Biology of Bone Sarcomas and New Therapeutic Developments

Bone sarcoma genesis can be explained by a conjunction between a minimum of one oncogenic event and an adequate microenvironment leading to the emergence of cancer, followed by its growth and potential migration to distant organs. Oncogenic events at the gene expression level (e.g. mutation, duplication, translocation) occurring during MSC differentiation increase the risk of their transformation to cancerous cells and result in the emergence of malignant osteoblastic or chondroblastic malignant cells. Indeed, osteosarcoma and chondrosarcoma cells express runx2 and sox9 in a similar manner than their non-malignant homologues [8‐11]. This expression of master genes in addition to their embryologic origin and their morphology strongly establish their close relationship with MSCs (Fig. 1). In this context, osteosarcoma cells originate from MSCs that are more or less committed to the osteoblast differentiation programme in which the oncogenic events occur. Consequently, osteosarcoma cells can express osteoblastic markers such as alkaline phosphatase, osteocalcin or bone sialoprotein and show a strong capacity to form osteoid tissue and induce the mineralisation of extracellular matrix. Chondrosarcoma cells share common features with chondrocytes and express chondrocyte markers such as type II collagen or aggrecan (Fig. 1). Because chondrosarcoma cells are cytologically and phenotypically related to chondrocytes, they are able to produce cartilaginous matrix into which malignant chondrocytes become encased. Chondrosarcoma can form benign lesions in which the Hedgehog signalling pathway (such as EXT1 and EXT2 involved in the endochondral ossification) is dysregulated and evolve into malignant entities [12‐14]. While osteosarcoma and chondrosarcoma can be considered as the result of a disturbed differentiation programme of MSCs, the origin of Ewing sarcoma is more controverted. Indeed, Ewing sarcoma cells are characterised by the expression of a fusion protein resulting from a chromosomal translocation between the EWS gene on chromosome 2 and a gene of the ETS family and consequently have been initially associated with the primitive neuroectodermal family of tumours [15]. However, the main frequent location of Ewing sarcoma in bone and the functional consequence of EWS‐FLi1 silencing in Ewing sarcoma cells fed the controversy and put a label of mesenchymal origin on Ewing sarcoma [15]. Indeed, Tirode et al. showed that the EWS‐FLI1 silencing in different Ewing cell lines resulted in the differentiation of sarcoma cells into mesenchymal lineages and more particularly into adipogenic and osteogenic lineages [16]. To date, its origin remains elusive with three potential hypotheses: neural crest stem cells [17], embryonic osteochondrogenic progenitor cells [18] or MSCs [16, 19]. Numerous pre-clinical models based on in vitro approaches and in vivo investigations (e.g. rat, mouse, zebrafish) mimicking the human disease have been proposed and are currently used to study the pathogenesis of bone sarcomas and/or for screening new drugs [20‐28].

Osteosarcoma, Ewing sarcoma and chondrosarcoma are separated into three different clinical entities identifiable by the patient populations affected, their localisation and their biological characteristics (Table 1). Osteosarcoma is the most frequent malignant primary bone tumour with a higher incidence in adolescent and young adults. Two peaks of incidence are conventionally described: (i) a main peak at 18 years and (ii) a second peak at 60 years with poor prognosis corresponding frequently to secondary osteosarcoma developed after radiotherapy or after Paget disease of bone [2, 3]. All osteosarcomas are characterised by the presence of a mineralised osteoid matrix produced by cancer cells and which results in the typical radiographic appearances called “sunburst” pattern [4, 29]. Osteosarcoma are very heterogeneous tumours (intra- and inter-tumoural heterogeneity) as revealed by the multiple histological subtypes according to the degree of cancer cell differentiation and consequently the quality of the extracellular matrix secreted (e.g. osteoblastic, chondroblastic, fibroblastic, telangiectatic osteosarcoma). The main affected areas of osteosarcoma are the metaphysis of the long bones with a preference to the proximal end of the tibia/fibula corresponding to the location of the growth plate. Genetic analyses confirmed the high heterogeneity of osteosarcoma [30‐32]. Bousquet et al. identified for instance more than 80 point mutations and some deletions related to more than 80 genes [30]. Kovac et al. interestingly identified a BRCAness signature in osteosarcoma which could be exploited as a new therapeutic targeting [31]. The overall survival of osteosarcoma patients is very dependent on their metastatic status at the time of diagnosis with a survival rate for patients with localised disease of around 65% after 5 years; however, when lung metastases are detected, survival drops to 30% (Table 1). Around 10–20% of patients show clinically detectable metastases at time of diagnosis and 85–90% are located in the lungs.

Ewing sarcoma is the second main represented bone sarcoma with 0.3/100,000/year. This bone sarcoma subtype accounts for 2% of childhood cancers, is more predominant in male than female with a sex male/female ratio around 1.5 and has a peak of incidence at 15 years. Sixty percent of Ewing sarcomas develop in flat bones and 40% affect the metaphysis of long bones (Table 1). Similar to osteosarcoma, the overall survival is also associated with the metastatic status of patients. For localised tumours, the overall survival is 50–60% at 5 years, which drops to only around 20% for metastatic sarcoma. At time of diagnosis, 20–25% of patients show clinically detectable metastases [33‐35]. Although Ewing sarcoma is the most homogeneous entity among bone sarcomas, composed of undifferentiated round cancer cells characterised by CD99-, FLI1-, HNK1- and CAV1-positive immunostaining associated with limited stromal components [36], recent work demonstrated in contrast their heterogeneity [37‐40]. Previous studies highlighted only a few recurrent somatic mutations in Ewing sarcomas (TP53, STAG2, CDKN2) [38, 41, 42]. However, more recent studies by Zhang et al. used next-generation sequencing (Ion AmpliSeq™ Cancer Hotspot Panel v2) to identify a series of five new mutations (KDR, STK11, MLH1, KRAS and PTPN11) related to a higher proliferation index and revealing a higher tumour heterogeneity than initially suspected [37]. This heterogeneity is not restricted to the genetic patterns but can be extended to epigenetic profiles [39]. Indeed, Sheffield et al. showed heterogeneous DNA methylation profiles between different tumours, which could reflect a continuum between mesenchymal and stem cell signatures in link with the EWS–FLI1 signature [39]. In addition, the expression levels of EWS–FLi1, which are variable in a tumour tissue, have a functional impact on cell migration. EWSR1–FLi1^high cells are characterised by high proliferation activity, while EWSR1–FLi1^low have a marked propensity to migrate, invade and metastasise [40].

Chondrosarcoma is the third entity of bone sarcoma in term of incidence with around 0.2 new cases per 100,000 each year and similar incidence between male and female (Table 1). Similar to all bone sarcomas, several subtypes can be identified according to their histological characteristics [43‐46] and are classified as low, intermediate or high grade on the basis of histopathological features [47]. Chondrosarcomas are characterised by a tumour chondrocyte-derived hyaline-like extracellular matrix, which eventually encases the cancer cells. The tumour tissue is organised in a mosaic of lobules separated by fibrous tissue. In addition, chondrosarcomas exhibit low vascularisation in contrast to osteosarcoma and Ewing sarcomas. Heterogeneity is also a hallmark of chondrosarcomas, which are associated with a complex cytogenetic signature [48, 49]. Thus, somatic mutations in isocitrate dehydrogenase (IDH)-1 or -2 are frequent (around 56%) in central and periosteal cartilaginous tumours and absent in endochondroma [50]. In addition to mutations in IDH1, IDH2, EXT (exostosin) and more conventional genes associated with cancer progression such as TP53 or Rb1, Tarpey et al. identified COL2A1 mutations (insertions, deletions and rearrangements) in the third cases [51]. The principal localisations of chondrosarcomas are pelvic bone, scapula and long bones (Table 1). While high-grade chondrosarcomas can be associated with metastases, these tumours are characterised by a high rate of local recurrence and consequently by a high morbidity [52, 53]. Osteosarcoma, Ewing sarcoma and chondrosarcoma are then characterised by a marked heterogeneity at the histological, genetic and epigenetic levels.

In addition to c-fos which has been associated with osteosarcoma formation due to its contribution in osteoblast differentiation [54, 55], some genetic predispositions have been linked with osteosarcoma development in hereditary syndromes such as Li-Fraumeni (p53 mutation) [56], Rothmund-Thompson [57], Werner [58] or Bloom syndromes (mutations of helicase genes) [59, 60], or retinoblastoma familial cancers [61]. Hereditary multiple exostoses (familial osteochondromatosis or diaphyseal aclasis) is an inherited genetic disease associated with osteochondromas and with EXT1 and EXT2 mutations [14, 62]. Even if several studies evaluated the risk of malignant transformation of multiple exostoses, the most recent study identified this risk at relatively low level (2.7%) with the development of low-grade chondrosarcomas [63]. However in most of the cases, patients do not show any predisposition genes and bone sarcomas are sporadic cases which could be explained by a close relationship with their local microenvironment altered during the malignant transformation process [64‐68]. The “seed and soil” theory proposed by Stephen Paget at the end of the nineteenth century gives a partial explanation of bone sarcoma formation [69]. At the early stage of the disease, proliferation of bone sarcoma cells in the bone environment leads to the dysregulation of the balance between osteoblasts and osteoclasts, in favour of an exacerbated osteoclast differentiation and local bone resorption. In turn, resorptive osteoclasts release pro-tumoral factors (e.g. cytokines, extracellular matrix components) initially trapped into the organic matrix of bone tissue [70]. The demonstration of this vicious cycle between osteoclasts and bone sarcoma cells has stimulated numerous pre-clinical and clinical investigations that revealed the decrease of tumour bone sarcomas after targeting of osteoclasts using anti-resorptive agents [71‐75]. In addition to their anti-resorptive activities, nitrogen-containing bisphosphonates could have a direct anti-proliferative activity on cancer cells [76, 77]. On the contrary, Endo-Munoz et al. showed the deleterious effect of osteoclastogenesis inhibition after zoledronic acid treatment which was associated with an increase of lung metastases in an osteosarcoma model [78]. The role of osteoclasts in bone sarcoma development is still unclear and osteoclasts could act as a pro-tumoral factor in the early stage of the disease due to their pro-angiogenic activity [79] and could exert the opposite role at a later stage of the disease [80].

Bone sarcoma development could be explained by the conjunction of multiple factors: (i) one or more oncogenic events from which the malignant transformation is initiated. The risks of genetic aberrations at the gene expression level (e.g. mutation, deletion, amplification) could increase with the proliferation rate of the cells of interest such as MSCs/osteoblasts during bone growth. A first mutation could lead to a chromosomal instability and consequently to the appearance of new oncogenic events [31]. (ii) A favourable microenvironment is a prerequisite for the growth of cancer cells. The differential repartition of bone sarcomas according to their subtypes are in favour of this theory. Furthermore, numerous studies demonstrated that MSCs induce pro-proliferative effects on bone sarcoma and promote osteosarcoma stemness strengthening the “seed and soil” theory [81, 82]. Local acidosis derived from the tumour growth and tumour-associated osteolysis has in return a strong impact on the stemness of MSCs [83, 84]. The bilateral dialogue established between cancer cells and their neighbours is a central aspect of bone sarcoma development. The diverse modes of communication include soluble factors (e.g. chemokines, cytokines), direct cell interactions and extracellular vesicles [64‐66]. Gap junctions are intercellular channels composed of transmembrane proteins named connexons that allow direct intercellular communication between two adjacent cells. Recent data investigated at the single-cell level showed intercellular communications through gap junctions between osteosarcoma cells and various other cell types [85]. Functional gap junctions have been observed between osteosarcoma cells and MSCs depending on their differentiation levels, and between cancer cells and endothelial cells. In contrast, while all bone cells express gap junctions, no gap junction-dependent communication has been demonstrated with macrophages, osteoclasts or osteocytes [86‐88]. Gap junctions are clearly involved in the tumour development and the loss of connexin43 expression in Ewing sarcoma cells favours the development of the primary tumour growth [89]. Another way of cell communication is transfer of extracellular vesicles loaded with proteins, mRNA and microRNA. Thus, it has been suggested that osteosarcoma cells are able to resist the effects of chemotherapeutic treatment such as doxorubicin by transferring exosomes carrying specific multidrug resistance factors (e.g. MDR-1, Pgp) from resistant to non-resistant cancer cells [90]. Recently, Baglio et al. described the education of MSCs by tumour-secreted extracellular vesicles [91]. These authors demonstrated the ability of osteosarcoma cells to incorporate TGF-β into extracellular vesicles which induced production of IL-6 in MSCs. IL-6 is in turn associated with an increase of tumour growth [92]. A vicious cycle is then established between MSCs and sarcoma cells through the release of extracellular vesicles.

The bone sarcoma microenvironment is not restricted to MSCs but is a very complex and dynamic environment (Fig. 2). This environment can be described as “niches” including bone, vascular and immune niches and more specific niches such as muscles and lung parenchyma for invading and metastatic cells. Even though there is no evidence of the correlation between the vessel density and the metastatic process in bone sarcomas, endothelial cells are strongly involved in the intra/extravasation of cancer cells. Recently, new regulators including brain, neuronal network and neurotrophic factors should be added to the list. It is now well recognised that the brain can act as a master regulator of bone mass [93, 94]. Bone remodelling is indeed regulated by a rich innervation, which is the source of neurotrophic factors, hormones and neurotransmitters [95]. Released locally or into the blood stream, these soluble factors could target bone sarcoma cells [96, 97]. The most recent evidence has been given by Punzo et al. who showed the anti-proliferative, pro-apoptotic and anti-invasive effects of endocannabinoid and endovanilloid systems in osteosarcoma [98] (Fig. 2). The bone environment is relatively specific to bone sarcomas and bone cells have been suspected to contribute to their development. Indeed, as described above, the blockade of bone resorption by bisphosphonates inhibits the tumour growth in pre-clinical models of osteosarcoma [92] and Ewing sarcoma [71] and slows down recurrent tumour progression after intralesional curettage in chondrosarcoma [76, 99]. Unfortunately, the results of a phase III clinical trial associating conventional chemotherapy and bisphosphonate (zoledronate) do not recommend this therapeutic strategy in osteosarcoma [100]. The lack of significant efficacy can be explained by the disparity of bisphosphonate or RANKL-blocking antibody efficacy observed using the parameters of bone remodelling in different mouse strains [101]. Alternatively, bisphosphonates could modulate macrophage differentiation through complex mechanisms. Tumour-associated macrophages (TAMs) can be subdivided in two types of populations, M1-polarised macrophages considered as antitumour effectors and M2-polarised macrophages, which are defined as pro-tumour modulators due to their positive impact on the neoangiogenic process [102]. In breast cancer models, it has been shown that cancer cells secrete soluble factors modulating macrophages towards the M2 state. Zoledronate counteracts this differentiation and favours a cytotoxic immune response linked with the differentiation of TAMs towards the M1 subtype [103]. In mesothelioma, zoledronate impairs the polarisation of TAMs to the M2 phenotype but leads to the accumulation of immature myeloid cells, which could reduce its effects [104]. In bone sarcoma, TAMs also appeared as key effectors of the pathogenesis [105‐107]. Indeed, the macrophage infiltration in osteosarcoma is correlated with metastatic suppression [105] and osteosarcoma cells dysregulate the balance of M1/M2 macrophages [106]. An abundant M2 macrophage infiltrate is consequently in favour of a metastatic profile [106]. In Ewing sarcoma, the targeting of TAMs by liposome-encapsulated clodronate that inhibits simultaneously M1 and M2 macrophages leads to a decrease of tumour growth [107]. Overall, these results demonstrate the key role of macrophages which regulate the development of bone sarcoma according to their number and M1/M2 phenotype. The role of the immune niche in bone sarcoma development is not restricted to TAMs and is also controlled by dendritic cells, tumour infiltrating lymphocytes or mast cells [108‐110].

Although current conventional treatments are relatively similar for osteosarcoma and Ewing sarcoma combining chemotherapy and surgery, the mainstay of local tumour control in chondrosarcoma is surgery with adequate margins (margins of normal tissue). Indeed, chemotherapy and radiotherapy are ineffective in the treatment of local and advanced chondrosarcoma patients. Consequently, both therapeutic approaches have limited impact in the management of these patients [111]. Unfortunately, adequate margins can only be achieved in 45–75% of patients. Inadequate margins are related to a high risk of local recurrence. Recent work validated the cryosurgery after intralesional curettage for low-grade chondrosarcoma. The technique appears safe and effective in selected patients [111]. Chemotherapy is recommended for high-risk chondrosarcoma and dedifferentiated chondrosarcoma but there is no recognised consensus defining the protocol and time schedule. The conventional therapeutic approach to osteosarcoma and Ewing sarcoma combines surgery (preoperative or neoadjuvant) and after chemotherapy (postoperative or adjuvant) and long-term (6–12 months) polychemotherapy [112‐115]. The conventional cocktail used in osteosarcoma is composed by a minimum of three drugs (reference combination: doxorubicin, cisplatin, methotrexate). Ifosfamide is the fourth drug used in osteosarcoma. Radiotherapy can be used when adequate surgery is impossible and for high-risk locations (e.g. spine); however, osteosarcomas are usually considered as radioresistant. In Ewing sarcoma, chemotherapy includes vincristine, ifosfamide, doxorubicin and etoposide. In addition, patients will receive radiotherapy since Ewing sarcoma responds relatively well to irradiation [116]. Several on-going clinical trials are studying new regimens of high doses of chemotherapy in combination with radiotherapy (Table 2). However, most conventional chemotherapy commonly results in relatively poor therapeutic responses, which has led to the development of new compounds with new therapeutic targets (Tables 2, 3; Fig. 3).

In the field of bone sarcoma, giant cell tumours (GCTs) have a special status. Indeed, GCTs are benign tumours with no nuclear cytologic aberration, intensively damaging the host bone and the cells can spread to the soft tissue in a similar manner to a malignant tumour [139‐141]. Indeed, high-grade malignant neoplasm can be identified at the time of diagnosis or subsequent surgery (secondary malignancy in GCT) or radiotherapy. Giant cell tumours of bone are rare tumours with an incidence of around 1 new case per 100,000 people per year and affect mainly young adults on the second and third decade. The ratio male/female of 1:2 is in favour of the female. The tumour tissue is characterised by three main cellular components: (i) giant multinucleated cells (osteoclast-like cells), (ii) mononuclear macrophages and (iii) mononuclear stromal cells (Fig. 4). Stromal cells secrete numerous pro-myeloid factors such as M-CSF and pro-osteoclastic factors such as RANKL resulting in monocyte/macrophage proliferation and osteoclastogenesis. Indeed, osteoclast precursors have monocytic/macrophagic origin and can proliferate, fuse and differentiate in the presence of M-CSF and RANKL (Fig. 4). RANKL is mandatory for osteoclastogenesis. RANKL binds to three distinct receptors: (i) RANK: a transmembrane receptor expressed at the surface of osteoclasts and their precursors and is responsible for osteoclast differentiation; (ii) OPG: a soluble decoy receptor blocking the binding of RANKL to RANK and therefore considered as an anti-bone catabolic agent, (iii) LGR-4 expressed at the cell membrane of osteoclasts and which negatively regulates osteoclast differentiation (Fig. 4) [142]. The origin of giant cell tumours of bone has been controversial for a long time. Nowadays, it is widely accepted that the stromal component is “the tumoural” element of the tissue and its dysregulation leads to the recruitment, proliferation and differentiation of macrophages. The clinical consequence is massive local bone destruction (Fig. 4). The current treatment is based on a resection surgery but unfortunately frequent recurrences associated with a high morbidity are observed. This is followed by a possible malignant transformation with a metastatic profile after up to 20 years.

Similar to other bone sarcomas, the local microenvironment is crucial in the tumour development and the osteolytic process. In this context, anti-bone resorption agents have been assessed in clinical trials with great success [143, 144]. A phase II clinical trial (NCT01564121) has assessed zoledronic acid in 24 patients [144]. The patients were treated with extensive intralesional curettage followed by five courses of bisphosphonate. Unfortunately, even if short adjuvant treatments with zoledronic acid were associated with a low rate of recurrence, the study did not show any significant impact on local recurrence. Denosumab, a humanised blocking antibody against RANKL, is currently evaluated in a series of 586 patients in a phase II clinical trial (NCTNCT00680992) [102]. Denosumab was administered subcutaneously at a dose of 120 mg every 4 weeks and a loading dose of 120 mg s.c. on study days 8 and 15. The intermediate results showed the safety of the drug and first clinical benefit with at least 90% of tumour necrosis after denosumab administration (estimated completion date: end 2017). Preoperative pretreatment is currently in discussion to facilitate the surgical resection in patients with aggressive tumours with high-risk location (e.g. spine).