Bone Tropism in Cancer Metastases

Abstract

Bone is a frequent site of metastases in many cancers. Both bone properties and the tumor-intrinsic traits are associated with the metastatic propensity to bone (i.e., the bone tropism). Whereas an increasing body of mechanistic studies expanded our understanding on bone tropism, they also revealed complexity across the bone lesions originated from different cancer types. In this review, we will discuss the physical, chemical, and biological properties of bone microenvironment, identify potential players in every stage of bone metastases, and introduce some of the known mechanisms regulating the bone colonization. Our objectives are to integrate the knowledge established in different biological contexts and highlight the determinants of bone tropism.

Metastasis is the direct cause of more than 90% of cancer-related deaths (Seyfried and Huysentruyt 2013). In many cases, tumors relapse in distant organs years or even decades after resection of the primary tumor (Obenauf and Massagué 2015). Bone is one of the most frequent sites of metastases. In the clinic, bone metastasis remains an incurable disease, which is responsible for morbidity in most advanced breast and prostate cancers, and characterized by severe pain, impaired mobility, pathologic fractures, numbness, paralysis, anemia, and hypercalcemia leading to coma and death (Coleman 2006). Bone tropism is the propensity and capacity of tumor cells to spread and eventually become clinically evident in bone (Fatatis 2011). Understanding the bone tropism will help us to specify the behavior and mechanism of bone metastases from that of the primary tumor and other metastases, and pave the way for effective clinical applications.

BONE TROPISM IN DIFFERENT CANCER TYPES

The incidence of bone metastases and median-survival of bone metastasis patients vary greatly among different cancer types, which is summarized in Table 1 (Selvaggi and Scagliotti 2005; Macedo et al. 2017). Bone metastases also vary in different cancer types by specific mechanisms to disrupt bone homeostasis. The osteolytic metastases, characterized by destruction of bone structure and loss of bone mass, are detected mostly in breast cancer, lung cancer, multiple myeloma, melanoma, and renal cell carcinoma. In contrast, the osteoblastic metastases, featured by excessive bone formation, are found predominantly in prostate cancer, as well as a minority of breast and lung cancers. In fact, lesions exhibiting both osteolytic and osteoblastic activities are often observed in patients with advanced breast cancer and prostate cancer (Suva et al. 2011). Thus, even within the same type of cancer, bone metastasis features may vary widely.

Table 1.

Incidence of bone metastases in cancer

Primary cancer type	Relative incidence in bone	Median survival from diagnosis
Breast	65%–75%	19–25 mo
Prostate	65%–75%	12–53 mo
Lung	30%–40%	6 mo
Thyroid	40%–60%	48 mo
Bladder	40%	6–9 mo
Renal	20%–25%	12 mo
Melanoma	14%–45%	6 mo

Data in the table based on Suva et al. (2011) and Macedo et al. (2017).

Breast Cancer

In the clinic, metastasis organ tropism and latency vary significantly across different molecular subtypes of breast cancer. For instance, luminal A/B subtypes show a lower recurrence rate (27.8%–42.9%) than the HER-2⁺ tumors (51.4%) as well as a longer metastasis-free survival compared to HER2 and basal subtypes (1.6–2.2 yr vs. 1.3 and 0.7 yr), indicating that the luminal subtype is less aggressive in general. Nevertheless, luminal tumors exhibit a higher incidence of bone metastasis (62.1% and 64.5%) than the HER2⁺ subtype (47.7%) and basal subtype (32.2%) (McGuire et al. 2015). The luminal subtype is characterized by the positive expression of estrogen receptors (Wiechmann et al. 2009). Consistently, the overall rate of bone metastasis is much higher in patients with ER⁺ tumors than those with ER⁻ tumors (71% vs. 47%) (Smid et al. 2008; Zhang et al. 2009; Kennecke et al. 2010). In spite of the lower bone relapse rate, ER⁻ tumors often develop bone relapse within a shorter time than ER⁺ tumors (5 yr vs. 10 yr) (Kennecke et al. 2010), indicating that the mechanisms of metastasis may differ between different subtypes. It remains unclear whether the difference in bone tropism is due to a stronger advantage conferred on the luminal cancer cells by the bone microenvironment, or a consequence indirectly caused by incompetency of luminal cancer cells in quickly colonizing visceral organs. In other words, the basal subtype may have equal or even stronger capacity to form bone metastasis, but only lacks sufficient time to manifest—as patients usually succumb to other metastases first. Further studies are needed to address these different possibilities.

Prostate Cancer

Prostate cancer exhibits a strong preference for bone metastasis. In autopsies of patients with advanced prostate cancer patients, metastases were much more frequently found in bones than in other organs (Bubendorf et al. 2000). Prostate cancer skeletal metastases are most often osteoblastic (characterized by increased mineral density), which are the opposite to osteolytic metastases predominantly occurring in breast cancer (Macedo et al. 2017). Androgen-deprivation therapy (ADT), the elimination of testosterone by medical or surgical castration, is the major treatment of prostate cancer (Andriole 2009). However, many patients will eventually develop the castration-resistance within 18–24 mo after ADT. As the disease progressed, more than 90% of patients with metastatic castration-resistant prostate cancer (mCRPC) will develop bone metastases (Frieling et al. 2015). The potential mechanistic links between resistance to ADT and bone-tropic metastases remain poorly studied. Preclinical data derived from xenograft models suggested that the loss of prostate-specific antigen (PSA) and androgen receptor (AR) might be associated with the osteolytic phenotype in prostate cancer bone metastases (Dai et al. 2016), which is yet to be clinically validated.

Lung Cancer

Lung cancer is the third most common form of cancer to spread to bone, after breast and prostate cancer (D'Antonio et al. 2014). The median survival of a patient with lung cancer bone metastasis is 6 mo, which is the shortest among all cancer types metastasizing to bone (Vicent et al. 2015). Notably, as one of the major indicator of lung cancer, the epidermal growth factor receptor (EGFR) status strongly correlates with the occurrence and pathologies of bone metastases (Bethune et al. 2010). In non-small-cell lung cancer (NSCLC), EGFR⁺ patients display a strong predisposition to bone metastases, which justifies prospective bone metastasis screening on EGFR⁺ patients (Kuijpers et al. 2018). Moreover, although NSCLC bone metastases are osteolytic in general, osteoblastic lesion is more commonly found in patients with mutant EGFR lung adenocarcinomas, especially after TKI therapy (Ansén et al. 2010; Pluquet et al. 2010). These observations indicate functional connections between oncogenic EGFR activities and bone tropism of metastasis.

The diversity and complexity of bone metastases in the clinic pose a challenge to basic research due to limited choices of cell or animal models. This limitation may underlie observed incongruence among published results and the discrepancy between clinical and basic research (Prinz et al. 2011). Therefore, cautions should be used to interpret the findings in the context of specific cancer types/subtypes and to draw general conclusions about all bone metastases.

DETERMINANTS OF BONE TROPISM

Based on the “seed and soil” hypothesis made by Stephen Paget in 1889, the determinants of metastasis tropism include tumor-intrinsic traits (the “seed”) and the characteristics of the host organs (the “soil”) (Zhang et al. 2019). One hundred and thirty years after the hypothesis was proposed, a growing body of evidence indicates that the interactions between cancer cells and the distant organs is more than natural selection for the fittest. Rather, to successfully colonize, disseminated tumor cells (DTCs) must adapt to the new microenvironment, which involves a series of dynamic crosstalk that may lead to reprogramming of both “seed” and “soil.” In the context of bone, the metastatic colonization process is arguably more complicated, due to the prolonged temporal course and the uniqueness of the bone environment.

Physiochemical Properties of Bone: The Soil

Bone is a mineralized tissue featured by intense vascularization, low oxygen level, high local calcium concentration, and acidosis (Johnson and Suva 2018). All these characteristics make the skeleton a unique and perhaps challenging environment for DTCs to colonize, survive, and proliferate.

Sinusoid Structure

Being arrested in the capillary bed is the first step of successful bone colonization (Mundy 2002). Due to the discontinuous and fenestrated endothelium of sinusoids, extravasation is not expected to be a rate-limiting step, making bone one of the most accessible organs to the circulating tumor cells (Nguyen et al. 2009; Esposito et al. 2018). Perhaps because of this, although breast cancer cells are capable to colonize multiple distal organs, bone is often the first site of metastasis (Suva et al. 2009). Even for the cancers that rarely develop overt bone metastases such as gastrointestinal cancer, DTCs can yet be found in the bone marrow (Hiraiwa et al. 2008).

Vasculature and Niches

Bone is a highly vascularized tissue. Accumulating evidence shows that blood vessels in the bone may provide an initial harbor for tumor cells and contribute to the subsequent processes of bone metastasis (Raymaekers et al. 2015). The perivascular niche, mainly formed by endothelial cells and perivascular stromal cells, is essential for dormancy maintenance in cancer cells (Ghajar et al. 2013; Price et al. 2016). In contrast, osteogenic niche (or endosteal niche), which is derived from mesenchymal stem cells and comprises osteoblasts and progenitor cells, has been shown to be a niche fueling proliferation (Wang et al. 2015).

How cancer cells transit between the endothelial niche and osteogenic niche, and how such a transition alters the fate of tumor cells remain to be elucidated. Some studies indicated that migration of cancer cells from the perivascular niche to the osteoblastic niche is mediated by various expression of cytokines (CXCL12, RANKL, PTHrP) in different niches (Esposito et al. 2018). Alternatively, the niche components may undergo trans-differentiation to switch niche functions. In prostate bone metastasis, the endothelial cells could be converted to osteoblasts via BMP4 secreted by cancer cells (Lin et al. 2017). Therefore, the seemingly different niches may actually share the same cell-of-origin. Consequently, cancer cell fate may be altered without spatial movement.

Of note, the bone metastatic niches may be the hematopoietic stem cell (HSC) niche. The endosteal niche is found important for both DTCs and HSCs by physical adhesion and attachment (Adams et al. 2007; Wang et al. 2015). One study on bone metastasis in prostate cancer showed that DTCs compete against HSC for the same niches in the bone marrow (Shiozawa et al. 2011). Mechanistically, DTCs and HSCs may similarly use the CXCL12/CXCR4 signaling to home and survive (Broxmeyer 2008; Teicher and Fricker 2010; Shiozawa et al. 2011). Considering these commonalities, further studies on HSC niche may generate valuable insights into bone metastasis, and vice versa.

Low Oxygen Level

In spite of its high vascularization, bone is an organ of extensive hypoxia. Oxygen tension in normal tissues falls between 2% and 9% (14–65 mmHg); in the periosteum and cortical bone, oxygen levels range from 4.2% to 7% (30–50 mmHg). In the bone marrow where bone metastases initiate and develop, the oxygen level is usually lower than 2% (pO2: Trabeculum: 0%–2%; Endosteum: ∼1.8%; Sinusoidal regions: ∼1.3%) (Johnson et al. 2017). Previous studies demonstrated that under low oxygen condition, cancer cell secretome may be altered to facilitate metastatic colonization and osteoclast differentiation (Gilkes 2016). In breast cancer, hypoxia (HIF-1 pathway) alone or together with TGF-β signaling promote the development of osteolytic bone metastases, which could be pharmacologically inhibited by anti-HIF1 treatment (Hiraga et al. 2007; Dunn et al. 2009). Additionally, the HIF-1 pathway also confers chemotherapy resistance on triple-negative (TN) breast cancers, supporting the rationale to combine HIF inhibitors with conventional cytotoxic chemotherapies (Samanta et al. 2014). Notably, most evidence for bone metastasis-specific impact of hypoxia was derived from fast-growing cells (e.g., human MDA-MB-231 cell lines and murine 4T1 cell lines), which largely bypassed the indolent growth period of bone metastasis. Studies using more latent tumor models (e.g., ER⁺ breast cancer and osteoblastic prostate cancer) are of urgent need to deepen our understanding of hypoxia in the context of bone metastases (Johnson et al. 2017).

Calcium

Bone is a mineralized tissue. Compared to the normal physiological calcium level (∼1.1–1.3 mmol/L), the local calcium level in bone varies widely from 2 mmol/L (nonresorbing bone) to as high as 8–40 mmol/L (resorbing bone) (Liao et al. 2006). Therefore, the high local calcium level might render cancer cell addictive to calcium signaling. This is exemplified by the prevalent up-regulation of calcium-sensing receptor (CaSR) in bone metastases from breast cancer, prostate cancer and renal cell carcinoma (Sanders et al. 2000; Liao et al. 2006; Joeckel et al. 2014; Frees et al. 2018). The extracellular calcium is not the only accessible resource to cancer cells. During the early bone colonization of breast cancer, osteogenic cells and cancer cells establish a physical connection through gap junctions which directly transfer calcium influx from the osteoblasts to the cancer cells. The calcium influx can promote the calcium signaling like NFAT and MEF2 in cancer cells and hence leads to progression of bone metastasis (Wang et al. 2018). In addition, in advanced bone metastases with skeletal muscle weakness, increased bone destruction and associated elevations in TGF-β activity can oxidize RyR1 and therefore, lead to Ca²⁺ leakiness from the muscle (Waning et al. 2015), which serves as another possible calcium reservoir for cancer cells.

Mechanisms of Bone Metastasis: How the Seeds Fit in the Soil

The journey of metastases is a stepwise cascade that includes (1) local invasion, (2) intravasation, (3) survival in circulation, (4) arrest in a distant organ, (5) extravasation, (6) micrometastasis, and (7) macrometastases (Obenauf and Massagué 2015). The preseeding steps 1–3 are likely to be shared by all distant metastases. In this section, we will focus on the postseeding steps by reviewing the unique biology of bone metastases.

Seed Preselection and Premetastatic Niche

Bone metastasis tropism may emerge even before cancer cells leave primary tumors. This may be driven by selective pressure exerted by the microenvironment of bone-like characteristics. A role of the breast tumor stroma in preselecting bone-tropism cancer cells was revealed in TN breast cancer. Cancer-associated fibroblasts (CAFs) in TN breast tumors produce CXCL12 (also known as SDF-1) and IGF1, two of the cytokines abundantly found in the bone microenvironment. Chronical exposure to these cytokines led to the enrichment of cancer cells with elevated c-Src activity (Zhang et al. 2013). Should these metastatic seeds arrive in the bone marrow, the c-Src activity would confer survival advantages through CXCL12–CXCR4 and IGF1–IGF1R pathway. Therefore, metastatic seeds with bone tropism are preselected in primary tumors due to the resemblance of the microenvironment in certain tumors to bone marrow.

Bone metastasis tropism may also emerge by preparing “soil” before metastatic seeds arrive. This concept is referred to as “premetastatic niche”: the primary tumors secrete factors and extracellular vesicles into circulation, which may alter the microenvironment to facilitate metastatic seeding in distant organs (Peinado et al. 2017). A recent study showed that exosomes from lung-, liver-, and brain-tropic cancer cells preferentially fuse with resident cells in their targeted organs, thereby establish a favorable microenvironment for distant metastasis via expression of specific integrins (Hoshino et al. 2015). The exosome-induced bone tropism was not elucidated in this study, which needs to be further explored in the future. Other reports demonstrate that integrins expressed by cancer cells, such as α2β1, α5β1, and αvβ3, mediate bone metastasis (Weilbaecher et al. 2011). Whether exosomes carrying these integrins create premetastatic niche remains to be tested. In addition to integrins, exosomal miRNA may also play roles in the premetastatic niche of bone metastases. For instance, miR-940 can be delivered by exosomes from primary prostate cancer to the bones, and induce extensive osteoblastic lesions by facilitating osteogenic differentiation (Hashimoto et al. 2018). Similarly, exosomal miR-141-3p from MDA PCa 2b cells promoted osteoblast activity, making it more permissive for bone metastases of prostate cancer (Ye et al. 2017).

The lysyl oxidase (LOX) may promote the premetastatic niche in bone. Through circulation, LOX secreted by primary breast tumors facilitates osteolytic lesion formation in the skeleton of ER⁻ patients (Cox et al. 2015). Additionally, the LOX-induced osteoclastogenesis is driven by the nuclear translocation of NFATc1, which is independent of the conventional RANKL signaling. Another study reported that colorectal cancer-derived LOX also promotes osteoclast differentiation albeit via a RANKL-dependent mechanism (Reynaud et al. 2017).

In a more complicated fashion, lung cancer cells form a feedback loop with distant bone cells, to drive tumor progression. Preclinical and clinical studies reveal that lung cancers increase bone stromal activity by secreting soluble receptor for advanced glycation end products (sRAGE) to the bone. Consequently, osteocalcin-expressing (Ocn⁺) osteoblastic cells are activated, which facilitate the outgrowth of lung cancer by remotely supplying a distinct subset of tumor-infiltrating SiglecF^high neutrophils (Engblom et al. 2017). In future study, it will be interesting to validate whether the increased Ocn⁺ osteoblastic cells can represent a premetastatic niche of secondary bone metastasis.

While secretome-induced premetastatic niche is intriguing, it should be noted that that evidence based on mouse models should be interpreted with caution. In mice, the (primary) tumor/body ratio could reach as high as 1:20 or even 1:10 by weight. In contrast, in human patients, the primary tumor (e.g., breast cancer) is typically <5 cm in diameter (Narod 2012), which should not exceed 1/1000 of the total body weight. Therefore, experiments based on adoptive transfer of exosomes and other secreted molecules should be cautious to avoid supraphysiological concentration.

Dormancy

Metastases relapse years after diagnosis or resection of primary tumors. In prostate cancer, the metastatic cells showed a strong bone tropism, and may stay asymptomatic for over years (Sturge et al. 2011). In breast cancer, despite that cancer cells may disseminate to bone at a very early stage, a significant proportion of patients do not manifest overt bone metastases until years later (Chen et al. 2017). Therefore, dormancy represents one of the most important characteristics of metastatic progression in the bone.

There are two different types of dormancy genes. The first type genes may inhibit proliferation but are essential for retention and survival of DTCs specifically in the bone. For example, E-selectin and CXCL12 in the perivascular environment maintain disseminated breast cancer cells in a dormant state (Price et al. 2016). E-selectin inhibition effectively blocks the homing of cancer cells, whereas the inhibition of CXCL12/CXCR4 interaction repels the dormant micrometastases into circulation. The significance of E-selectin and CXCL12/CXCR4 signaling for bone metastases is further supported by other studies (Müller et al. 2001; Kang et al. 2003; Stübke et al. 2012), highlighting the notion that dormancy needs to be coupled with long-term survival and bone retention for later outgrowth to be possible.

The second type of dormancy-related genes may be merely growth inhibitory. Ghajar et al. found that TSP-1 expression around microvessel stalks induces sustained quiescence of breast cancer cells in the perivascular niche. Interestingly, the suppression was relieved when new vessels sprout, which stimulates outgrowth of micrometastases (Ghajar et al. 2013). This type of dormancy genes may inversely correlate with clinical recurrences. Examples include MSK1 (Gawrzak et al. 2018), LIFT (Johnson et al. 2016), BMP7 (Buijs et al. 2012; Kobayashi et al. 2012), and KAI1/CD82 (Bandyopadhyay et al. 2006). All these genes are expressed by cancer cells, and their suppression promotes bone metastases. It has become evident that the dormancy mechanism may be diverse and vary in different biological contexts. While the list of dormancy genes may continue to grow, provocative questions remain: how the expression of this gene is maintained or evolutionally selected throughout the prolonged dormancy if they do not confer a growth advantage on cancer cells, and how the expression is finally suppressed to terminate dormancy and resume metastatic outgrowth.

Early Colonization after Dormancy

The mere presence of DTCs in bone marrow does not necessarily lead to successful colonization. In order to survive and outgrow in bone, cancer cells need to adapt to the initially nonpermissive environment and hijack bone-specific cellular and molecular mechanisms to gain access to growth factors and cytokines. For example, CXCR4 is one of the most up-regulated genes in bone metastasis of breast cancer, and this is probably due to strong selective or adaptive pressure exerted by bone niche cells that express abundant CXCL12/SDF-1, the ligand of CXCR4. This interaction further elevates the c-Src signaling, which is strongly associated with long-term survival, the exit of dormancy, and therapeutic resistance (Myoui et al. 2003; Zhang et al. 2009; Chiu et al. 2017).

Cancer stemness and the epithelial-to-mesenchymal transition (EMT) are widely recognized to be important for distant metastasis (Kang and Pantel 2013). Specifically for bone metastasis, EMT has also been subjected to the intensive investigation (Gao et al. 2014). In prostate cancer, a loss of PTEN is accompanied with RAS/MAPK signaling activation and induces EMT, leading to macrometastasis in bone with 100% penetrance (Mulholland et al. 2012). In breast cancer, DTCs found in the bone marrow aspirates are enriched with CD44⁺/CD24⁻/low cells (Balic et al. 2006). The CD44⁺/CD24⁻ subpopulation is also more mesenchymal and stem-like (Mani et al. 2008), suggesting that EMT may be essential for early dissemination and homing to the bone. Another study demonstrated that expression of DKK1 endow mesenchymal-like DTCs capacity to evade immune surveillance by inhibiting Wnt signaling in local immune cells (Malladi et al. 2016), supporting that EMT may play roles in the long-term survival of DTCs before colonization starts.

After EMT-mediated dissemination and homing are accomplished, the ultimate colonization of the distant metastasis requires the restoration of epithelial status (namely, mesenchymal-to-epithelial transition [MET]) (van Denderen and Thompson 2013; Esposito et al. 2019). The epithelial trait is selected in colonization maybe because it allows cancer cells to adhere to other cells or extracellular matrix, and gain momentum of proliferation. Multiple proteins and pathways may mediate the cell–cell direction and alter the fates of DTCs (Fig. 1), including VCAM-1/integrin (α4β1) (Lu et al. 2011), integrin (α9β1)/Tenascin C (San Martin et al. 2017), Jagged 1/Notch (Sethi et al. 2011), E-cadherin/N-cadherin (Wang et al. 2015), OB-cadherin (Chu et al. 2008), Connexin 43 (Wang et al. 2018), and Glg1/E-selectin (Esposito et al. 2019). Thus, the static “epithelial” or “mesenchymal” state may not be the key determinant of metastasis capacity. Rather, the “plasticity” to swing between these states may be the essential property (Li and Kang 2016; Pastushenko et al. 2018). Indeed, cells expressing hybrid mesenchymal (e.g., vimentin) and epithelial (e.g., E-cadherin) markers were shown to be therapeutic resistant and exhibit enriched cancer stemness (Creighton et al. 2009). The dependency of cell–cell adhesion may be particularly strong in the bone because of tissue rigidity and stiffness, which make anchor-dependent growth a rate-limiting process (Johnson and Suva 2018).

Figure 1.

Epithelial-to-mesenchymal/mesenchymal-to-epithelial transition and functional molecules for attachment in early bone colonization. After cancer cells undergo EMT to facilitate their distal migration, they have to retrieve the epithelial trait for successful colonization in bone. During this process, cancer cells express different attachment proteins to form a physical connection with multitypes of resident cells.

In different cancer types/subtypes, the exit of dormancy and outgrowth of macrometastases may follow distinct mechanisms and growth kinetics. Animal studies comparing bone metastases of ER⁺ MCF7 cell and TN MDA-231 cell disclosed dichotomous patterns at early colonization. Whereas MDA-231 cells closely interact with osteoclasts throughout the colonization process, MCF7 lesions are initially interacting with osteogenic cells (mainly MSCs and osteoblasts) before osteolytic growth starts (Lu et al. 2011; Wang et al. 2017). The distinct dependency on the osteogenic niche in different cancer cells might explain the inconsistent roles of osteogenic cells observed in different publications (Wang et al. 2015; Rossnag et al. 2018; Ren et al. 2019). Consequently, MDA-231 shows faster progression in bone than MCF7, which is in line with the much more rapid growth kinetics in the bone metastases of ER⁻ breast cancer in the clinic (Zhang et al. 2009; Kennecke et al. 2010).

Overt Proliferation

Osteolytic Growth

The paradigm of late-stage, osteolytic bone colonization can be summarized as “osteolytic vicious cycle.” It refers to the mutual activation between cancer cells and osteoclasts: while cancer cells can, directly and indirectly, promote osteoclast maturation, osteoclasts drive osteolysis, which releases growth factors embedded in the bone matrix and reciprocally fuel cancer cell proliferation (Guise et al. 2006; Weilbaecher et al. 2011). Multiple pathways have been implicated in this cycle. Specifically, cancer cells secrete factors such as parathyroid hormone (PTH) or PTH-related peptide (PTHrP), interleukin (IL-1, IL-6, and IL-11), TNF-α, which act on osteoblasts to modulate the expression of genes including RANKL (up-regulation) and OPG (down-regulation) via NFкB pathway (Cappellen et al. 2002). Macrophage-stimulating protein (MSP), CCL2 (Sørlie et al. 2002; Andrade et al. 2018), VEGF-A (Park et al. 2012) as well as extracellular matrix proteinases and transcriptional factors like GLI2 also regulate the osteolytic vicious cycle (Alexaki et al. 2010). The augments of these factors, in turn, boost osteoclast maturation and accelerate bone resorption, leading to widespread bone destruction. In this process, many growth factors deposited in the bone matrix—including TGF-β, IGF1, EGF, and calcium ions—are released and reciprocally stimulate tumor growth (Fig. 2, left; Mundy 2002; Park et al. 2007).

Figure 2.

Overt osteolytic or osteoblastic progression in bone. Cancer cells give rise to osteolytic (left) or osteoblastic (right) outgrowth in bone. The osteolytic metastasis is fueled by reciprocal interactions between cancer cells, osteoblasts, and osteoclasts. This causes bone destruction and tumor progression, which is referred as a “vicious cycle.” The mechanism of osteoblastic metastasis is yet to be further developed. The determinants that tilt the osteoblast–osteoclast balance are not fully understood, either.

Kang et al. (2003) conducted an unbiased study to identify genes driving bone metastasis through in vivo selection on triple-negative breast cancer cells. The major discoveries include chemokine receptor CXCR4 for bone-homing, angiogenic factor fibroblast growth factor-5 (FGF5), osteoclast stimulating factor interleukin-11 (IL-11), osteopontin (OPN), matrix metalloproteinase-1 (MMP1), integrins, and metalloproteinase with thrombospondin motifs 1 (ADAMTS1). The functions of these genes converge on activation of osteoclasts and the vicious cycle, supporting the central role of this process in late-stage bone colonization. One remaining question is whether the expression of these osteolysis-promoting genes is inherent in primary tumors or gradually acquired during the metastatic progression. The study of single-cell progenies of MDA-231 suggested that the osteolysis propensity was genetically determined before cancer cells arrive in bone (Minn et al. 2005). Nevertheless, we cannot rule out the possibility that alternative mechanisms such as epigenetic reprogramming may also contribute to the acquisition of osteolytic phenotype, especially for those indolent bone metastases (Valastyan and Weinberg 2011).

Beyond the protein-coding genes in bone metastases, Ell et al. (2013) unveiled a group of tumor-induced miRNAs as regulators and biomarkers of osteolytic bone metastasis. Specifically, the expression of miR-33a-5p, miR-133a, miR-141-3p, miR-190, and miR-219-5p were found to be suppressive for bone metastases by targeting osteoclast differentiation genes. In contrast, miR-16 and miR-378, which are up-regulated during osteoclastogenesis, were proposed to serve as serum biomarkers for diagnosis of bone metastasis in breast cancer.

Osteoblastic Growth

Osteoblastic bone metastases, characterized by abnormal new bone growth, occur mostly in prostate cancer. Many molecules and pathways mediate both osteoblastic prostate cancer and osteolytic breast cancer, such as TGF-β, fibroblast (FGF), insulin-like (IGF), vascular endothelial (VEGF), and platelet-derived (PDGF) growth factors, Wnt signaling, and bone morphogenetic proteins (BMPs) (Obenauf and Massagué 2015). In fact, this is expected because osteoblast activities are also required in osteolytic metastases, and most of these pathways regulate the growth and differentiation of osteoblasts. Only a few factors are known to be uniquely needed by osteoblastic bone metastases, which will be discussed below. However, the determinants that tilt the osteoblast–osteoclast balance and cause the discrepancy between overt osteoblastic and osteolytic metastases are yet to be identified (Fig. 2, right).

PSA: The prostate surface antigen (PSA) is an important serum biomarker used to monitor prostate cancer tumorigenesis as well as metastatic recurrence (Manca et al. 2017). Active PSA in the bone microenvironment enhances tumor growth by altering signaling specifically in osteoblasts including up-regulation of TGF-β and RANKL, down-regulation of osteoprotegerin, enhancement of Runx2 expression (a transcription factor essential for osteoblast differentiation) and elevation of the Wnt signaling (Chirgwin and Guise 2006). Particularly, PSA cleaves PTHrP, disrupts the indigenous bone resorption, and in turn, allows the enrichment of osteoblasts, which may drive the osteoblastic metastases instead of the osteolytic outgrowth (Iwamura et al. 1996).

Endothelin 1 (ET1): Endothelin-1 is reported to stimulate bone formation and osteoblast proliferation. A strong correlation between the ET1 expression and osteoblastic bone metastases is observed in prostate cancer patients (Nelson et al. 1995). Interestingly, even in breast cancer bone metastases, which are usually osteolytic, the expression of ET1 predisposes an osteoblastic growth (Yin et al. 2003).

Osteomimicry: The Metabolic Adaptation

Recently a growing body of evidence shows that cancer cells rewire their metabolic program to adopt the metastatic organs. Upon colonization, cancer cells tend to resemble the metabolism of local tissues for survival, which is often associated with the adaption to the energy resource, nutrition availability and oxygen level at the metastatic organs (Schild et al. 2018). In bone metastasis, this specific adaption process is termed as osteomimicry, which is proposed to mimic the metabolism of resident cells in bone, especially osteoblast. Osteoblasts favor to metabolize glucose into lactate even in the presence of sufficient oxygen, a process known as aerobic glycolysis or Warburg effect (Karner and Long 2018). Similarly, bone metastases of breast cancer cells are also observed to increase the glycose uptake by up-regulation of glucose transporters (e.g., GLUT1) and release a larger amount of lactate—the products of aerobic glycolysis, compared with lesions in other organs (Hanahan and Weinberg 2011; Lemma et al. 2017). Consistently, pharmacological inhibition of the lactate transporter MCT-1 impairs lactate-induced bone resorption and hence blunts the bone metastases progression.

The reprogrammed metabolism in bone metastases is further exemplified by the high expression of OPN. OPN helps anchor osteoclasts to the mineral matrix of bone and activate glucose and lipid signaling (Shi et al. 2015), indicating an important role in sugar homeostasis and glucose metabolism in the skeleton. The up-regulation of OPN was observed in bone metastases of advanced nasopharyngeal carcinoma as well as breast cancer, and may predict bone metastases in other cancer types (Carlinfante et al. 2003; Kruger et al. 2014; Hou et al. 2015).

Escaping the Immune Surveillance

The role of bone marrow-derived cells, especially the immune cells in metastasis regulation, was long overlooked due to limited clinical samples and lack of suitable experimental models. Bone represents a specific microenvironment with constant bone repair and regreneration but rare microorganism infection. Therefore, it has been traditionally speculated that the innate immune response other than the adaptive response may dominate in the bone microenvironment (Charles and Nakamura 2014). Additionally, the bone microenvironment contains very high numbers of MDSCs and immunosuppressive regulatory T cells (Treg), thereby making bone a permissive environment for DTCs to hide from immunosurveillance (Fujisaki et al. 2011; Zhao et al. 2012). The local immune response might be further blunted as DTCs hijack HSC niches for survival, because the HSC niche is immune-privileged, offering protection from immunological insults (Fujisaki et al. 2011).

Despite the potential immune privilege of the bone environment, evidence also emerges to suggest possible roles of adaptive immune cells in bone metastasis. This is exemplified by a clinical breast cancer analysis highlighting the enrichment of CD56⁺ CD8⁺ T cells and memory CD4⁺ T cells in bone marrow aspirates when DTCs are present in the patients (Feuerer et al. 2001). In another study, depletion of the infiltrating plasmacytoid dendritic cells (pDCs) dampens the development of bone metastasis by expanding CD8⁺ T cells (Bidwell et al. 2012; Sawant et al. 2012). These results urge for a more thorough characterization of the immune landscape and a deeper understanding of the functional roles of various immune cells in different stages of bone colonization.

CONCLUDING REMARKS

Bone tropism is about how cancer cells adapt to bone microenvironment in every key step of bone colonization. The specificity of bone environment exerts unique selection pressure on metastatic cells and hence endows them with special traits and phenotypes for outgrowth in bone (Fig. 3). While the conceptual framework of bone metastasis is well established, the complexities at the molecular level are frequently observed in different biological contexts. This prompts us to develop new insight to synthesize and integrate the existing knowledge.

Figure 3.

Metastatic traits for bone colonization.

Perhaps one useful way to reflect on bone tropism is to use convergent evolution as an analogy: the functionally similar physiological traits may be selected based on structures of different biological ontogeny. For instance, in the evolution of flight, animals from different ancestors independently evolve the “wing structure” to fly (Stayton 2015). Likewise, cancer cells originated from distinct cancer types may be selected to acquire similar traits to colonize bone. Similar to different structural solutions of the wing in different species, cancer cells evolve to use the different toolkit to build up their bone-tropic traits at the molecular level (Fig. 4). This analogy might help us better understand the commonality and diversity of metastatic mechanisms across different cell models and cancer types. Indeed, recent studies have started to characterize metastases from perspectives of evolutionary biology (Turajlic and Swanton 2016; Ullah et al. 2018). These new perspectives may help us nominate the major and common players in bone metastases, accelerate the translation to clinical practice, and revolutionize our understanding of these devastating diseases.

Figure 4.

Bone metastases from different origins are analogous of convergent evolution. (Left) Seagull, bat, and butterfly independently evolved the capacity of flight based on distinct structures of wings. (Right) Likewise, cancer cells originated from different organs may acquire comparable capacity for bone colonization underlying different molecular mechanisms.