Factors involved in cancer metastasis: a better understanding to “seed and soil” hypothesis

EMT (epithelial to mesenchymal transition) represents a shift toward the mesenchymal state, allowing cells to adopt migratory and invasive behavior [9], while the reverse process is referred as mesenchymal to epithelial transition (MET). EMT has been implicated in the process by which cancer cells enter the circulation and seed metastases [10]. Yu et al. analyzed the EMT in CTCs from breast cancer patients and found that EMT plays a critical role in the bloodborne dissemination of human breast cancer [11].

Although EMT was thought to be important in tumor progression, it is inconsistent with the fact that metastatic lesions share the epithelial nature of primary tumors [12]. To explain this apparent paradox, it was proposed that EMT is reversible [13]. Notably, there are a few studies also supporting a role for MET in distant sites. MET was implicated in the formation of clinically significant metastasis in bladder cancer [14]. In addition, accumulating experimental evidence showed the requirement of MET in the colonization and metastasis of carcinomas [15, 16], which suggests implications for future therapies against metastasis. Targeting EMT alone might be counterproductive, inhibiting both EMT and MET could be promising therapeutic strategy.

Most of the observations exploring the role of EMT in tumors have relied on cell culture-mediated loss-of-function and gain-of-function trials. However, more recently, two papers published in Nature provided intriguing evidence that EMT is not required for metastasis in vivo [17, 18]. While these results are interesting, researchers also suggest potential limitations to these studies, such that tracking cells based on the expression of a single gene may miss ongoing EMT events [19], and these genetic manipulations may fail to suppress the expression of versions of EMT and completely suppress activation of EMT [20, 21]. Hence, in the future, further evidence is required to support the conclusion that EMT is not required for metastatic colonization.

The cancer stem cell (CSC) theory suggests that many types of solid tumors are hierarchically organized and sustained by a distinct subpopulation of CSCs [22]. In the cancer stem cell model, cancer stem cells are described as a reservoir of cells within the tumor that have the ability to self-renew and to provide the heterogeneous lineages of cancer cells that constitute the tumor [23, 24]. Alternative term in literature is described as “tumorigenic cell”, or “tumor initiating cell”. It has been known purely a minority of cancer cells have the ability to form tumor [25, 26] and metastatic colonization is a highly inefficiency process that only a small subpopulation of disseminated tumor cells accomplished [27].

In 2005, the concept “migrating cancer stem cell” was first established, which support the existence of mobile cancer stem cell [28]. A subset of cells from human brain tumors showing stem cell properties were identified in vitro and in vivo [29], thus providing strong support for the CSC theory. In particular, this model integrated two decisive features, stemness and EMT. Evidence was provided that CSCs derived from metastatic breast tumor cells exhibit significantly higher tumorigenic and metastatic capacities than low metastatic cells [30]. Most intriguingly, the EMT program was indicated to play a pivotal role in facilitating the entrance of non-stem cells into stem cell states. For example, in a mammary tumor progression model, it was suggested that the acquisition of stem and tumorigenic property is stimulated by EMT induction [31]. Collectively, based on these observations, cancer stem cells have the capacity of long-term self-renewal and therapy-resistant, making it more possible to transform from the primary tumor, survive in the circulation and colonize distant sites. In other words, the cancer stem cell would be the most tenacious “seed” to successfully colonize in a foreign “soil”.

Autophagy is an intracellular degradation system, by which cells breakdown cytoplasmic materials in the lysosome [32]. Although autophagy has long been postulated to be involved in cancer metastasis for many years [33], the exact role of autophagy and underlying molecular mechanisms is still controversial.

On one hand, evidence indicated that autophagy showed an anti-metastatic effect [34]. For example, Brahma et al. showed that rottlerin (a protein kinase C-delta inhibitor) can stimulate autophagy, resulting in cell death in pancreatic CSCs [35]. Similarly, caffeine (neuroactive compounds) induced autophagy and promoted apoptosis in various cell lines, including cancer cells. These results are consistent with previous studies on the use of caffeine to treat human tumors [36].

On the other hand, emerging experimental data support the idea that autophagy plays a pro-metastatic role in cancer metastasis, for its involvement in regulating tumor invasion [37], anoikis resistance [38], CSCs viability [39‐41], EMT program [42, 43], and tumor colonization [38, 44]. Clinically, pancreatic cancers exhibit high basal levels of autophagy, and thus, a phase II clinical trial (https://​clinicaltrials.​gov: NCT01273805) and translational study of hydroxychloroquine (HCQ), an inhibitor of autophagy, in patients with previously treated metastatic pancreatic cancer was started. However, inconsistent autophagy inhibition was achieved and demonstrated negligible therapeutic efficacy [45]. Taken together, the role of autophagy in cancer metastasis still warrants further investigation, and clinical translation by targeting autophagy remains to be achieved.

A clinical phenomenon has been recognized for years that many patients relapse with metastatic disease months or years after primary tumor treatment because residual tumor cells can enter a dormant state and become refractory to therapies. Tumor dormancy is described as a lag time between dissemination and metastatic outgrowth, in which disseminated tumor cells (DTCs) maintain quiescence, which is a stable, non-proliferative cellular state. Tumor cells can disseminate to distant sites and enter a dormant state for long periods, only to then give rise to metastasis [46].

With regard to cancer cell itself, dormant cancer cells retrieved from metastasis-free distant sites retain their metastatic ability [47]. Recently, numerous genes have been shown to be correlated with dormancy in many types of cancer [48‐50]. For example, in a bone metastasis dormancy model, VCAM-1 expression by cancer cells is shown to support reactivation of indolent cancer cells and bone metastasis by interacting with the microenvironment [50]. These evidences showed the crucial role of cancer cells in mediating tumor dormancy and reactivation.

According to experimental and clinical evidence, tumor dormant state can also be regulated by microenvironmental factors in certain organs. For instance, evidence was provided that BMP7 (bone morphogenetic protein 7), secreted by bone stromal cells, plays a key role in dormancy and recurrence of prostate cancer [51]. TGF-2β and p38 signaling was shown to keep DTCs in a dormant state in bone marrow but not in lung [52]. In a breast cancer model, the BMP inhibitor Coco was demonstrated to promote breast cancer cell dormancy escape [53].

The findings discussed above suggest that dormancy and reactivation are governed by complex interactions between DTCs and the microenvironment of the target organ. Intriguingly, it was indicated that the induction of autophagy is involved in the induction and survival of dormant ovarian cancer cells [54]. Future research requires more endeavors to find out the link between autophagy and tumor dormancy, which will provide insights into novel therapeutic strategies to metastatic cancer.

EVs are secreted vesicles that include exosomes (30–100 nm diameter), microvesicles (MVs 100–1000 nm diameter), and newly identified “large oncosomes” (1–10 μm diameter) [55]. In recent years, EVs have been shown to play a critical role in mediating the interaction between tumor cells and host cells, which prepares the pre-metastatic niche for the formation of secondary sites [56]. In particular, cancer cells secrete considerably more EVs than normal cells, leading to a substantial increase in detectable EVs circulating in the blood. Increasing evidences have been shown that tumor-secreted EVs, containing DNA, RNA, proteins, and other molecule components, such as hypoxia-inducible factor-1α (HIF1α) [57, 58], are capable of mediating cell-cell communication and playing an essential role in inducing metastasis. Tumor-secreted exosomes are small membrane vesicles (30–100 nm) derived from the luminal membranes of multivesicular bodies and released into the extracellular milieu by fusion with the membrane. On one hand, tumor-secreted exosomes are shown to be involved in enhancing the metastatic traits of cancer cells and remodeling the primary microenvironment. In other words, educating the primary “soil” for a tumor-permissive microenvironment and metastasis. For instance, it was reported that breast cancer cells with increasing metastatic potential secrete exosomes that proportionally increase cell motility, which suggests that exosomes play a dynamic role in mediating metastasis [59]. MCF7 and MDA-MB-231 cells secrete exosomes that induce the differentiation of adipose-derived mesenchymal stem cells into myofibroblasts, increasing their secretion of factors such as SDF-1, VEGF, CCL5, and TGFβ, which are involved in regulating tumor progression and metastasis [60]. On the other hand, tumor-secreted exosomes are capable of educating distant “soil”. In pancreatic ductal adenocarcinomas (PDACs), macrophage migration inhibitory factor (MIF) was found to be overexpressed in PDAC-derived exosomes, and its blockade impeded liver pre-metastatic niche formation [61]. Exosomes from mouse and human organotropic cancer cells were found to preferentially fuse with resident cells at their predicted destination. Further, the uptake of tumor-derived exosomes by organ-specific cells was also shown to promote the pre-metastatic niche [62]. Consistently, small nuclear RNAs from primary tumor-derived exosomes were found to activate alveolar epithelial TLR3, leading to chemokine secretion and neutrophil infiltration [63], and to contribute to the pre-metastatic niche formation. In a recent study, elegant work by Guoguang and colleagues validated a novel mechanism concerning liver-tropism metastasis, that is, the remodeling of the liver microenvironment by EGFR-containing exosomes derived from tumor cells mediated gastric cancer metastasis [64]. Taken together, these studies indicate that primary tumor exosomes contribute to cancer metastasis by promoting the metastatic traits of cancer cell and, more importantly, by educating the primary soil and distant soil. Microvesicles (MVs), also known as “microparticles” [65] and/or “metastasomes” [57], owing to their ability to merge with and transfer a repertoire of bioactive molecular content (cargo) to recipient cells, are increasingly regarded as mediators of intercellular communication [58]. Accumulating evidence showed that many cancer-related cell biological processes correlate with accelerated rates of MVs secretion from tumor cells [66]. Particularly, MVs are also shown to play a crucial role in metastasis [67]. It was reported that osteopontin expressed by tumor-secreted MVs plays an important role in bone marrow-derived cell mobilization and colonization of tumors [65]. Large oncosomes (LO) are large (1–10 μm diameter) cancer-derived extracellular vesicles (EVs), originating from the shedding of membrane blebs. In human prostate cancer, it was found that LO that contains metalloproteinases, RNA, caveolin-1, and the GTPase ADP-ribosylation factor 6 and are biologically active toward tumor cells, endothelial cells, and fibroblasts, suggesting a mechanism that LO is involved during the formation of pre-metastatic niche [68].

It has been well established that cytokines and chemokines derived from cancer cells can selectively attract and activate different cell types. The diverse functions of these factors establish them as key mediators of communication between cancer cells and microenvironment. Thus, these factors are always associated with multiple aspects of cancer metastasis. To understand how cancer cells affect the tumor microenvironment, Kim et al. conducted a biochemical screen for macrophage-activating factors secreted by metastatic carcinomas; the results indicated that Lewis lung carcinoma produced factors that activated myeloid cells via TLR2 and that induced TNF-α secretion by myeloid cells, thus enhancing metastatic growth by preparing the pre-metastatic niche [69]. CCL2-expressing breast tumor cells engaged CCR2⁺ stromal cells of monocytic origin, including macrophages and preosteoclasts, to facilitate breast cancer metastasis to lung and bone [70], respectively. Similarly, tumor cell-secreted IL6 causes Stat3 phosphorylation in lymphatic endothelial cells (LECs), inducing CCL5 expression in LECs and accelerating triple-negative breast cancer (TNBC) cell metastasis [71]. In summary, these factors indeed play a pivotal role in promoting metastasis by preparing the formation of the pre-metastatic niche.

However, it is important to note that these factors mediate a complex interplay between various host cell types and tumor cells, and pharmacological inactivation of these molecules or their receptors to reduce metastasis should be cautiously employed. For instance, the inhibition of CCL2–CCR2 signaling blocks the recruitment of inflammatory monocytes, inhibits metastatic seeding and prolongs the survival of tumor-bearing mice [72]. However, a paper published in Nature reported a paradoxical effect of CCL2 in four syngeneic mouse models of metastatic breast cancer. Surprisingly, the interruption of CCL2 inhibition led to an overshoot of metastases and accelerated death [73].

Other molecular components also play a pivotal role in mediating metastasis. For instance, in a previous study [74] with in vitro and in vivo models using MDA-MB-231 human breast cancer cells, it was demonstrated that tumor-derived osteopontin (OPN) induces mesenchymal stem cells (MSC) production of CCL5, mediating metastasis. Likely, breast cancer cell-derived tenascin C (TNC) promotes the survival and outgrowth of pulmonary micrometastases [75]. In hepatocellular carcinoma (HCC), it was found that a tumor-derived protein secretory clusterin (sCLU) contributed to HCC migration and EMT in vitro and metastasis in vivo [76]. Together, certain proteins derived from tumor cells also play a key role in mediating metastasis. Interestingly, a recent study reported that nucleotides released from the highly metastatic breast cancer cell also contribute to pre-metastatic niche formation by mediating lysyl oxidase secretion, collagen crosslinking, and monocyte recruitment [77].

As discussed above, the metastatic traits of seed indeed play a pivotal role during cancer metastasis; however, there is a big gap between the seeding cells and the formation of secondary tumor. Because most tumor dissemination occurs through the blood, CTCs that have been shed into the vasculature are of obvious interest [78]. Metastasis is a multistep process that includes local invasion by cancer cells, intravasation, arrest at a distant organ, extravasation, adapting to a new environment and colonizing distant organ, which is also regulated by the delivery and survival of CTCs in circulation and the ability of intravasation and extravasation [79].

After entering the circulation, the survival and transportation of CTCs depends on the physical interactions, mechanical forces and the microenvironment they encountered [80]. At the cellular and molecular levels, CTCs adhesion is a complex process involving dynamic cell-cell interactions. For instance, endothelial adhesion is necessary for CTCs dissemination, evidence indicated that monocytes promoted metastatic breast cancer cell adhesion to endothelium under flow [81]. By using in vitro models of vasculature, platelets have been shown to be an important ally for CTCs survival and extravasation [82]. More recently, CTCs clusters were demonstrated to greatly contribute to the spread of cancer [83]. In addition, physical trapping of CTCs and optimal circulation pattern are necessary for metastasis formation. CTCs were thought to preferentially arrest in the microvasculature that appeared to be more curved, branched, and stretched [84]. By imaging the establishment of brain metastases in vivo, capillary branching was revealed to be efficient in trapping CTCs [85]. Moreover, Leonard Weiss et al. suggested that metastatic colonization sites also correlated with blood flow patterns, as revealed by autopsy. Furthermore, CTCs that successfully survived and crossed the physical barrier imposed by the endothelium will eventually seed the distant organ [86].

Altogether, the metastatic cascade is driven by a sequence of both mechanical and molecular factors. Both favorable soil and suitable biomechanics are necessary for the eventual formation of a secondary tumor [87]. Additionally, it is self-evident that CTCs have great potential value for clinical applications, such as the early detection and novel treatments for metastases, as CTCs provide the possibility of targeting metastasis in real time [78].

The primary tumor microenvironment has been known to play a pivotal role in the regulation of cancer metastasis [88]. Numerous studies have focused more on the seed-to-soil signaling events that explain the mechanism by which the seed remodels the microenvironment. However, the soil-to-seed signaling events have been largely ignored. In the past decades, data have shown that signals provided by the primary tumor microenvironment are important modulators of the capacity of tumor cells to invade, access the vasculature, and metastasize. A variety of stromal cells and other molecular components surrounding the primary tumor have been identified to provide signals to enhance the invasive properties of cancer cells in many experimental models (Table 1). In addition, hypoxia in the primary microenvironment also plays a pivotal role in the regulation of tumor metastasis.

Tumor-associated macrophages (TAMs) are alternatively activated cells that are induced by interleukin-4 (IL-4)-releasing CD4⁺ T cells. It was demonstrated that TAMs enhance the invasiveness of breast cancer cells via exosome-mediated delivery of oncogenic miR-223 [89]. Chen et al. showed that CCL18 derived from breast TAMs promotes the invasive capacity of cancer cells by triggering integrin clustering and promoting their adherence to the extracellular matrix [90]. Similarly, in pancreatic cancer, CCL20 released from TAMs enhances the invasiveness of cancer cells via its unique receptor CCR6 [91]. More recently, TAM-secreted lipocalin-2 (Lcn2) was found to promote cancer cell dissemination by regulating EMT, resulting in increased cancer cell motility [92].

Mesenchymal stem cells have been shown to localize to breast carcinomas, where they integrate into the tumor-associated stroma. It was proven that MSCs within the tumor stroma enhance the metastatic ability of cancer cells, which is dependent on CCL5 signaling via its chemokine receptor CCR5 [93, 94]. To elucidate mechanisms involved in cancer cell migration, CCL9 produced by MSCs was also shown to play a critical role in promoting cancer cell invasion [94]. More recently, Gonzalez and colleagues reported that MSC-derived DDR2, a unique receptor tyrosine kinase activated by fibrillary collagen in the primary tumor, endows cancer cells with growth and migratory advantage through alignment with collagen [95].

PHD proteins serve as oxygen sensors and may modulate oxygen delivery, and haplodeficiency of PHD2 normalized the endothelial lining and vessel maturation, which promoted tumor perfusion and oxygenation and inhibited the migration ability of cancer cells [96]. Furthermore, it is revealed that endothelial cells within the prostate cancer microenvironment secreted IL-6, resulting in the downregulation of androgen receptor (AR) signaling in prostate cancer cells, which enhance the invasion of cancer cells [97]. More recently, it is important to note that these studies showed a novel role of endothelial cells in promoting metastasis by chaperoning circulating tumor cells and protecting them from anoikis [98].

In breast cancer, data indicated that CAFs promote tumor growth and angiogenesis through their ability to secrete SDF-1 (stromal cell-derived factor 1)in large part [99]. Further, Stroma associated with cancer metastases is enriched in Cav1-expressing CAFs, in vitro and vivo, and it was demonstrated that fibroblast expression of Cav1 (caveolin-1) favors migration and invasion of cancer cells by regulating p190RhoGAP, and stromal Cav1 remodels microenvironments to facilitate tumor invasion [100]. Intriguingly, exosomes secreted by CAFs also were shown to promote breast cancer cell protrusive invasion and motility through Wnt-planar cell polarity (PCP) signaling [101].

Recently, increasingly more evidence supports the concept that adipocytes are associated with tumor growth and metastasis, and several molecules derived from adipocytes in the tumor microenvironment are associated with metastasis [102, 103]. Evidence also indicated that adipocytes promote cancer cell invasion [104] and the EMT program [105]. In prostate cancer, it was shown that pre-adipocytes promote metastasis by modulating miR-301a/AR/TGF-β1/Smad/MMP9 signals [106]. More recently, data indicated that IGFBP-2 secreted by mature adipocytes increases breast cancer cell invasion [107]. Together, these observations support the point that signals provided by adipocytes may play a key role in mediating metastasis.

Other molecular signaling within the primary microenvironment can also regulate tumor metastasis by increasing the metastatic potency of tumor cells and modulating the microenvironment [108, 109]. Platelets have long been recognized as contributing to tumor metastasis [110]. Data indicated that platelet-derived TGFβ and platelet–tumor cell contacts synergistically activate signaling pathways in cancer cells, resulting in their transition to an invasive mesenchymal-like phenotype and thus promoting metastasis in vivo [111]. Similarly, adenine nucleotides derived from tumor cell-activated platelets regulate the opening of the endothelial barrier to allow transendothelial migration of tumor cells via the P2Y₂ receptor and therefore promotes cancer cell extravasation [112]. Together, these observations suggest that signals from platelets promote several steps of tumor metastasis. By using genetic mouse models and pharmacological inhibitors, pericyte depletion was demonstrated to be associated with enhanced hypoxia, EMT and Met receptor activation, which provided evidence that pericytes within the primary tumor microenvironment likely serve as important gatekeepers against metastasis [113]. Most intriguingly, data suggest that immune cells within the primary microenvironment provide signals that promote tumor metastasis. For instance, B lymphocytes were found to be associated with tumor cell dissemination and EMT activity. More recently, infiltrating CD4⁺T cells could promote tumor metastasis by enhancing cancer cell invasion via modulation of FGF11/miRNA-541/AR/MMP9 signaling [114].

Because of the rapid growth of tumor cells, the tumor often outpaces its blood supply, resulting in substantial hypoxia. Hypoxia has long been recognized to promote aggressive phenotypes of tumor and induce tumor invasiveness and metastasis. To investigate the molecular mechanism of Notch-ligand activation by hypoxia in primary soil, it was found that hypoxia activates Jagged2 in breast cancer cells, induces EMT and enhances cell survival in vitro [115]. Moreover, intratumoral hypoxia promotes metastasis by activating hypoxia-inducible factors (HIFs). Data showed that HIFs regulate molecular signaling between breast cancer cells and MSCs to stimulate metastasis [116]. More recently, a paper demonstrated that the HIF-1a/CCL20/IDO axis plays a crucial role in accelerating cancer metastasis in hepatocellular carcinoma by inducing EMT and an immunosuppressive tumor microenvironment [117]. Moreover, it is worth noting that hypoxia in the primary tumor is also involved in facilitating pre-metastatic niche formation in secondary organs by providing cytokines and growth factors that create a congenial microenvironment via the recruitment of myeloid cells and a reduction in cytotoxic effector functions of the NK cell population [118].

Previous studies indicated that some molecules are specially induced in pre-metastatic lung endothelial cells and microphages that transmit signals to cancer cells to promote colonization and tumor growth [119‐121]. For example, in the leukocyte-rich microenvironment of the lungs, macrophage binding to breast cancer cells via the α4-integrin-VCAM-1 interaction transmits survival signals, which stimulates lung-specific metastasis [120]. Inflammatory chemo-attractants and myeloid cell recruitment have been suggested to be associated with pre-metastatic niche formation [122]. Mechanistically, myeloid cells were shown to remodel the pre-metastatic lung into an inflammatory and proliferative environment, to diminish immune protection, and to induce EMT of cancer cells [123, 124]. Stroma-derived periostin (POSTN), a component of the extracellular matrix within the secondary sites lung is shown to recruit Wnt ligands and thereby increases Wnt signaling, which enables cancer stem cell maintenance and thus promotes tumor colonization and metastasis [27]. In addition, other immune cells in the lung microenvironment are also involved in promoting metastasis. For instance, neutrophils play a key role in inflammatory responses. In a mouse breast cancer model, Stefanie et al. demonstrated that neutrophil-derived leukotrienes mediate the colonization of distant lung by selectively expanding the subpool of cancer cells that retain high metastatic potential [125]. More recently, evidence supported that T-cell-intrinsic expression of the oxygen-sensing prolyl-hydroxylase (PHD) proteins function in T cells to coordinate immunoregulatory programs within the pre-metastatic lung that are permissive to cancer metastasis [126].

Evidence indicated that hepatic stellate cells (HSCs) play a crucial role in modulating the pro-metastatic niche [127]. A recent study supported that the secretion of granulin by metastasis-associated macrophages (MAMs) stimulates resident hepatic stellate cells (hStCs) into myofibroblasts that release periostin, thus leading to a fibrotic microenvironment in the liver that sustains metastatic tumor growth [128]. Similarly, macrophage migration inhibitory factor (MIF) derived from human hepatic sinusoidal endothelial cells (HHSECs) instead of colorectal cancer (CRC) cells, induced migration and EMT and promoted proliferation and apoptotic resistance in CRC cells [129]. Additionally, angiopoietin-like 6 protein, a soluble factor accumulated in hepatic blood vessels, was found to interact with circulating cancer cells, which induced liver colonization of CRC cells [130]. Liver sinusoidal endothelial cell lectin (LSECtin) expressed in the liver microenvironment is shown to directly correlate with CRC progression, including adhesion and metastasis [131].

Bone marrow-derived hematopoietic progenitor cells that express vascular endothelial growth factor receptor 1 (VEGFR1; also known as Flt1) form cellular clusters before the arrival of tumor cells, and the expression pattern of fibronectin is shown to provide a permissive niche for incoming cancer cells [132]. Furthermore, bone colonization is induced by the osteogenic niche that is mediated by heterotypic adherens junctions including osteogenic N-cadherin and cancer-derived E-cadherin, which activate the mTOR pathway in cancer cells and consequently drive tumor progression [133]. In addition, the osteocytes within the bone microenvironment were found to promote cancer invasion and growth by secretion of CCL5, MMP and extracellular ATP and adenosine in prostate cancer and breast cancer [134, 135].

Astrocytes were suggested to be associated with brain metastasis [136]. Recently, several studies reported that astrocytes within the brain microenvironment indeed play a pivotal role in mediating metastasis. For instance, IL-23 derived from astrocytes upregulates the secretion of the matrix metalloproteinase MMP2 and enhances the metastatic potential of brain metastasizing melanoma cells [137]. In particular, elegant work by Lin Zhang et al. demonstrated that astrocytes in the brain microenvironment induced the loss of PTEN in tumor cells by secretion of exosomal miRNA, which in turn created a permissive metastatic niche for cancer cells. Mechanistically, cancer cells receive signals from the brain microenvironment that lead to an enhanced secretion of the chemokine CCL2, which recruits myeloid cells that reciprocally stimulate the outgrowth of brain metastatic cancer cells through enhanced proliferation and reduced apoptosis [138].

As discussed above, both seed and soil factors play a pivotal role in mediating metastasis. The seed depends on supportive soil; however, it is important to note the dynamic interplay between seed and soil affect each other and evolve together. As a best proof, CSCs are thought to be the most tenacious seed, and TAMs are particularly abundant in “soil” among the immune cells present in the tumor site [139]. CSCs require a supportive niche to maintain a balance between self-renewal and differentiation; moreover, the dynamic interplay between CSCs and TAMs may affect the functional role and phenotype of each other [140].

Accumulating data indicated that distinct subsets of TAMs were found in different tumor microenvironments and that these TAMs were classified into two macrophage classes: M1 phenotype and M2 phenotype [141]. Furthermore, distinct subpopulations of TAMs represent distinct functional roles in the tumor microenvironment [142]. Additionally, TAMs were also shown to have the ability to regulate the adaptive immune response [143]. Increasing evidences have validated the key role of TAMs in supporting tumor growth, angiogenesis and metastasis by regulating the tumor microenvironment [144, 145]. Of note, previous studies indicated that TAMs are “educated” by myeloid-derived suppressor cells and present a distinct state of macrophage polarization, which enhances the tumor-supportive role of stromal cells. This study highlights the crosstalk between TAMs, MSCs, and tumor cells and supports the idea that TAMs within the tumor microenvironment evolve together with tumor cells. More recently, experimental data showed that host-produced histidine-rich glycoprotein inhibited tumor growth and metastasis by skewing TAMs polarization away from the M2- to a M1-like phenotype. Promising therapeutic strategies may involve targeting the crosstalk between TAMs and CSCs. However, more research is required in this area to harness the great opportunities that emerging knowledge offers. Similarly, tumor-associated neutrophils in lung cancer were shown to have the ability of polarizing to either an “N1” or “N2” phenotype that inhibits or promotes tumor progression, respectively [146].

With the goal in mind that treatment of metastatic cancer is to achieve eradication of cancer cells and hence seed factors—including unique genes expressed in tumor cells [148], factors affecting the EMT program [9], the existence of CSCs [149], autophagy [150], tumor dormancy state[151] and tumor-secreted factors that play pivotal roles in mediating metastasis are ideal targets for therapeutic interventions.

By targeting seed factors, it is likely to provide a curative effect that inhibits tumor growth and reduces metastasis. For instance, in a preclinical model, schisandrin B (Sch B), a naturally occurring dibenzo cyclooctadiene lignan with very low toxicity, could inhibit cancer metastasis by suppressing TGF-β-induced EMT of tumor cells [152]. More recently, targeting metastasis-initiating cells via fatty acid receptor CD36 resulted in almost complete inhibition of metastasis in a mouse model [153]. A humanized monoclonal antibody targeting αv integrins, which are involved in cell-to-extracellular matrix and cell-to-cell interactions, was used in a multicenter phase 1 study to inhibit prostate cancer metastasis [154]. A phase II trial of AS1411 (a novel nucleolin-targeted DNA aptamer) in metastatic renal cell carcinoma [155] has been completed, which suggests a novel way to target cancer cells at the molecular level and to improve treatment.

Although a plethora of clinical trials that targeted seed factors have had substantial successes in metastatic cancer therapy [156‐158], one problem plaguing the use of therapeutics targeting seed factors is drug resistance, which is determined by not only the complexity of the genomic aberrations that tumor cells harbor but also the properties of the tumor microenvironment [159]. It is also worth noting that targeting seed factors is not always effective, for example, to determine whether metastatic cancers that overexpress Her-2 (a gene found in both normal cells and cancer cells) or CEA (a protein present mostly in cancer cells) can be treated effectively with lymphocytes (white blood cells) that have been genetically engineered to contain an anti-Her-2 protein or anti-CEA protein. Two clinical trials were conducted (https://​clinicaltrials.​gov:NCT00924287,NCT00923806), respectively. However, both studies were terminated due to the suboptimal side effect. In particular, two recent studies published in Nature demonstrated that, in early lesions before any apparent primary tumor masses were detected, a subpopulation of early cancer cells was indicated to be invasive and to be able to spread to distant organs [160, 161]. These observations remind us that targeting at seed may be less effective.

Considering the abovementioned observations, disseminated cancer cells from early and later stages have metastatic potential. Therefore, therapies targeting the seed of metastasis need to address issues of heterogeneity. Taken together, targeting seed of metastasis currently seems to be challenging and less effective. Future research focusing on uncovering the mechanisms involved in drug resistance and the complex link between EMT, CSCs, autophagy and metastatic dormancy may shed light on novel treatments that involve combined targeting of these factors.

Based on data presented in this review, primary soil-derived factors often confer tumor cells with the ability of invasion and tumor growth. It is likely that interventions targeting these factors will inhibit cancer metastasis. For example, a phase I study of monoclonal antibody F19 targeting a cell-surface protein of tumor stromal fibroblasts was conducted [162]. More recently, in a multicenter, randomized, placebo-controlled, phase 3 clinical trial, regorafenib—the first small-molecule multi-kinase inhibitor—was used to treat metastatic colorectal cancer by blocking various signaling pathways implicated in promoting tumor progression [163]. However, it is naive to think that individual cellular or molecular components function in isolation in a complex system. For instance, despite inhibition of CCL2 or CCR2 within the tumor microenvironment was indicated to be beneficial in inhibiting metastasis [72]. Based on this preclinical model, to determine the safety and effectiveness of blocking CCL2 and CCR2 in metastatic patients, two clinical trials have been conducted, respectively (https://​clinicaltrials.​gov: NCT00992186, NCT01015560). However, the results proved to be less effective possibly due to the highly complicated interaction between chemokines and chemokine receptors, as various chemokine ligands do not exclusively bind to one chemokine receptor [164].