What’s new in bone forming tumours of the skeleton?

Osteoid osteoma and osteoblastoma typically present during the second decade of life, with men being overrepresented (male to female ratio 2:1) [4]. Osteoid osteoma is usually located at the long bones in the lower extremity, but other commonly described sites involve the spine, upper extremity, hands, feet and pelvis [4, 5, 7]. The most prominent clinical symptom of osteoid osteoma is frequent and severe night pain that responds adequately to nonsteroidal anti-inflammatory drugs (NSAIDs) [4, 5]. Osteoblastoma is larger in size, and the majority is localized in the posterior column of the spine [4, 5, 8], resulting in neurologic symptoms as a recurring sign [4]. Pain is frequently present, but in contrast to osteoid osteoma, it does not respond to administration of NSAIDs [4, 5]. Both osteoid osteoma and osteoblastomas have no malignant potential, although osteoblastoma can behave as a locally aggressive tumour [4]. For radiologists, the diagnosis of osteoid osteoma is usually straight forward, showing a characteristic oval radiolucency (nidus) with surrounding sclerosis, while osteoblastoma can be accompanied by a more broad differential diagnosis depending on its location, including aneurysmal bone cyst, giant cell tumour of bone and osteosarcoma [4, 9].

Osteoid osteoma and osteoblastoma are histologically indistinguishable [10] (Fig. 1a, b). Both tumours are composed of irregular trabeculae of woven bone, lined with active osteoblasts. In osteoid osteoma, the central area of the lesion (nidus) is sharply demarcated and surrounded by hyper-vascularized sclerotic bone. In between the trabeculae, there is loose vascularised stroma, and small osteoclast-like giant cells are frequently seen [7, 11]. Osteoblastoma can show slightly more haphazardly arranged trabeculae [6]. Additional aneurysmal bone cyst (ABC)-like changes can be present, especially in larger tumours [4]. The term epithelioid osteoblastoma is reserved for osteoblastomas with the presence of large osteoblasts with an epithelioid appearance. Surrounding cytoplasm is abundant, and nuclei are hyperchromatic or show prominent nucleoli [4]. The most important differential diagnosis includes osteoblastoma-like osteosarcoma, that is distinguished from osteoblastoma based on the presence of host-bone infiltration and lack of differentiation towards the periphery [12]. However, this can be difficult to appreciate in small biopsies or curettage specimens. Definitive diagnosis is always made based on radiological and clinicopathological correlation.

Before the elucidation of the genetic background of osteoid osteoma and osteoblastoma, clonal chromosome aberrations were reported in two osteoblastomas, with structural alterations involving 22q13.1 [13], and only non-recurrent rearrangements were found using cytogenetic studies [14]. In 2018, in a quiet genomic background with paucity of somatic alterations, recurrent FOS and—to a lesser extent—FOSB rearrangements were found in both osteoid osteoma and osteoblastoma using RNA sequencing, demonstrating that both tumours were similar at the molecular level. In 5 out of 6 cases, FOS rearrangements were present, while the remaining case showed rearrangements involving its paralogue, FOSB. All FOS breakpoints were exonic and involved exon 4. Rearrangement partners were both introns of others genes (ANKH, KIAA1199, MYO1B) or intergenic regions [2]. Equivalent to FOS rearranged epithelioid hemangioma [15, 16], stop codons were encountered at, or early after the break points, leading to truncation of the protein with retention of the leucine zipper, and therefore its function as a transcription factor. Functional studies in epithelioid hemangioma demonstrated that the truncated protein was more resistant to degradation [17]. In the FOSB rearranged osteoblastoma, rearrangement resulted in an in frame fusion connecting PPP1R10 to FOSB, leading to altered signalling, due to promotor swapping [2]. Strikingly, FOSB fusions were also involved in pseudomyogenic hemangioendothelioma and atypical epithelioid hemangioma, resulting in promoter swapping [18, 19]. As genetic alterations in these vascular tumours are identical to those found in osteoid osteoma and osteoblastoma, one can speculate that a comparable molecular mechanism of tumorigenesis is operable in osteoid osteoma and osteoblastoma.

These novel molecular findings have provided new tools to improve diagnostic accuracy, as both fluorescence in situ hybridization (FISH) and immunohistochemical staining can detect FOS rearrangements (Fig. 1c, d). FISH was performed in an independent cohort and showed in the majority of cases rearrangements involving FOS and to a lesser extent FOSB [2]. In a follow-up study, immunohistochemistry showed strong and diffuse nuclear staining in the majority (79%) of osteoid osteomas and osteoblastomas, using a FOS antibody against the N terminus [20]. However, a previously published small study cohort demonstrated that osteoid osteoma and osteoblastoma lacked strong nuclear expression of FOS, indicating variability in sensitivity between different antibodies [21]. In terms of specificity, strong nuclear expression of FOS has been detected in a subset of other bone forming tumours and was only rarely present in osteosarcoma [2, 21]. Notably, in mouse models, the c-fos oncogene caused osteosarcoma, when fused with a highly active promotor and the v-fos 3’ untranslated region [22]. This is intriguing as in human tumours FOS and FOSB rearrangements have so far only been identified in vascular and bone forming tumours lacking malignant potential [15, 16, 18, 19].

Primary high-grade osteosarcoma occurs most often in young children and adolescents, but there is a second peak at a later age. In the latter group, osteosarcoma can occur secondary to radiation or Paget’s disease [25]. Osteosarcoma has a slight male predominance [26]. Patients with osteosarcoma often show signs of localised deep pain, especially manifest at night, developed over a longer period of a few weeks to months. This could also be in combination with limited mobility or localised warmth. A palpable mass can be present, which is tender during physical examination [27].

For diagnosis of conventional osteosarcoma, a radiograph is made in two planes, in which the lesion appears as lytic, sclerotic or mixed lytic and sclerotic. This lesion often expands into the surrounding soft tissue, with periosteal reaction and destruction of cortical bone [28]. MRI or CT imaging may provide additional information, guiding the subsequent biopsy of the lesion [28].

The presence of osteoid, the unmineralized extracellular matrix produced by the tumour cells, is the hallmark of osteosarcoma and visible as a pink dense structure in haematoxylin and eosin stained sections (Fig. 2a). Mineralization can occur. Osteosarcoma can arise in the medulla (central) or at the bone surface. Different osteosarcoma subtypes are recognised, based on their clinical presentation in combination with histological and molecular features (Table 2) [26]. High grade central osteosarcoma is the most common subtype, and most papers published over the last decade, as well as this review, focus on this subtype.

Certain hereditary syndromes predispose to osteosarcoma, such as Li-Fraumeni syndrome (mutations in TP53 or, less frequently, CHEK2), retinoblastoma (mutations in RB1) and Rothmund-Thomson syndrome (mutations in RECQL4) [29‐31]. Other hereditary syndromes with germline mutations in RecQ-like helicases, including RAPADILINO syndrome, Baller-Gerold syndrome, Werner syndrome and Bloom syndrome, also have an increased risk for osteosarcoma [32]. Another hereditary syndrome in which a helicase is mutated is ATR-X syndrome (alpha-thalassemia mental retardation syndrome). Patients with ATR-X syndrome show intellectual disability and skeletal abnormalities. Recently, two patients have been reported with ATR-X syndrome that developed osteosarcoma [33, 34].

High-grade osteosarcoma is characterised by a complex karyotype with many amplifications, deletions and (random) translocations (Fig. 2b). This complex genome hampers identification of the driver genes causing genome instability: very few recurrent alterations have been identified in osteosarcoma. One mechanism explaining the genomic instability in osteosarcoma is chromothripsis, the shattering of one or a few chromosomes into small fragments that are stitched together in a random order and orientation [35]. It was first discovered by Stephens et al. in chronic lymphocytic leukaemia, chordoma and osteosarcoma [35]: chromothripsis occurs in 3% of all cancers and in 30% of osteosarcomas. A more recent study confirmed chromothripsis in osteosarcoma, but showed a higher percentage—nearly 90%—where tumours with chromothripsis also frequently harbour amplifications [36]. The discrepancy may be attributed to the uncertain definition of chromothripsis. Exome sequencing shows a relatively low mutational burden in osteosarcoma ranging from 0.3–1.2 mutations per mega base; however, there is a pattern of localised hypermutation called kataegis in 50% of the tumours [3, 37]. These point mutations are non-recurrent, haphazard and cannot be considered as driver mutations. Further hampering the identification of driver genes is that no benign precursor of osteosarcoma is known. This is in contrast with for instance colorectal cancer, in which a benign precursor can be used to investigate multi-step progression behind tumorigenesis. Nevertheless, recent next-generation sequencing studies have revealed known and novel recurrent genetic alterations in osteosarcoma (Table 3). Most genes that were found to be altered are involved in maintaining genomic stability. Among the most commonly altered genes in osteosarcoma are the main players in maintaining genome stability: TP53 and RB1.

Mutations in TP53 can be found in germline or can be sporadic. Previously, using immunohistochemistry or sequencing of the DNA binding domain of TP53, mutations were detected in only 20% of osteosarcomas [44]. Interestingly, the more sensitive whole genome sequencing studies can detect more sub clonal mutations and reveal a much higher percentage (47–90%) of osteosarcomas harbouring TP53 alterations [3, 36‐38, 45]. Furthermore, many TP53 alterations involve structural alterations, most often consisting of translocations in the first intronic region of TP53, which is 10 kb in length. These alterations can only be detected with whole genome sequencing [46].

The second most frequently altered gene in osteosarcoma is RB1 (retinoblastoma 1), involved in blocking cells from entering S phase of the cell cycle [47]. Loss of Rb function in osteosarcoma therefore leads to a loss in Rb blockade of cell division. In addition to germline mutations, somatic mutations in RB1 were identified in 29–47% of osteosarcomas [3, 38].

The importance of TP53 and RB1 in osteosarcoma genesis is illustrated by the fact that patients with germline mutations in TP53 and RB1 are highly susceptible to cancer and frequently develop sarcomas. Different in vitro and in vivo studies confirm the important role of TP53 and RB1 mutations in sarcoma genesis [48, 49]. For example, homozygous deletion of TP53 and RB1 in osteogenic differentiated murine MSCs gives rise to osteosarcoma when injected into mice [49], while heterozygous deletion of TP53 is sufficient to induce osteosarcoma in a mouse model [48].

MDM2 (mouse double minute 2 homologue) regulates p53 activity by ubiquitinating p53 protein leading to proteasomal degradation of p53 [50]. Up to 12% of high-grade osteosarcomas have amplification of the MDM2 gene at 12q13-15, but this is higher in low-grade central osteosarcoma and parosteal osteosarcoma, with around 29% and 67–79% MDM2 amplification, respectively [41, 51] (Table 2). The CDK4 gene (cyclin-dependent kinase 4) is located within the same region at 12q13-15 [52] and regulates Rb activity by phosphorylating Rb, resulting in deactivation of Rb. CDK4 and MDM2 are often co-amplified and overexpressed in osteosarcoma. CDK4 is amplified in 67% of parosteal osteosarcomas, but rarely in high-grade osteosarcoma (9%) [41, 53]. As the percentage of CDK4 and MDM2 amplifications in low-grade central osteosarcoma and parosteal osteosarcoma are much higher than in high-grade osteosarcoma, most likely the CDK4/MDM2 amplified high-grade tumours represent progression from low grade osteosarcoma [53].

Rb activity is also regulated by p16, which normally inhibits both CDK4 and CDK6. P16 is encoded by the CDKN2A gene at chromosome 9p21.3, that also encodes for p14. Homozygous deletion of the CDKN2A locus, which is associated with poor prognosis in osteosarcoma, eradicates both expression of p16^Ink4A and p14^ARF, of which the latter is a negative regulator of MDM2 [38, 54‐56]. Therefore, deletion of p16 and p14, similar to co-amplification of CDK4 and MDM2, leads to inactivation of both the p53 and Rb pathway.

In addition to the p53 and Rb pathway, also other pathways involved in maintaining genome stability can be affected by mutations, both in sporadic as well as hereditary osteosarcoma. For instance, ATRX mutations can be found both as germline or somatic mutations [57], which is in contrast to mutations in RecQ-like helicases where only germline mutations have been identified. Around 29% of osteosarcomas harbour somatic mutations in ATRX [3]. The role of ATRX mutations in osteosarcoma genesis is largely unknown. ATRX is involved in chromatin remodelling and plays an important role in maintenance of chromosome stability [58]. Loss-of-function mutations in ATRX can lead to activation of the alternative lengthening of telomeres (ALT) pathway, maintaining the length of chromosome ends [59]. ALT is found in 59% of osteosarcomas, which is much higher as compared with other cancers such as carcinomas (5–15%) [60].

DNA repair is essential in maintaining genome stability. For instance, homologous recombination, the DNA repair pathway in which BRCA plays an important role, is crucial in maintaining genome stability. A recent whole exome sequencing (WES) study revealed a subset of osteosarcomas resemble features of BRCA mutant tumours [38]. These tumours show loss of heterozygosity, genomic instability and a mutation signature of substitutions and deletions that is also found in breast cancers with BRCA1/2 mutations. Around 80% of osteosarcomas show this BRCAness signature [38]. As this signature is linked to defects in homologous recombination, this vulnerability might be exploited with PARP inhibitors based on the principle of synthetic lethality. Indeed, different in vitro studies with osteosarcoma cell lines show that osteosarcoma cells are sensitive to PARP inhibitors [61, 62]. These results are promising, suggesting a possible new therapeutic strategy for osteosarcoma. However, further investigation on homologous recombination deficiency and PARP inhibitor sensitivity in osteosarcoma is needed.

Although the genes that play a role in genome stability are among the most frequently mutated genes in osteosarcoma (RB1, TP53, CDK4, MDM2, ATRX), these genes function in essential cell survival pathways. Therefore, these genes are difficult to specifically target in the treatment of osteosarcoma. Fortunately, also mutations in other genes are frequently found that are easier to target as they are involved in hormonal pathways. For example, mutations in genes involved in IGF (insulin-like growth factor) signalling, including the IGF1 receptor (IGF1R), were identified in around 7–14% of osteosarcomas, with many of these genes having altered activity compared with normal human osteoblasts or mesenchymal stem cells [36, 63]. The IGF signalling pathway is known to be important in normal bone growth, bone development and bone metabolism, and it is therefore not surprising that it might also play a role in osteosarcoma pathogenesis [64, 65]. These findings provide a rationale to explore anti-IGFR therapy as a treatment strategy for a subset of osteosarcomas.

The oestrogen hormonal pathway is also altered in osteosarcoma. Healthy osteoblasts normally express oestrogen receptor alpha (ERα), but this is lacking in osteosarcoma [66]. Until recently, the mechanism behind the inactivation of oestrogen receptor in osteosarcoma was not known. In a recent study, it was found that ERα was hypermethylated in osteosarcoma, which can be ameliorated by the DNA methyltransferase inhibitor DAC [67]. DAC could re-express ERα and subsequently restored defective osteogenic differentiation and inhibited proliferation in osteosarcoma cells. This study illustrates that epigenetic alterations such as hypermethylation of genes are also important in osteosarcoma genesis.

Although recent sequencing efforts did not identify specific druggable osteosarcoma driver genes, they did reveal new and known recurrent genetic events involved in osteosarcoma that shed light on its initiation (Table 3). Most of the mutated genes function in genome maintenance pathways and the majority of osteosarcomas show genome instability in the form of chromoanagenesis [36]. Therefore, it is reasonable to assume a connection between these specific genetic mutations and chromoanagenesis, especially chromothripsis.

A main player in maintaining genome stability, TP53, was linked to chromothripsis in patients with Li- Fraumeni syndrome [68]. Furthermore, rats with a heterozygous deletion of TP53 developed osteosarcomas, with chromothripsis and other complex structural rearrangements [69]. As cells with aberrant TP53 have impaired cell cycle control [70], TP53 loss-of-function alterations can facilitate chromothripsis by allowing cell cycle progression despite DNA damage [68]. Thus, cells with inactive TP53 and DNA damage from chromothripsis will proliferate uncontrollably. Moreover, mutations in the DNA binding domain of TP53 can cause a neomorphic gain-of-function, that could very well contribute to the initiation of chromothripsis itself [71].

However, TP53 alone cannot explain all tumours with chromothripsis, as is evident from studies that illustrate there are tumours with functional TP53 with chromothripsis, and tumours with aberrant TP53 without chromothripsis [72, 73]. Genes functionally similar to TP53 might also be able to initiate and/or maintain chromothripsis. Whether this is the case for osteosarcomas needs to be further elucidated.

It is striking that for osteosarcoma, many different genes have been identified and with each new sequencing study, the list of potential driver genes is ever-growing. Whether a genetic alteration—in TP53 or other genes—is a cause or consequence of chromothripsis remains unknown. One could argue whether the identified altered genes in osteosarcoma should be called “driver events” if these genetic alterations are the consequence of a single catastrophic event, such as chromothripsis. Therefore, perhaps the answer to what causes osteosarcoma could be found by discovering what causes chromothripsis. Different mechanisms have been proposed to what causes chromothripsis, such as micronuclei formation with DNA damage, telomere attrition and chromosome pulverisation by DNA damaging agents [74, 75]. Which event is the ultimate cause of osteosarcoma, is not yet known.