Main

Parosteal and low-grade central osteosarcomas are rare low-grade primary malignant bone tumors comprising <4% of all osteosarcomas. In contrast, fibrous dysplasia is one of the most common benign fibro-osseous tumor-like lesions in medullary bone. Morphologically, there is considerable overlap between these lesions all showing variably shaped bony trabeculae surrounded by spindled-shaped cells, with little to no cytologic atypia.1 In the majority of cases, imaging studies can easily differentiate parosteal osteosarcoma from fibrous dysplasia and central low-grade osteosarcoma: parosteal osteosarcoma is a surface tumor that may secondarily invade bone marrow, and low-grade central osteosarcoma and fibrous dysplasia are centrally based. Occasionally, fibrous dysplasia arises eccentrically in the marrow space, presenting as an exophytic surface-based lesion. Such lesions are referred to as fibrous dysplasia protuberans.2 Hence, distinguishing fibrous dysplasia protuberans from parosteal osteosarcoma can be challenging and differentiating between low-grade central osteosarcoma and fibrous dysplasia can also be difficult.

Parosteal osteosarcoma typically harbors one or more supernumerary ring chromosomes with amplification of the MDM2, SAS, and CDK4 genes.3, 4 MDM2 amplification is reported to occur in 93% (14/15) of low-grade central osteosarcomas3, 5 and in 79% (68/86) of parosteal osteosarcomas.5, 6 Dedifferentiated parosteal osteosarcoma is a distinct tumor variant in which a (high-grade) sarcoma coexists with a conventional parosteal osteosarcoma. Dedifferentiation affects 16–43% of parosteal osteosarcomas with the characteristic gene amplifications being retained and can be present at the first presentation (synchronous type) or at the time of recurrence (metachronous type).7, 8, 9 Bone tumors analyzed previously for GNAS alterations include fibrous dysplasia (N=405),10, 11, 12, 13, 14, 15, 16 ossifying fibroma (N=65),10, 12, 15, 16 osteofibrous dysplasia (N=19),10, 12, 13 low-grade central osteosarcoma (N=12)10, 13, 14, and parosteal osteosarcoma (N=10).10 GNAS mutations have been reported in fibrous dysplasia and have not been described until recently in any other fibrous osseous lesions with the exception of a single low-grade central osteosarcoma.14 These reported GNAS mutations in fibrous dysplasia include p.R201C,13 p.R201H,13 p.R201S,17 and p.Q227L.13 A GNAS mutation in codon 201 (p.R201G) has been found in lesions in McCune–Albright syndrome.18 Other GNAS mutations including p.R201P, p.R201L, p.R201S, p.Q227R, p.Q227H, and p.Q227K have been detected in sporadic endocrine tumors.19, 20 As fibrous dysplasia may be a component of the mosaic disorder in McCune–Albright syndrome, it would not be surprising if any of the above mutations were detected occasionally in fibrous dysplasia. Notably, GNAS mutations have been reported to occur in up to 90% of fibrous dysplasias with p.R201 involving 95% of cases; mutations p.R201H and p.R201C occurring in 57 and 38%, respectively, and p.Q227L being detected in ~5%.13 The GNAS p.K233N has only been detected in one case of colon adenocarcinoma, described in the Catalogue of Somatic Mutations in Cancer (COSMIC) with the mutation ID COSM192557. More recently, GNAS R201 substitutions have been reported in 88% of intraductal papillary mucinous tumors of the pancreas (88%) and in 16 of 32 low-grade appendiceal mucinous neoplasms.21

The mutual exclusiveness of GNAS mutations and MDM2 amplifications has been analyzed only in a small study by Tabareau-Delalande et al.10 who demonstrated the absence of GNAS mutations in 10 parosteal osteosarcomas that harbored MDM2 amplifications.10 It is therefore of particular interest that Carter et al.22 reported that 5/9 parosteal osteosarcomas harbored GNAS mutations. Specifically, they identified GNAS p.R201C mutations in four cases and p.R201H in another case. Of the five parosteal osteosarcomas, two additionally revealed MDM2 amplifications.22 This recent publication questions the reliability of using GNAS alterations for supporting a diagnosis of fibrous dysplasia and excluding a diagnosis of low-grade central and parosteal osteosarcomas. The aim of this study was to extend the study undertaken by Carter et al. to assess the frequency of GNAS alterations in a larger set of low-grade osteosarcomas.

Materials and methods

Patients and Samples

The study included review of the histology and clinical notes from 40 patients with osteosarcoma (29 parosteal, 6 periosteal, and 5 low-grade central) from the files at the Royal National Orthopaedic Hospital, and from 50 patients with osteosarcoma (26 parosteal, 5 periosteal, and 19 low-grade central) from the University Hospital Basel (Table 1). Ethical approval was obtained from the Cambridgeshire 2 Research Ethics Service (reference 09/H0308/165), the University College London Biobank for Health and Disease ethics committee (covered by the Human Tissue Authority license 12055: project EC17.1), and from the Ethikkommission beider Basel (reference 274/1).

Table 1 Clinical parameters of the two cohorts of patients with low-grade osteosarcomas
Full size table

DNA Extraction

Representative formalin-fixed paraffin-embedded tissue blocks were selected following review by at least two experienced bone tumor pathologists. DNA was extracted from tumor-rich areas (>60% tumor). Five to 10 consecutive 10-μm thick sections were cut and placed in eppendorf tubes (Eppendorf, Hamburg, Germany). If tumor enrichment was required, unstained sections were microdissected manually using a 14 g hypodermic needle.23 Genomic DNA was extracted from the formalin-fixed paraffin-embedded material using QIAamp DNA FFPE Tissue Kit (Qiagen, Hilden, Germany). DNA was extracted from separate blocks containing conventional and dedifferentiated components of parosteal osteosarcoma. DNA concentration and quality were assessed by Nanodrop spectrophotometer (ThermoFisher, Wilmington, DE, USA) and the Qubit dsDNA HS Assay Kit (Life Technologies, Carlsbad, CA, USA).

MDM2 Amplification by Fluorescent In Situ Hybridization

Fluorescent in situ hybridization for MDM2 amplification was performed as previously described on 5-μm sections cut from representative tissue blocks of all cases (Royal National Orthopaedic Hospital and University Hospital Basel).24 The fluorescent in situ hybridization probe for MDM2 on chromosome 12 with its respective chromosome enumeration probe was purchased from Zytovision (Zytovision GmbH, Bremerhaven, Germany), Zytolight SPEC MDM2/CEN12 (centromere 12; Z-2013) Dual Colour Probes. Non-informative cases were excluded from the analysis. The number of each signal for MDM2 (green) and CEN12 (red) was scored by counting a minimum of 50 non-overlapping nuclei per case, and the average number of MDM2 and CEN12 signals was then calculated. A ratio of >2.0 was considered to represent MDM2 amplification, a ratio of ≤2 was considered to be non-amplified. All fluorescent in situ hybridization was reviewed by at least two experienced individuals.

Next Generation Sequencing

Genomic DNA (≥10 ng) was used for multiplex polymerase chain reaction (PCR) amplification (50 amplicons panel). The resulting amplicons were barcoded (Fluidigm, South San Francisco, CA USA) and the DNA libraries were generated using the Fluidigm Access Array System (Fluidigm). Emulsion PCR was performed using the Ion Personal Genome Machine Template OT2 200 prep kit and the One Touch 2 System (Life Technologies). Templates were enriched to at least 70% of template-positive Ion Sphere particles and the enrichment was checked with the Ion Sphere Quality Control Kit (Life Technologies). Sequencing was performed on a 316 Chip v2 (Ion Sequencing 200 kit v2, Life Technologies) and run on the Ion Torrent Personal Genome Machine PGM (Life Technologies). Primers were designed with the Primer3 software (http://bioinfo.ut.ee/primer3-0.4.0/primer3/) for GNAS mutation hotspots including the missense mutations in codons p.R201, p.Q227, and p.K233 (Supplementary Table 1). Reads were aligned to the reference genome hg19 and BAM files were generated using Ion torrent suite version 4.0.1. Reads were visualized using Integrative Genomics Viewer (IGV v 2.2 Broad Institute, Boston, MA, USA) (https://www.broadinstitute.org/igv/) with the appropriate BED files. A mutation was deemed present when there was ≥5% mutation present in a region with ≥500 reads. Six cases of fibrous dysplasia known to harbor GNAS mutations were included as control samples.

Sanger Sequencing

Bidirectional Sanger sequencing for GNAS codons p.R201 in exon 8 and codon p.Q227 in exon 9 was performed for all cases. Each PCR was carried out with 0.4 μM of GNAS p.R201 exon 8 and GNAS p.Q227 exon 9 primers (Supplementary Table 1). The PCR cycling parameters included initial denaturation at 94 °C for 2 min, 40 cycles of denaturation at 94 °C for 30 s, annealing at 62 °C for 30 s, and extension at 72 °C for 1 min, followed by a final extension at 72 °C for 10 min. The PCR products of GNAS p.R201 exon 8 (159 bp) and GNAS p.Q227 exon 9 (174 bp) were separated on a 6% polyacrylamide gel and visualized by ethidium bromide staining. The DNA products were purified with the DNA Clean-up & Concentrator (Zymo Research, Irvine, CA, USA) and aliquots of 30 ng of the PCR products were used for the bidirectional Sanger sequencing (Beckman Coulter). The sequencing chromatograms were analyzed using the FinchTV software (Perkin Elmer, Waltham, MA, USA).

Mutation-Specific Restriction Enzyme Digest for GNAS Mutations

Cases from the Royal National Orthopaedic Hospital cohort exhibiting inconclusive and/or equivocal Sanger Sequencing results were also tested with the mutation-specific restriction enzyme digestion assay as reported previously.13 PCR products were used to screen for the two previously reported GNAS codons p.R201 and p.Q227 mutations in which 10 substitutions are reported to occur. In brief, for five of the mutations (p.R201G, p.R201S, p.Q227R, p.Q227H, and p.Q227K) the nucleotide substitution directly altered the cutting site of the restriction enzyme. For each of the other five mutations (p.R201C, p.R201H, p.R201P, p.R201L, and p.Q227L), a mismatch primer was designed containing a specific nucleotide substitution allowing size discrimination between alleles containing the wild type or mutant nucleotide following digestion of PCR products with the appropriate restriction enzyme.13 Products were separated through 8% polyacrylamide gels and visualized using the Bio-Rad Gel Doc 2000TM system (Bio-Rad Laboratories, Hercules, CA, USA).

Assessment of Amplifiable DNA by Droplet Digital PCR

To assess the number of amplifiable DNA molecules of the 36 parosteal osteosarcoma samples from the Royal National Orthopaedic Hospital cohort, we performed droplet digital PCR on the Q × 200 droplet digital PCR system (Bio-Rad Laboratories). The reaction was performed in duplicate with 25 ng of DNA and tested using 0.1 μM of the GNAS p.R201 exon 8 primer (Supplementary Table 1) and EvaGreen Supermix (Bio-Rad Laboratories) in order to count the number of amplifiable DNA molecules that would have been added to the Fluidigm next generation sequencing library preparation. Highly purified Human Genomic DNA (Bioline Reagents, London, UK) was used as a positive control.

Results

MDM2 Amplification by Fluorescent In Situ Hybridization

Table 2 shows the percentage of MDM2 amplifications in low-grade osteosarcomas. Of the informative samples analyzed by fluorescent in situ hybridization from the Royal National Orthopaedic Hospital tumor set, MDM2 amplification was detected in 28/34 (82%) parosteal osteosarcomas, in 2/5 (40%) low-grade central osteosarcomas, and in none of the 6 periosteal osteosarcomas. Of the 24 informative samples from the University Hospital Basel, 6/9 (67%) parosteal osteosarcomas, 3/12 (25%) low-grade central osteosarcomas, and none of three periosteal osteosarcomas showed MDM2 amplification. Fluorescent in situ hybridization analysis of the seven parosteal osteosarcoma cases, from the Royal National Orthopaedic Hospital, with conventional and dedifferentiated components revealed MDM2 amplification in both components.

Table 2 MDM2 amplification in the low-grade osteosarcoma cases informative by fluorescent in situ hybridization
Full size table

Next Generation Sequencing

GNAS substitutions involving p.R201, p.Q227, and p.K233 codons were not found in 47 samples from 40 patients with osteosarcoma from the Royal National Orthopaedic Hospital cohort that were analyzed by next generation sequencing (Table 3). Seven of these samples represented the dedifferentiated component of a matching conventional parosteal osteosarcoma (Figures 1b and c). All cases, apart from one, generated reads at high depth, averaging 2800 filtered reads (Figures 1b and c), with the outlying case generating only 523 reads. All six GNAS positive controls (p.R201C and p.R201H) generated an average of 3200 filtered reads and revealed the relevant mutations (Figure 1a).

Table 3 GNAS mutational status of the Royal National Orthopeadics Hospital low-grade osteosarcoma cases by next generation sequencing
Full size table
Figure 1
figure 1

Mutational analyses of GNAS in parosteal osteosarcoma and fibrous dysplasia by next generation sequencing and Sanger sequencing. (a) Haematoxylin and eosin (H&E) image of a fibrous dysplasia (original magnification: × 20). (a1) Representation of the reads aligned to the reference genome (Integrative Genomics Viewer software—IGV v 2.1, Broad Institute) for a GNAS mutation p.R201C (CGT>TGT) in exon 8 in a fibrous dysplasia. (a2) Sanger sequencing chromatogram of a GNAS mutation p.R201C (CGT>TGT) in a fibrous dysplasia. (b) H&E image of a conventional parosteal osteosarcoma (original magnification: × 20). (b1) Representation of the reads aligned to the reference genome for a GNAS exon 8 wild-type in a conventional parosteal osteosarcoma. (b2) Sanger sequencing chromatogram of a GNAS exon 8 wild-type in a conventional parosteal osteosarcoma. (c) H&E image of a dedifferentiated parosteal osteosarcoma (original magnification: × 20). (c1) Representation of the reads aligned to the reference genome for a GNAS exon 8 wild-type in a dedifferentiated parosteal osteosarcoma. (c2) Sanger sequencing chromatogram of a GNAS exon 8 wild-type in a dedifferentiated parosteal osteosarcoma.

Full size image

Sanger Sequencing

Bidirectional Sanger sequencing was performed for GNAS codons p.R201 and p.Q227: mutations were not detected in any of the 57/62 parosteal (p.R201) and in 58/62 parosteal (p.Q227) osteosarcoma samples from 55 individuals as well as in none of the included 24 low-grade central osteosarcomas and 11 periosteal osteosarcomas (combined from both centers). In total, 20 cases exhibited inconclusive sequencing results by Sanger sequencing, 14 cases from the University Hospital Basel and 6 cases from the Royal National Orthopaedic Hospital. The six samples from the Royal National Orthopaedic Hospital were re-sequenced in both directions by Sanger sequencing, after which five of six samples provided interpretable calls in both directions revealing the absence of GNAS mutations. These results were further validated by demonstrating failure to detect a GNAS substitution in these 5/6 samples by mutation-specific restriction enzyme digestion (data not shown). The only Royal National Orthopaedic Hospital case that failed to give a conclusive result by Sanger sequencing and/or mutation-specific restriction enzyme digestion was a parosteal osteosarcoma for which the DNA revealed low depth of coverage by next generation sequencing (523 reads) and it was also noteworthy that this case revealed poor quality DNA (vide infra—Assessment of Amplifiable DNA by Droplet Digital PCR) and was therefore excluded from the analysis. Sanger sequencing of codon p.R201 and codon p.Q227 failed in eight and six University Hospital Basel cases, respectively (Table 4). These were not analyzed by mutation-specific restriction enzyme digestion.

Table 4 GNAS mutation status of the low-grade osteosarcoma cases by Sanger sequencing
Full size table

Assessment of Amplifiable DNA by Droplet Digital PCR

Assessment of the quality of amplifiable DNA of the parosteal osteosarcomas from the Royal National Orthopaedic Hospital was undertaken by droplet digital PCR. We considered any sample with at least 100 amplifiable copies of the GNAS exon 8 region per next generation sequencing run to be satisfactory. The only case to have <100 amplifiable copies was the parosteal sample that failed both bidirectional Sanger sequencing and mutation-specific restriction enzyme digestion (data not shown).

Discussion

In this study, we assembled a large series of low-grade osteosarcomas from two independent institutions and failed to detect a single GNAS mutation in 55 parosteal, 24 low-grade central, and 11 periosteal osteosarcomas. All cases were thoroughly re-evaluated at a histological level by experienced bone tumor pathologists and the diagnosis was confirmed. In addition, fluorescent in situ hybridization analyses revealed 34/43 (79%) informative parosteal osteosarcomas to harbor MDM2 amplifications which is in line with previous reports.5 Bidirectional Sanger sequencing was used as the ‘gold standard’ for GNAS mutational analysis as this technology has been ‘tried and tested’ for many years. Furthermore, Carter et al.22 had employed this technique to detect GNAS mutations in 5/9 parosteal osteosarcomas.22 The lack of GNAS mutations detected by Sanger sequencing was confirmed in 46 osteosarcoma samples analyzed by next generation sequencing—a more sensitive technology—whereas in the positive controls the mutations were identified. Furthermore, the absence of specific mutations was confirmed by mutation-specific restriction enzyme digestion as a third technology in a smaller subset of cases.

Of the five cases of parosteal osteosarcomas that Carter and colleagues found to harbor a GNAS mutation, three dedifferentiated during the course of the disease. Interestingly, the mutations were detected in both the conventional and dedifferentiated components in one case, only in the conventional component in another case, and only in the dedifferentiated component in a third case.22 We therefore analyzed both the conventional and dedifferentiated components of seven dedifferentiated parosteal osteosarcomas for the presence of GNAS mutations but failed to detect any GNAS mutation. The periosteal osteosarcomas included in our study neither showed GNAS mutations nor MDM2 amplifications. Although periosteal and parosteal osteosarcomas both develop on the surface of bones and belong to the group of juxtacortical osteosarcomas, periosteal osteosarcomas do not share MDM2 amplifications as a common feature which is line with previous reports.25

DNA isolated from formalin-fixed paraffin-embedded tissue is often highly fragmented and chemically degraded, especially in decalcified samples. As this might also explain our failure to detect GNAS mutations, we tested a subset of our samples for DNA quality: the DNA was found to be of sufficient quality in all but a single case that was excluded from the analysis. Consequently, the absence of GNAS mutations in low-grade osteosarcomas cannot be attributed to the lack of amplifiable DNA.

The absence of GNAS alterations in 24 low-grade central osteosarcomas (5/17 with MDM2 amplification) provides strong evidence that these mutations are mutually exclusive in fibrous dysplasia and low-grade central osteosarcoma. However, GNAS mutations per se do not exclude malignancy as malignant transformation of fibrous dysplasia can rarely occur, especially in polyostotic disease. However, in the few reported cases and in our experience (not reported), malignant transformation of fibrous dysplasia usually shows high-grade morphology and does not raise diagnostic difficulties. Hence, interpretation of genetic alterations must always be performed in the context of morphology and imaging.26, 27

Our findings support prior observations that in the appropriate context, the detection of a GNAS mutation in a primary bone tumor is supportive of a diagnosis of fibrous dysplasia, and GNAS mutations rarely, if ever, occur in parosteal osteosarcoma. The reason for the findings by Carter and colleagues, however, remain elusive. DNA contamination in a laboratory is always a concern when using PCR-based technologies and strict controls and practices must be adhered to. An alternative explanation is that the DNA was degraded and a PCR artifact occurred. On the basis of this study, we argue that the detection of GNAS mutations still remains a useful molecular tool for distinguishing fibrous dysplasia from low-grade osteosarcoma.