CTL can help maintain organismal hemostasis through directly eliminating pathogens (19). Also, CTL as the main subtype of lymphocytes in the tumor microenvironment originated from activated naïve T cells to exert anti-tumor activity(20). Emerging data demonstrated that the CTL effector function was correlated with its intratumoral density (21, 22). Previous studies showed that the infiltration of CTLs predicted clinical outcomes as well as response to immunotherapy(23). Our study also revealed that osteosarcoma with higher CTL score had a better survival outcome than those with lower CTL score, suggesting that CTL may be fundamental in killing cells in the microenvironment of osteosarcoma. Also, the DEGs between high and low CTL score groups was enriched in several pro-inflammatory pathways, indicating the infiltration of CTLs was predictive of inflamed subtype of osteosarcoma. Subsequently, the TIICs between high and low CTL score patients were characterized, further supporting the inflamed subtype of osteosarcoma can be marked by infiltration of CTLs. Finally, a CTL-related optimal risk model was developed from CTL-related DEGs by LASSO and multivariate COX regression analyses. In this model, high risk score, calculated by the expression levels of six CTL-related DEGs, was associated with poor outcome.
It was shown that the number of CTLs is an important biomarker to classify tumor microenvironment (16). Tumors with a high level of CTLs were more likely to respond to immunotherapy. Immune-excluded tumors is characterized by CTLs in the marginal region, immune-inflamed tumors with CTLs is identified by CTLs in the center region, and immune-desert tumors is absent of CTLs either in the marginal region or the center region (24, 25). Previous studies largely about melanoma have demonstrated that inflamed tumors are most responsive to immunotherapy (26, 27). However, the predictive value of CTLs in osteosarcoma is still to be addressed. A phase II trial to evaluate the efficacy of pembrolizumab in patients with osteosarcoma showed 1 partial response and 6 stable disease out of the 22 patients (28). Considering that the response rate is still unsatisfying, biomarkers to predict immunotherapy efficacy in osteosarcoma are badly needed. From our study, we find that CTLs were strongly associated with antitumor immunity pathways such as antigen processing and presentation pathway and MHC protein complex pathway (8), suggesting that CTLs might serve as a potential immunotherapy biomarker in osteosarcoma.
Tumor-associated macrophages (TAM) have two phenotypes in the microenvironment: anticancer M1 phenotype or pro-tumor M2 phenotype (29). M1-like TAMs induced by interferon-γ can activate Th1-type immune response with a high capability of antigen presentation. Several pro-inflammatory cytokines such as interleukin (IL)-1β, IL-6, IL-23, CXCL9 and CXCL10 (30, 31). The CXCL9 and CXCL10 secreted by M1-like TAMs can function as key CTL chemoattractant cytokines (32, 33). However, M2-like TAMs controlled by several anti-inflammatory cytokines such as IL-10 and transforming growth factor (TGF)-β were involved in activating Th1 immune responses and upregulates pro-tumor cytokines and chemokines, including TGF-β, IL-10, CCL17, CCL18, CCL22 and CCL24 (34). Additionally, M2-like macrophages might hamper the function of CTLs, leading to tumor progression and drug resistance (35, 36). In our study, we found that the percentage of M1-like TAMs in the osteosarcoma microenvironment was significantly higher in the CTL score high cohort, indicating the fundamental role of TAMs in regulating the infiltration of CTLs in osteosarcoma.
Finally, we developed a six genes signature for overall survival constructed on the DEGs between high and low CTL score osteosarcoma patients. Four of these genes (APBB1IP, TNFSF8, PPARG and LILRA6) were highly expressed in the high CTL group, while two of these genes (PDK1 and CBS) were highly expressed in the low CTL group. The AUC values of 1-, 3-, and 5- year for this risk model were 0.73, 0.86, and 0.84, respectively, indicating an excellent predicative value of the six-gene signature. Moreover, the high risk cohort showed significantly lower overall survival compared with low risk cohort. Therefore, high-risk patients according to our risk model need more frequent follow-ups and aggressive treatments.
There are some limitations of our study to be addressed. Firstly, the sample size of the study is limited and some other clinical pathological characteristics were also lacked. Second, the risk model of our study should be tested in other osteosarcoma datasets. Finally, all the results were generated from public datasets and need to be further validated by real clinical experiments.
In summary, the CTLs were predictive of overall survival in osteosarcoma and could be a potential immunotherapy biomarker. The predictive capability of the six-gene risk model is high, which could help clinicians assess the prognosis of osteosarcoma patients and selecting appropriate treatment strategies.