2.1 Production and elimination of ROS

ROS, as byproducts of oxygen consumption and cellular metabolism, are formed by the partial reduction of molecular oxygen and exist in both cancer cells and normal cells.⁵⁴ It mainly comprising superoxide anions (O₂⁻), hydroxyl radicals (•OH), and hydrogen peroxide (H₂O₂).⁵⁵ Cancer cells exhibit an altered redox state compared to normal cells,⁵⁶ which depends on ROS production efficiency and scavenging capacity.⁵⁷ ROS accumulation is attributed to the exogenous environment and endogenous metabolism.⁵⁸ Exogenous ROS can be generated secondary to exposure to many risk factors, including radiation, cigarettes, metals, asbestos, and air pollutants.^59-61 Endogenous ROS mainly originate from the mitochondrial respiratory chain and NADPH oxidases (NOXs).⁶² Discrete subcellular organelles, including the endoplasmic reticulum, peroxisomes,⁶³ and nuclei, also generate ROS.^{62, 64, 65} When excessive accumulation of ROS occurs, the antioxidant system functions as a ROS eliminator to maintain redox homeostasis.⁶⁶ This system includes endogenous antioxidant enzymes,⁶⁷ including superoxide dismutase,⁶⁸ catalase,⁶⁹ glutathione peroxidase,⁷⁰ glutathione reductase,⁷¹ and peroxiredoxins,⁷² and antioxidant molecules such as glutathione, coenzyme Q, ferritin, and bilirubin.^73-75 However, the balance between the oxidant and antioxidant system is generally broken in cancer cells.

2.2 ROS participate in tumor metastasis

In normal cells, the well-balanced function of the antioxidant system enables intracellular ROS concentration in a dynamic equilibrium.⁷⁶ However, this balance is overturned in cancer cells,^{2, 77, 78} promoting tumor progression or cell death, due to the imbalance between oxidant and antioxidant systems. High ROS can cause mitochondrial membrane potential depolarization and damage the respiratory chain, leading to more O₂⁻ production by mitochondria,⁷⁹ which initiates a vicious cycle of ROS production in cancer cells. ROS can regulate critical metabolic pathways by activating or inhibiting ROS-sensitive transcription factors and modifying central metabolic enzymes. Nuclear factor erythroid 2-related factor (NRF2) and BTB and CNC homology 1 (BACH1) are key transcription factors that regulate intracellular redox activity through interconnected but distinct mechanisms.⁸⁰ As a central regulator of the cellular antioxidant response, NRF2 is negatively regulated by Kelch-like ECH-associated protein 1 (KEAP1).⁸¹ BACH1 participates in heme homeostasis by heterodimerizing with MAF proteins and binding to NF-E2 gene regulatory sites, suppressing NRF2-dependent oxidative stress response.⁸² NRF2 and BACH1 are often mutated in cancer cells,⁸³ resulting in transcriptional repression of other genes, such as nuclear factor kappa-B (NF-кB) and FOXO.^{84, 85} For tumor cells undergoing metastasis, remodeling of their oxidant and antioxidant systems usually exerts their biological effect accumulatively over time.^{18, 86} Finding the underlying cellular or molecular mechanisms is essential for understanding the role of ROS in metastasis.

3.1 Epigenetic remodeling under oxidative stress

Epigenetics refers to multiple covalent modifications made to nucleic acids and histones, or the production of specific ncRNAs, which could control gene expression without altering the DNA sequence.^{182, 183} This field primarily involves mechanisms such as DNA methylation, histone modifications, ncRNAs, and m⁶A modifications. DNA methylation and histone modifications can modulate gene transcription by altering the compaction state of chromatin, while ncRNAs and m⁶A modifications mainly interact with mRNA, influencing its stability and translational efficiency. The regulation and functional activity of epigenetic modifications are based on multiple enzymes and proteins known as epigenetic “writers,” “erasers,” and “readers,” which are responsible for the initial attachment, subsequent movement, and real-time recognition of various modifications, respectively.^{184, 185} It is worth noting that oxidative stress can also serve as an essential modulator of epigenetic modifications, directly altering the level of epigenetic modification, or indirectly impacting on certain writers, erasers or readers.^186-188

3.2 ROS-regulated DNA methylation reprogramming modulates metastasis

ROS-induced DNA methylation alterations show multifaceted metastasis effects (Figure 2). Generally, cancer cells enhance their antioxidant ability during metastasis to defend against excessive ROS,^254-256 whereby DNA methylation plays an irreplaceable role. For example, a study focusing on the state of DNA methylation in non‑small cell lung cancer (NSCLC) revealed that 79 of 121 NSCLC patients exhibited demethylation in the peroxiredoxin‑5 promoter region. Notably, hypomethylation of peroxiredoxin‑5 activated the Nrf2 signaling pathway and thus significantly decreased E‑cadherin and increased vimentin.²⁵⁷ Several earlier studies found that decreased catalase expression is also associated with tumor metastasis. For instance, ROS-induced hypermethylation in the catalase promotor is responsible for downregulating catalase during cancer development,²⁵⁸ while DNMT inhibitor treatment successfully upregulates catalase.²⁵⁹ Interestingly, ROS can facilitate their accumulation during metastasis by regulating DNA methylation, rather than upregulating the antioxidant system to circumvent oxidative damage. A study in 2021 revealed that ROS upregulate the expression of FOXC1 by activating ROS-ERK1/2-p-ELK1 signaling, leading to DNA hypermethylation of the cystathionine γ-lyase promoter by increasing DNMT3B expression, finally leading to a decrease in cysteine levels and an increase in ROS levels. This positive feedback loop induces further ROS accumulation, ultimately promoting metastasis.²⁶⁰

Additionally, the silencing of tumor suppressors is partly caused by DNA methylation under oxidative stress. ROS-induced DNA damage products, including 8-OHdG and DNA double-strand breaks, indirectly induce DNA hypermethylation-related transcriptional silencing of tumor suppressor genes by recruiting chromodomain helicase DNA-binding protein 4 and thus recruiting DNMTs, promoting the proliferation, invasion, and metastases of colorectal cancer cells.²⁶¹ Moreover, it is imperative to recognize that ROS are also capable of conducting epigenetic remodeling on metastasis-related factors, including E-cadherin, N-cadherin, Vimentin, Snail, Slug, ZEB1, and ZEB2, to control metastasis.^262-267 A comparative study in hepatocellular carcinoma demonstrated that ROS trigger hypermethylation of the E-cadherin promoter by snail-regulated DNMT1, leading to silencing of E-cadherin and thus promoting metastasis.²⁶⁸ A sublethal dosage of hydrogen peroxide treatment also decreased E-cadherin in breast cancer by enriching DNMT1, H3K9me3, and H3K27me3 in the E-cadherin promoter.²⁶⁹ Consistently, exposure to oxygenated polycyclic aromatic hydrocarbons (Oxy-PAHs) can prompt the downregulation of E-cadherin by DNA methylation, consequently facilitating lung cancer metastasis.²⁷⁰ Interestingly, it has been reported that ROS is indispensable to (NH₄)₂SO₄-modulated lung cancer metastasis.²⁷¹ Prolonged exposure to (NH₄)₂SO₄ induced ROS accumulation and activated HIF-1α, resulting in hypermethylation of E-cadherin promoter regions and facilitating EMT.

Living in a high-intensity oxidative stress environment, CTCs are more susceptible to ROS.^{272, 273} The results of single-cell sequencing of isolated CTCs in 2017 showed that DNA methylation patterns in CTCs resembled those of epithelial-like cells. Notably, methylation at the promoter of the microRNA-200 family was remarkably higher in prostate CTCs, indicating that the distribution of methylation manifests spatial specificity, which is probably related to the organotropism of CTCs colonization.²⁷⁴ Another study concerning the DNA methylation pattern of CTCs in peripheral blood from patients with breast cancer revealed that, compared with healthy individuals, the promoter methylation of tumor suppressor and metastasis suppressor genes, including CST6, BRMS1, and SOX17, was at a higher level.²⁷⁵ Notably, surviving CTCs are characterized by immune escape, which can be partly attributed to DNA methylation remodeling. Recently, Guo et al. found by high-resolution genome-wide single-cell DNA methylation sequencing that 40 core domains are uniformly hypomethylated from early prostate malignancy to CTCs.²⁷⁶ Interestingly, hypomethylation of these regions silenced gene transcription, which contradicts the conventional view that hypermethylation of DNA suppresses transcription. One explanation is that the sites of DNA methylation determine transcriptional outcome, which means that hypomethylation in CpG islands of promotors promotes gene expression. In contrast, hypomethylation in long-range hypomethylation regions inhibits transcription. Subsequent analysis demonstrated that these transcriptionally silenced genes are mainly immune-related, including five CD1 genes that present lipid antigens to NKT cells.²⁷⁶ Therefore, hypomethylation-induced transcription inhibition contributes to the immune escape of CTCs, where oxidative stress might serve as an initiating factor. Collectively, DNA methylation in the promoter regions of key tumor genes responds to the regulation of oxidative stress, which is an important biological event during tumor metastasis.

3.3 Histone modification remodeling caused by ROS affects metastasis

Histone modifications, akin to DNA methylation, are both subjected to modulation by ROS and serve as epigenetic switches for gene transcription,^{277, 278} which also regulate metastasis (Figure 3). ROS induced by Oxy-PAHs simultaneously promoted DNA methylation of E-cadherin and histone acetylation of snail by separately activating DNMT3a and suppressing HDAC1. Consequently, E-cadherin inhibition and snail activation facilitate lung cancer metastasis.²⁷⁰ Consistently, H₂O₂ treatment also augmented DNA methylation and histone deacetylation by activating the ERK pathway and thus stimulating DNMT and HDAC1, leading to transcriptional silencing of E-cadherin.²⁶⁹ However, ROS can also play a regulatory role in some cases, dependent on histone acetylation but not DNA methylation. For instance, exposure to hexavalent chromium, a common human carcinogen that can generate ROS, strengthened the binding of HDAC1 to the E-cadherin gene promoter but did not change DNA methylation levels.²⁷⁹

Notably, histone methylation remodeling is essential for tumor adaptation to oxidative stress. Hypoxia in pancreatic cancer facilitated ROS generation in a HIF-1-dependent manner and thus limited the activation of lysine-specific demethylase 5A (KDM5A). Subsequently, increasing methylation of histone H3 facilitated the transcription of snail and tumor metastasis.²⁸⁰ Another study revealed that H3K9 methylation by radiation-induced ROS downregulated E-cadherin expression and upregulated N-cadherin expression, thus promoting metastasis.^{281, 282} Conversely, it has been reported that hesperetin administration resulted in the accumulation of H3K4me3 and H3K27me3 but a decrease in H3K9me3, finally increasing E-cadherin expression and decreasing N-cadherin expression. In this regard, the methylation sites determine whether histone methylations function as transcription promotors or inhibitors. Specifically, H3K4me3 and H3K36me3 are associated with active transcription, whereas H3K27me3, H3K9me2/3, and H4K20me3 are correlated with repressed gene expression.^{283, 284} Moreover, oxidative stress-induced ECM remodeling is tightly associated with tumor metastasis.^285-287 It has been reported that oxidative stress stimulates HDACs and thus upregulates the expression and activity of MMPs,^288-290 facilitating the degradation of the endothelial glycocalyx.¹¹⁰ Notably, ROS-induced fragmentation of the glycocalyx favors entry into the circulating system for tumor cells.²⁹¹ Moreover, loss of endothelial glycocalyx is indispensable for CTCs colonization. Circulating lung cancer cells facilitate their adhesion to the brain endothelium by secreting factors to destroy the endothelial glycocalyx.²⁹² Taken together, ROS-mediated histone modification remodeling regulates multiple biological events associated with tumor metastasis, including EMT, ECM remodeling, and distal colonization.

3.4 Metastasis involves different ncRNA regulation patterns under oxidative stress

MiRNAs have been observed to participate in cancer metastasis under oxidative stress (Figure 4). Generally, various miRNAs, including miR-99a,²⁹³ miR-373,²⁹⁴ miR-372,²⁹⁵ miR-142,²⁹⁶ and miR-212,²⁹⁷ are involved in metastasis by directly or indirectly regulating ROS generation. More importantly, ROS can also regulate the transcription and biogenesis of miRNAs and thus influence metastasis. It has been reported that the antioxidant molecule Nrf2 modulates miR-206 to promote tumorigenesis.²²⁶ Meanwhile, miR-206 also participated in metastasis, which suppressed TGF-β signaling and thus controlled EMT, migration, and invasion.^{298, 299} Similarly, downregulated miR-4521 suppresses gastric carcinoma metastasis under hypoxia conditions by regulating IGF2 and FOXM1.³⁰⁰ Nonetheless, another study showed that ROS play a carcinogenic role by upregulating miR-21 and thus downregulating programmed cell death 4 protein (PDCD4) in patients with gastric cancer.³⁰¹ Noticeably, some ROS-sensitive transcription factors, including P53, NF-κB, and HIF-1, are also involved in tumor metastasis by regulating miRNAs. ROS resulting from total parenteral nutrition indirectly activate p53 by phosphorylating p38 and thus drive the transcription of miR200, which mediates EMT.^{223, 302} Likewise, the function of miRNAs in metastasis partly depends on NF-кB. For example, TGF-β-activated NF-кB is involved in the binding of Smad and miR21.³⁰³ ROS-inhibited NF-κB dampened miR-21, thus suppressing apoptosis and indirectly driving metastasis.³⁰⁴ Consistently, miR-19a accumulation secondary to ROS also dampened apoptosis by inactivating NF-κB.³⁰⁵ In addition, it has been reported that, under hypoxia conditions, HIF-1α indirectly influences tumor metastasis by regulating miR210, miR382, miR-421, and miR191.^306-309 Overall, miRNAs can respond to oxidative stress to directly or indirectly mediate the process of tumor metastasis.

In addition to miRNAs, other ncRNAs exhibit ROS response properties during tumor metastasis (Figure 5). PiRNA is a kind of ROS sensor that plays a dual role in tumor progression.^{310, 311} For example, piR-823 showed tumor oncogenic activity in esophageal squamous cell carcinoma by upregulating DNMT3B.³¹² However, another study indicated that piR-2158 suppressed IL11 expression and IL-11 secretion by competing with the AP-1 transcription factor subunit FOSL1, thus blocking JAK-STAT3 signaling to inhibit tumorigenesis, invasion, and EMT in breast cancer.³¹³ It should be noted that some snoRNAs, including SNORD32A, SNORD33, and SNORD35A, are found to be under the control of oxidative stress.^{229, 314} Studies have reported that these snoRNAs are closely correlated with tumor metastasis.^315-318 Additionally, transient receptor potential melastatin 8 (TRPM8) supported hepatocellular carcinoma metastasis by upregulating SNORA55 to induce nuclear and mitochondrial dysfunction.³¹⁹ It is essential that the activation of TRPM8 is under the control of oxidative stress in tumors.^{319, 320} Therefore, it is conceivable that ROS may facilitate tumor metastasis by modulating snoRNAs, although further investigations are warranted to support this postulation. In addition to the short ncRNAs, growing studies have suggested that circRNAs participate in ROS-regulated metastasis. In solid tumors, circRNAs mainly serve as oncogenic components in metastasis, the function of which is often dependent on miRNAs.^321-324 Nevertheless, circRNAs can inhibit tumor metastasis in some specific metastatic tumors. For example, circLMO1 suppresses cervical cancer metastasis by competing with miR-4291 to drive ferroptosis.³²⁵ Consistently, circPTK2 inhibits TGF-β-induced EMT and metastasis by regulating transcriptional intermediary factor 1 γ (TIF1γ) in NSCLC.³²⁵ Similar to miRNAs, circRNAs are generally regarded as ROS regulators in ROS-related tumors.³²⁶ Interestingly, oxidative stress can also change cell fate by altering circRNA levels. It has been reported that oxidative stress-induced circHBEGF supports ECM production by sponging miR-646 and thus activating the downstream EFGR/AKT/STAT3/ERK pathway.³²⁷ In addition, circFoxo3 promoted FOXO3 expression and pulled down high levels of mouse double minute 2 homolog protein, facilitating apoptosis under oxidative stress.³²⁸ However, the roles of ROS-induced circRNAs in tumor metastasis remain largely elusive. Therefore, further investigations are needed to explore the function of ROS-responsive piRNAs, snoRNAs, or circRNAs involved in tumor metastasis.

LncRNAs are also engaged in tumor metastasis, serving as competing endogenous RNAs (ceRNAs) of miRNAs.^{329, 330} Studies have reported that lncRNAs modulate tumor metastasis as effector molecules of oxidative stress. For example, ROS resulting from PM2.5 exposure upregulated lncRNA loc146880 expression, causing malignant metastasis of lung cancer cells.³³⁰ LncRNA-XIST drives osteosarcoma metastasis under oxidative stress. Mechanistically, oxidative stress induces osteosarcoma cell invasion, migration, and EMT by regulating the XIST-miR-153-SNAI1 axis, where XIST is caused by oxidative stress and miR-153 is inhibited by XIST.³³¹ In addition, an analysis based on RNA-seq datasets identified aberrantly expressed lncRNAs.³³² These results demonstrated that two lncRNAs, H19 and HULC, were increasingly expressed under the treatment of hypoxic or inflammatory factors. Further study identified that these lncRNAs, upregulated by oxidative stress, drive cell migration and invasion by targeting IL-6 and CXCR4 in ceRNA patterns of sponging let-7a/let-7b and miR-372/miR-373.³³² Consistently, another study revealed that the HIF-2 activation response to hypoxia can induce lncRNA RAB11B-AS1 upregulation, which mediates angiogenesis and breast cancer metastasis.³³² It can be seen that the mechanism by which ROS affect tumor metastasis through regulating ncRNA function is complex, and more in-depth exploration is needed to explore the cross-talk between different ncRNAs in the context of oxidative stress.

3.5 Metastasis relies on precise m⁶A reprogramming under oxidative stress

The remodeling of m⁶A methylation is an adaptive mechanism, which makes it possible for tumor cells to utilize oxidative stress signals while circumventing oxidative damage. For example, the downregulation of YTHDF1 in NSLCC under chemotherapy stress reduces the m⁶A level of KEAP1, thus upregulating Nrf2 to attenuate oxidative stress.³³³ Another study revealed lung adenocarcinoma cells upregulated METTL7B and enhanced antioxidant capacity, resulting in resistance to EGFR-tyrosine kinase inhibitors.³³⁴

The reprogramming of m⁶A modifications under various oxidative stress might also serve as an “engine” driving tumor metastasis (Figure 6). Exposed to the stress of alcohol, C/EBP β boosts the expression of KIAA1429 (an RNA methyltransferase), thus accelerating pancreatic cancer metastasis.³³⁵ Another study indicated that WZ35-induced ROS imbalance attenuated the function of ALKBH5 by regulating the YAP1-AXL axis, subsequently upregulating the m⁶A modification levels of glutaminase 2 (GSL2) and then leading to GSH depletion.³³⁶ ALKBH5 suppressed the expression of γ-glutamylcyclo transferase 1 (CHAC1) to modulate intracellular GSH and ROS, which conferred adaption to cisplatin-induced oxidative stress onto gastric cancer cells.³³⁷ Within the hypoxic tumor microenvironment, serine/threonine kinase receptor-associated protein mediated the ubiquitination degradation of FTO, subsequently liberating metastasis-associated protein 1 (MTA1) and promoting metastasis.³³⁸ Notably, hypoxia can strengthen the migration ability of tumor cells by reducing the m⁶A level of target gene mRNA. Under normal oxygen levels, METTL14 promotes the degradation of SLC7A11 mRNA in a YTHDF2-dependent pathway, finally facilitating the ferroptosis of hepatocellular carcinoma. However, the function of METTL14 can be suppressed under hypoxia conditions, facilitating the expression of SLC7A11 and then supporting the survival and migration of hepatocellular carcinoma.³³⁹ Consistently, another study also found that hypoxia remarkedly remodeled the patterns of m⁶A modification in integrin subunit beta 1 (ITGB1) mRNA and attenuated its degradation mediated by the YTHDF2 protein, subsequently upregulating ITGB1 and resulting in the generation of proto-oncogene proteins, thus driving lymph node metastasis.³⁹ Moreover, hypoxia can activate the expression of ALKBH5 in a manner dependent on HIF-1α and HIF-2, which subsequently facilitates SOX2 expression through demethylation of SOX2 mRNA.³⁴⁰ Upregulated SOX2 maintains the stem-like state of tumor stem cells, which is conducive to tumor metastasis.³⁴¹ In addition to affecting mRNA's m⁶A methylation, hypoxia can greatly influence specific ncRNAs’ m⁶A methylation and then orchestrate tumor metastasis. Under a hypoxic microenvironment, HIF-1α-induced METTL16 directly binds to Lnc-CSMD1-7 and then decreases the RNA stability of Lnc-CSMD1-7 through m⁶A methylation, ultimately facilitating hepatocellular carcinoma metastasis.³⁴² However, it has also been reported that hypoxia-modulated ncRNAs are capable of regulating the patterns of m⁶A methylation, contributing to tumor metastasis. For instance, HIF-1α upregulation transcriptionally promotes lncRNA NT-AS1 expression. Then it activates METTL3 to increase ITGB1 expression via m⁶A modification, thereby activating the MAPK/ERK/PD-L1 signaling pathway and ultimately promoting the immune escape of pancreatic cancer cells.³⁴³ Likewise, m⁶A modification-mediated macrophage reprogramming orchestrates an antitumor immune response.³⁴⁴ Specifically, METTL3 ablation suppresses SPRED2 mediated by YTHDF1, thus modulating cytokine responses, promoting Treg infiltration, and enhancing tumor-promoting macrophages. In short, oxidative stress impacts tumor metastasis, where m⁶A modification remodeling serves as an important bridge.

The role of ROS as either a promoter or inhibitor during metastatic progression remains inconclusive, contingent upon the ROS levels, specific downstream targets, and cancer types involved. According to the available evidence, the influence of ROS-regulated epigenetic modifications on tumor metastasis is highly variable, with even identical modifications exerting divergent effects on different tumor types. Tumor metastasis is a complex, multi-stage cascade, and the oxidative stress levels within tumor cells fluctuate dynamically. Consequently, it is crucial to analyze epigenetic profiles at different stages of metastasis, particularly focusing on the regulatory patterns associated with tumor driver genes. However, obtaining epigenetic information from CTCs presents significant challenges due to their low abundance in the bloodstream and existing technical limitations. Additionally, the high heterogeneity of tumors leads to variability in the epigenetic regulation of the same gene across different tumors, and these effects can even be contradictory. Furthermore, within tumor tissue from the same patient, oxidative stress levels may differ significantly between sites, resulting in spatial variations in the epigenetic status of the same gene. Addressing these gaps requires in-depth investigations to fully elucidate the role of oxidative stress-induced epigenetic modifications in driving tumor metastasis.

4.1 Targeting DNMTs against metastatic tumors

Targeting DNA methylation is a promising cancer therapeutic strategy. Generally, inhibition of DNMTs results in either the upregulation or downregulation of essential genes involved in cancer metastasis.³⁴⁵ 5-Azacitidine (5-aza) and 5-aza-2′-deoxycytidine (DAC) are two major DNMT inhibitors (DNMTis) approved by the US Food and Drug Administration and European Medicines Agency for the treatment of hematological malignancies.^{346, 347} Several phase I/II clinical trials have investigated the potential of 5-aza against metastatic cancer.^347-350 It has been reported that 5-aza administration drives EMT and facilitates the CSC phenotype, finally promoting bone metastasis of prostate cancer.³⁵¹ DAC can also combat tumor metastasis. The results of a phase I-II study in the 1990s indicated the possibility of using DAC against metastatic carcinoma, including lung cancer and prostate cancer.^{352, 353} In recent years, researchers have tried to elucidate the underlying DAC-related mechanism and exploring new strategies to widen their clinical usability. For example, it has been reported that DAC suppressed colorectal peritoneal metastasis by regulating macrophage-dependent T-cell activation in the premetastatic microenvironment.³⁵⁴ Moreover, the combination of DAC with other drugs, namely, everolimus, cisplatin, and 5-fluorouracil, has been extensively explored to combat tumor metastasis.^355-357 Some other DNMTis, including zebularine,³⁵⁸ 5-fluoro-2′-deoxycytidine,^{359, 360} SGI-110 (guadecitabine),^{361, 362} CP-4200,³⁶³ 4′-thio-20-deoxycytidine,^{364, 365} and RX-3117,³⁶⁶ have been developed to improve therapeutic effects with fewer side effects. Due to the function of DNA methylation in immunogenic modulation, some DNMTis have exhibited potential in the tumor immunotherapy. In mouse models and clinical studies, the use of methyltransferase inhibitors can induce the reprogramming of DNA methylation in exhausted T cells, preventing their generation.³⁶⁷ For tumor cells, DNMT inhibitors can enhance the natural immune response of tumor cells from multiple dimensions.³⁶⁸

It has been reported that 5-aza attenuates H₂O₂-induced senescent phenotype in tumor cells by inducing autophagy, suggesting that oxidative stress may be beneficial for the antitumor effect of 5-aza.³⁶⁹ Consistently, 5-aza significantly attenuated the tumorigenic potential of HK-2 cells under chronic oxidative stress.³⁷⁰ Therefore, strategies combining DNMTis and ROS-generators might achieve better therapeutic effect. Ruthenium (Ru) complex-mediated apoptosis coupled with decitabine-regulated DFNA5 demethylation induced immunogenic cell death to suppress gastric tumor growth and metastasis.³⁷¹ Several studies also indicated that the combination of DAC and photodynamic therapy (a treatment based on ROS generation) remarkedly potentiated antitumor immune responses.^372-374 In addition, DAC combined with paracetamol effectively limited the progression of head and neck squamous cell carcinoma by induction of oxidative stress. DAC could also downregulate antioxidant responses to support ROS accumulation.³⁷⁵ Some natural products with excellent antioxidative or prooxidative properties can restrict tumor malignant progression by dampening DNMTs.^{376, 377} Resveratrol elevated ROS accumulation by inducing mitochondrial destruction and thus upregulated DLC1 (tumor suppressor gene) expression by inhibiting DNMT1, finally promoting cellular senescence in breast cancer.³⁷⁸ Another in vivo study indicated that resveratrol management effectively inhibited lung metastasis of MDA231 human breast cancer cells by attenuating TGF-β-induced EMT.³⁷⁹ Similarly, epigallocatechin-3-gallate (EGCG) downregulated ubiquitin-like containing PHD and Ring finger 1 (UHRF1) and DNMT1 in a ROS-dependent manner, which silenced tumor suppressors such as p16.³⁸⁰ In addition, high concentrations of curcumin triggered DNA damage and induced DNA demethylation by suppressing DNMT1 binding, partly responsible for the attenuated migration ability of human gastric cancer cells.³⁸¹

Collectively, DNMT inhibitors have shown great potential against tumor metastasis. Given that oxidative stress drastically alters DNA methylation remodeling in tumors, DNMT inhibitors, to some extent, deprive tumor cells of the adaption to oxidative stress, thus disrupting redox homeostasis and limiting metastasis. Some antioxidants or oxidants target DNMTs by modulating ROS levels, remodeling tumor cell DNA methylation patterns, and inhibiting metastasis.

4.2 HDAC intervention relieves tumor metastasis

HDACis can enhance the acetylation of lysine residues on histone and nonhistone proteins. For many years, HADCis have been widely investigated in metastatic tumors. Studies have reported that monotherapy with vorinostat (suberoylanilide hydroxamic acid, SAHA) yielded good effectiveness against metastatic tumors, including invasive human NSCLC³⁸² and metastatic breast cancer.³⁸³ Furthermore, SAHA is effective against some metastatic tumors in combination with other drugs, namely, brigatinib,³⁸⁴ carboplatin, paclitaxel,³⁸⁵ and pembrolizumab.³⁸⁶ Despite this, the results of several phase I/II trials indicated that SAHA with or without other drugs yielded only limited therapeutic effects in solid tumors, such as metastatic breast cancer, metastatic thyroid carcinoma, and metastatic transitional cell carcinoma.^387-389 Similar to hydroxamic acid, entinostat in combination with chemotherapeutics or targeted drugs, including pembrolizumab, exemestane, enzalutamide, and capecitabine, has also been evaluated in various metastatic tumors.^390-392 It is worth mentioning that entinostat, combined with entinostat, generally gained good tolerance and increased antitumor activity against metastatic breast cancer.³⁹⁰ Romidepsin is another HDACi based on cyclic peptides. Combined therapy with romidepsin and chloroquine led to improved survival for breast cancer patients with refractory metastases.³⁹³ In addition, some short-chain fatty acid HDACis, including sodium butyrate and valproic acid, have also yielded effective therapeutic effects against metastatic cancers in combination with other treatments.^394-397 It can be seen that HDACis have shown great potential in clinical therapy against metastasis. However, several clinical trials based on romidepsin failed to achieve effective therapy in metastatic tumors, namely, metastatic head and neck cancer, metastatic prostate cancer, and colorectal cancer.^{398, 399}

Mechanistically, monotherapy with HDACis might deprive tumor cells of their tolerance to the harsh microenvironment to some extent,^{400, 401} but this may not be enough to completely prevent malignant metastasis. Like DNMTis, some HDACis have shown potential for immunotherapy.⁴⁰² It has been reported that HDACis and DNMTis can not only alter the gene expression patterns of tumor cells but also reshape the killing ability of T cells, transforming the tumor microenvironment into a hot tumor type that is more responsive to immunotherapy.^{403, 404} Based on this, a series of epigenetic therapies combined with immune checkpoint therapy combinations have been approved for clinical trials (NCT01928576, NCT03220477, NCT01928576, NCT03233724, NCT03220477, NCT03576963, NCT02901899, NCT02397720, and NCT04296942), reflecting the great potential of this combination therapy.

Strategies simultaneously targeting HDACs and oxidative stress can also produce synergistic antitumor effect.^{405, 406} For instance, valproic acid-modulated HDAC inhibition coupled with excessive intracellular ROS induced by doxorubicin effectively inhibited the progression of hepatocellular carcinoma.⁴⁰⁷ Codelivery of lactonic sophorolipids (HDAC inhibition) and ganetespib (Hsp90 inhibition) enhanced ROS accumulation and then effectively limited the metastasis of NSCLC.⁴⁰⁸ Besides, the combination of tubastatin A (HDAC6-specific inhibitor) and palladium nanoparticles (inducing oxidative stress) synergistically induced the apoptosis of breast cancer cells.⁴⁰⁹ Recently, several natural products with the ability to modulate oxidative stress were found to inhibit metastatic tumors by targeting HDAC. Curcumin, a natural HDACi, effectively retarded the metastatic process of various solid tumors. For example, curcumin markedly downregulated the expression of tumor markers, thus sensitizing breast cancer cells to chemotherapy.⁴¹⁰ Additionally, resveratrol acetylated and reactivated PTEN to limit tumor metastasis by targeting and suppressing MTA1/HDAC.⁴¹¹ Quercetin treatment also partially potentiated cell death in leukemia by suppressing HDAC.⁴¹² Consistently, curcumin, in combination with other HDACis, including vorinostat and panobinostat, greatly enhanced the therapeutic effect compared with HDAC inhibitor monotherapy.⁴¹³

4.3 Therapeutic strategies based on ncRNAs

NcRNA-based therapies have been extensively explored in recent years.⁴¹⁴ In general, there are two main methods for targeting ncRNA against tumor metastasis: targeting ncRNAs that function as facilitators in tumor metastasis with ncRNA inhibitors, such as antagomirs (targeting mRNA)⁴¹⁵ and miRNA sponges,⁴¹⁶ and restoring the normal function of ncRNAs that are downregulated in metastatic tumors via synthetic ncRNA-like molecules,⁴¹⁷ such as miRNA mimic agents. Growing numbers of antagomirs have been investigated for the treatment of tumor metastasis. For example, systemic treatment with a miR-10b antagomir effectively inhibited lung metastasis but did not reduce primary mammary tumor growth.⁴¹⁸ Similarly, antagomirs targeting miR-1246 and miR-20a also limited the progression of metastatic tumors.^{419, 420} Although artificial miRNA or circRNA sponges are capable of silencing miRNA, their clinical utility is still lacking. It is worth mentioning that some natural products can also inhibit metastasis by regulating ncRNAs. An early study indicated that curcumin significantly changed the expression profiles of miRNAs under oxidative stress.⁴²¹ Curcumin effectively suppressed the lung metastasis of colorectal cancer by activating a ROS/KEAP1/NRF2/miR-34a/b/c cascade.⁴²² Besides, curcumin administration greatly limited lncRNA H19-induced EMT in tamoxifen‑resistant breast cancer cells.⁴²³ Similarly, another natural antioxidant, resveratrol, showed great potential application in cancer management by targeting miRNAs or lncRNAs.⁴²⁴ On the other hand, the dysfunction of ncRNAs serving as tumor suppressors can be restored by synthetic ncRNA-like molecules such as miRNA mimic agents.^425-427 For instance, miR34, serving as a suppressor in the metastasis of various solid tumors, has been taken to clinical trials. The results showed that the liposomal miR34 mimic has great potential to be applied in anticancer therapy.⁴²⁸ Moreover, the administration of miR-708-5p to mice with lung cancer resulted in efficient antitumor activity. miR-708-5p suppressed lung cancer invasion and metastasis by inhibiting the cytoplasmic localization of p21.⁴²⁹ Given this, artificial analogs of other miRNAs such as miR-503, miR-152, miR-145, miR-133a, miR-193b, miR-148b, miR-433, miR-127, and miR-200 still need to be explored and applied in the treatment of metastatic tumors.^430-434

Importantly, with the development of nanodelivery technology, some ncRNAs with antitumor activity have been encapsulated into nanoparticles to achieve accurate delivery which is promising for use in overcoming tumor metastasis.^435-437 A hydrogel-embedded, gold nanoparticle-based delivery vehicle successfully carried miR-96/miR-182 and achieved efficient local, selective, and sustained release, amplifying the antimetastatic ability of exogenous miR-96/miR-182.⁴³⁸ In addition to miRNAs, siRNA, and shRNA have also carried by delivery vehicles to realize antitumor therapy. For example, lipid nanoparticles carrying β3 integrin siRNA alleviated primary tumor growth and significantly inhibited metastasis in triple negative breasr cancer.⁴³⁹ Codelivery of siRNA and miRNA by liposome-polycation-DNA nanoparticles also significantly limited the lung metastasis of melanoma.⁴⁴⁰ In another study focusing on delivering lncRNA, plasmids encoding tumor suppressor lncRNAs were encapsulated into liposomes to prepare nanoparticles targeting renal cell carcinoma, which markedly inhibited tumor angiogenesis and metastasis.⁴⁴¹ To further improve antitumor efficacy, many nano-drugs can combine ncRNA functional regulation with redox homeostasis regulation, such as the simultaneous delivery of miRNA and antioxidants (Table 1, Figure 6).

4.4 Therapeutic strategy targeting m⁶A methylation modulators

Under oxidative stress, dysfunction of certain m6A methylation regulators frequently results in abnormal reduction or increase of overall levels of m6A methylation, thus affecting tumor metastasis. Therefore, targeting these m⁶A modulators has great potential in therapeutic applications against tumor metastasis.^503-505 Recently, the potential of targeting m⁶A regulators has been explored in many studies, most of which have shown satisfactory effects against tumor metastasis. For instance, the administration of small-molecule inhibitors targeting m⁶A regulators can slow down or even reverse metastasis-associated biological events. The inhibitor of FTO, Dac51, can effectively block tumor immune evasion by activating CD8⁺ cells.⁵⁰⁶ Similarly, Alk-04, a selective inhibitor of ALKBH5, significantly enhanced melanoma therapy by decreasing suppressive immune cell accumulation.⁵⁰⁷ A new sodium channel blocker, imidazobenzoxazin-5-thione MV1035, has been found to attenuate the migration and invasive ability of glioblastoma cells by inhibiting ALKBH5.⁵⁰⁸ MiRNA mimics can also be applied as potential inhibitors targeting m6A modulators. It has been reported that miR-186 inhibited hepatoblastoma metastasis by inhibiting the METTL3-modulated Wnt/β-catenin pathway.⁵⁰⁹ Moreover, miR-135 was also reportedly to suppress the EMT of breast cancer cells by silencing ZNF217 and inactivating METTL3.⁵¹⁰

Some drugs with certain antioxidant abilities specifically target m⁶A modulators, thus limiting tumor metastasis. For example, maclofenamic acid remarkedly increased m⁶A methylation levels by suppressing FTO activity, subsequently inhibiting the survival of cancer stem cells.^{511, 512} Another drug with antioxidant capacity, simvastatin, downregulates m⁶A methylation of EZH mRNA by targeting METTL3, limiting EMT in lung cancer.⁵¹³ It was also reported that resveratrol can attenuate Aflatoxin B1-induced ROS accumulation, thus leading to the reprogramming of m6A and eventually impacting hepatic function.⁵¹⁴ Therefore, resveratrol might serve as a kind of potential antitumor drug, or at least as a drug to prevent tumorigenesis. Given that oxidative stress participates in the modulation of the m⁶A regulator during tumor metastasis, the combined intervention of these regulators and oxidative stress-associated signals might contribute to improving the antitumor therapy effect. For example, compared to targeting IGF2BP3 alone, combined suppression of IGF2BP3 and HIF-1α further prevented stomach cancer angiogenesis and metastasis.⁵¹⁵ Codelivery of YTHDF1 siRNA with photosensitizer remarkedly enhanced photodynamic therapy efficiency by modulating the m6A modification level of HSP70 and lysosomal proteases.⁵¹⁶

4.5 The challenges and limitations of epigenetic therapy

Currently, a major challenge for epigenetic therapies lies in achieving therapeutic specificity, extending beyond mere target selectivity. This lack of specificity can lead to the overactivation of oncogenes while simultaneously inhibiting tumor suppressor genes, resulting in genomic instability.⁵¹⁷ Moreover, poor efficacy is notably evident in solid tumors, as DNMT inhibitors rely on DNA incorporation, which is primarily effective in actively dividing cells, rather than in solid tumors with relatively slow division rates. Additionally, HDACis display tissue-dependent effects: while HDAC1 overexpression correlates with poor prognosis in lung and pancreatic cancer, it may paradoxically improve survival outcomes in breast cancer.⁵¹⁸ Such observations underscore the importance of tissue-specific target validation for achieving successful outcomes. Additionally, cellular mechanisms affecting drug uptake and metabolism may lead to resistance, posing another significant limitation.⁵¹⁹

NcRNA therapies have shown promise both in vitro and in vivo, yet their clinical application remains hindered by challenges related to immunogenicity, specificity, and delivery.⁵²⁰ Ongoing clinical trials for miRNA therapies highlight immune-related adverse effects as a major concern. For instance, a Phase I trial of MRX34, a liposomal miR-34a mimic, was halted after severe immune-related events led to the deaths of four patients.⁴²⁸ Serious adverse events, including sepsis, hypoxia, cytokine release syndrome, and liver failure, predominantly occurred in the later treatment stages, indicating immune-mediated toxicity. Off-target effects are another pressing issue, as nontarget cell uptake or excessive dosing can lead to aberrant expression beyond physiological protein levels.⁵²¹ The issue of dose of RNA therapy is also an important consideration in the off-target problem, where overdose can saturate the system and inhibit the function of endogenous miRNAs.⁵²² Ensuring efficient delivery of ncRNA to target organs and cells while ensuring intracellular RNAi activity remains a significant challenge. A study by Sugo et al. successfully delivered siRNA targeting myostatin to mouse skeletal muscle using anti-CD71 antibody conjugation, yet the siRNA was trapped in endosomes, limiting its efficacy.⁵²³

Targeting dysregulated m6A regulators with small molecule agonists or inhibitors holds potential as a therapeutic strategy in cancer treatment. However, research on the underlying mechanisms by which m6A regulators influence tumorigenesis remains limited.⁵²⁴ The functions and in-depth mechanisms of m6A regulators in cancer remain largely unestablished and need future investigations. Moreover, findings regarding the roles of m6A regulators (such as METTL3, METTL14, FTO, and ALKBH5) in tumors are contradictory; some factors may act as both tumor suppressors and oncogenes. This complexity reduces the clarity of m6A as a singular therapeutic target, delaying the development of related small-molecule drugs. Notably, combining m6A modification inhibitors with chemotherapy, radiotherapy, and immunotherapy is expected to overcome the problem of tumor chemotherapy resistance,⁵⁰⁵ but further basic research and clinical cohort studies are still needed.