To explore a prognostic nomogram of tumor is not only critical to know of the probability of the survival outcome but also the selection appropriate treatment options. Despite the progress made in lung cancer OS predictive models, most of which are constructed by gene expression, DNA methylation, radiographic, genome mutation or variant, new predictive rule based on novel biomarkers still need to be done to discover a more precise forecast model independently or by combining with previous nomograms. A-to-I editing has been widely implicated in cancer as reported recently, which produces a great potential for cancer prognostic prediction. Out study successfully developed an A-to-I editing-based nomogram for LUSC. The C-index, calibration plot and decision curve demonstrated a good predictive performance of this nomogram in both the training group and validation group. To the best of knowledge, this is the first study reporting a nomogram for cancer with regard to A-to-I editing.
Our study first identified OS-related A-to-I sites for LUSC according to transcriptome-wide A-to-I editing analysis based on the TCGA data, following by Lasso regression to determine an A-to-I biosignature including TMEM120B(A2588I), HMOX2(A224I), CALCOCO2(A2603I), MIR548AE2(A113641I), ZNF440(A3942I), CLCC1(A2315I), and CHMP3(A23735I), which was significantly associated with OS of LUSC patients. Besides, this biosignature was positively correlated with T stages and clinical stages. With regard to the corresponding genes the seven A-to-I sites betiding, HMOX2 is a biomarker for tumor initiating cells of lung cancer and a potentially therapeutic target, suppression of which will significantly increase cancer survival. CALCOCO2 is an autophagy receptor that contributes to autophagy addiction in K-Ras driven lung cancer[32, 33]. CLCC1 acts as a cell-surface biomarker for tumor initiating cells. CHMP3 is a possible tumor suppressor with downregulated expression across a wide range of human cancers and its high level predicts a better survival outcome of breast cancer patients. These evidences supported the functional underpinnings of the A-to-I risk biosignature. Nevertheless, the biological effects of TMEM120B, MIR548AE2 and ZNF440 on LUSC have not been studied.
A-to-I editing may lead to non-synonymous amino acid mutations, mis-regulation of alternative splicing, disturbance codon preference, and microRNA-mRNA redirection or RNA-binding protein-mRNA redirection, thereby influencing gene expression and function. Among the seven A-to-I editing sites, TMEM120B(A2588I), CALCOCO2(A2603I), ZNF440(A3942I) and CLCC1(A2315I) are located in the 3’-UTR of their host genes, and HMOX2(A224I) is located in the 5’-UTR of HMOX2, thus they have a great potential to affect expression of host genes. Indeed, the editing level of ZNF440(A3942I) and CLCC1(A2315I) were negatively correlated with the expressions of ZNF440 and CLCC1, respectively. HMOX2(A224I) editing level might be also related to HMOX2 expression with a clear trend to be statistically significant. These results suggested that the three editing sites are functional. However, no association was observed between the expression and the editing level of other two genes (TMEM120B and CALCOCO2), suggesting an unconventional regulatory mechanism across them. In future, we will focus on verifying the biological functions of these A-to-I sites. Wonderingly, MIR548AE2(A113641I) and CHMP3(A23735I) are located in the intron of MIR548AE2 and CHMP3, respectively. We guessed they may belong to pre-mRNA sequences or non-coding RNA molecules that have not been identified by now.
Moreover, a difference in the editing level of TMEM120B(A2588I), CLCC1(A2315I), and CALCOCO2(A2603I) was observed between tumor tissues and normal tissues, suggesting a role of these sites on affecting the occurrence of LUSC. Therefore, these sites may be alternative biomarkers for LUSC diagnosis.
To construct a prognostic model for LUSC, we developed a nomogram incorporating the seven A-to-I sites, age at diagnosis, gender, and TN stages. The nomogram exerted a medium accuracy on predicting OS of LUSC, which is similar with those were reported in other studies. Besides, the nomogram had better overall net benefit than the T, N stating system at 1 and 3 years. Although the nomogram showed well predictive performance in our study, to apply the nomogram to the "real-world" clinic, more LUSC cohorts, especially the prospective cohorts, are needed to assess the robustness. In addition, the current study have several limitations. First, we lacked an external group to validate the A-to-I biosignature and the nomogram. Second, therapeutic schedule is important to show the application of prognostic nomogram for selection of appropriate treatment options, but the data is unavailable in the TCGA database. Finally, there may be some bias in the process of subjects’ recruitment and data analysis.
In conclusion, we found the first risk prognostic biosignature based on A-to-I RNA editing in LUSC, which was associated with OS and clinical stages of LUSC patients. We further constructed a nomogram incorporating the A-to-I biosignature and clinicopathological features. The nomogram exerted well predictive performance and reliability for LUSC OS. Large prospective cohorts are warranted to validate the robustness of this model to assess the application value in "real-world" clinic.