OS is the most common, yet rare, bone malignancy in adolescents and young adults, with an incidence of 3.3 per 1000,000 [4]. Owing to its rarity and clinical heterogeneity, it is extremely difficult to build firm evidence based on clinical trials. Therefore, genetic analyses of OS are fundamental to deepening our understanding of this disease, developing treatment strategies, and improving clinical outcomes. However, whole-exome sequencing and multi-omics analyses are still not always feasible, and CGP is relatively widely available in outpatient clinics.

In this study, CGP data revealed that CCNE1, MCL1, MYC, RB1, CDK4, CDKN2A/B, H3F3A, KMT2D, MDM2, RAC1, and SETD2 mutations are associated with age (Table 1). CCNE1, CDKN2A, CDKN2B, and STED2 were also associated with age in the AACR Project GENIE cohort. A large array-based next-generation sequencing study has demonstrated that copy number variants were more frequent in CCND3, AURKB, CCNE1, GID4, and MYC in younger patients with OS, whereas MDM2, CDKN2A/B, and FRS2 were more frequently altered in older patients with OS [20]. Although the cutoff ages differed (40 and 30 years), both studies revealed that CCNE1 and MYC amplifications occurred predominantly in younger patients, whereas MDM2 amplification and CDKN2A/B deletions occurred in elderly patients. In previous studies, CCNE1 and MYC were amplified in 7–33% and 7–52% of OS samples, respectively [21,22,23,24], and co-amplification of CCNE1 and MYC was a rare event (1.1%) with an aggressive clinical course [25]. MDM2 was amplified in 7–26% and CDKN2A was deleted in 7–25% of OS samples [21,22,23,24]. In contrast, two studies using cutoff ages of 18 and 21 years showed no genomic differences between younger and older age groups [21, 22]. These differences among studies may be due to the different cutoff ages and the relatively small number of patients. Consistent with our speculations, one study showed that OS patients < 18 years of age have significantly more clustered rearrangements associated with chromothriptic regions than patients ≥ 50 years of age. However, no statistical difference was observed between patients aged < 18 and those aged 18 to 50 [26]. Taken together, these findings suggest that heterogeneous OS develops through multiple distinct oncogenic mechanisms and that these mechanisms are, to some extent, age-related.

Several drug resistance factors, such as ABC transporters, DNA repair factors, non-coding RNA, and cancer stem cells, have been reported in OS [27]. RNA sequencing data for 43 primary OS samples from the TARGET-OS database revealed that chemoresistant OS is characterized by the upregulation of osteogenic markers and downregulation of immune response markers [28]. Two genomic sequencing studies involving 25 and 48 samples have shown that chemo-responders have higher frequencies of COSMIC3 signatures associated with homologous recombination repair deficiency than those of non-responders [26, 29]. Although these three studies compared chemotherapy responses based on histological evaluations, we did not have histological data and employed reported data based on the RECIST criteria for each chemotherapy regimen. Owing to the targeted sequencing procedure and differences in methods for assessing chemosensitivity, we were unable to identify osteogenic, immune-related, or DNA repair markers as chemosensitivity indicators. However, gene alteration signatures, based on unsupervised gene clustering, tended to reflect the chemotherapy response. Previous studies have shown that MDM2 amplification is mutually exclusive of TP53 alterations [21, 22]. Additionally, an unsupervised clustering analysis has shown that TP53, MDM2-CDK4, and CDKN2A/B alterations form distinct groups [30]. Consistent with these previous findings, our unsupervised clustering revealed distinct CCNE1 and TP53 alteration-positive and MDM2 and CDKN2A/B alteration-positive groups, with different age distributions. Therefore, the chemotherapy-responsive subset characterized by these distinct gene alteration signatures may be generalizable to the OS population. We also found that MYC amplification was significantly correlated with the IFO response (Table 6). Several studies have reported that MYC amplification is associated with cisplatin and methotrexate resistance and a poor prognosis [24, 31, 32]. This discrepancy in responses to IFO and other drugs has led to controversial results in clinical trials [5,6,7,8]. Notably, in this study, age showed the strongest correlation with chemotherapy response. Additionally, the gene alteration signature and MYC amplification associated with chemotherapy response were significantly associated with age. These results suggest that age-biased gene alterations may contribute to the high chemotherapy sensitivity observed in young patients with OS. Further studies are needed to validate our findings and to explore the utility of the CGP test in selecting a chemotherapy regimen.

In Japan, the CGP test is covered by public insurance for patients with advanced disease after the completion of standard treatment. However, the CGP test serves various purposes; it is used for diagnosis, prognostic prediction, and the identification of cancer predisposition, in addition to guiding genome-informed therapies. Therefore, the optimal timing for testing must be determined by a specialist based on the specific disease context. Clinical practice guidelines recommend that the timing of the CGP test should not be determined solely by the line of treatment [33]. Earlier CGP testing may improve access to genotype-matched clinical trials because previous treatment lines sometimes impede eligibility [34]. Furthermore, a decline in overall health is a common reason for not undergoing genotype-matched therapy [35]. Our results also suggest that earlier CGP testing may contribute to favorable outcomes. The accumulation of evidence from CGP testing, beyond the selection of genome-matched therapy, may lead to the earlier adoption of CGP tests in the future, especially in rare cancers.

This study had several limitations. First, this study had an inherent selection bias, as it consisted mainly of patients with poor prognoses because the CGP test is indicated for advanced cases. Second, we used the FoundationOne® CDx; accordingly, our analyses were limited to 324 genes related to oncogenesis. Therefore, we did not investigate gene alterations related to other factors, such as osteogenesis or immunosuppression. Third, with respect to chemotherapy sensitivity, a central review was not conducted, and responses were assessed by each treating physician. Additionally, owing to the lack of dose intensity data for each regimen in the C-CAT database, we were unable to evaluate chemotherapy sensitivity in relation to the intensity of chemotherapy. Furthermore, we found a relatively weak correlation between gene clustering and response to chemotherapy. Therefore, further studies using CGP tests that encompass a broader range of gene alterations are necessary to identify robust predictors of chemotherapy sensitivity. Despite these limitations, the relatively large patient cohort and the use of a readily available CGP test are key strengths of this study.

In conclusion, our analysis of CGP data revealed distinct age-related distributions of genetic alterations in patients with OS. The combination of these gene alterations may be valuable for predicting sensitivity to chemotherapy. Early CGP testing could be beneficial in selecting an optimal treatment strategy.