha (PGC-1α), breast cancer cells drastically enhance mitochondrial biogenesis to counteract electron leakage and ROS generation (Lebleu et al., 2014). Further antioxidant strategies include the upregulation of peroxiredoxin 2 (PRDX2), a small antioxidant protein (Arriba et al., 2012), as well as the overexpression of dihydropyrimidinase-like 4 (DPYSL4), a protein facilitating oxygen consumption (Higuchi et al., 2018). Exposure to oxygen in the lung ME may represent a contrast to bone, which is generally hypoxic. How this difference contributes to organ-specific metastasis will be an interesting question to tackle.

Brain

Brain tissues have the highest energy demands of all organs. Although glucose is the primary source of energy, other sources can also be utilized. In brain metastasis, when glucose is limiting, cancer cells display metabolic flexibility that enables them to exploit locally available nutrients such as acetate, glutamine, amino acids and glutamine (Ebert et al., 2003). Brain metabolic mimicry can extend to catabolism of γ-aminobutyric acid (GABA). For example, some brain metastatic cells up-regulate GABA receptors and transporters to better utilize GABA as an energy source for protein synthesis (Neman et al., 2014). How cancer cells adapt to the unique metabolic environment in the brain and acquire the ability to utilize brain-specific pathways remains to be further investigated.

Overall, it is tempting to hypothesize that cancer cells evolve to resemble the metabolic phenotype of the relevant distal organ during colonization. Further investigations of tissue-specific metabolic rewiring of metastatic cells at different stages of colonization will be required to test this hypothesis. This is technically very challenging, but will substantially enrich our knowledge of metastatic organ-tropism.

Immune ME and organotropism

Recent breakthroughs in characterizing the tumor immune ME and cancer immunotherapy highlight the importance of immune cells in the formation of metastasis (Kitamura et al., 2015). Immunosuppressive cells not only help cancer cells evade immunosurveillance and improve their survival in the circulation, but also actively modify the ME of target organs to prepare a premetastatic niche (Liu and Cao, 2016; Peinado et al., 2017).

Different organs harbor unique tissue-resident immune cells that are involved in metastasis. Macrophages are the major type of tissue-resident immune cells and have been extensively studied in variable metastatic settings. For example, VCAM1-expressing micrometastases were found to recruit integrin α4β1-positive osteoclasts, tissue-resident macrophages in bone, to accelerate their micro- to macro-metastasis transition (Lu et al., 2011), indicating that enrichment of osteoclasts in the tumor ME contributes to cancer cell awakening and progression to macrometastasis. Kupffer cells are liver-specific tissue-resident macrophages, which exert phagocytic and cytotoxic activity towards DTCs (Kolios et al., 2006). However, they can also trap CTCs at the liver sinusoid, increasing their chances of colonizing the liver (Bayón et al., 1996). Early depletion of Kupffer cells increased metastatic burden, but depletion at a later stage of tumor growth decreased metastasis(Wen et al., 2013). The alveolar macrophages in lung are also a double-edged sword for CTCs. The tumoricidal activity of alveolar macrophages has been long recognized; however, recent studies suggest that these cells can also promote lung metastasis. Alveolar macrophages have been shown to suppress T cell responses and to generate the pro-inflammatory mediator Leukotriene B4 to facilitate lung metastasis (Nosaka et al., 2018; Sharma et al., 2015). Microglia are the tissue-resident macrophages in the brain. Their cytotoxic function induces cancer cell apoptosis (He et al., 2006). However, similar to other tissue-resident macrophages, microglia can also promote metastasis to their host organ. Pukrop et al. showed that microglia enhance invasion and colonization of breast cancer cells in the brain via a WNT-dependent pathway (Pukrop et al., 2010). Also in breast cancer cells, loss of the lncRNA XIST led to increased secretion of exosomal miR503, which induced M2 polarization of microglia to promote metastasis (Xing et al., 2018). Indeed, it appears that primary tumor/CTC/metastatic lesions are able to re-educate tissue-resident macrophages to promote metastasis progression in their host organs. Understanding the genes and signaling pathways implicated in this process, and determining how to interrupt and reverse these processes will require further investigation. Furthermore, recent studies suggest that tissue-resident macrophages in different organs have different origins (embryonically-derived and/or monocyte-derived), which may correlate with different functions (Epelman et al., 2014). Although macrophages from disparate origins may carry out similar functions, their epigenomic landscape remains distinctive, and may influence their capacity to be educated by cancers.

Immune cells can also be systemically recruited by tumors to aid organ-specific metastasis, primarily due to their immunosuppressive functions and ability to establish the pre-metastatic niches. Below we present examples of how these immune cells affect organ-specific metastasis.

Bone

The HSC niche, which is likely the niche harboring DTCs, may have immune privilege through recruitment of immunosuppressive cells. Sawant and colleagues found that bone marrow-derived plasmacytoid dendritic cells recruited myeloid-derived suppressor cells (MDSCs) and T regulatory cells (Tregs) to bone metastases of breast cancers to inhibit tumor-specific cytolytic CD8+ T cells and promote metastatic colonization (Sawant et al., 2012). In a breast cancer model, MDSCs were found in a greater number in bone metastases than in primary tumors and lung metastases (Bidwell et al., 2012). Silencing of interferon-7 (IRF7) mediated type I IFN signaling in breast cancer cells facilitated bone metastasis by further restricting immunosurveillance, and restoration of IRF7 significantly decreased MDSCs but increased CD8 T cells and NK cells which inhibit bone metastasis (Bidwell et al., 2012). In addition, tumor-educated CD4+ T cells have been shown to promote bone metastasis by secreting RANKL and inducing pre-metastatic osteoclastogenesis (Monteiro et al., 2013). Indeed, one intriguing aspect of bone (and bone marrow) is the fact that it is where most immune cells are generated, but typically in immature or naïve states. It is therefore conceivable that tumor-induced immune responses may be unique in the bone, although this is understudied and remains to be elucidated.

Liver

Liver organotropism is also mediated by specialized immune cells beside Kupffer cells. For example, Zhao et al. found that a myeloid cell subset (CD11b/Gr1 mid) recruited by a CCL2/CCR signaling pathway promoted colorectal cancer liver metastasis (Zhao et al., 2013). Both bone marrow-derived macrophages and neutrophils also help prepare the liver premetastatic niche (Costa-Silva et al., 2015). Metastasis-associated macrophages (MAMs) can secrete granulin to activate hepatic stellate cells. Upon activation, hepatic stellate cells develop into periostin-producing myofibroblasts, which produce a fibrotic ME that sustains metastatic tumor growth (Nielsen et al., 2016). The tissue inhibitor of metalloproteinases-1 (TIMP-1) recruits neutrophils to increase liver metastases in a CXCL12/CXCR4 dependent manner (Seubert et al., 2015). Neutrophils, in turn, induce the liver pre-metastatic niche through fibroblast growth factor 2-dependent angiogenesis (Gordon-Weeks et al., 2017). Neutrophils also play a role in liver-specific colonization by promoting lung cancer cell adhesion to liver sinusoids. Macrophage-1 antigen (Mac1) was found to be responsible for interactions between neutrophils and CTCs (Spicer et al., 2012). Importantly, the immune tolerance of liver is unique, due to the liver’s ability to metabolize a wide variety of xenobiotic compounds (Tiegs and Lohse, 2010). Therefore, in the liver, DTCs can take “shelter” from the immune system. For example, liver sinusoidal endothelial cell-derived lectin (LSECtin) inhibited the hepatic T cell immune response and enhanced cancer cell migration, thereby promoting colorectal cancer liver metastasis (Tang et al., 2009; Zuo et al., 2013).

Lung

The role of different immune cells in organotropism has been extensively studied in lung metastasis. For instance, CCL2 produced by cancer cells and myeloid cells recruits inflammatory monocytes and CD206+/Tie2+ macrophages to pulmonary metastatic sites and pre-malignant lesions, respectively, to orchestrate breast cancer lung metastasis. Inflammatory monocytes secrete VEGF to facilitate tumor cells extravasation, whereas CD206+/Tie2+ macrophages downregulate E-cadherin and cancer cell adhesion to promote dissemination and intravasation (Linde et al., 2018; Qian et al., 2011). macrophages recruited to lung metastasis can also bind to cancer cells through an interaction with integrins α4/VCAM1, which subsequently activate Ezrin-PI3K/Akt signaling to prolong the survival of cancer cells in the lung (Chen et al., 2011). Neutrophils support lung metastasis initiation by secreting leukotrienes which selectively expand clones with high tumorigenic potential (Wculek and Malanchi, 2015). In addition, neutrophil extracellular traps (NET) induced by inflammation are required for dormant cancer cells to awake in the lung. Mechanistically, two NET-associated proteases, neutrophil elastase and matrix metalloproteinase 9 cleaved laminin and subsequently induced integrin α3β1 signaling-dependent proliferation of dormant cancer cells (Albrengues et al., 2018). MDSCs recruited to breast tumors secrete the cytokines TGF-β, IL-6, and IL-23, which induce the accumulation of T-helper cell 17 (Th17). Th17 cells, in turn, secrete IL-17 which promotes further recruitment of MDSCs and lung metastasis (Novitskiy et al., 2011). CCR4+ Treg cells have also been implicated in lung metastasis by inhibiting NK cells through secreting β-galactoside-binding protein (βGBP) (Olkhanud et al., 2009). Of note, loss of RON kinase promotes a CD8+ T cell response that specifically inhibits the outgrowth of lung metastasis (Eyob et al., 2013). Unlike other organs, the lung is constantly exposed to the external environment. On one hand, pathogens need to be immediately eliminated, entailing an efficient and rapid innate immune response. On the other hand, these responses need to be tightly regulated to avoid excessive inflammation and tissue damage. As such, lung provides a distinctive immune ME for metastasizing tumors. Further studies are needed to investigate how the tightly controlled inflammatory milieu influences metastatic colonization in the lung.

Brain

The brain may be another organ with an immunoprivileged ME due to the protection of the BBB (Peinado et al., 2017). However, increasing evidence indicates that the interaction between BBB and circulating immune cells is not uniform and various types of immune cells can indeed cross the BBB in different physiological and pathological conditions (Schwartz et al., 2013). Moreover, brain metastatic progression may disrupt the BBB to allow increased immune cell influx. However, the roles of immune cells in brain metastases remain at this time poorly understood.

In summary, the immune landscape of metastasis reflects the combined effects of tumor intrinsic pathways (leading to variable secretomes), the local environment (tissue-resident immune cells and vascular barriers), and the systemic host environment. Primary tumors may alter the local environment in distant organs before metastatic seeding occurs. The local environment may critically influence the seeding process, when cancer cells are still few in number and likely under stress in the foreign milieu. As cancer cells successfully survive and progress in distant organs, they may evolve the ability to educate local immune cells and to systemically recruit new immune cells, which may promote further progression. Due to different immune contexts, specific organs may impose different selective pressures on metastatic cancer cells and entail organ-specific mechanisms. Comparative analyses of the immune ME of metastases in different organs will likely provide novel insights into these mechanisms.

EMT and organotropic metastasis

EMT is critical for metastasis as discussed elsewhere (Lambert et al., 2017; Yeung and Yang, 2017). However, a recent study also implicates EMT in metastatic organotropism (Reichert et al., 2018). In this study, deletion of P120CTN, a stabilizer for membranous E-cadherin, in metastatic PDAC models, shifted the metastatic burden from the liver to the lung. Additional experiments showed that liver, but not lung metastasis required at least one copy of p120ctn. Furthermore, lung organotropism of the p120ctn homozygous PDAC cells could be shifted to the liver by transfection of p120ctn isoform 1A. These phenotypes were associated with P120CTN mediated E-cadherin expression and epithelial integrity as demonstrated by the selective pressure for E-cadherin-positive liver metastasis and E-cadherin-deficient lung metastasis in the Cdh1 heterozygous PDAC mice.

In addition to direct regulation of organotropism, EMT pathways may affect organ-specific metastasis through regulating cancer cell metabolism. By analyzing metabolic gene expression in 978 human cancer cell lines, Shaul et al. found that mesenchymal cell lines share 44 up-regulated metabolic genes. Among them, dihydropyrimidine dehydrogenase (DPYD), a pyrimidine-degrading enzyme, was highly expressed upon EMT induction (Bierie et al., 2014). SNAIL, a key transcriptional repressor of EMT, reprogramed glucose metabolism in breast cancer by repressing phosphofructokinase PFKP and fructose-1,6-bisphosphatase 1 (FBP1), which pushed the glucose flux towards the pentose phosphate pathway and granted cancer cell survival advantages (Dong et al., 2013; Ryu et al., 2017). Therefore, EMT may interact with metabolic reprogramming during cancer cell metastasis and contribute to organotropism.

Tumor-infiltrating immune cells such as tumor-associated macrophages (TAMs) and MDSC can promote EMT (Chockley and Keshamouni, 2016). On the other hand, accumulating evidence suggests that EMT can regulate the tumor immune ME. Snail-induced EMT increased the number of Tregs and reduced the number ofdendritic cells, which promoted melanoma metastasis to the lung (Kudo-Saito et al., 2009). In lung cancer cells, EMT activator ZEB1 repressed the EMT suppressor miR200, thereby relieving the miR200-mediated inhibition of PDL1, an immune-checkpoint protein, thus leading to CD8 T cell immunosuppression and metastasis (Peng et al., 2014). Furthermore, EMT-induced modulation of E-cadherin and cell adhesion molecule 1 (CADM1) regulated NK cell-mediated metastasis-specific immunosurveillance (Chockley et al., 2018). The link between EMT and tumor immune ME suggests that EMT could regulate organ-specific metastasis through modulating immune cells.

Recent studies suggest that epithelial plasticity of tumor cells cannot be classified simply as either epithelial or mesenchymal. Instead, cancer cells may possess a spectrum of intermediate states (Pastushenko et al., 2018). This hybrid epithelial/mesenchymal state has been termed “partial EMT” (P-EMT) and may be regulated by epithelial protein internalization, in contrast to the transcriptional repression of epithelial genes in complete EMT (Aiello et al., 2018; Grigore et al., 2016). P-EMT PDAC cells preferably traveled as clusters of CTCs whereas PDAC cells that underwent a complete EMT invaded as single CTCs (Reichert et al., 2018). Furthermore, CTC clusters showed specific DNA hypomethylation at binding sites of several EMT transcription factors, such as OCT4, NANOG, SOX2, and SIN3A (Gkountela et al., 2019). However, the formation of CTC clusters required cell adhesion components such as plakoglobin that are also epithelial markers (Aceto et al., 2014), further supporting the P-EMT status of CTC clusters. Given the importance of epithelial plasticity in organotropism, the role of P-EMT and CTC clusters in organ-specific metastasis needs to be considered and warrants further investigation.

Conclusions and Perspectives

With a few exceptions, the “seed and soil” hypothesis still provides a conceptual framework for our understanding of metastasis organotropism. Within this framework, an increasing number of molecular and cellular mechanisms have been discovered in recent studies. The specificity of the metastatic process in each organ is largely based on the unique ME, which consists of unique resident cell types, ECM, and secretomes. Cancer cells that either evolve fitness or adapt to the local ME exhibit tropism toward the particular organ. Their “fitness” or adaptability are typically encoded by epigenomic programs and manifested by the expression or activation of specific genes and pathways that allow cancer cells to exploit the specific aspects of the ME in which they reside, to endure a hostile ME, and to ultimately remodel the ME to fuel aggressive colonization.

One significant advance in our understanding of the metastatic ME is the “rediscovery” of various “niches” within each organ with highly organized spatial structures, specialized resident cells, and often well-defined physiological functions, as previously shown for other processes such as hematopoiesis. Examples include the perivascular and osteogenic niches in the bone, which appear to dictate different cellular fates of DTCs. To better understand the organization and cell interactions within these microenvironmental niches, analyses at the single-cell level, combined with cutting-edge microscopic techniques will be needed.

The recent renaissance of tumor immunology and metabolism have extended our definition of “seed and soil” to new dimensions. Pioneering studies have begun to reveal how metabolism of metastases may differ from that of primary tumors. Considering the vastly varying oxygen levels, acidity, and metabolite profiles of different organs, it would not be surprising to find unique tumor characteristics matching organ-specific metabolic environments. These environments may play particularly important roles in the early stages of colonization, before cancer cells acquire the ability to remodel the host environment. Similarly, baseline immune profiles may represent another major difference among organs, which may be reflected in the types of cancer cells able to invade the organ. The combination of specific invading cancer cells and the local immune landscape may lead to engagement of different “defense” systems, which may be utilized to tailor immunotherapies.