In this study, we proposed a novel classification method for sarcomas based on the enrichment scores of 14 pathways, which were involved in immune, stromal, DDR, and oncogenic signatures. In three datasets for bulk tumors and a scRNA-seq dataset, we reproducibly identified three sarcoma subtypes: Imm-C, Str-C, and DDR-C. Imm-C had the strongest anti-tumor immune signatures and the lowest ITH; Str-C showed the strongest stromal signatures, the highest genomic stability and global methylation levels, and the lowest proliferation potential; DDR-C had the highest DDR activity, expression of the cell cycle pathway, tumor purity, stemness scores, proliferation potential, and ITH, the most frequent
TP53 mutations, and the worst survival. It is interesting to observe that there was no significantly different TMB between DDR-C and Imm-C, while their immune infiltration levels were significantly different. Two possible reasons could explain this observation: (1) DDR-C had more frequent arm-level copy number amplifications and deletions than Imm-C; and (2) DDR-C had higher ITH than Imm-C, since both CNAs [
41] and ITH [
18] are negatively correlated with anti-tumor immune response. DDR-C displayed worse clinical outcomes than Imm-C and Str-C. It could be attributed to the high proliferation potential, stemness, ITH, and genomic instability in DDR-C. Meanwhile, the lowest ratios of immune-stimulatory/immune-inhibitory signatures in DDR-C indicate the least activation of anti-tumor immune response in this subtype. It could also be a factor that leads to the worst prognosis in DDR.
The TCGA Research Network analyzed six types of adult soft tissue sarcomas: DDLPS, LMS, UPS, MFS, MPNST, and SS [
1]. We found that 52%, 38%, and 10% of DDLPS tumors belonged to Str-C, Imm-C, and DDR-C, respectively (Additional file
7: Fig. S4A). It indicates that most DDLPS tumors are Str-C or Imm-C. In contrast, 80% of SS and 67% of ULMS tumors are DDR-C, indicating that SS and ULMS are dominated by DDR-C. In addition, 0%, 7%, and 8% of SS, ULMS, and STLMS tumors belonged to Imm-C, compared to 55% of UPS tumors being Imm-C; 5% of UPS tumors were Str-C, compared to 52% of DDLPS and 53% of STLMS tumors belonging to Str-C. Taken together, these data suggest that the different types of adult soft tissue sarcomas are dominated by different subtypes we identified. In another study [
42], Thorsson et al. identified six immune subtypes of TCGA pan-cancer, including wound healing, IFN-γ dominant, inflammatory, lymphocyte depleted, immunologically quiet, and TGF-β dominant. We found that the wound healing sarcomas were mainly Str-C and DDR-C, and only 11% were Imm-C; the IFN-γ dominant sarcomas were mainly Imm-C and DDR-C, and only 16% were Str-C; the inflammatory sarcomas were dominated by Str-C (71%), compared to 10% of the lymphocyte depleted tumors were Str-C (Additional file
7: Fig. S4B). This study revealed that the inflammatory subtype and lymphocyte depleted subtype had the best and worst prognosis, respectively. It is in accord with our results of the worst survival in DDR-C in that 10% of the inflammatory sarcomas and 61% of the lymphocyte depleted sarcomas belonged to DDR-C, respectively. In addition, Gibault et al. identified five subtypes (Clusters A–E) of soft tissue sarcomas based on gene expression profiles [
6]. Cluster B was a subgroup of sarcomas with a favorable metastasis outcome in multivariate analysis [
6], which constituted the least proportion (9%) of DDR-C among the five clusters (Additional file
7: Fig. S4C). It is consistent with the worst prognosis of DDR-C among the three subtypes we identified. Clusters C–E harbored poorly differentiated LMS, UPS, MFS, and pleiomorphic sarcomas of the limbs and displayed combinations of expression of genes involved in invasion, extracellular matrix, or inflammatory processes, which were predominant in DDR-C, Str-C, and Imm-C, respectively. Hence, our subtyping method demonstrated a clearer separation of sarcomas with respect to pathways and biological processes compared to previous subtyping methods.