Markers for bone sarcomas

Effects of preanalytic conditions on markers in bone sarcomas

The effects of decalcification protocols on immunohistochemical and molecular biomarkers of bone tumors should be a constant concern in routine practice not only of bone specialized but of all pathologists. Decalcification remains necessary for routine diagnosis but can alter both proteins and nucleic acids. Use of strong acid such as hydrochloric acid has the advantage of limiting the duration of the decalcification process, but recent data showed their deleterious effect on DNA and RNA integrity as well as on protein preservation [ , ]. Hydrochloric acid showed quickly and irremediable split of RNA, quickly making any genetic study for search of gene fusions impossible by in situ hybridization or by RNA sequencing, but also DNA and protein alteration, leading to difficulty of interpretation for routine molecular genetics and immunohistochemical studies and to false negative results. Thereby, a nondecalcified sample should be performed when possible. General recommendations of the authors for decalcification of bone samples include an adequate fixation before any decalcification and a daily control of decalcification progress. Fine-needle biopsies, surgical open biopsies, and curettage should be decalcified as much as possible using a weak acid ( as formic acid ) with short cycles, or EDTA, particularly if immunohistochemical and genetic techniques are considered. One or two samples dedicated to immunohistochemistry and molecular analyses should be performed on surgical specimens for tumor pathology and not decalcified if possible or decalcified with formic acid or EDTA. The remaining tissue and surgical specimens for nontumoral pathology can be decalcified with hydrochloric acid.

Markers in high-grade osteosarcomas

From a genetic point of view, the molecular abnormalities of osteosarcomas involve variable genes and various signaling pathways, few of which currently represent possible therapeutic targets and none of these abnormalities constitute a specific diagnostic or prognostic marker that can be used routinely.

It is estimated that 20% of patients declaring osteosarcoma before age 25 have a genetic predisposition, including germline abnormalities of TP53 (Li–Fraumeni syndrome), RB1 (hereditary retinoblastoma), RecQ-like helicases (Werner syndrome, Rothmund–Thomson syndrome, RAPADILINO syndrome, Baller–Gerold syndrome, and Bloom syndrome), or ATR-X syndrome [ ].

Somatic alterations seen in high-grade osteosarcomas are complex. These tumors are characterized by multiple genomic abnormalities and a high heterogeneity between tumors, within the same tumor and between primary tumor and metastases [ , ]. These alterations include complex karyotypic modifications with structural and numerical chromosomal anomalies, gene copy number variations, and/or deleterious variants involving oncogenes or tumor-suppressor genes.

The rise of high-throughput sequencers in recent years has improved knowledge of cellular function alterations involved in the tumorogenesis of osteosarcomas.

Interestingly, most genes altered in osteosarcoma are involved in maintaining genomic stability. TP53 and RB1 alterations are the most frequent anomalies observed in high-grade osteosarcomas (up to 90% [ ] and to 64% [ ], respectively) but several others have been reported that can impact cell cycle (such as MDM2, CDK4, CDKN2A, C-MYC, etc.), DNA damage response (ATR-X, PALB2, FANCA, FANCC, CDKN1A, etc.), signaling pathways such as IGF (IGF1, IGF1R, IGF2R, etc.), PI3K–AKT–mTOR (PTEN, TSC2, PIK3CA), or Ras/Mek (NF1, NF2, MAP2K4, etc.), receptors with tyrosine kinase activity (PDGFRA, FGFR1, ALK, ERBB4, MET, etc.) and several genes involved in the nuclear chromatin, cytoskeleton, or matrix management as well as cellular metabolism.

One mechanism explaining the genetic instability observed in high-grade osteosarcoma is chromotripsis, a phenomenon observed in 30%–90% of osteosarcoma but rarely in carcinomas (<5%), where one or a few chromosomes undergo a massive and random rearrangement [ , , ]. Recent sequencing data suggest that chromotripsis is more frequent in younger patients and that oncogenesis may be more driven by this catastrophic event in young osteosarcoma patients as compared to older adults [ ].

Markers of benignity/malignancy in osteogenic tumors: new data

Until recently, no routine marker for malignancy has been identified in bone-forming lesions. For example, studied markers such as ezrin, galectin-1, HLA-G, and p63 have been shown to be expressed in both osteoblastoma and osteoblastic osteosarcoma but can be reliable tools to differentiate chondroblastic osteosarcoma from conventional chondrosarcoma [ , ].

The recent discovery of recurrent FOS (an AP-1 transcription factor) and its paralogue FOSB genes rearrangements in osteoblastoma and osteoid osteoma (almost 90%) represent a valuable maker for benignity in some instances, where it can be difficult to differentiate osteoblastoma from osteosarcoma [ ]. Fusions of FOS and FOSB can be detected by in situ hybridization or by molecular genetic techniques and its overexpression can be detected by immunochemistry ( Fig. 39.1 ) . Amary and contributors found in a series of 337 osteogenic lesions (84 osteoblastomas, 33 osteoid osteomas, 215 osteosarcomas, and five samples of reactive new bone formation) that the majority of benign osteogenic tumors (83% of osteoblastomas and 73% of osteoid osteoma) showed significant expression of c-FOS in the osteoblastic tumor cell, whereas a few osteosarcomas (14%) showed focal or patchy c-FOS expression, with no FOS rearrangement found by FISH [ ]. These data were confirmed by another team, with a warning on the decrease in the intensity of immunostaining parallel to the increase in the duration of decalcification [ ]. Besides a cutoff 50% of immunostained cells was used to distinguish between OO/OB and other mimics. FISH seems to have more specificity.

Markers in low-grade osteosarcomas

The diagnostic challenges raised by low-grade osteosarcomas (LGOSs) are different. Whereas high-grade osteosarcomas are very complex and polymorphic tumors, LGOSs are quite simple and monomorphic tumors. Parosteal LGOS, arising on the surface of bone, and central LGOS, arising in the medullary cavity, are slow-growing tumors occurring in young adults (aged 20–30 years), like their high-grade counterparts, developed in most cases in long bones, with a predilection for the distal femur. After complete surgical removal, 5-year overall survival in LGOS is over 90%.

Parosteal and central LGOS share a similar histological appearance, characterized by the association, in various amounts, of hypocellular and cytologically regular spindle cell proliferation, suggesting a desmoid tumor, with quite well-differentiated bone trabeculae sometimes mimicking fibrous dysplasia.

In contrast with high-grade osteosarcomas, LGOSs have a simple genetic profile characterized by the presence of a supernumerary ring and/or giant rod chromosomes on the conventional karyotype. FISH and CGH studies have shown that these chromosomes contain amplified sequences of the region q13-15 of chromosome 12, including notably the oncogenes mouse double minute 2 ( MDM2 ) and cyclin-dependent kinase 4 ( CDK4 ) [ , ]. The biological consequence of the 12q14-15 amplicon is the deregulation of the cell cycle with both a decrease in apoptosis and an increase in cell proliferation. MDM2 inhibits p53 and therefore decreases apoptosis. CDK4 phosphorylates the RB1 gene product, which no longer interacts with E2F transcription factors, allowing the cell cycle to proceed though the G1-S checkpoint.

In specialist centers, diagnosis of LGOS can be made in most cases on microscopic features correlated with imaging data. However, diagnosis of LGOS may be difficult and, according to the literature, misdiagnosed as benign in 30% of cases. These difficulties arise from the fact that LGOS histologically mimic a group of fibrous or fibro-osseous benign lesions involving the medullary canal or the bone surface, including in particular fibrous dysplasia, desmoplastic fibroma, and myositis ossificans, when it secondarily adheres to the periosteum.

In the studies by Dujardin and coworkers and Yoshida and coworkers [ , ], MDM2 and CDK4 protein expression was assessed by immunochemistry in, respectively, 72 and 23 cases of LGOS, some of them with misleading morphological features, and 107 and 40 cases of benign fibrous or fibro-osseous lesions of bone or para-osseous tissue.

Immunohistochemical expression of MDM2 and CDK4 was present in 89% and 100% of the LGOS studied and was absent in all cases of benign fibrous or fibro-osseous lesions except for one case of Nora's lesion included in the series of Yoshida and coworkers. The gene amplification status in this case was unfortunately unknown, as no molecular or cytogenetic studies were carried out ( Fig. 39.2 ).

In the series by Dujardin and coworkers, a study on frozen tissue specimens by Array-CGH in 18 cases of LGOS showed an amplification of chromosome 12q13-15 in all cases. In conclusion, assessment of the MDM2/CDK4 amplification/overexpression status is a sensitive and highly specific marker in the differential diagnosis between LGOS and benign mimics.

The value of MDM2 overexpression in immunochemistry and of MDM2 amplification by genetic techniques were since confirmed by several other studies [ , ].

On the other hand, whereas low-grade parosteal osteosarcomas are usually characterized by MDM2 anomalies, no MDM2 / CDK4 coamplification is found in intermediate-grade periosteal osteosarcoma. In a series of 27 periosteal osteosarcomas from 20 patients [ ], Righi and coworkers found no MDM2 and CDK4 overexpression in all tested cases except one, and no MDM2/CDK4 coamplification in the 10 tested tumors.

In about 15% of cases, LGOS may dedifferentiate, which means progress toward a high-grade sarcoma, with the appearance, juxtaposed to the low-grade osteosarcomatous component, of foci most of the time consisting in either spindle cells/pleomorphic cells undifferentiated high-grade sarcoma, or in high-grade osteosarcoma.

Interestingly, the foci of dedifferentiation, observed in five of the cases in the series by Dujardin and coworkers, also showed amplification/overexpression of MDM2 and CDK4, in contrast to a control group of high-grade conventional osteosarcomas. These results were confirmed by Yoshida and coworkers who found, in a series of 81 primary and 26 recurrent/metastatic osteosarcomas [ ], a coexpression of MDM2 and CDK4 on immunochemistry in 7 tumors. MDM2 and CDK4 amplification was confirmed by the genetic study and careful microscopic examination showed low-grade osteosarcomatous component in 6 cases. Proving amplification/overexpression of MDM2 and CDK4 in a high-grade osteosarcoma should raise the question of the dedifferentiation of an LGOS. Although today conventional high grade osteosarcomas and dedifferentiated LGOSs receive the same chemotherapy, distinguishing both entities may be important for targeted molecular therapy in the future.

Finally, investigating guanine nucleotide-binding protein/α-subunit (GNAS) mutations, involved in the pathogenesis of fibrous dysplasia, may constitute a valuable complementary diagnostic tool for the differential diagnosis between LGOS and fibrous dysplasia, GNAS variants being found in about 50% of cases [ , ] but not in LGOS [ , ].

Malignant transformation can rarely occur in fibrous dysplasia, with around 100 cases reported to date. This malignant development most often takes the appearance of osteosarcoma. Interestingly, osteosarcoma arising from a fibrous dysplasia can show GNAS mutation [ , ]. Proving the presence of GNAS variant in an osteosarcoma with fibrous dysplasia-like or more generally with a fibro-osseous component should raise the question of the malignant transformation of a fibrous dysplasia ( Fig. 39.3 ).

Markers in central and periosteal chondrosarcomas

Diagnostic markers

Differential diagnosis between conventional cartilaginous tumors and other bone tumors

IDH isoforms mutations

The current oncogenic paradigm of central and periosteal conventional cartilaginous tumors involves early alterations of the isocitrate dehydrogenase 1 ( IDH1 ) and isocitrate dehydrogenase 2 ( IDH2 ) genes. These alterations which have been identified in low-grade, high-grade, as well as dedifferentiated tumors link the benign and malignant counterparts of central and periosteal conventional cartilaginous tumors in a spectrum of disease. The isoforms IDH1 and IDH2 encode isocitrate dehydrogenase 1 and 2 enzymes, respectively, involved in the conversion of isocitrate to α-ketoglutarate (αKG).

IDH mutations generate the formation of the oncometabolite D-2-hydroxyglutarate (D2-HG), which is a competitive inhibitor of several α-ketoglutarate-dependent dioxygenases, including epigenetic regulators such as TET dioxygenases and histone demethylases, resulting in a DNA hypermethylated profile [ ]. The inhibitory effects of D2-HG lead to a hypermethylated DNA profile and cause deregulated chondrocyte differentiation and the stabilization of hypoxia-inducible factor 1α, which promote tumorigenesis. Nevertheless, these abnormalities are neither necessary nor sufficient to induce oncogenesis of cartilaginous tumors [ ].

IDH1 and IDH2 mutations were initially described in 2011 in cartilaginous tumors of patients with enchondromatosis, who are also more likely to develop other tumors such as gliomas and acute myeloid leukemia. At that time, gliomas were already known to harbor IDH genes mutations ( as observed in 70%–80% of diffuse gliomas and secondary glioblastomas [ , ]). A high rate of 87%–90% of IDH1 mutations was found in cartilaginous tumors of patients with Ollier's disease and Maffucci syndrome [ , ].

The incidence of these mutations was then evaluated in sporadic conventional central and periosteal tumors of appendicular and axial skeleton, with 52% of IDH genes mutations in low-grade central cartilaginous tumors ( enchondromas or grade 1 chondrosarcomas ) and 58.9% of central high-grade chondrosarcomas, most of them (88%) involving IDH1 [ ]. Acral tumors showed a high mutation rate (90%) whereas tumors involving the long bones of the appendicular skeleton and the flat bones showed a lower IDH mutation rate (respectively, 53% and 35%). These abnormalities were also found in periosteal cartilaginous tumors (71% of periosteal chondromas and 15%–100% of periosteal chondrosarcomas [ , ]). IDH mutations are dominant and involve the exon 4 of the gene, which leads to the substitution of an amino acid on the active site of the protein. According to Amary et al., the most frequent mutation is an R132C transition (39.5%), followed by R132G transversions (19.7%), R132H (17.3%), R132L (7.4%), and R132S (7.4%). A few IDH2 R172S transversions (8.6%) are also found, and more rarely R140 [ ] ( Fig. 39.4 ).

Interestingly, chondrosarcomas arising in some localizations such as the facial bones, the vertebrae, the sternum, or the upper airways seem to be rarely IDH mutated [ , ].

Unfortunately, in routine practice, the only antibody available against IDH mutants is the R132H antibody, which is a highly sensitive and specific marker for this mutant allele and observed in the majority of IDH-mutated gliomas (85%–93%) [ , ], but less commonly in cartilaginous tumors [ ].

The discovery of IDH gene mutations in conventional central cartilaginous tumors is a valuable molecular tool for the differential diagnosis of these tumors and has gained interest since the introduction of several clinical trials targeting IDH-mutated proteins in various tumors including chondrosarcoma, for which classical chemotherapy and radiotherapy have shown limited efficiency in the treatment of advanced high-grade tumors.

Although the search for IDH1 / IDH2 variants in cartilaginous tumor still does not resolve the question of whether it is benign or malignant, the potential diagnostic applications involved in finding these genetic abnormalities remain numerous.

Indeed, IDH mutations are not found in other mesenchymal tumors, especially in osteosarcoma and in other cartilaginous tumors such as chondroblastoma, chondromyxoid fibroma (CMF), and peripheral, mesenchymal, and clear cell chondrosarcoma.

Thereby, IDH1 / IDH2 mutation analysis can be a useful marker for distinguishing chondroblastic osteosarcomas from chondrosarcomas. DA Kerr and coworkers [ ] found, in a series including 25 predominantly high-grade chondrosarcomas and 65 chondroblastic or partially chondroblastic osteosarcomas, that the presence of IDH1 / IDH2 mutations could be useful in distinguishing chondroblastic osteosarcomas from chondrosarcomas. They showed, in agreement with the literature, that IDH1/IDH2 mutations were found in 61% of the chondrosarcomas but were absent from all cases of osteosarcoma. The presence of IDH1/IDH2 mutations in a sarcoma with cartilaginous differentiation thus strongly favors the diagnosis of chondrosarcoma over chondroblastic osteosarcoma.

Similarly, detecting an IDH1 / IDH2 mutation may help to distinguish a dedifferentiated chondrosarcoma with osteosarcomatous differentiation from an osteosarcoma that occurs de novo or from an undifferentiated pleomorphic sarcoma (UPS) of bone [ , , ].

When located in the skull base, chondrosarcomas may be difficult to distinguish from chordomas, partly because of the small size of biopsy samples but also because in this location chondrosarcomas frequently show myxoid changes and chordomas may contain a hyaline cartilaginous matrix, the so-called chondroid chordoma.

Sox9 is expressed in both notochordal and cartilaginous tumors and is not useful for differential diagnosis between chordoma and chondrosarcoma [ ].

Arai M and coworkers [ ] investigated the IDH1 / IDH2 mutation status of a series of intracranial tumors including 13 chondrosarcomas and 10 chordomas. The results were as expected regarding IDH1 / IDH2 mutations, with a predominant IDH1 R132C type in 46.1% of chondrosarcomas and in no cases in the chordomas. Another recent series including 30 skull base chondrosarcomas confirmed the presence of IDH mutations in this location, with a high rate of 85.7% [ ].

Brachyury

Brachyury immunostaining is very specific of notochordal cells and is useful in routine practice in the differential diagnosis between intracranial chondrosarcoma and chordoma ( see section “ Chordoma ”).

Podoplanin

Several studies have also demonstrated that D2-40 (podoplanin) is a true chondroid marker which may be useful for the differential diagnosis between chondrosarcomas and chordomas. In the series by Daugaard S and coworkers [ ], all 25 chondrosarcomas studied showed a positive staining reaction for D2-40 whereas all 5 chordomas included were negative. In another series of 22 chordomas, 20 chondrosarcomas, and 12 enchondromas, Huse JT and coworkers [ ] showed that D2-40 robustly and reliably immunostained 100% of enchondromas and 94% of grades I and II chondrosarcomas but did not stain chordomas.

Another diagnostic application that needs further exploration is the differential diagnosis between chondromyxoid fibroma (CMF) and chondrosarcoma with myxoid changes. In a series of cartilaginous tumors by Damato and coworkers, IDH1 mutations were not identified in any of the 19 cases of CMF studied [ ]. More recently, Nord and coworkers [ ] found from a whole-genome sequencing analysis of 20 CMFs that most tumors (90%) share a GRM1 rearrangement due to promoter swapping or gene fusions involving variable genes partners (such as COL12A1 , TBL1XR1 , or BCLAF1 ) and resulting in a massive increase in expression levels of GRM1 compared to control tissues.

Synovial chondromatosis represents another potential differential of chondrosarcoma but is now known to harbor frequent rearrangement of Fibronectin 1 ( FN1 ) with activin receptor 2A ( ACVR2A ) or alternative fusion partner(s) [ , ].

Differential diagnosis between enchondromas and atypical cartilaginous tumors/grade 1 chondrosarcomas

No marker is available in routine practice for the differential diagnosis between enchondromas and atypical cartilaginous tumors/grade 1 chondrosarcomas.

Lai and coworkers [ ] have performed an analysis of paraffin-embedded low-grade chondrosarcoma and enchondroma tissue samples by liquid chromatography–tandem mass spectrometry in order to uncover novel protein biomarker candidates. The proteomics analysis highlighted periostin, a cell adhesion matrixial protein that interacts with integrins via its fasciclin-like domains (FAS1) and hydrophilic C-terminal region that interacts with ECM proteins such as collagens, fibronectin, tenascin, or heparin [ ], as a potential marker from 17 candidates, and found a periostin expression in 14/23 low-grade chondrosarcomas versus 4/31 enchondromas.