There is plenty of evidence pointing out that the immunosuppressive TME exhausts T cells and renders them anergic. This subsequently enables tumor cells to evade host immune-mediated elimination.[34] Costimulatory molecules, especially the immune checkpoints, expressed on cancer cells or tumor-infiltrating lymphocytes play vital roles in regulating the anti-tumor immune response. Further, the blocking antibody targeting PD-L1/PD-1 has directly prolonged survival in patients with metastatic cancer.[35, 36] Presently, the costimulatory molecules mainly consist of two major families: the B7-CD28 family and the TNF family.[37] In this study, we simultaneously detected the expression pattern and clinical significance of 60 costimulatory molecules in patients with LUAD. Based on the significant genes, we developed a novel survival prediction model (CMS) based on the expression of five costimulatory molecular features in the TCGA dataset. The CMS score was found as an independent risk factor for patients with LUAD. Furthermore, the CMS was well validated in eight different public GEO datasets and 77 cases from frozen tissues with qRT-PCR data. Interestingly, through prognostic meta-analysis, we proved that our CMS had better prognostic value than the previous costimulatory molecule-related signature. We also explored the immune panorama—including immune cell distribution and inflammatory activities—in CMS high- and low-risk patients. More importantly, we found that CMS high-risk patients were considered the optimum candidates for immune checkpoint-based immunotherapies. To our knowledge, this is the first and most comprehensive study to date to describe the prognostic and immunotherapy response prediction value of a CMS in patients with LUAD.
To get the whole picture of costimulatory molecule expression in patients with LUAD, we collected the 13 members from the B7-CD28 family and the 47 members from the TNF family into our analysis.[13, 16] After the univariate Cox proportional hazards regression analysis and stepwise Cox proportional hazards regression model, we found that all five selected genes (CD40LG, TNFRSF6B, TNFSF13, TNFRSF13C, and TNFRSF19) belonged to the TNF family. This indicated that costimulatory signals and pathways in the TNF family had a more important prognostic value than those in the B7-CD28 family in patients with LUAD. CD40LG—also known as CD40L, TNFSF5, or CD154—is a membrane-bound protein belonging to the TNFSF family. CD40LG has been a therapy target in cancer treatment because of its ability to trigger Th1-type immune responses.[38] The expression and prognostic states of the CD40LG-CD40 axis was previously reported in lung cancer.[39] TNFRSF6B, a soluble decoy receptor, is also known as Decoy receptor 3 (DcR3), belongs to the TNFRSF family.[40] TNFRSF6B inhibits apoptosis and promotes angiogenesis through binding with FASL, LIGHT, and TL1A.[41, 42] Moreover, studies found that DcR3 is a potential immunotherapy target for cancer treatment.[43] TNFSF13, also known as APRIL and CD256, is a proliferation-inducing ligand that plays an important role in B cell development.[44] The clinical significance of TNFSF13 in several cancers was previously revealed and included NSCLC,[45] breast cancer,[46] B-cell chronic lymphocytic leukemia,[47] and other tumor types. TNFRSF13C (BAFFR or CD268), a receptor of BAFF, is a crucial regulatory factor in B cell proliferation, development, and maturation.[48] Hong Qin et al. reported that a novel anti-BAFFR antibody may be a promising strategy for drug-resistant B-cell malignancies.[49] TNFRSF19, also known as TROY or TAJ, is a member of the TNFRSF family and demonstrates complex and pleiotropic functions in different cellular contexts.[50] Present evidence displayed that TNFRSF19 acted as a tumor suppressor in patients with lung cancer.[51] Although the expression details of these five members in various cancer types have been described, the combination and functions of these molecules still warrants further exploration.
To verify the robustness of CMS, we reproduce the model in nine different cohorts, and the significance of CMS was finally confirmed by prognosis meta-analysis. It is worth mentioning that the number of validation cohorts in our research was larger than that of any other studies in the LUAD population. This made our signature more reliable and clinically feasible. Before our study, a signature based on the expression of costimulatory molecules from the B7-CD28 family was constructed.[31] Through meta-analysis, we obtained two crucial conclusions: the CMS signature had prognostic significance across these public datasets, although some of the P-values were not statistically significant and our CMS model demonstrated an advantage over the reported B7-CD28 model. These conclusions are consistent with our finding that the TNF family has a more important prognostic value for patients with LUAD.
Through analysis, the most related genes of CMS, the potential mechanisms of CMS in LUAD was proved to be closely associated with the immune-related process. Hence, the details of CMS-specific immune profiles were further analyzed. We found that there were higher proportions of DCs, NKs, and Tregs in CMS high-risk patients TME. Simultaneously, inflammatory metagene analysis revealed that CMS score was negatively related to monocyte/myeloid lineage- and T cell-specific functions (HCK and LCK). What's more, CMS score was also found negatively related to the antigen-presenting process of T cells (MHC-I and MHC-II) in LUAD. Thus, CMS high-risk patients appear to exhibit a high immune cell infiltration microenvironment while in an immune-suppressive state.
Interestingly, this research highlighted the potential role of CMS in predicting the response to immunotherapy in patients with LUAD. Because the immune checkpoint targets (PD-L1 and PD-1) are costimulatory molecules, CMS may have the ability to predict the response to ICIs-based immunotherapy. Due to the lack of details regarding mRNA expression in cases with immunotherapy, we had to evaluate the relationship indirectly. We collected TMB, the number of neoantigens, the protein level of PD-L1, and the TIDE scores. TMB is one of the classic biomarkers for immunotherapy response, and neoantigen burden is always increased by TMB. This will be useful for T cell recognition.[52, 53] The PD-L1 expression level was another well-known biomarker for ICIs in lung cancer.[54] The TIDE score is a newly-developed method for immunotherapy response prediction, and considered a more accurate biomarker than TMB or PD-L1 expression.[30] Collectively, high-risk patients exhibited high TMB and PD-L1 expression. From a mechanical standpoint, this resonated with the results of the immune profile analysis. By comparing the CMS scores with these different verified biomarkers, we confirmed that CMS high-risk patients were more suitable for immunotherapy. These findings give us additional confidence that the CMS scores may act as a novel predictive biomarker for immunotherapy response.
There are some limitations to this study that warrant consideration. Firstly, although we tried our best to include as many independent datasets as possible for validation, this study was retrospective. Secondly, the CMS-specific immune landscape was realized through bioinformatic methods with RNA-seq data. This analysis may have been influenced by noise. Thirdly, because the mRNA expression data from patients with immunotherapy was not available, the prediction ability of CMS for immunotherapy response was estimated indirectly. Future prospective studies could affirm the complete prediction ability and a molecular picture of the CMS signature.