Radiotherapy (RT) plays a crucial role in cancer treatment and more than half of all cancer patients receive RT during the course of disease (Baskar et al., 2012). In last 20 years, the outcomes of RT have been improved dramatically due to the developments of highly conformal RT techniques, such as intensity-modulated RT (IMRT), intensity-modulated arc therapy (IMAT) and stereotactic RT (SRT) (Minniti et al., 2021). However, unfortunately the outcomes of therapy are not fully satisfactory. The response to radiotherapy varies widely between individuals and some patients are resistant to RT treatment. Radioresistance is considered as a main factor impeding efficacy of radiotherapy. Although radioresistance has been implicated to be associated with several biological alterations of the tumor cells, such as tumor metabolism (Tang et al., 2018), cell cycle arrest (Chen et al., 2017), oncogene and tumor suppressor alterations (Pitroda et al., 2009), microenvironment (TME) change (Suwa et al., 2021), autophagic regulation (Chang et al., 2014), cancer stem cells (CSCs) generation (Ning et al., 2013), and DNA damage response (DDR) and repair (Huang and Zhou, 2020; Sun et al., 2020), the mechanisms underlying resistance to radiation are still largely obscure. Therefore, uncovering the processes and charactering the molecules associated with regulation of radioresistance may lead to improved, efficient treatment for cancer patients.

Previous studies revealed that the difference in the status of mutation and expression profile of gene including microRNAs (miRNAs), is associated with radioresistance (Mercatelli et al., 2008; Huang et al., 2013; Mueller et al., 2013; Hatano et al., 2015; Mao et al., 2016; Xu et al., 2016; Tao et al., 2018; Xu et al., 2018). MicroRNAs are small non-coding single-stranded RNAs of 19–23 nucleotides in length, which play a critical role in post-transcriptional regulation by degrading or preventing the translation of their target messenger RNA (mRNA). They play important roles in tumor development and metastasis. Emerging evidences demonstrated that miRNAs may be involved in regulation processes associated with response to radiation. For examples, previous studies revealed that some microRNAs, such as miR-95 (Huang et al., 2013), miR-221, miR-222 (Mercatelli et al., 2008), and miR-106b (Hatano et al., 2015) enhanced radioresistance in cancer cells, while miR-30a (Xu et al., 2016), miR-16 (Tao et al., 2018), miR-449 (Mao et al., 2016), miR-17 (Xu et al., 2018), and miR-100 (Mueller et al., 2013) enhanced the radiosensitivity. Several miRNAs regulate DNA damage response (DDR) pathway. For example, miR-101 can regulate the expression of ataxia-telangiectasia mutated (ATM) gene and DNA-PK. ATM is a central regulator of DNA damage response (Maréchal and Zou, 2013). Mutation and inactivation of ATM can lead to increased instability of genome and impaired repair ability for DNA double-strand breaks. It has been experimentally demonstrated that up-regulation of miR-101 reduced the protein levels of DNA-PK and ATM, rendering tumor cells more sensitive to radiation (Yan et al., 2010). MiR-223 is a regulator of maturation and differentiation of hematopoietic stem cells. Up-regulation of miR-223 will reduce the expression of ATM and make U87 cells sensitive to radiation in vitro (Liang et al., 2014). MiR-375 is a negative regulator of p53 which is a key tumor suppressor gene that inhibits cell growth by activating cell cycle arrest or apoptosis (Vousden and Prives, 2009). Increase of miR-375 was detected in recurrent gastric cancer (Zhang et al., 2011), and the increased miR-375 interacts with the 3′UTR of p53 gene that negatively regulates p53 and its downstream pathway genes, resulting in radioresistance of cells to radiation (Liu et al., 2013).

In the era of precision medicine, a biomarker of intrinsic radiosensitivity would be extremely valuable for selecting patients in whom will benefit from RT, adjusting individual dosing, and aiding decision making. Previous studies have demonstrated that several miRNAs were associated with radioresistance or radiosensitivity, implying they could serve as promising biomarkers for prediction of RT response. Besides, miRNAs serving as biomarkers have several advantages. Firstly, unlike mRNAs, miRNAs remain largely intact in routinely collected, formalin-fixed, paraffin-embedded (FFPE) clinical tissues. Therefore, the detection for miRNA levels could be conveniently performed in clinical practice. Secondly, miRNAs have been called ‘the master regulators’ of gene expression since a single miRNA can regulate several hundreds of mRNA targets. Researches have indicated that many cell phenotype or subtype is likely governed by a miRNA regulatory network (Yang D. et al., 2013), suggesting some miRNAs may sever as the possible major determinants of cellular phenotype (including radioresistant phenotype) (McDermott et al., 2017). In additional, cell lines are usually used as models for radioresistance research in previous reports. However, tumors from patients are heterogeneous. Thus, the factors related to RT response are more complex than in cell lines. On the other hand, the results from cellular experiments also need to be confirmed with clinical samples. Therefore, we proposed that miRNA profile analysis based on clinical samples could provide a direct assessment for their performances as biomarkers for prediction of RT response. In this study, we performed a large-scale analysis of miRNA expression data collected from The Cancer Genome Atlas (TCGA) from those who treated with RT across eight cancer types. By analyzing the expressional difference between RT response and progressive disease samples, the miRNA signatures for predicting RT response were obtained and their performance for prediction of RT response were estimated.

Currently, only few biomarkers have been evaluated for their radiotherapy-specific predictive value. Several research groups have reported that miRNAs were involved in regulation of radioresistance, hinting they have the potential serving as biomarkers for prediction of RT response. However, most of them was identified based on cellular experiments in previous studies. The evaluation based on clinical samples should be performed. In this study, we evaluated the predictive value of miRNA signatures for predicting RT response by using data of clinical samples. Comparison of miRNA expression of radioresistant and radiosensitive tumors led to the identification of 47 miRNAs. Most of them showed to be predictive for RT response (AUC ranging from 0.606 to 0.889). Some of them showed the high specificity for the prediction. For instance, miR-101, miR-30d, and miR-196s has AUC of 0.889, 0.889, and 0.865, respectively. To further improve the prediction performance, we developed a combinatorial model in each cancer type. In these models, the best combinations of multiple miRNAs could be obtained, which leads to an improved discriminatory power. The results showed that the most of AUC in each cancer type was greater than 0.8 when the combinatorial models were applied. Particular, the AUC value reach to 0.992 by applying the combinatorial model in BLCA. The prediction models and miRNAs identified here have the potential clinical application. As these combinatorial models contained 2-13 miRNAs, it is convenient to develop digital PCR and real-time qPCR based clinical test in routine practice (Hindson et al., 2013).

Identifying genetic clues to the molecular basis of radioresistance is a major challenge. Our study identified a series of miRNA that were differentially expressed between RT good responders and poor responders, providing useful clues for further functional assays to demonstrate a possible regulatory role in radioresistance. Among DEMs, several miRNAs were known to be involved in the regulation of DNA damage repair or radiosensitivity. For example, our results showed that miR-34a was upregulated in S group and was downregulated in R group in SARC. MiR-34a is one of the family members of miR-34 (Misso et al., 2014), and is a key regulator of tumor suppression. The overexpression of miR-34a makes cells sensitive to radiation by inhibiting several targets of DNA damage repair pathway (Lacombe and Zenhausern, 2017), such as Bcl-2 (Liu et al., 2011), LyGDI (Duan et al., 2013), Notch-1 (Kang et al., 2013), Rad51 (Cortez et al., 2015). MiR-29a is a member of the miR-29 family (Patel and Noureddine, 2012), and miR-29 is a tumor suppressor that can promote cell senescence and differentiation (Martinez et al., 2011; Chuang et al., 2022). The studies found that overexpression of miR-29a enhanced radiosensitivity, and promoted apoptosis in radiation resistant CaSki and c33a cells (Martinez et al., 2011). Overexpression of miR-29a induces a significant decrease in cell migration speed through K-ras/c-raf/p38 signal pathway, and may reduce metastasis of lung cancer (Chuang et al., 2022). In this study, we found that miR-29a in SARC tended to be highly expressed in S group. MiR-214 involved in radioresistance have been demonstrated in multiple cancer types (Zhang and Zhang, 2017; Hu et al., 2018; Li et al., 2019). In ovarian cancer, the expression level of miR-214 rises after ionizing radiation, which activates P13K/Akt pathway by targeting PTEN, resulting in the increased radioresistance of cell lines (Zhang and Zhang, 2017). In colorectal cancer, miR-214 is significantly down-regulated in cells after ionizing radiation, which resulting in the increased sensitivity of cells to radiation (Hu et al., 2018). MiR-214 is overexpressed in osteosarcoma tissues and is a negative regulator of phlda2, maintaining radioresistance of osteosarcoma cells to apoptosis (Li et al., 2019). We found that miR-214 tended to be highly expressed in the R group, implying it may play a role in radioresistance in BLCA. Besides, miR-101, miR-146a, miR-196a, miR-143, miR-222 (Shi et al., 2019), and mir-130a (Ha Thi et al., 2019), has been reported to be involved in the regulation of DNA damage repair or radiosensitivity in previous studies.

As several miRNAs previously reported to be associated with radiotherapy sensitivity (e.g., miR-34, miR-101, miR-29a, miR-214, miR-146a, miR-196a, miR-143, miR-222, and mir-130a), we hypothesized that the signatures identified in here would identify additional radiotherapy sensitivity-related miRNAs. Our gene function enrichment analysis showed the many miRNAs were involved in EMT pathway. EMT is associated with characteristics of cancer stem cells, including radioresistance and chemoresistance. Several miRNAs were reported to directly target multiple key components of EMT pathway. For example, the miR-200 family inhibits EMT and tumor metastasis, inhibits self-renewal of cancer stem cells, and enhances radiosensitivity of several types of cancer (Baumann et al., 2008). Overexpression of miR-200c regulates oxidative stress response genes and increases radiosensitivity of lung cancer cells (Magenta et al., 2011). Zheng et al. found that the down-regulation of miR-200c was related to radiation tolerance in esophageal squamous cell carcinoma (Zheng et al., 2017). Zhang et al. found miR-205 can suppress EMT by targeting the EMT-inducing transcription factor ZEB1 (Zhang et al., 2014). Function enrichment analysis indicated that 14 miRNAs, including miR-34a, miR-215, miR-221, miR-30d, miR-194-2, miR-let-7, miR-194-1, miR-129-1, miR-29b-2, miR-29b-1, miR-29a, miR-192, miR-150, and miR-143, were associated with EMT pathway in our study. This enrichment implied that these miRNAs might be involved in radioresistant phenotype through EMT pathway. Previous studies reported that miR192 regulated the EMT pathway (Khella et al., 2013). Song et al. confirmed that miR‐192 promoted EMT of gastric cancer, migration and invasion by targeting RB1 (Song et al., 2022). In additional, the previous study has demonstrated that miR-192 was significantly upregulated in cisplatin-resistant lung cancer cells, and miR-192 induced cisplatin resistance through activating the NF-κB pathway (Li et al., 2022). Moreover, miR-192 could influence 5-fluorouracil resistance (Boni et al., 2010). Cisplatin and 5-fluorouracil were genotoxic and their cytotoxic mode of action prominently involves the generation of DNA lesions followed by the activation of the DNA damage response and the induction of mitochondrial apoptosis. Zhai et al. found miR-143 suppressed epithelial–mesenchymal transition and inhibited tumor growth of breast cancer through down-regulation of ERK5 (Zhai et al., 2016). Up-regulating miR-143 enhances E-cadherin-mediated cell-cell adhesion ability, reduces mesenchymal markers, and decreases cell proliferation, migration, and invasion in vitro (Zhai et al., 2016). Yang et al. found that up-regulated miR-143 represses EMT in esophageal cancer cells (Yang et al., 2019). Our results showed that miR-143 was upregulated in S group and was downregulated in R group in esophageal cancer samples. We suggested that miR-143 might play a regulatory role in radiosensitivity through influencing EMT pathway in esophageal cancer. The further experiments will be required for function assay. We also noticed that miR-150 tended to be highly expressed in the RT response samples, implying it may play a role in radiosensitivity. Previous report showed that the expression of miR-150 decreased significantly in serum after irradiation in animal studies (Jia and Wang, 2022). Recently, miR-150-5p were confirmed to target ZEB1 and caused mRNA degradation, thus blocking EMT (Lu et al., 2017). Together, these multiple clues suggested that these miRNAs (miR-192, miR-143, and miR150) might serve as putative regulators of radiosensitivity through EMT pathway.

In conclusion, this work showed the miRNA signatures could serves as biomarkers to classify the RT response patients, and further investigations with larger numbers of patient samples are currently underway to validate the utility of using these biomarkers. Finally, additional work will be required to determine the role of these miRNAs in radioresistance of tumor.