Introduction

Advanced cancers are prone to metastasize [1–3]. Although metastatic cells could theoretically intrude any organ, clinical experience demonstrates that they have preference for lung (20–54% incidence) [4], liver (30–70% incidence) [5], bone (20% incidence) [6, 7], brain (15–72% incidence) [8] and adrenal gland (10–50% incidence) [9]. However, the homing to bone is much higher for certain cancers, including breast, prostate, lung and thyroid carcinomas, for which the likelihood to develop bone metastases increases considerably and the incidence is as high as 70% [10].

Visceral metastases are more likely to be fatal, with a long-term survival falling from 90 to around 5% [11]. In contrast, patients with only metastases to bone can survive up to 10 years or more [12–16]. In some patients, bone metastases develop many years after the surgical removal of the primary tumour, suggesting that the osteotropic malignant cells may have a long period of quiescence before developing the secondary lesion [17, 18]. Nevertheless, patients with overt bone metastases present with severe symptoms, including intractable bone pain, nerve compression syndromes, hypercalcaemia and pathological fractures, which considerably reduce the quality of life [10].

Effective treatments for bone metastases are not yet available. Bisphosphonates have demonstrated clinical utility in the palliative treatment of patients with bone metastases. They decrease skeletal morbidity (bone pain, pathological fractures), leading to an improvement of the quality of life but, unfortunately, they do not provide a life-prolonging benefit to patients with advanced cancer [19, 20]. Therefore, development of new therapeutics is required and, to achieve this goal, profound insights into the molecular mechanisms underlying the formation of bone metastases should be provided. Here, we describe cellular features associated with bone metastases, analyse the main molecular determinants known to impact on bone metastasis formation, and discuss our perspective for future molecularly-targeted therapeutic approaches.

Classification and cellular features of bone metastases

In bone metastases, metastatic cells actually intrude the bone marrow cavity where they grow forming a secondary lesion [21]. The mineralised nature of the bone tissue would theoretically prevent growing metastatic cells from forming wide tumours. This circumstance is however circumvented by a tight relationship between metastatic cells and bone cells which leads to microenvironmental changes that promote the enlargement of the bony cavity, thus creating more space suitable for tumour growth [21].

Bone metastases are classified as osteolytic, osteosclerotic or mixed lesions [21, 22] (Fig. 1). Osteolytic metastases are typical of breast cancer. They are caused by tumour-derived factors (Table 1) that stimulate the activity of bone-resorbing cells, the osteoclasts, leading to enhanced bone destruction [21, 25]. Radiographically, osteolytic lesions appear as radiolucent areas, frequently located in the skull and proximal ends of the long bones. Histologically, tumour cells reside in the bone marrow, and are surrounded by a number of osteoclasts, actively degrading bone. The progression of osteolytic lesions ultimately leads to the complete destruction of the bone wall and tumour cells can then extrude the bone cavity infiltrating the surrounding tissues. These osteolytic areas frequently fracture even in the absence of traumas [21, 22].

Fig. 1
figure 1

Classification of bone metastases. Left panel: cartoon depicting a normal bone, in which osteoclasts and osteoblasts function in a concerted manner. Osteoclasts remove the old bone matrix, which, after a poorly defined reverse phase, populated by mononuclear cells (reverse cells) participating to the coupling between osteoclast and osteoblast activity, is replaced by new osteoid released by active osteoblasts, which eventually mineralises. Middel panel: in osteolytic bone metastases, an incredibly high number of osteoclasts are formed which resorb the mineralized matrix destroying the tissue. Right panel: in osteosclerotic bone metastases, numerous osteoblasts appear forming new trabeculae that occlude the bone marrow. This bone matrix has the histological and biochemical features of woven bone and is inordinately deposited in the medullary cavity. Mixed bone metastases (not shown) have both increased osteoclasts and osteoblats in close vicinity

Full size image
Table 1 Factors implicated in osteolytic and osteosclerotic metastases [23, 24]
Full size table

Osteosclerotic metastases are more typical of prostate cancer and are caused by cancer-derived factors (Table 1) that stimulate the differentiation and activity of bone-forming cells, the osteoblasts, thus leading to increased bone formation [21, 22]. Radiographically, osteosclerotic lesions appear as dense areas, often located to the axial skeleton and, particularly, in vertebral bodies and pelvis. Histologically, tumour cells residing in the bone marrow are surrounded by a high number of osteoblasts that form wide trabeculae of woven bone similar to that observed in primary ossification. Tumour-associated woven bone has however a poorly organised microstructure, increasing again the risk of pathological fractures [21, 22].

Bone resorption and bone formation are almost always coupled [26]. This coupling is a dynamic process, which is altered in cancer, thereby leading to skeletal lesions that are predominantly osteolytic or osteoblastic. However, in many instances, bone metastases may consist of mixed lesions [21]. Indeed, it is believed that a bone metastasis may evolve from an osteoblastic to an osteolytic pattern through a continuous process of which we only have a static representation at the time of the radiographical or histological assessment.

Bone metastasis formation consists of a series of interrelated steps that begins with the tropism of cancer cells to the bone through specific migratory and invasive processes, then follows with the growth of cancer cells in the bone marrow which requires that these cells acquire “bone-like” or osteomimetic properties, and ends with bi-directional interactions between cancer cells, osteoclasts and osteoblasts which determine whether the subsequent bone metastasis is osteolytic or osteoblastic. Molecular mechanisms involved in each of these steps are gradually being unravelled, and are potential therapeutic targets for the prevention and treatment of bone metastases. Further complexities are introduced by the fact that the bone microenvironment does not only include the cellular architecture of the bone tissue but also bone marrow-derived haematopoietic progenitors and cancer stem cells that altogether constitute a niche supporting the development of metastases.

Bone tropism

Different molecular mechanisms are responsible for the propensity of cancer cells to metastasize to bone. The chemokine receptor CXCR4 controls the metastatic destination of breast cancer cells in certain organs (lung, liver and bone marrow) where its ligand, the chemokine CXCL-12, is produced in high quantity [27]. Consistent with this, the blockade of CXCR4 using antibodies or a synthetic peptidic antagonist reduces the formation of experimental lung and bone metastases caused by CXCR4-expressing breast or prostate cancer cells [28, 29]. However, the inhibition of chemokine receptors in vivo only partially blocks metastasis formation, suggesting that additional factors are involved in the bone tropism of cancer cells. Indeed, bone-derived cytokine RANKL triggers the migration of RANK-expressing cancer cells in vitro, and osteoprotegerin (OPG), a natural inhibitor of RANK–RANKL interaction, blocks the bone tropism of these cancer cells in vivo [30]. There is also a growing body of evidence from preclinical research showing that integrins mediate metastasis to specific organs. For instance, we have recently shown that αvβ3 integrin overexpression in breast cancer cells enhances bone metastasis incidence in animals, and that a nonpeptide αvβ3 integrin antagonist causes a profound and specific inhibition of bone colonisation by αvβ3-expressing cancer cells in vivo [31]. In a similar vein, bone colonisation by prostate cancer cells has been reported to be mediated by α2β1 integrin [32]. Tumour αvβ3 and α2β1 integrins mediate the attachment of cancer cells to extracellular matrix proteins (BSP and type-I collagen, respectively). It is therefore possible that these integrins act in concert with CXCR4 and RANKL to promote the bone colonisation by cancer cells.

Another important determinant for bone tropism, also linked to integrin functions, is the proto-oncogene c-Src, a non-receptor tyrosine kinase, homologous to the viral oncogene v-Src [33]. It plays a role in cell growth, cytoskeletal remodelling, adhesion and motility [34]. Although ubiquitously expressed, c-Src deficiency appears to affect only the skeleton with no apparent effects on other organs [35]. In many tumours, c-Src is upregulated or hyperactivated thus affecting cancer cell properties linked to proliferation, motility and responses to growth factors [36]. Interestingly, comparing the transcriptomes of human breast cancer bone metastases versus visceral metastases, we observed that a subset of up-regulated genes are under the control of c-Src (MetaBre unpublished observations). In a similar vein, a clone of the parental MDA-MB-231 breast cancer cell line, which shows increased capacity of bone metastasis, also exhibits elevated c-Src protein and tyrosine phosphorylation of c-Src [37]. In addition, reduced c-Src activity in breast cancer cells decreases their malignant phenotype and osteotropism in experimental metastases [37, 38], a circumstance that leads to the administration of c-Src inhibitor or biologic agent, which successfully slowed down the metastatic process [38–40].

Osteomimicry

The definition of osteomimicry is the acquisition by tumour cells of bone cell-like properties, which improve homing, adhesion, proliferation and survival in the bone microenvironment [41, 42] (Fig. 2). This is due to the ability of osteotropic malignant cells to express transcription factors (Runx2, MSX2) [43, 44] that are master regulators of osteoblast differentiation and strong inducers of the expression of bone proteins. For instance, bone matrix proteins, including osteopontin [45], osteocalcin [46], osteonectin [47] and bone sialoprotein II [48], are frequently highly expressed in breast and prostate cancers, which represent tumours with the highest propensity to colonise bone. In addition, these proteins are also highly expressed in human breast cancer cell lines that form experimental bone metastases when injected in immunocompromised mice. Minn et al. [49] have demonstrated that single cell populations obtained from the MDA-MB231 human breast cancer cell line, selected for their high osteotropism, express a unique set of genes, among which transcripts typical of the osteoblast phenotype are prominent. In addition, the ability of these single cell populations to induce bone metastases is greatly enhanced when an osteomimetic gene, the osteopontin, is co-expressed along with either one of the genes of the osteotropic signature. Conversely, the blockade of Runx2 transcription factor in MDA-MB-231 cells, an inducer of osteopontin expression, inhibits breast cancer bone metastasis formation [50].

Fig. 2
figure 2

Multigenic programme of osteomimicry. Osteotropic tumour cells acquire the ability to grow in the bone microenvironment because of their bone cell pseudo-phenotype due to the high expression of transcription factors (Runx2, MSX2), extracellular matrix proteins [osteopontin, osteonectin, osteocalcin, bone sialoprotein II (BSP II)], proteases (cathepsin K), and other bone-related factors (COX2, Cx43, CDH11) that, under physiological conditions, regulate osteoblast differentiation and osteoclast activity

Full size image

A global transcriptome analysis of another osteotropic MDA-MD231 cell variant, BO2, harvested from an experimental bone metastasis after injection of the parental cells in nude mice, has confirmed that several genes among those mostly up- or down-regulated relative to the parental cell line, correspond to genes whose expression is associated with the osteoblast differentiation process [51]. Most interestingly, the proteins encoded by a set of genes, including CDH11, COX-2, CTGF, Cx43 and SPARC, which are overexpressed by BO2 cells and are known to be up-regulated during osteoblast differentiation, are also selectively overexpressed in human breast cancer bone metastases relative to the primary tumour and visceral (liver) metastases. Likewise, proteins encoded by genes underexpressed in BO2 cells and downregulated during osteoblast differentiation, are also selectively underexpressed by human breast cancer bone metastases relative to the primary tumours and visceral (liver) metastases [51].

In addition, a global gene profiling analysis performed on human metastatic tissues from breast carcinomas has established that there is a unique set of genes overexpressed in bone metastases compared to any other type of visceral metastases (liver, lung and brain) examined so far (MetaBre unpublished results). Among these genes, many have an osteomimetic significance and, for instance, the already known osteomimetic bone sialoprotein II is >100-fold overexpressed in bone versus visceral metastases (MetaBre unpublished results). Collectively, these data point to the osteomimetic properties of malignant cells as a key event that favours the development of a secondary lesion in the bone/bone marrow microenvironment.

The question however remains as these osteomimetic properties refer only to the osteoblast phenotype or if they can be extended to the osteoclast phenotype as well. For instance, others and we have shown that cathepsin K, a typical and highly specific osteoclast gene, is overexpressed in human breast cancer cells that metastasize to bone [52, 53]. Therefore, it is conceivable that a multigenic mimicry programme is indispensable for a tumour cell to develop in the bone, and that this programme includes both osteoblastic and osteoclastic genes.

Microenvironmental factors

What makes a tumour cell with a multigenic osteomimicry programme capable of developing a secondary bone lesion is probably associated with the favourable microenvironment [54]. Bone tissue is subjected to a continuous bone remodelling cycle in which osteoclasts resorb the old and damaged bone, and osteoblasts replace this bone with newly formed matrix [55–57]. These cycles are repeated life-long, therefore it is believed that the entire skeletal matrix is replaced several times during life. Many factors regulate bone remodelling, including systemic hormones and local factors, among which interleukins, cytokines, colony-stimulating factors, eicosanoids, the RANKL/OPG axis and PTHrP, play relevant roles [55–58]. In addition, many growth factors are synthesised by the osteoblasts and embedded into the bone matrix, mostly as inactive peptides, during the bone formation phase [55–57]. These factors, which include Transforming growth factor β (TFGβ), Platelet derived growth factor (PDGF), Bone morphogenetic proteins (BMPs) among others, are then released from the matrix during bone resorption and activated both by the low pH created by the osteoclasts to remove the bone mineral, and by a set of proteases present in the microenvironment [59]. Therefore, during bone resorption, the bone/bone marrow microenvironment is enriched by a plethora of agents regulating many cellular activities. In addition, it has to be noted that the bone and the bone marrow are tightly linked, and that bone cells and haematopoietic cells are reciprocally regulated and interconnected in their function [60, 61].

In 1889, Stephan Paget [62] had proposed that environmental factors provide a fertile ground (the soil) in which tumour cells (the seed) can grow. This “Seed and Soil” theory has been largely demonstrated by many studies over the century and is particularly true for osteotropic cancers [63]. Indeed, in bone, years ago the groups of Yoneda et al. [21], have recognized that a vicious cycle is established during the formation of bone metastases, consisting in the perturbation of the microenvironment initiated by the tumour cells that produce many factors stimulating the osteoclasts, with the end point of an increased bone resorption (Fig. 3). In addition, cancer cells secrete bone morphogenetic and Wnt protein antagonists (noggin, DKK-1) that inhibit osteoblast activity which, in turn, enhance the osteolytic pattern of bone metastases [32, 64]. Conversely, cancer cells (especially in the prostate) may release endothelin-1, which stimulates bone formation and inhibits bone resorption, leading to the formation of osteosclerotic lesions [65, 66]. Additional factors like Fibroblast growth factor (FGFs), PDGF and BMPs may be involved as well in the osteoblastic pattern of bone metastases [22]. As stated above, upon bone resorption, osteoclasts then release tumour-seeking factors from the matrix, which in turn stimulate proliferation, survival and migration of the tumour cells. In addition, many bone matrix-derived factors, including TGFβ, PDGF and BMPs, have the ability to induce the epithelial-mesenchyme transition of cancer cells, a key event that greatly enhances their malignant phenotype [67]. Finally, tumour cells have the ability to induce angiogenesis through the secretion of Vascular endothelial growth factor (VEGF) and FGFs [25]. It is therefore clear that the osteomimetic properties of cancer cells and their adaptation to survive in such an enriched environment, are key determinants for the development of bone lesions.

Fig. 3
figure 3

The vicious cycle. The cartoon illustrates the many factors that reciprocally stimulate on one hand osteoclast, osteoblast and vascular cells activity, and on the other hand tumour cell growth and survival in the bone microenvironment

Full size image

The impact of the stem cell niche

It is known that a well-defined stem cell hierarchy is responsible for normal tissue regeneration, which ends up with the repair of the tissues and replacement of worn-out cells in the organs [68]. Stem cells are also implicated in the development of cancer [69, 70]. Genome and population evolution in tumours is quite complex and requires multistep events. Random genetic and epigenetic changes are likely to occur in normal tissues by continuous proliferation, environmental stress, physiological changes and others. This is believed to lead to development of genetic heterogeneity, with mutations that in most cases remain neutral, and cells eventually die or persist without undergoing further transformation. However, some cells may harbour non-neutral mutations that stratify them into stem/progenitor clones or differentiated clones. Further mutations, perturbing the balance of self-renewal over quiescence in the former, or inducing loss of cell cycle control in the latter, may lead to a tissue that contains pre-tumour stem cells. These stem cells may generate a benign tumour or, if subjected to further genetic hits with mutations giving an advantage within the tissue niche, they give rise to a malignant tumour that retains deregulated stem cells. These cells have metastatic potential and, with further genetic hits and mutations that give an adaptive advantage at the secondary site, eventually form secondary tumours with niche dominance of specific clones [71]. Interestingly, it is believed that metastatic cells that achieve the bone marrow, are likely to remain dormant for many years before forming an overt bone metastases [72]. This seems to represent the backdrop why patients with certain cancers, including breast and prostate, develop bone metastases decades after the surgical removal of the primary tumour [73–76]. The molecular determinants influencing tumour stem cell quiescence in the bone marrow environment are yet to be elucidated. Understanding these mechanisms may help devising strategies to cure the disease or at least to induce persistent remission.

It is interesting to note that osteoblasts play a fundamental role in the quiescence of the long-term haematopoietic stem cell, that through cell–cell, cell–matrix and paracrine interactions with a subset of Spindle-shaped, N-cadherin positive osteoblasts (SNO) are kept associated to the endosteal bone surface and prevented to proliferate, until microenvironmental changes promote their detachment from SNO and the progression toward the myeloid and lymphoid lineages [77, 78]. SNO cells are thus suspected to provide a niche for haematopoietic stem cells, which, in combination with specific environmental factors, paracrine/autocrine signals and cellular determinants, provide inhibitory stimuli maintaining the long-term haematopoietic stem cell in a quiescent status [77, 78]. In contrast, the so-called vascular niche, represented by yet to be defined cells residing in the bone marrow sinusoidal system, appears to be involved in the rescue of intense proliferation and transition toward an active status [78]. It is thus tempting to hypothesise that a tumour stem cell intruding the bone marrow could be recruited by the SNO or SNO-like cells and kept quiescent, until environmental changes may restore their ability to enter the cell cycle (Fig. 4). It has to be noted that among the factors released by osteoclasts during bone resorption, there are molecules typically involved in the epithelial-mesenchyme transition (TFGβ, PDGF, BMPs). Therefore, osteoclast activation may play a critical role in the awakening of dormant tumour cells. Many environmental changes could elicit osteoclast activation, among which inflammation is most likely an important condition, which is known to increase bone resorption and reduce bone formation [79–81]. Should this circumstance be confirmed, it is imaginable that administration of anti-inflammatory drugs could represent a valid devise to keep tumour stem cell quiescent and induce persistent remission of the bone metastatic disease. In keeping with this hypothesis, COX2 inhibitors are effective for the treatment of experimental bone metastases [82], therefore we believe plausible that one mechanism of action could be prevention of stem cell activation.

Fig. 4
figure 4

The role of the osteoblast niche in the control of tumour cell growth. Upper panel: among the many osteoblasts lining the bone surface, a few are Spindle-shaped N-cadherin positive osteoblasts (SNO) implicated in the maintenance of haematopoietic stem cell quiescence (osteoblast niche). In this circumstance, scattered tumour cells reaching the bone marrow could be sequestered by the SNO and kept dormant. Middle panel: in the event that environmental conditions change and induce osteoclast formation and bone resorption, factors involved in the epithelial-mesenchyme transition may be released from bone matrix (or directly by the osteoclasts) and activate the quiescent tumour cells. RANKL and inflammatory cytokines are known to be potent osteoclast-inducing factors and could play a role in this context. Lower panel: activated tumour cells are then induced to proliferate progressing into an overt metastases. Whether or not the so-called vascular niche may play a role in tumour cell activation is presently unknown, but would match with the knowledge that the vascular niche is involved in activation of the haematopoietic stem cells and that tumour cells strongly depend on neoangiogenesis for their survival

Full size image

Therapeutic implications and future perspectives

As mentioned above, there are many new exciting pathways that can be harmed to inhibit the development of bone metastases and more will be identified in the near future due to the tremendous effort in the field. New conventional drugs as well as innovative therapeutics are expected to provide effective tools to combat the disease (Table 2), cooperating with bisphosphonates or replacing them in new treatment protocols.

Table 2 Therapeutic agentsa
Full size table

Besides anti-resorptive (bisphosphonates, cathepsin K inhibitor), anti-COX2, anti-CXCL-12, anti-integrin αVβ3, anti-c-Src tyrosine kinase and anti-Runx2 experimental treatments already described above, it is worth mentioning here a few more therapies which have potential for future applications. For instance, denosumab is a fully human monoclonal antibody to RANKL that has high affinity and specificity for RANKL. Its mechanism of action is similar to OPG, which prevents RANKL from binding RANK, thus inhibiting osteoclast formation and reducing the incidence of osteolytic metastases [83]. It is a human IgG2 molecule with a long circulatory residence time, resulting in rapid and sustained decrease of bone resorption. After positive trials in which it was proven to reduce bone loss in postmenopausal osteoporosis, denosumab is now in phase II clinical studies for breast cancer and multiple myeloma bone metastases. It shows efficacy similar to that of bisphosphonates but with a better compliance and ease of treatment [84, 85].

Promising expectation is also provided by preliminary studies showing the efficacy of inhibitors of the TGFβ superfamily, which include natural inhibitors, soluble forms of the receptors, blocking antibodies and small chemical inhibitors directed towards the TGFβ family itself or their receptors [86, 87]. These inhibitors are being tested in a number of diseases whose pathogenesis is associated with misregulation of TGFβ family members, including cancer, muscular dystrophy, obesity and bone diseases, among which bone metastases appear good targets due to the prominent role of TGFβ in the development of both osteolytic and osteoblastic lesions.

Promise also emerges by the use of anti-angiogenetic agents. These have the advantage not to be restricted to specific tumour histotypes or sites of secondary lesions. They target endothelial cells, which are easy to reach by systemic treatment, and have limited side effects because physiologic angiogenesis occurs in adult only in certain circumstances such as the ovarian/uterin cycle and wound healing. It is, therefore, a selective therapy and the targeted endothelial cells are genetically stable, therefore they are unlikely to develop drug resistance. Two groups of compounds have been approved as antiangiogenic monotherapy for solid tumours: small-molecule kinase inhibitors, and humanized anti-VEGF monoclonal antibody [88]. Although they are not being specifically tested for bone metastases in clinical trials, PTK787 (a VEGF receptor tyrosine kinase inhibitor) decreases the formation of osteoblastic lesions in animals bearing C4-2B prostate tumours [89], suggesting these anti-angiogenic drugs have the potential to inhibit tumour spreading to bone similar to their action in other organs. Also interesting is the possibility of anti-angiogenic therapy preventing tumour stem cell activation, which is currently being tested in experimental models of xenograft tumours [90].

Finally, endothelin receptor antagonists are in phase II and III clinical trials for a wide range of solid tumours [91], including prostate cancer. Endothelin 1 is one of the most relevant inducer of prostate cancer osteosclerotic metastases, stimulating osteoblasts and potentiating the effect of growth factors [23]. It is thus possible that current clinical trials may unravel the potential for endothelin receptor antagonists to combat prostate cancer bone metastases.

Concluding remarks

In conclusion, bone metastases are likely to rely on: (i) the ability of cancer cells to exhibit specific receptor–ligand interactions that direct their homing to bone, (ii) the osteomimicry which, by exploiting a bone pseudo-phenotype, leads to tumour development in the bone/bone marrow tissue; (iii) the microenvironmental factors which establish a self-perpetuating vicious cycle and (iv) the stem cell niche which could maintain tumour stem cells dormant until permissive conditions arouse them. Future developments are expected to improve our knowledge on molecular determinants that are critical for bone metastasis formation and develop new therapeutics to combat cancer-induced bone diseases.