Noncoding RNAs Controlling Oxidative Stress in Cancer

Simple Summary Mitochondria in cancer cells produce reactive oxygen species, inducing a vicious cycle between oxidative stress, genomic instability, and cancer development. However, too high oxidative stress kills tumor cells. Cancer cells protect themselves directly by increasing antioxidant expression. In addition, shifting the phenotype of immune cells from ant-oncogenic to pro-oncogenic reduces reactive oxygen levels within the tumor microenvironment to maintain tumor growth. This review shows how noncoding RNAs may regulate mitochondrial function, antioxidant expression, and immune cell reprogramming. Interestingly, noncoding RNAs in microvesicles secreted by mesenchymal stem cells, cancer-associated fibroblasts, and cancer cells contribute to this immune cell reprogramming. Further research on the role of noncoding RNAs in the communication between cell types in the tumor microenvironment is warranted. Abstract Mitochondria in cancer cells tend to overproduce reactive oxygen species (ROS), inducing a vicious cycle between mitochondria, ROS, genomic instability, and cancer development. The first part of this review deals with the role of noncoding RNAs in regulating mitochondrial ROS production and the expression of antioxidants in cancer cells, preventing the increase of ROS in the tumor microenvironment. In addition, cytotoxic T and natural killer cells release high levels of ROS, inducing cell death, while anti-immune regulatory T cells, tumor-associated M2 macrophages, and myeloid-derived suppressor cells, at least at the initial stage of tumor growth, release low levels of ROS supporting tumor growth. Therefore, this review’s second part deals with noncoding RNAs’ role in regulating the metabolic reprogramming of immune cells about ROS release. Furthermore, the enrichment of noncoding RNAs in microvesicles allows communication between cell types in a tumor and between a tumor and tumor-adjacent tissues. Therefore, the third part illustrates how noncoding RNA-containing microvesicles secreted by mesenchymal stem cells and primary tumor cells may primarily aid the shift of immune cells to a pro-oncogenic phenotype. Conversely, microvesicles released by tumor-adjacent tissues may have the opposite effect. Our review reveals that a specific noncoding RNA may affect oxidative stress by several mechanisms, which may have opposite effects on tumor growth. Furthermore, they may be involved in mechanisms other than regulating oxidative stress, which may level out their effects on oxidative stress and tumor growth. In addition, several noncoding RNAs might share a specific function, making it very unlikely that intervening with only one of these noncoding RNAs will block this particular mechanism. Overall, further validation of the interaction between noncoding RNAs about cancer types and stages of tumor development is warranted.


Introduction
In 2000, Hanagan and Weisberg defined six hallmarks of cancer: self-sufficiency in growth signals, insensitivity to growth-inhibitory signals, evasion of programmed cell death, unlimited replicative potential, sustained angiogenesis, and tissue invasion and metastasis [1]. The controlled release of reactive oxygen species (ROS) is crucial for tumor growth in hypoxia [2]. Mitochondria in tumor cells tend to overproduce ROS, particularly by NADPH oxidases (NOX). ROS induce genomic instability, and mitochondrial and

Figure 1.
Noncoding RNAs and mitochondrial function and oxidative stress. Controlled release of ROS is required to entertain tumor growth, while excessive ROS production leads to cell death. Upregulated noncoding RNAs are in red.
An increase of tumor-suppressive let-7 family and miR-34a-5p led to a decrease of MYC and perturbation of mitochondrial function. Knockdown of the let-7 family or miR-34a-5p restored MYC levels and mitochondrial function [34]. MiR-30a promoted mitochondria-dependent intrinsic apoptosis [35]. MiR-124 induced apoptosis via the intrinsic mitochondrial pathway in human oral squamous cell carcinoma cells [36]. MiR-125 targeted cancer cells and intracellular mitochondria, altering cellular bioenergetics, lipid, and glucose metabolism, inducing apoptosis in human tongue squamous carcinoma cells [37]. MiR-128 suppressed SIRT1 expression, promoting the production of ROS and apoptosis in TRAIL-treated colorectal cancer cells [38]. The mitochondrially localized growth arrestspecific transcript 5 (GAS5) could disrupt mitochondrial membrane potential and promote BAX, BAK, cleaved-caspase 3, and cleaved-caspase 9 expressions in epithelial ovarian can cells [39] and inhibit associated mitochondrial metabolic enzymes in breast cancer cells in response to nutrient stress [40] (Figure 1). An increase of tumor-suppressive let-7 family and miR-34a-5p led to a decrease of MYC and perturbation of mitochondrial function. Knockdown of the let-7 family or miR-34a-5p restored MYC levels and mitochondrial function [34]. MiR-30a promoted mitochondria-dependent intrinsic apoptosis [35]. MiR-124 induced apoptosis via the intrinsic mitochondrial pathway in human oral squamous cell carcinoma cells [36]. MiR-125 targeted cancer cells and intracellular mitochondria, altering cellular bioenergetics, lipid, and glucose metabolism, inducing apoptosis in human tongue squamous carcinoma cells [37]. MiR-128 suppressed SIRT1 expression, promoting the production of ROS and apoptosis in TRAIL-treated colorectal cancer cells [38]. The mitochondrially localized growth arrest-specific transcript 5 (GAS5) could disrupt mitochondrial membrane potential and promote BAX, BAK, cleaved-caspase 3, and cleaved-caspase 9 expressions in epithelial ovarian can cells [39] and inhibit associated mitochondrial metabolic enzymes in breast cancer cells in response to nutrient stress [40] (Figure 1).
In aggregate, the activation of cytotoxic T cells, NK cells, and dendritic cells increase ROS. Conversely, the activation of Treg and Th2 cells decreases ROS. The activation of Th1 cells increases ROS by activating dendritic cells but decreases ROS by activating Treg cells. The Th17 cells could also have divergent effects by activating MDCs and Th1 cells ( Figure  3).

Figure 3. Control of T cell function and oxidative stress by noncoding RNAs. Cancer cells evade immune suppression by downregulating the activity of cytotoxic CD8 + T cells and NK cells, suppressing cytokines, and increasing Treg cells. Hypoxia-induced activation of Wnt/β-catenin ablates
CD8 + T-cell function and suppresses anti-cancer immunity. In addition, hypoxia induces HIF1A, ADAM metallopeptidase domain 10, and MHC class I polypeptide-related sequence A (MICA), which, together with TGF-β, decreases NK cell activity. NK cells share properties with adaptive T cells, and together they induce cancer cell death through IFN-γ and tumor necrosis factor (TNF-α). Thereby, IFN-γ is synergistically enhanced by IL-4, IL-2-and IL-12 and stimulated by platelet-derived growth factor (PDGF)-DD. CD4 + T cells consist of T-helper (Th) 1, Th2, Th17, and Treg cells. DC-derived IL-12 favors the differentiation of Th1 cells. Th1 cells produce IL-2, maintaining Treg cells. In contrast, Th2 cells produce IL-4, IL-10, and IL-13, stimulating M2 TAM differentiation. Th17 cells promote tumor growth by IL-17, inducing angiogenesis and recruiting MDSCs. In contrast, Th17 cells cause anti-tumoral immunity by recruiting DCs, CD4 + T, and CD8 + T cells, activating CD8 + T cells and inducing plasticity toward Th1 cells, producing IFN-γ and TNF-α, associated with cancer Figure 3. Control of T cell function and oxidative stress by noncoding RNAs. Cancer cells evade immune suppression by downregulating the activity of cytotoxic CD8 + T cells and NK cells, suppressing cytokines, and increasing T reg cells. Hypoxia-induced activation of Wnt/β-catenin ablates CD8 + T-cell function and suppresses anti-cancer immunity. In addition, hypoxia induces HIF1A, ADAM metallopeptidase domain 10, and MHC class I polypeptide-related sequence A (MICA), which, together with TGF-β, decreases NK cell activity. NK cells share properties with adaptive T cells, and together they induce cancer cell death through IFN-γ and tumor necrosis factor (TNF-α). Thereby, IFN-γ is synergistically enhanced by IL-4, IL-2-and IL-12 and stimulated by platelet-derived growth factor (PDGF)-DD. CD4 + T cells consist of T-helper (Th) 1, Th2, Th17, and T reg cells. DCderived IL-12 favors the differentiation of Th1 cells. Th1 cells produce IL-2, maintaining T reg cells. In contrast, Th2 cells produce IL-4, IL-10, and IL-13, stimulating M2 TAM differentiation. Th17 cells promote tumor growth by IL-17, inducing angiogenesis and recruiting MDSCs. In contrast, Th17 cells cause anti-tumoral immunity by recruiting DCs, CD4 + T, and CD8 + T cells, activating CD8 + T cells and inducing plasticity toward Th1 cells, producing IFN-γ and TNF-α, associated with cancer cell death. T reg cells are a subset of FOXP3-positive CD4 + T cells strongly inhibiting anti-tumor immune responses mediated by CD4 + CD8 + T cells. Hypoxia interacts with FOXP3 and enhances the T reg cells through TGF-β1. In aggregate, the activation of cytotoxic T cells, NK cells, and dendritic cells increase ROS. Conversely, the activation of T reg and Th2 cells decreases ROS. The activation of Th1 cells increases ROS by activating dendritic cells but decreases ROS by activating T reg cells. The Th17 cells also have divergent effects by activating MDCs and Th1 cells. Pro-oncogenic ROS are in green, anti-oncogenic ROS are in red. Upregulated noncoding RNAs are in red; down-regulated in green. Arrowheads reflect activation; hammerheads reflect inhibition.
In aggregate, the activation of cytotoxic T cells, NK cells, and dendritic cells increase ROS. Conversely, the activation of T reg and Th2 cells decreases ROS. The activation of Th1 cells increases ROS by activating dendritic cells but decreases ROS by activating T reg cells. The Th17 cells could also have divergent effects by activating MDCs and Th1 cells ( Figure 3).

Noncoding RNAs and T Cells
MYC expression is deregulated in various cancer types. In breast cancer, MYC is overexpressed in 30-50% of high-grade tumors [68]. Inactivation of MYC oncogene sustained regression of invasive liver cancers [69]. MYC is frequently overexpressed in both sporadic and colitis-associated colon adenocarcinomas [70]. Furthermore, MYC is an adverse prognostic marker in colorectal cancer [71]. Multiple myeloma is considered a plasma cell malignancy associated with MYC deregulation [72]. MYC exerts a context-and cell-dependent function. For example, MYC acts early in T cell activation [73]. However, amplification of MYC alone is insufficient for tumor development in vivo; co-amplification of the PVT1 gene downstream of the MYC gene is required. In colorectal cancer samples, PVT1 gene expression increased 10-fold, associated with an enrichment of CD8 + T cell subsets in colorectal cancer lesions. In addition, the expression of PVT1 transcripts with the open reading frame in target T cells correlated with IFN-γ production [74]. MiR-150 hampered the activation of CD8 + T cells [75], and overexpression of miR-150 reduced inducible NKT cells [76]. However, the upregulation of PVT1 in hepatocellular carcinoma silenced miR-150, leading to the overexpression of the hypoxia-inducible protein 2 [77], which induced a cytotoxic T-cell response in patients with metastatic renal cell carcinoma [78]. MiR-155 is needed for CD8 + T cell responses to cancer cells. In the absence of miR-155, the number of effector CD8 + T cells was reduced, and miR-155-deficient CD8 + T cells were ineffective at controlling tumor growth by the upregulation of suppressor of cytokine signaling-1, causing defective cytokine signaling through signal transducer and activator of transcription (STAT)-5 in melanoma [79]. Furthermore, NEAT1 or LDHA knockdown promoted the secretion of CD8 + T-lymphocyte factors, including TNF-α and IFN-γ, enhancing the anti-tumor effects [80]. In contrast, UCA1 attenuated the killing effect of cytotoxic CD8 + T cells by targeting miR-148a [81] (Figure 3).
Hypoxia-induced GAS5 promotes the anti-tumor effect of NK cells by sponging miR-18a in gastric carcinoma, increasing the secretion of IFN-γ, TNF-α [82], and targeting miR-544, upregulating the RUNX family transcription factor 3 in liver carcinoma [83]. In contrast, increased levels of miR-20a in ovarian cancer tumor cells may indirectly suppress NK cell cytotoxicity by downregulating MICA/B expression [84].
Furthermore, miR-155 is induced upon T-cell activation and promotes Th1 differentiation when over-expressed in activated CD4 + T cells. Antagonism of miR-155 leads to the induction of IFN-gamma receptor alpha-chain [85]. MiR-17 and miR-19b control Th1 responses, for example, by inducing the release of IFN-γ and suppressing T reg cell differentiation by hampering FOXP3-mediated activation [86].
MiR-21 regulates the glycolysis of CD4 + T cells through the PTEN/PI3K/AKT pathway to accelerate the cell cycle, thereby facilitating CD4 + T cell polarization toward Th2 cells, releasing IL-13 [87]. Multiple myeloma is tightly dependent on the inflammatory bone marrow microenvironment, and IL-17-producing Th17 cells sustain multiple myeloma cell growth. However, early inhibition of miR-21 in naive T cells impaired Th17 differentiation [88]. MiR-130b overexpressed in diffuse large B-cell lymphoma was associated with Th17 cell activation with IL-17 release [89]. MiR-155, increased in cervical cancer tissues, inhibited the expression of target gene SOCS1, promoting the differentiation of Th17 cells and increasing IL-17 [90].
Hypoxia and MYC downregulate miR-34a and increase CCL22, recruiting T reg cells in hepatocellular carcinoma [91]. Upon T cell receptor stimulation, CD4 + T helper lymphocytes release miR-containing extracellular vesicles. A significant increase in miR-21 and a significant reduction in miR-155 resulted in increased T reg differentiation in pediatric acute lymphoblastic leukemia [92]. MiR-124 induced the STAT3 pathway, reversed the glioma cancer stem cell-mediated immunosuppression of T-cell proliferation, and induced FOXP3-positive Treg cells. Treatment of T cells from immunosuppressed glioblastoma patients with miR-124 induced marked effector response, including upregulation of IFN-γ and TNF-α [93]. The HOXA cluster antisense RNA 2 (HOXA-AS2) promotes the differen- tiation of T reg cells by sponging miR-302a and upregulating lysine demethylase 2A and jagged 1 in glioma cells [94] (Figure 3).
In contrast, LPS and IFN-γ switch M2 to M1 TAMs [20,112]. Finally, NO produced by CAFs appears to have a critical role in the reprogramming process of M2 to M1 phenotype and the tumor environment [113] and causes oxidative/nitrosative stress [114] (Figure 4).

Noncoding RNAs and MDSCs, and Macrophages
MiR-30a in B-cell lymphoma targeted suppressor of cytokine signaling 3 (SOCS3), activating JAK-STAT3 signaling and promoting MDSC differentiation and immunosuppression [115]. HOTAIR, highly expressed and associated with poor prognosis in hepatocellular carcinoma, promoted the secretion of CCL2 in hepatocellular carcinoma cells and increased the proportion of macrophages and MDSCs [116] (Figure 4).

Role of Noncoding RNA-Containing Microvesicles
Microvesicles facilitate the communication between different tissues and cell types within these tissues [25]. For example, miR-9 secreted in exosomes from triple-negative breast cancer cells induces the differentiation of fibroblasts to cancer-associated fibroblasts, increasing cell motility [130]. Melanoma cell-secreted exosomes enriched in miR-155 induced the reprogramming of fibroblasts into CAFs, triggering a proangiogenic switch by promoting the expression of proangiogenic factors, including VEGFa, fibroblast growth factor 2, and MMP9, by directly targeting SOCS1 and activating JAK2/STAT3 [131]. Interestingly, 8-OHD provoked ROS overproduction by downregulating JAK/STAT, PI3K/AKT, and oxidative phosphorylation [132]. In addition, human melanoma-derived exosomes reprogramed human adult dermal fibroblasts, increasing aerobic glycolysis and decreasing oxidative phosphorylation via miR-155 and miR-210 [133] (Figure 5). In addition, cancer cell exosome-derived miR-9 and miR-181a promote the expansion of MDSCs from granulocytes in breast cancer by activating the JAK/STAT signaling pathway via targeting SOCS3 and the protein inhibitor of activated STAT3, PIAS3 [134]. Tumor cells may inhibit M1 polarization and promote cancer growth by selective shuttling of miR-21 enriched microvesicles [135,136]. The cancer cell exosomal ELFN1 antisense RNA 1 (ELFN1-AS1) was high in patients with advanced osteosarcoma. In addition, overexpression of ELFN1-AS1 significantly promoted the proliferation, migration, and invasion of osteosarcoma cells, while knockdown of ELFN1-AS1 exhibited the opposite effects. Meanwhile, exosomal ELFN1-AS1 could be transferred from osteosarcoma cells to macrophages, promoting macrophage M2 polarization by sponging miR-138-5p and miR-1291, which may facilitate osteosarcoma progression [137]. In contrast, exosomes isolated from 4T1 breast cancer cells delivering a miR-33 mimic into IL-4 induced M2 macrophages converted them to the M1 phenotype as indicated by an increase in expression of M1 markers, including IRF5, NOS2, and CD86, and a decrease in M2 markers including ARG, Ym1, and CD206. Furthermore, the secretion of TNF-α and IL-1β decreased, while the secretion of IL-10 and TGF-β increased [138].
MiR-150 and miR-7 were required for mesenchymal stem cells to decrease NK cell activities and recruit T reg cells [139]. In addition, miR-150 in lung cancer-secreted EVs reduced the expression of CD226 on NK cells, creating a favorable immunosuppressive microenvironment by releasing IL-10 [140]. Hypoxia-primed CAFs secrete exosomes enriched in miR-21-5p, promoting macrophage M2 polarization and reducing apoptosis. Injection in immunocompromized mice significantly increased tumor growth, cancer cell proliferation, intra-tumoral angiogenesis, and M2 polarization of macrophages [135].
IL-4-activated macrophages co-cultivated with breast cancer cells without direct secreted miR-223-enriched microvesicles promoting the invasion of breast cancer cells via upregulating Mef2c [141]. Its expression depends on the overexpression of the antioxidant enzyme SOD1 [142]. In addition, exosomes derived from hypoxic macrophages enhanced the malignant phenotype of epithelial ovarian cancer cells enriched in miR-223 that induced PI3K/AKT signaling, which is crucial for controlling oxidative stress [143].
In addition, CAFs and cancer cells may secrete exosomes promoting the polarization of macrophages to an M1 (anti-tumoral profile), expressing more NOS2 or M2 (pro-tumoral profile) phenotype, expressing more arginase [146].
However, not all noncoding RNAs possibly interfering with oxidative stress in cancer have been mentioned. Ultimately, we made a selection of well-documented noncoding RNAs. This selection explains why circular RNAs have not been discussed. Furthermore, the quality of this review paper depends on that of the original studies, and shortcomings in the original studies will affect the content of this review. First, functional and clinical studies are mostly limited to a single noncoding RNA, although noncoding RNAs function in networks, as illustrated in this review. Furthermore, the outcome of silencing a single lncRNA is attributed chiefly to one miR's effect, although no broader omics search is performed to unravel all potentially involved miRs. In addition, the nonspecific effects of silencing experiments are mostly disregarded, and often reconstitution experiments are not performed. These shortcomings may explain contrasting results. In addition, meta-analyses are limited in number and confined to only a few noncoding RNAs and cancer types. However, these are required to obtain reliable predicting networks, discriminating between cancer types in which one noncoding RNA may have opposite effects. Unfortunately, information about the sequence of changes in expression profiles of noncoding RNAs at different stages of disease progression is lacking. Finally, we lack algorithms to determine if noncoding RNAs have any clinical value in addition to phenotypic, therapeutic, behavioral, and social data in a predicting model. Artificial intelligence or machine-learning methods may be applied to fit vast amounts of expression data combined with phenotypic, therapeutic, behavioral, and social data.

Conclusions
Noncoding RNAs regulate the oxidative stress in the tumor microenvironment by regulating mitochondrial function, antioxidant expression, and reprogramming of immune cells. Herein, noncoding RNAs in microvesicles contribute to the communication between different cell types in the tumor microenvironment and inflamed tumor-adjacent tissues. In particular, this communication warrants further investigation.