Bone microenvironment signals in osteosarcoma development

Abstract

The bone is a complex connective tissue composed of many different cell types such as osteoblasts, osteoclasts, chondrocytes, mesenchymal stem/progenitor cells, hematopoietic cells and endothelial cells, among others. The interaction between them is finely balanced through the processes of bone formation and bone remodeling, which regulates the production and biological activity of many soluble factors and extracellular matrix components needed to maintain the bone homeostasis in terms of cell proliferation, differentiation and apoptosis. Osteosarcoma (OS) emerges in this complex environment as a result of poorly defined oncogenic events arising in osteogenic lineage precursors. Increasing evidence supports that similar to normal development, the bone microenvironment (BME) underlies OS initiation and progression. Here, we recapitulate the physiological processes that regulate bone homeostasis and review the current knowledge about how OS cells and BME communicate and interact, describing how these interactions affect OS cell growth, metastasis, cancer stem cell fate and therapy outcome.

Bone formation and OS pathogenesis

Most bones develop through the process of endochondral bone formation. In this process Mesenchymal stem/progenitor cells (MSC) differentiate into chondrocytes which are responsible for the expansion of the primary ossification centers and control the formation of osteoblasts, the production and mineralization of the bone matrix and the attraction of blood vessels and haematopoietic cells (Fig. 1a). Chondrocyte differentiation and proliferation are regulated by several endocrine, paracrine and autocrine factors including parathyroid hormone-related peptide (PTHrP), Indian hedgehog (IHH), fibroblast growth factors (FGF), bone morphogenic proteins (BMP), sex-determining region Y-box 9 (SOX9), Runt-related transcription factor 2 or growth hormone (GH) [1]. As bones continue to grow, a multilayered cartilaginous structure responsible for bone elongation forms between primary and secondary ossification centers (Fig. 1b). This structure, known as growth plate (GP), disappears after puberty when chondrocyte proliferation decreases and bone completely replaces cartilage [2].

Fig. 1.

Endochondral bone formation. a MSC differentiate to chondrocytes through the formation of cell mass condensations (i). This primary ossification centers expand through the proliferation of chondrocytes (ii) and the subsequent hypertrophy of post-proliferative chondrocytes (iii) which direct the mineralization of the surrounding matrix and control the conversion of progenitor perichondral cells to osteoblasts that form the bone collar. After that, hypertrophic chondrocytes undergo apoptosis and the newly formed osteoblasts invade the cartilage mold and form the definitive bone matrix (iv). This mold also creates access to blood vessels within the matrix. b GP are multilayered structures divided into: (1) a reserve layer of resting cells that assure the germinal structure, (2) a proliferative layer of rapidly proliferating chondrocytes, (3) a hypertrophic layer of non-proliferating chondrocytes synthesizing collagen X and proteoglycans, (4) a degenerative layer where chondrocytes undergo apoptosis and (5) an ossification layer (primary spongiosa) where MSC-derived osteoblasts replace the matrix with mineralized bone. The marrow space formed below the GP provides niches for hematopoiesis. Most OS develop during puberty in the GPs of long bones as a result of the oncogenic transformation of MSC or their derived osteogenic progenitors

OS is an intra-osseous malignant neoplasm that preferentially arises in areas of actively growing bone. Hence, the majority of these tumors develop during puberty in the metaphysis of long bones in close proximity to the GPs (Fig. 1b). OS is characterized by a severe chromosomal instability, which leads to the presence of abundant genetic alterations, the mutations in the retinoblastoma (RB) and the P53 tumor suppressor genes being the most common and significant alterations [3, 4]. Experimental evidence strongly suggests that either MSC or MSC-derived osteogenic cell types may represent the cell of origin that acquires these genetic alterations and initiate OS under the influence of proper microenvironmental signals [5–9] (Fig. 1b). The relevance of the BME signals, defining the OS phenotype, was recently evidenced in a comparison of the sarcomas that originated from P53 ^−/− and P53 ^−/− RB ^−/− MSC upon ectopic or orthotopic inoculation. These cells give rise to leiomyosarcoma-like tumors when injected subcutaneously into immunodeficient mice [8, 10]. However, intra-bone injection or inoculation into an ectopic model of bone-mimicking material, together with BME factors like BMP2, consistently generated osteoblastic OS, thereby demonstrating that BME signals play a role in defining the sarcoma phenotype [11]. In addition, these kinds of models of OS developed from the putative cells of origin (transformed MSC) have been used to track the distribution of OS cells in the BME during different stages of disease progression (Fig. 2).

Fig. 2.

In vivo distribution of OS cells within the BME. Histological sections of OS generated orthotopically in immunocompromised NOD/SCID mice, stained with hematoxylin and eosin (HE) or specific antibodies (anti-GFP; anti-human vimentin). a Images of incipient OS originated from GFP-expressing murine P53^−/− RB^−/− MSC, which initiate OS with an incidence of 80 % [11], located in either tibial metaphysis or epiphysis (upper panels) or in tibial diaphysis (lower panels). OS cells (GFP-positive cells) at this stage distribute peripherally to BM cavities, in close relation with host bone, osteoblasts, blood sinusoids and blood vessels. b Images of invasive OS generated from human primary OS cells, taken at the tibial diaphysis level whose integrity is damaged by the invading tumor. Tumor cells inside the mouse setting were identified using an anti-human vimentin antibody, known to react with this particular tumor. This immunohistochemistry shows that tumor cells can be found in close proximity to host-derived tumor stroma and blood vessels. c Formation of metastasis in the lungs of mice intra-bone inoculated with GFP-expressing P53^−/−RB^−/− MSC. The metastatic process includes the extravasation of GFP+ tumoral cells from blood vessels (left panel) and the migration, homing and proliferation of metastatic GFP+ cells (middle panel). Original magnification is shown. B bone, BM bone marrow, BV blood vessels, T tumor mass, M muscle, 1 osteocytes, 2 osteoblasts, 3 blood sinusoids, 4 erythroid precursors, 5 granulocytes, 6 megakaryocytes, 7 tumor stroma

An interesting hypothesis postulates that the incidence pattern of OS could be related to an imbalance between the demand of progenitor cells in periods of increased bone formation and remodeling and the limited expansion capacity of normal stem cells [12]. This would ultimately lead to prevalent proliferation of pre-malignant stem cells, which would occupy the same specific niches as the cells involved in maintenance of bone homeostasis. In normal conditions, osteoblasts, osteoclasts and bone marrow (BM) mesenchymal precursors invade the primary ossification center in response to factors secreted by the GP chondrocytes, which finely regulate the behavior of these invading cells. Likewise, abnormal precursors could be subject to a variety of environmental stimuli, which have the ability to modulate tumor progress within GP.

The fact that an increased incidence of OS correlates with periods of accelerated bone growth has suggested a role for growth-related factors, such as GH, in OS generation. In addition to GH systemic release, a local production of this hormone has been described in metaphyseal region of normal growth plates and in 25 % of canine OS samples [13]. Although a direct action of GH on stem cells has been proposed, GH is believed to operate through local IGF-1 production, which seems to promote expansion of proliferating cells and osteogenic differentiation in an autocrine/paracrine manner [14].

Additional regulatory agents present in the GP milieu, such as IHH, FGF, BMP or wingless-type MMTV integration site (WNT) factors have been linked to OS development (see below). Furthermore, high-mobility group box 1 protein, secreted by hypertrophic chondrocytes, facilitates invasion by endothelial cells, osteoclasts and osteoblasts. This factor has been shown to induce OS proliferation and invasion and to promote OS resistance to chemotherapy [15]. On the other hand, negative regulators also present in the GP might counterbalance a putative OS-supporting microenvironment. Accordingly, anti-angiogenic pigment epithelium-derived factor, mainly expressed in avascular areas of the GP, plays a role in hampering OS invasion and directly promotes OS cell apoptosis [16].

Dysregulation of bone homeostasis by bone cell–tumor interactions

The rich BME constitutes a fertile “soil” that favors the growth of both primary and metastatic tumoral “seeds” [17]. A paradigm of the interaction between bone niche and cancer cells is the so-called “vicious cycle”. In physiological conditions, the bone remodeling process couples osteoclast-mediated bone resorption and osteoblast-promoted bone formation to maintain bone homeostasis. In this tightly regulated mechanism, osteoclast differentiation and activation are controlled through the balanced expression of the receptor activator of nuclear factor κB ligand (RANKL) on osteoblasts and stromal cells, and its receptor (RANK) on the osteoclast surface [18]. However, the development of primary bone tumors, including OS, or the invasion of the bone tissue with distant metastasis of OS and other tumors, disrupts this balance and induces a “vicious cycle” between osteoclasts, stromal cells, osteoblasts and cancer cells. OS or metastatic cells release factors like PTHrP and Interleukin (IL) 11 which may act either by directly activating osteoclasts or by stimulating the expression of RANKL in osteoblasts [17, 19, 20]. OS or metastatic cells may also respond to environmental conditions like hypoxia or acidosis by producing osteoblast-stimulatory factors like vascular endothelial growth factor (VEGF), platelet-derived growth factor (PDGF) and endothelin-1 (EDN1) [21]. In addition, factors released by cancer cells or resulting from bone lysis, such as IL1, IL6, tumor necrosis factor α, PTHrP or TGFβ, mediate RANK expression on the osteoclast surface. In these conditions, the interaction between RANKL and RANK mediates the activation and maturation of osteoclasts [19], resulting in a dysregulated bone lysis activity and the release of bone matrix growth factors such as TGFβ, IGF1, FGF or BMP, which in turn promote tumor cell proliferation and further bone destruction [22–24] (Fig. 3). This “vicious cycle” between cancer cells and the BME was first described in bone metastasis, but there is evidence supporting the notion that OS cells mediate bone destruction by stimulating osteoclast development and that this osteoclast activity is related to primary OS aggressiveness. These data therefore suggest that such “vicious cycle” also operates at primary OS sites [20, 25]. This OS-induced paracrine response seems to be mediated by both RANKL–RANK-dependent and independent mechanisms [26–28] and osteoclastic cells might reciprocally modulate the osteoblastic behavior of OS cells [29].

Fig. 3.

Interrelation between bone environment cell types and OS. The presence of OS metastatic cells with osteolytic potential in the bone microenvironment initiates a “vicious cycle” in which tumor cells produce factors (PTHrP, TGFβ or IL11) that stimulate the activation of osteoclast through a RANKL–RANK-mediated interaction between osteoblast and osteoclast. This activation results in a dysregulated bone lysis and increased release of growth factors from the bone matrix (BMP, TGFβ, IGF1 or FGF) which, in turn, promotes tumor cell proliferation (blue arrows). In a similar way, primary OS cells might also induce a similar cycle of osteoclastic activity and bone lysis associated with enhanced tumor aggressiveness. OS cells may also induce the production of pro-tumorogenic molecules (lactate, VEGF and IL6) from the bone environment MSC (magenta and orange arrows). In addition, immune-stimulatory (M1) and immune-suppressive (M2) tumor-associated macrophages and other immune cells play a role in OS progression through cytokine- and chemokine-mediated signaling (gray and red arrows). Physical conditions of the tumor microenvironment also play relevant roles in OS development (green arrows). Thus, hypoxia (through HIF1-mediated signaling) and acidic pH favor angiogenesis, stemness and metastatic behavior, stimulate OS cells to produce osteoblast-stimulatory factors (VEGF, PDGF and EDN1) and may also cooperate with TGFβ signaling to potentiate the “vicious cycle” of bone tumors. Finally, extracellular calcium-mediated signaling also contributes to the “vicious cycle” through the induction of PTHrP production or the generation of EMV containing pro-osteoclastic cargo. The final effect of the different bone cells–tumor interactions on OS development is indicated according to the above-indicated color code

RANK ligand-stimulated osteoclast activity is a specific microenvironmental factor related to bone tumors and its high expression in OS has been related with a poor response to chemotherapy [30]. However, it has been reported that targeting RANKL with a gene silencing strategy in rodents does not affect tumor growth, but may increase the response to chemotherapy [31]. Another approach consists of direct inhibition of osteoclast development using bisphosphonates [32]. In this case, zoledronic acid administration has been related to a better response to chemotherapeutics [33] and it is now under clinical trials as adjuvant treatment in combination with chemotherapy (clinicaltrials.gov identifier NCT00691236). In addition, the administration of the RANKL-decoy receptor osteoprotegerin (OPG) by gene transfer in preclinical models of OS was effective in preventing the formation of osteolytic lesions and reduced tumor incidence and growth, eventually resulting in improved survival rates [23]. However, OPG is also able to efficiently bind TRAIL, preventing its association with the transmembrane receptors and therefore counteracting its pro-apoptotic activity [34]. Therefore, OPG secreted by BME cell types may help tumor cells to evade apoptotic cell death mediated by TRAIL produced by host immune cells [35]. Accordingly, the potential detrimental effects of inhibiting TRAIL-mediated tumor cell apoptosis should be kept in mind when designing OPG-based therapies. In this regard, the use of the soluble chimeric protein RANK-Fc, which blocks RANKL-mediated bone resorption but does not interact with TRAIL and is effective in preventing OS development in mouse models, may represent an interesting alternative therapeutic strategy [36]. More recently, the use of a monoclonal antibody to RANKL (denosumab), used to inhibit bone resorption by osteoclasts in osteoporosis and bone metastasis, in combination with the tyrosine kinase inhibitor sorafenib, has achieved complete remission in a case of unresectable OS [37]. However, a novel role for osteoclasts in the prevention of metastasis in OS has also been reported [38]. The authors reconcile these seemingly conflicting observations about the role of the osteoclast in OS progression, suggesting that in initial phases of the disease, osteoclasts may promote tumor growth through the release of growth factors in the “vicious cycle” process that accompanies OS-induced bone remodeling. Thus, OS cells may be discouraged from leaving the primary site due to the abundance of growth factors released by osteoclasts during this period and, therefore, the persistence of osteoclasts at this stage could be predicted to suppress metastasis. In later stages, OS cells would have gained further genetic/epigenetic alterations leading to the acquisition of the ability to inhibit osteoclastogenesis, related to an increased metastatic potential. Thus, the stimulation of osteoclastogenesis in advanced OS would prevent metastasis [39].

Cell–cell interactions mediated by the erythropoietin-producing hepatoma (EPH) receptor tyrosine kinases and their ligands called ephrins (EPH receptor interacting proteins) also regulate relevant bone-related processes like MSC migration and bone remodeling [40]. Different members of these families of ligands/receptors mediate the communication between osteoblast and osteoclast lineages and may act as facilitators (“coupling stimulators”; e.g., ephrin B2) or antagonizers (“coupling inhibitors”; e.g., ephrin A2) of the transition of bone resorption and bone formation [40]. Genome-wide microarray analysis of OS samples revealed an increased expression of the EPHA2 receptor and its ligand ephrin A1 [41]. Likewise, ephrin B1 expression in OS cells was correlated with poor prognosis, whereas the staining pattern of ephrin A4 could associate either with poor or favorable prognosis when tumors display a cytoplasmic or cytoplasmic/nuclear pattern, respectively [42, 43].

The generation of extracellular membrane vesicles (EMV) could constitute an effective form of horizontal communication between OS tumor cells and their microenvironment. The formation of EMV has been recently detected in a bioluminescent OS orthotopic mouse model as well as in the human OS cell lines 143B and HOS [44]. These vesicles, generated by tumor cells in response to increased intracellular calcium levels, contain bioactive pro-osteoclastic cargo, including matrix metalloproteinases 1 and 13 (MMP1 and 13), TGFβ, CD-9 and RANKL, and could contribute to the bone destruction and the tumor–bone “vicious cycle” [44]. Cell fusion represents another form of interaction between cancer cells and their microenvironment. This phenomenon was observed in co-cultures of human OS cell lines and mouse embryonic fibroblasts-derived myofibroblasts at a rate of 1–2 % and the authors suggest that these could be the origin of the multinucleated giant cells observed in OS [45].

Role of tumor-associated stroma on OS development

Mesenchymal stem/progenitor cells are able to migrate to tumor sites where they may differentiate into tumor-promoting cell types such as cancer-associated fibroblasts/myofibroblasts or macrophage-like cells [46–48]. MSC and their derived cell types may support tumor growth, angiogenesis, motility, metastasis, chemotherapy resistance as well as constitute a niche for CSCs in a wide variety of tumor types [49]. The immunomodulatory properties of MSC may also play a role in favoring a tumor-promoting environment [46], although human MSC lose their immune suppressive properties when they become transformed and acquire the ability to initiate sarcomas [50].

Bone marrow-derived MSC have been isolated at high frequencies from human OS samples, demonstrating that these cells represent a major component of the OS microenvironment. Nevertheless, these OS-derived MSC do not show tumor-related chromosomal aberrations and are non-malignant [51]. Despite the absence of such alterations, several reports suggest a tumor-promoting activity of MSC associated with the OS microenvironment. Thus, intravenously inoculated human MSC were able to target orthotopic xenografts developed from the human OS cell line Saos-2 in a mechanism involving tumor-produced CXCL12. Importantly, the presence of MSC in the tumor environment promotes OS growth and metastasis, at least in part, through the production of CCL5 [52]. To explain this pro-tumorogenic activity of MSCs, a cross talk mechanism was proposed, in which OS-produced TGFβ played a role in the inhibition of the osteogenic differentiation of MSC and induced the production of IL6 and VEGF. These MSC-produced cytokines are in turn capable of enhancing OS tumorogenic properties. Thus, secreted IL6 acts on OS cells by activating the signal transducer and activator of transcription 3 (STAT3), which stimulates proliferation, migration, metastasis and apoptosis resistance [52–55] (Fig. 3). Using rat OS models, several groups have also found evidence of the role of MSC in promoting the emergence of pulmonary metastases [56, 57] or in the modulation of the hypoxia-induced desensitization of β2-adrenergic receptors in OS cells, thus contributing to support tumor progression [58].

In addition, MSC play a role as OS cell feeders after being metabolically reprogrammed by OS cells to a glycolytic phenotype (Warburg metabolism). These reprogrammed MSC enhance lactate secretion though an increased expression of monocarboxylate transporter (MCT) 4, and OS cells upload lactate through MCT1, eventually raising their mitochondrial respiration to facilitate migration [59] (Fig. 3).

Cross talk signaling between BME and OS

Bone tissue is composed of a mineralized ECM and specific cell types, under the control of local and systemic factors that are released during the bone resorptive phase, including many that act as growth factors for OS cells, such as, TGFβ, IGF, FGF and BMP. Conversely, OS cells act reciprocally on the composition and activity of surrounding microenvironment elements to promote tumor growth. Here, we focus on the main environmental signals that potentially regulate the fate of OS.

TGFβ/BMP signaling

TGFβ released from the disrupted ECM is involved in the establishment of the “vicious cycle” between cancer cells and tumor microenvironment (see above). Similarly, TGFβ promotes the growth of OS cell lines in vitro [60], suggesting a further role for this factor in OS development. Accordingly, levels of TGFβ1 are selectively increased in high-grade OS [61], and TGFβ3 expression tightly correlates with OS progression [62]. Likewise, SMAD family members 1 and 2, intracellular effectors for BMP and TGFβ, respectively, are upregulated in 70 % of human OS samples [63]. Conversely, inhibitory SMAD7 induces primary tumor growth slowdown in animal models of OS. This antitumor effect is associated with the presence of lower levels of osteolytic factors such as RANKL, suggesting that Smad7 overexpression affects the “vicious cycle” between tumor cells and BME [64]. TGFβ also participates in the metastatic dissemination of OS, partially by regulating cell migration and invasion as shown in animal models [64]. In OS patients, expression of both TGFβ1 and TGFβ3 in primary tumor associates with higher incidence of metastasis [65], and those patients with metastatic disease present elevated levels of TGFβ in their serum [66]. On the other hand, TGFβ also induces a negative feedback loop in OS MG63 cells through a mechanism by which TGFβ induces the secretion of the heat shock protein HSP 90β, which in turn binds TGFβ1 latent complex and inhibits TGFβ1 signaling [67].

The role of BMP in OS development has been recently reviewed [68]. Despite the attempts to establish a relationship between the expression of BMPs and OS clinical outcome, a clear correlation with OS prognosis has not been determined so far. Also in this line, few studies have investigated the role of BMP signaling on OS progression and their results were somewhat contradictory. BMP2 fails to induce terminal differentiation in OS cell lines that harbor differentiation defects and instead promote tumor growth [69]. Similarly, BMP2 has also been reported to increase cell migration by either enhancing incorporation of integrin β1 into lipid rafts or modulating fibronectin-integrin β1 signaling in murine osteoblastic and OS cell lines [70]. On the other hand, BMP2 treatment decreases the tumorogenic potential of the OS cell line OS99-1 by inducing the upregulation of osteogenic genes and in vivo bone formation [71]. Similarly, BMP9 overexpression suppresses in vitro growth and invasion properties of OS cell lines 143B and MG63 [72].

Our data using primary transformed MSC show that BMP2 treatment enhances the expression of osteogenic markers through a mechanism mediated, at least in part, by WNT/β-catenin signaling [11]. Likewise, BMP2 is able to induce in vivo osteogenic differentiation and OS development upon ectopic subcutaneous inoculation of P53^−/−RB^−/− MSC without significant differences in tumor latency, thus mimicking the effects observed upon orthotopic inoculation of these cells [11]. These data suggest that BMP2 produced by cell types present in the BME acts on potential OS initiating cells to contribute to the acquisition of the OS phenotype without decreasing their tumorogenic potential.

WNT signaling

Like BMP, despite its role in physiological osteogenic differentiation, the role of WNT in OS development is currently under debate. On one hand, some authors report an increment in WNT activity in OS samples. Hence, an increase in positive WNT regulators like low-density lipoprotein receptor-related protein 5 and β-catenin correlates with lower event-free survival and increased lung metastasis, respectively, and WNT1-negative patients display a trend toward a better prognosis [73]. Furthermore, WNT1, WNT4, WNT5A, WNT7A and several Frizzled class receptor isoforms are upregulated in several OS cell lines, while WNT3A, β-catenin and lymphoid enhancer-binding factor 1 are specifically induced in human OS cells Saos2 [73, 74]. In vitro, WNT5A enhances the migration of human OS cells through WNT non-canonical pathway, by activating phosphatidylinositol-4,5-bisphosphate 3-kinase/V-Akt murine thymoma viral oncogene homolog (PI3K/AKT) [75]. In agreement with a positive role of WNT in OS, WNT-β-catenin inhibitors are often downregulated in OS. Thus, WNT inhibitory factor 1 (WIF1) is usually silenced in OS through epigenetic mechanisms, and the inhibitory factors Dickkopf protein (DKK) 3 and frizzled-related protein are downregulated in OS samples. In vitro data and preclinical models provide support for this notion, since forced expression of either DKK3 or WIF1 precludes in vivo OS growth and impairs invasion and mobility of tumor cells [73, 76, 77]. Finally, this pathway is involved in OS resistance to chemotherapy, as shown by β-catenin deletion leading to an increased sensitivity to methotrexate of Saos-2 cells [73]. Moreover, the inactivation of NOTCH and WNT pathways in OS cell lines yields sensitization to chemotherapeutic drugs [73]. Therefore, a number of therapies have been assayed to target these pathways using downregulation, inactivation or silencing techniques [78].

In sharp contrast, recent studies have reported the absence of nuclear β-catenin staining in OS samples and cell lines suggesting that the WNT signaling pathway is inactivated in human OS [79, 80]. In this regard, the WNT inhibitor DKK1 enhances tumorogenesis of MOSJ OS cells in orthotopic murine models, likely through the induction of aldehyde dehydrogenase (ALDH) 1 in these cells [81]. In line with this, neutralizing DKK1 antibodies inhibit proliferation of MG63 OS cells [74].

Ligands of tyrosine kinase receptors

Several ligands (IGF1/2, VEGF, PDGF and FGF) activate cell membrane tyrosine kinases receptors (IGF1/2R, VEGFR, PDGFR and FGFR) triggering shared signal transduction pathways [PI3K/AKT, mitogen-activated protein kinases (MAPK), and mammalian target of rapamycin (mTOR)], which are actively involved in the pathogenesis of several sarcomas. Substantial levels of expression of many of these receptors have been documented in primary patient OS cells and in OS cell lines, suggesting that these receptors may contribute to OS development [82].

FGF

Suggesting a positive role for FGF in OS development, it has been reported that growth factors of this family stimulate the proliferation of human OS cells [83] and that treatment with FGF inhibitors leads to decreased proliferation in mouse OS cells [84]. Notably, FGF and other signaling molecules, like leukemia-inhibitory factor, are highly expressed in OS stroma and suppress osteogenic differentiation through the hyperactivation of extracellular signal-related kinases (ERK) 1/2. These factors also enhance proliferation and migratory activity and confer drug resistance, thus contributing to maintain an immature and aggressive phenotype in OS [85]. In addition, ERK, C-Jun N-terminal kinase and p38 MAPK kinases have been found to regulate cell proliferation, differentiation invasion and migration in OS, and their inhibition with arsenic trioxide inhibits OS cell invasiveness [86]. Finally, it has been shown that parathyroid hormone interferes with FGF signaling to modify the proteoglycan ECM content and regulate OS cell migration [87].

IGF

OS tumors and/or cell lines express IGF1, IGF1R and IGF2R and appear to require IGF1 for growth [82, 88]. Importantly, IGF1 expression in surgical primary OS is associated with a more aggressive tumor type, metastasis, chemotherapy refractoriness and lower survival [89]. Moreover, inhibition of the GH/IGF1 axis prevented metastases in a mouse model of OS [90] and the use of inhibitors of IGF1R and its downstream pathways has shown promise in preclinical models of OS [91]. Therefore, phase I and II trials of the IGF1R antagonists figitumumab, cixutumumab, AMG479, R1507 and SCH 717454, either alone or in combination with other agents, are currently under clinical investigation for patients with sarcomas. Preliminary results show variability in responsiveness to these therapies and the biological basis for these differences in response remains unclear [92].

VEGF and angiogenic signaling

The binding of the pro-angiogenic protein VEGF with its receptor VEGFR results in the expression of proteins such as MMP and plasmin proteases, which degrade the ECM and allow the formation of new vessels. The amplification, protein overexpression and high serum levels of VEGF and VEGFR in OS have been associated with the presence of lung metastasis and poor overall survival [93]. In addition to increased angiogenesis, another possible explanation for the aggressive behavior of VEGF-positive OS is based on the detection of a subpopulation of highly aggressive OS cells, showing evidence for autocrine signaling through VEGFR1, which drives a feedforward loop resulting in upregulation of VEGF production and increased cell proliferation and survival [94]. MMP2, 9 and 19 and the erythroblastosis virus E26 oncogene homolog 1 have also been implicated in the ability of OS cells to invade [95, 96]. Within this context, RECK, a membrane-bound protein which inhibits MMP9 and MMP2 and is downregulated in OS, reduces the invasiveness of OS cell lines when overexpressed [97]. Moreover, OS cells are able to directly induce the proliferation of endothelial cells. For instance, human aortic endothelial cells proliferate and form capillary-like structures in response to a Ying Yang 1-dependent signal in the OS cell line SaOS [98]. In addition, vasculogenic mimicry, a phenomenon in which tumor cells generate non-endothelial microvascular channels that mimic the function of blood vessels, was observed in 23 % of OS samples, and may represent an unfavorable prognostic factor [99].

A number of agents directed against VEGF signaling as well as other antiangiogenic compounds, such as sorafenib, angiostatin, bevacizumab or endostatin, have been tested in OS with promising pre-clinical results, although they could not be reproduced in some instances within clinical trials [100].

PDGF

Finally, the coordinated expression of PDGF and its receptor has also been associated with OS progression [101]. Interestingly, it has been reported that OS cells have a platelet-aggregating activity and that these OS–platelet interactions induce the release of PDGF from platelets, which in turn promotes the proliferation of OS cells [102].

Hedgehog signaling

The hedgehog (HH) ligand IHH, its receptor patched 1 and the target gene glioblastoma-associated (GLI) 1 are highly increased in many primary OS, suggesting the existence of a ligand-dependent activation of the pathway in these tumors [103]. Similarly, the expression of GLI2 correlates with poor prognosis in these patients, and in vitro knockdown of GLI2 leads to an enhanced sensitivity of OS cell lines to cytotoxic agents [76, 103]. Additionally, mouse models with increased HH signaling in mature osteoblasts in a permissive background develop OS with high penetrance. These pro-tumorogenic effects of HH signaling are mediated by the overexpression of the hippo signaling effector Yes-associated protein 1 (YAP1) and the long non-coding RNA H19 [104]. Moreover, YAP1 is aberrantly expressed in human OS and its knockdown results in a decreased proliferation and invasion in OS cell lines [105].

NOTCH signaling

NOTCH signaling could also play a role as OS driver event [106] and the activation of this pathway seems to play a role in OS progression. Primary human OS show transcriptional up-regulation of NOTCH receptors, the NOTCH ligand jagged-1 and NOTCH target transcription factors hairy and enhancer of split (HES) 1 and HES-related protein 2 [107]. Furthermore, the expression of HES1 was inversely correlated with survival [108]. Compared to normal human osteoblasts and non-metastatic OS cell lines, metastatic OS cell lines display higher levels of NOTCH receptors, NOTCH ligand delta-like protein 1 and HES1, and NOTCH signaling seems to play a relevant role in the promotion of OS metastasis [108]. Importantly, the inhibition of the NOTCH pathway suppresses OS growth and metastasis [107–109].

Regulation of tumorogenesis by microRNAs

Changes in the expression profile of a number of microRNAs (miRs) with potential roles in tumor development have been reported in OS [110, 111]. Thus, the miR cluster at the chromosome 14q32 locus, involved in regulation of C-MYC oncogene, is frequently downregulated in OS. Expression of miRs belonging to this cluster is significantly associated with higher recurrence-free survival [112]. The expression of miR-34 family (mir-34a, miR-34b and miR-34c), transcriptional targets of P53, is also decreased in OS biopsies. miR-34a has been shown to inhibit OS cell proliferation and metastasis, possibly via downregulation of the C-MET gene [113]. In addition, miR-16, which inhibits the ERK1/2 pathway and silences IGF1R leading to cell growth arrest, is also diminished in OS cell lines and patient samples [111].

Some miRs appear to regulate tumor response to chemotherapy. Several reports have highlighted the existence of significant differences in miRs expression profiles between good and poor responders to chemotherapeutic agents in OS. For instance, miR-132 levels are reduced in good responders to ifosfamide treatment [114], while a significant increase in miR-221 and miR-140 are found in tumors refractory to cisplatin and methotrexate, respectively. Similarly, elevated miR-210 associates with low response to chemotherapy and poor survival rate [111].

Metastatic capacity of OS cells can also be modulated by changes in miRs expression profiles. Thus, miR-17-92 cluster overexpression in OS cell lines, induced by C-MYC and E2F activity, downregulates FAS expression and promotes OS lung metastasis [115]. Likewise, there is a correlation between reduction in miR-183 levels and the development of OS lung metastasis, possibly due to the role of this miR in the inhibition of Ezrin gene, a known regulator of cell motility and invasion in OS (see below) [116]. Similarly, both miR-335 and miR-340, whose expression is often decreased in OS, target the expression of ROCK, a kinase which antagonizes OS cell migration and invasion in vitro [117]. Coordinated regulatory networks, comprising miRs and transcription factors, have also been identified in proliferative OS cells. This is the case for miR-9-5p, miR-138, and miR-214 with SP1 and MYC transcription factors. These regulatory networks are involved in nuclear factor kB (NFκB), RB-mediated signaling and focal adhesion processes [118].

Finally, miRs also influence interactions between OS cells and the BME. Thus, a recent report showed that miR-19a can be transferred between OS cells to the osteoblast-like cell line MC3T3 via tunneling nanotubes [119]. The fact that this miR is upregulated in human OS suggests that this type of tumor-environment communication could play a role in OS development.

Role of immune system cells and cytokine-mediated signaling

Recent research has revealed several mechanisms of cross talk signaling between OS cells and cells of the immune system. Thus, IL6 and natural killer T cells have been found to play a rate-limiting role in the development of radiation-induced OS in mouse models [120]. Moreover, tumor expression of B7-H3, a coinhibitory T cell regulator, promotes OS cell invasion and correlates with patients’ survival and metastasis [121]. The interaction between immune cells and OS through IL17A-mediated signaling could also play a role in metastasis. Thus, the interaction of IL17A and its receptor induced the expression of VEGF, MMP9 and CXCR4 and promoted metastasis in OS cells in a STAT3-dependent manner [122].

Most chemotherapeutic drugs induce a transient state of systemic immunosuppression. In OS, early lymphocyte recovery after chemotherapy predicts better outcome [123]. In addition, patients with postoperative infections, and therefore activated immune response, show better survival rates [124]. Therefore, clinical observations suggest that immunotherapies could be potentially useful for OS. Accordingly, enhanced systemic immune responses reduce regulatory T lymphocytes in metastatic lesions and inhibit metastatic growth [125]. Another immune-based strategy, which is currently being developed to target OS, consists of the genetic modification of T cells to render them antigen specific or resistant to inhibitory factors or to increase their ability to home to tumor sites [126]. For instance, T cells modified to express human epidermal growth factor receptor 2-specific chimeric antigen receptor are able to target drug-resistant CSC subpopulations in OS cell lines [127].

The relationship between macrophages and OS has also been studied [39]. The ratio of immune-stimulatory M1 and the immune-suppressive M2 tumor associated macrophages in OS seems to play an important role in the development of metastatic behavior (Fig. 3). Thus, the constitutive expression of M1 macrophages seems to prevent metastasis development in the initial stages of the disease [39]. Therefore, macrophage activating drugs have been used in combination with chemotherapy resulting in improved survival [128].

Signaling through chemokines also plays a relevant role in the interaction between OS and its environment. The expression of several chemokine receptors, including CXCR4, CCR5 CCR7 and CCR10, has been detected in human OS samples, with CXCR4 being the receptor showing a stronger association with potential metastatic development and adverse clinical outcome [129]. Importantly, the interaction between CCL5 and its receptor, CCR5, promotes the migration of human OS though the ERK-mediated activation of NFκB, which eventually results in the activation of αvβ3 integrin [130]. Likewise, the CCL5/CCR5 axis induces VEGF expression in human OS, though the PKCδ/c-Src/hypoxia inducible factor (HIF)-1α signaling pathway and secreted VEGF, in turn, induce angiogenesis in the OS microenvironment [131].

Proteins of the CNN (connective tissue growth factor/cysteine-rich protein/nephroblastoma overexpressed gene) family are ECM cytokines that regulate important functions of resident cells of the BME [132]. Studies have shown that several components of the CNN family are overexpressed in OS samples, when compared to normal bone, and their expression levels correlate with poor prognosis (CNN1), tumor stage (CNN1 and CNN4) and risk of metastasis development (CNN1 and CNN3) [133, 134]. It could also be speculated that these CNN proteins secreted by primary bone tumors could promote the RANKL-induced osteolytic “vicious cycle” to favor tumor growth as described for CNN proteins secreted by bone metastasis of different types of tumors [135].

Signaling pathways specifically involved in metastasis of OS

OS cells that have acquired metastatic properties enter the vasculature system and can circulate throughout the organism, extravasate, home and metastasize preferentially in the lungs (Fig. 3c) [136]. Among the proteins with a more documented role in OS metastasis is ezrin, a member of the ezrin/radixin/moesin family of cytoplasmic peripheral membrane proteins, which links the actin cytoskeleton with membrane proteins. High ezrin expression in OS is associated with poor overall survival and increased risk of tumor evolution/relapse [137]. Erzin expression provides survival advantage against the stress experienced by metastatic OS cells during the initial interaction with the lung microenvironment [138]. During this adaptation process ezrin facilitates metabolic efficiency and regulates protein translation through signaling pathways, such as mTOR and PKC [137]. Thus, the active phosphorylated form of ezrin is observed at high levels upon OS cell arrival to the lung, but is lost later on in the metastatic process, suggesting a key role for ezrin in the earlier steps of metastasis [139]. Given the relevant function of ezrin in OS metastasis, targeting ezrin itself or erzin-associated targets may represent an efficient therapy to prevent metastatic progression. In this regard, the use of small molecule inhibitors of ezrin, the blocking of mTOR signaling and the inhibition of PKC activity can effectively reduce the metastatic burden in OS models [137].

Resistance to receptor-mediated cell death is one of the most common mechanisms developed by metastatic cell populations. Thus, the loss of FAS or its function in OS cells has been posed as an underlying mechanism by which OS cells evade host resistance in metastatic niches constitutively expressing FAS ligand (FASL), such as the lung [140]. Accordingly, FAS expression inversely correlates with metastatic potential in human OS [141] constituting a potential therapeutic target for metastatic OS. In this regard, it has been found that chemotherapeutic agents (gemcitabine) and histone deacetylase inhibitors (entinostat) induce the upregulation of FAS and the consequent regression of OS lung metastases [142, 143].

Finally, TGFα, a member of the epidermal growth factor family, also plays a role in OS metastasis. TGFα interacts with epidermal growth factor receptors on OS cells to elicit the activation of the PI3K/AKT/NFκB pathways, resulting in the expression of intracellular adhesion molecule 1 and the promotion of migration and metastasis [144].

Effect of environmental physical/chemical conditions on OS development

Physical/chemical factors such as oxygen tension, pH and extracellular calcium levels influence tumor proliferation, invasion, metabolism, and cellular viability. Bone is a hypoxic environment that could constitute a selective milieu that favors the growth of hypoxia-resistant cancer cells. Hypoxia deregulates a number of genes in tumor and stromal cells, mainly through transcriptional mechanisms involving HIF1α [24]. This transcription factor is expressed in a high percentage of metastatic OS and cell lines and its expression positively correlates with poor prognosis [145, 146]. The effect of a hypoxic environment on OS cells includes increased levels of angiogenesis, migration, invasion and proliferation [21, 24, 147]. Under hypoxia, HIF1α is able to bind to the promoter of the pro-angiogenic factor VEGF leading to its up-regulation and promotion of neoangiogenesis. In addition, hypoxia-induced HIF1α participates in the process of organ homing metastasis in the OS cell line SOSP-9607, through a process mediated by CXCR4 and its ligand CXCL12 [146]. Moreover, HIF1α and TGFβ cross talk signaling cooperates to drive the “vicious cycle” of bone tumors [21] (Fig. 2).

A low oxygen environment may condition the OS response to chemotherapy. For instance, OS cell lines are refractory to cisplatin-, doxorubicin- and etoposide-induced cell death when cultured under hypoxia, through a HIF1α-independent mechanism [148]. However, a recent report shows that doxorubicin resistance in OS cells is mediated by HIF1α-induced activation of the multidrug resistance-associated transporter ABCB1 [149]. In addition, hypoxia-induced chemoresistance in OS can be mitigated by inhibition of the WNT/β-catenin pathway [147]. This drug resistance may also be partially explained by a role for hypoxia in promoting the CSC phenotype within the tumor [24]. According to the relevant role of hypoxia in OS development, reoxygenation of the tumor microenvironment through transcutaneous application of CO_2, induces apoptosis in OS cells, downregulating the expression of HIF1α, MMP2 and MMP9, and decreasing in vivo tumor growth and pulmonary metastasis [150].

Regarding the extracellular pH, it is known that cancer cells may promote acidosis and that an acidotic bone environment is also able to potentiate the “vicious cycle” of bone metastasis [21]. Notably, it has been reported that the combination of hypoxia, acidosis and elevated interstitial fluid pressure in the OS environment additively influence tumor proliferation, invasion and angiogenesis in human OS cell lines [151].

Finally, calcium is the main inorganic component of the bone matrix and its release into the BME by osteoclastic bone resorption also contributes to the “vicious cycle” by different mechanisms involving the production of PTHrP [21]. Using OS tumor initiating cells embedded in calcium phosphate biomaterials, we have evidenced an active role for microenvironmental calcium substrates in the development of the OS phenotype [11]. In a similar way, OS cell lines grown onto calcified materials showed an osteoblastic phenotype and modified the expression pattern of cytokines, such as TGFβ, which favor OS growth [152].

Effect of bone environment on OS-CSCs subpopulations

Similar to normal stem cells, different signals from the CSC niche in OS may play a role in the regulation of their self renewal, differentiation, growth, drug resistance and metastatic potential [84, 153–155]. Among these factors, FGF increases the formation of CSC-enriched floating tumor-spheres in serum-free conditions and this sphere-forming fraction is enriched in both SOX2 and FGFR2 expression [84]. Remarkably, SOX2 has been reported to be required for self-renewal and tumorogenesis in OS-CSCs. Moreover, treatment with inhibitors of FGF signaling leads to decreased SOX2 expression and reduced proliferation in mouse OS [84].

Another BME factor, TGFβ, is able to induce sphere-forming capacity and other CSC features in non-CSC populations cultured in a hypoxic environment, indicating that TGFβ signaling is key to maintain OS cells in an immature state [156].

Conversely, microenvironmental signals with proven pro-osteogenic activities, like BMP or WNT signaling, seem to decrease CSC frequency in OS. Thus, BMP2 inhibits the tumorogenic potential of the OS cell line OS-991 by inducing the differentiation of the ALDH^br subpopulation of CSC, although the ALDH^lo non-CSC population displays a much higher differentiation response to BMP2 [71, 157]. Likewise, the formation of spheres from OS cells dramatically decreases when WNT signaling is induced, and conversely WNT signaling is inactive in OS sphere-forming cells [84]. Similarly, WNT signaling decreases SOX2 levels resulting in osteogenic differentiation. Therefore, WNT and FGF-SOX2 signaling appear to have opposing functions in the maintenance of CSC in OS [84, 154].

Finally, CXCL12 is another chemokine present in the BME, and OS-CSC have been found to overexpress its receptor, CXCR4, as compared to non-CSC tumor cells [153]. Given that the CXCL12/CXCR4 signaling pathway is involved in hematopoietic stem cell maintenance as well as in metastatic processes, a possible role of this signaling pathway in the maintenance and evolution of OS-CSC subpopulations has been suggested [155].

Perspectives and conclusions

Increasing evidence indicates that signaling pathways known to be functional in the BME may alter the properties of OS cells. Likewise, some of these signaling routes are reciprocal, being induced by OS cells and producing pro-tumoral signals in different microenvironmental cell types. Nevertheless, apart from the “vicious cycle” between cancer cells and the BME (which has been mainly characterized in bone metastasis rather than in primary OS), the interaction of OS cells with microenvironment supportive cells has barely been described, with limited in vivo data available. The most appropriate models to unravel the role of the cross talk between cancer cells and BME in OS development will be those which aim at integrating the cell of origin for OS combined with OS-predisposing oncogenic lesions in an orthotopic microenviroment. The orthotopic inoculation of relevant subpopulations of human OS-CSCs will also constitute a valuable tool to study microenvironmental connections, with therapeutic implications. The use of these models in combination with novel techniques of imaging, cell tracking, nanotechnology, genomics, proteomics, metabolomics and improved radiological or histological methods may help to decipher the mechanistic basis of the OS–BME interplay.

The overall survival rate for OS patients has now reached a plateau, with the current survival rate for metastatic disease remaining below 20 %. This fact highlights the need for novel, more effective therapeutic strategies, and targeting the pro-tumorogenic signaling of the BME could represent a relevant treatment approach. In this regard, new therapies including specific inhibitors, gene silencing strategies and antibodies designed to target neovascularization, osteoclast activation, the immune system, growth factor signaling pathways or environment-induced drug resistance are being tested alone or in combination with conventional treatments. Finally, given the dependence of CSC on their niche to maintain their stemness, progress in defining the OS-CSC niche could also provide new therapeutic strategies.

Abbreviations

AKT

V-Akt murine thymoma viral oncogene homolog

ALDH

Aldehyde dehydrogenase

BM

Bone marrow

BMP

Bone morphogenic proteins

CCL

Chemokine (C–C Motif) ligands

CSC

Cancer stem cells

CXCL

Chemokine (C–X–C Motif) ligands

DKK

Dickkopf proteins

ECM

Extracellular matrix

EDN1

Endothelin 1

EMV

Extracellular membrane vesicles

EPH

Erythropoietin-producing hepatoma

ERK

Extracellular signal-related kinases

FGF

Fibroblast growth factors

GLI

Glioma-associated oncogene

GH

Growth hormone

GP

Growth plate

HES

Hairy and enhancer of split

HH

Hedgehog proteins

HIF

Hypoxia-inducible factors

IGF

Insulin-like growth factors

IHH

Indian hedgehog

IL

Interleukin

MAPK

Mitogen-activated protein kinases

MCT

Monocarboxylate transporter

miRs

MicroRNAs

MMP

Matrix metalloproteinases

MSC

Mesenchymal stem/progenitor cells

mTOR

Mammalian target of rapamycin

NFkB

Nuclear factor kB

OPG

Osteoprotegerin

OS

Osteosarcoma

PDGF

Platelet-derived growth factor

PI3K

Phosphatidylinositol-4,5-bisphosphate 3-kinase

PTHrP

Parathyroid hormone-related peptide

RANK

Receptor activator of nuclear factor kappa B

RANKL

RANK ligand

RB

Retinoblastoma

SOX2

Sex-determining region Y-box 2

STAT3

Signal transducer and activator of transcription 3

TGFα/β

Transforming growth factor α/β

VEGF

Vascular endothelial growth factors

WIF1

WNT inhibitory factor 1

WNT

Wingless-type MMTV integration site family

YAP1

Yes-associated protein 1