It is generally believed that patients diagnosed with late stage lung ADC tend to have a poorer clinical prognosis [2, 6, 43, 44]. Therefore, seeking for a robust prognostic signature to distinguish high-risk from low-risk patients and a nomogram to predict each patient’s survival probability, is essential for determining the optimum therapeutic strategies and reversing the adverse prognosis for lung ADC patients. Accumulating evidence proved that abnormal m6A RNA methylation modifications were intimately related to the pathogenesis and development of various cancers [11, 23, 29, 31, 45–47], whereas the studies investigating their role in lung ADC are few. In this study, we discovered a significantly abnormal expression of m6A regulator genes in lung ADC. Besides, a robust five-gene prognostic signature was developed and regarded as an unfavorable prognostic factor for lung ADC. Based on the signature, we constructed a prognostic nomogram to predict each patient’s survival probability. To further assess the model performance, an exterior validation set (GSE72094) was used, and the nomogram was demonstrated with a satisfactory discrimination and calibration performance.
Our results revealed that seventeen out of twenty m6A regulator genes were observed with either up-regulation or down-regulation in lung ADC samples, indicating these genes might be involved in the oncogenic activities and prognosis for lung ADC patients. Unanimous with our findings, the abnormal expression of m6A regulator genes was also found in other cancers. For instance, it is reported that the writer WTAP was found to be up-regulated in various cancers, including hepatocellular carcinoma (HCC) [48], acute myelogenous leukemia (AML) [49], and glioblastoma [50], and was identified as an oncogene. The eraser ALKBH5 was reported to be highly expressed in ovarian cancer [51], and the silencing of ALKBH5 significantly improved autophagy and suppressed proliferation and invasion. The reader YTHDF1 was observed with an up-regulation in CRC [52], and its knockdown significantly inhibited the tumorigenicity and colonosphere formation ability. Therefore, expression dysregulation of m6A regulator genes was prevalently found in various cancers, and was strongly related to tumorigenesis, progression, and prognosis.
When GSEA was adopted, we found m6A regulator genes were significantly associated with some important malignancy-related pathways, including p53 signaling pathway, cell cycle, mismatch repair, nucleotide excision repair, and pathways in cancer. Comparable results were also reported in other studies. It is reported that m6A regulator genes were extremely important in several cancer-related pathways, such as p53 signaling pathway [29, 53], cell cycle [29, 30, 54], Ras [30], inflammatory response [29, 53], and PPAR signaling pathway [53].
In the present study, a five-gene prognostic signature (IGF2BP1, IGF2BP2, HNRNPA2B1, METTL3, and HNRNPC) was identified and displayed with a great prediction value for patients with lung ADC. Furthermore, a distinct survival difference was found between different risk groups, which denoted that clinicians could successfully stratify the lung ADC patients into low-risk and high-risk groups based on the risk scores derived from the prognostic signature, thereby conducing them to make wiser clinical decisions. For the five identified m6A regulator genes, IGF2BP1, IGF2BP2, HNRNPA2B1, and HNRNPC were determined as adverse prognostic genes, while METTL3 was determined as a favorable prognostic gene. Interestingly, all these genes were highly expressed in lung ADC patients. Yan M et al. demonstrated that HNRNPC was drastically up-regulated in non-small cell lung cancer (NSCLC) tissues, which was in agreement with our research. Then, the researchers also discovered the overexpression of HNRNPC dramatically accelerated the proliferation, migration, and invasion of lung cancer cells. Moreover, the elevated expression of HNRNPC was significantly related to advanced tumor stages, metastasis, and shorter survival time [55]. IGF2BP1, a target of miR-491-5p, was reported to be significantly increased in the expression of NSCLC cells, and promoted tumor cell proliferation, migration, and invasion [56]. The same trend was also discovered in liver cancer [57]. Zhu J et al. revealed that METTL3 was significantly up-regulated in lung adenocarcinoma, and patients with high expression of METTL3 were observed with significantly better OS [24]. Similarly, another study reported that METTL3 was regarded as a tumor suppressor gene for CRC, and the up-regulation of METTL3 could dramatically suppress tumor cell proliferation, migration and invasion. Furthermore, patients with elevated expression of METTL3 were observed with a significantly better survival [25, 45]. Overall, their findings were in agreement with ours. However, other studies demonstrated an entirely opposite function of METTL3 for bladder cancer [58], NSCLC [59], and liver cancer [24]. Du M et al. proved that METTL3 was targeted by miR-33a and the down-regulated METTL3 could inhibit the proliferation of lung cancer cells [59]. In addition, another study found METTL3 was significantly up-regulated in bladder cancer and its knockdown could significantly suppress bladder cancer cell proliferation, invasion, and survival in vitro and tumorigenicity in vivo [58]. These findings demonstrated that METTL3 plays different (or even opposite) roles in different cancers [60]. IGF2BP2, a direct target of miR-485-5p, was discovered to be significantly up-regulated in the expression of NSCLC, and its depletion significantly suppressed tumor cell proliferation and invasion [61]. Similarly, IGF2BP2 was overexpressed in pancreatic cancer and patients with the overexpression of IGF2BP2 were observed with significantly worse OS. Furthermore, the up-regulated IGF2BP2 expression promoted tumor cell growth by activating the PI3K/Akt signaling pathway [62]. HNRNPA2B1, involved in RNA-binding and pre-RNA processing, was high expression in NSCLC patients and associated with a worse prognosis. Furthermore, the overexpression of HNRNPA2B1 accelerated NSCLC cell growth by activating COX-2 signaling pathway [63].
Based on the multivariate Cox model, we developed a prognostic nomogram to predict each lung ADC patient’s 1-year, 3-year and 5-year OS probability. As far as we know, this was the first study incorporating the m6A regulator genes into the model to construct a prognostic nomogram for lung ADC patients. Notably, the nomogram performed well both in the training and validation set, indicating a robust prediction performance. To evaluate the clinical usefulness, we utilized DCA to ascertain whether the nomogram-based decisions could improve patients’ outcomes. The DCA showed that the threshold probabilities were within the range of 0.06–0.48, 0.18–0.81, and 0.42–0.80 for 1-year, 3-year, and 5-year OS, respectively. If the threshold probabilities of lung ADC patients were within the above ranges, adopting the nomogram to predict OS added more benefits than either hypothetical treat-all-patients or treat-none scenarios. In addition, we also found the nomogram had a higher predictive accuracy than the traditional prognostic index—AJCC stage. In summary, our five-gene signature-based nomogram was effective in predicting each lung ADC patient’s survival probability and could offer a better reference for treatment guidance than single traditional clinical index.
To explore the potential reasons for risk signature being an independent prognostic factor for lung ADC patients, we conducted a WGCNA to identify the related modules, physiological biology, pathways, and hub genes. Our results revealed that cell cycle, an important cancer-related pathway, was significantly associated with the risk signature. Furthermore, a total of six genes (CCNA2, CCNB1, BUB1B, BUB1, KIF2C, and KIF11) were determined as the real hub genes. Notably, these hub genes all played an important role in the proceeding of cell cycle [64–70]. As our prognostic signature was consisted of five m6A regulator genes (IGF2BP1, IGF2BP2, HNRNPA2B1, METTL3, and HNRNPC), we inferred that these five m6A-related genes might act with the six hub genes, thereby influencing the process of cell cycle. Further experimental studies were warranted to ascertain the real molecular mechanisms.
ICB therapy has revolutionized traditional treatment strategies for NSCLC and other cancers, and advanced patients treated with anti-PD-1 and anti-CTLA-4 agents have been demonstrated with better prognosis [71–73]. In addition, previous literatures have reported that m6A is extremely important in immunoregulation and autoimmune diseases [74, 75]. Therefore, it is essential to investigate the association of the developed m6A signature with immunity. To our knowledge, this is the first study assessing the relationship between m6A signature with immunity. It is reported that higher PD-L1 expression in tumor cells was closely associated with improved efficacy of immunotherapy [76, 77], and the stimulation of PD-1/ PD-L1 pathway inhibits CD8 + T cell proliferation and promotes its apoptosis [78, 79]. In this study, we found high-risk lung ADC patients had significantly higher expressions of PD-L1 and PD-L2, and lower proportion of CD8 + T, which suggested high-risk lung ADC patients had a superior response to ICBs. Additionally, we also found high-risk patients had significantly higher TMB. Several reports have demonstrated that TMB serves as a useful biomarker of response for PD-1/PD-L1 blockade across various cancers, and higher TMB is associated with superior progression-free survival and objective response rates [77, 80–82]. Finally, after using TIDE algorithm, we also found high-risk patients had a significantly higher proportion of ICB response than low-risk patients. All the above findings indicated that high-risk patients might be more sensitive to the ICB therapy and more likely to benefit from immunotherapy. With the aid of the signature, clinicians can be easier to identify the potential beneficiaries from immunotherapy.
Several limitations in the present study should be noted. First, all our data were derived from the existing public database. Although an excellent performance in the prediction of the survival for lung ADC patients was observed in our study, a prospective, multicenter study was still warranted to further validate our results. Second, as the study populations were mainly from the US, and we only focused on white and black patients, our findings might not be generalized to other countries and races. Third, the C-index of our nomogram was merely around 0.7 and the AUC of the signature was within 0.6–0.7. Therefore, incorporating some acknowledged prognostic factors, such as tumor grade, radiation, chemotherapy, operation modes, and immunotherapy, might be conducive to enhancing the prediction accuracy of the present model. Finally, all our results were based on the data mining, and consequently an experimental study is essential for better ascertaining the associated molecular mechanisms.