In 1863, Rudolf Virchow first observed that tumor tissues contain leukocytes, indicating an intimate correlation between inflammation and cancer (Balkwill and Mantovani, 2001). Over the past decades, studies involved in elucidating cancer-associated mechanisms have increased our understanding of the complex TME, which is composed of cellular and non-cellular components. The cellular components include fibroblasts and tumor cells, vascular endothelial cells, and immunosuppressive and antitumor immune cells; extracellular matrix (ECM), oxygen, and metabolites constitute the non-cellular components (Binnewies et al., 2018). The composition of the TME explains why traditional chemoradiotherapeutic approaches directly targeting tumor cells are often non-effective. Immunotherapy is an emerging clinical therapeutic approach that focuses on targeting immune cells. It is worth noting that a wide range of tumor patients exhibit resistance to immunotherapy. It is generally accepted that the efficacy of immunotherapeutic approaches and prognosis depend on the density and diversity of immune cells within the tumor site (Fridman et al., 2012). Accordingly, tumors are classified into hot (highly infiltrated) and cold (non-infiltrated) tumors based on the presence and absence of TILs, respectively. Hot tumors appear to have an effective response to anti-CTLA-4, anti-PD-1, and anti-PD-L1 immunotherapies, while cold tumors do not respond to these immunotherapies (Gajewski, 2015). Hot tumors are characterized by high levels of TILs, accumulation of proinflammatory cytokines such as IFN-γ, activation of inhibitory checkpoints (CTLA-4, PD-L1, etc.), genomic instability, presence of immunosuppressive factors such as indoleamine-pyrrole 2,3-dioxygenase 1 (IDO1), and the activation of major histocompatibility complex class I (MHC I). In contrast, cold tumors are characterized by poor lymphocyte infiltration inside the tumor and tumor stroma, absence of PD-L1, low mutational burden, and poor antigen presentation (loss of MHC I, IFN-γ defects, etc.) (Hegde et al., 2016). In 2009, Camus et al. (2009) described another type of tumors known as “altered tumors,” which contain stromal T cells, prevent T-cell infiltration inside of tumors, and present phenotypes that are between those of hot and cold tumors. Altered tumors are characterized by the activation of tumor-cell-intrinsic oncogene pathways such as Wnt/β-catenin and nuclear factor kappa-B (NF-κB); presence of tumor-soluble inhibitory mediators such as vascular endothelial growth factor (VEGF) and transforming growth factor-β (TGF-β); increased levels of immunosuppressive cells such as myeloid-derived suppressor cells (MDSCs), regulatory T cells (Tregs), and tumor-associated macrophages (TAMs); epigenetic changes in the TME; and metabolic competition (hypoxia, overconsumption of glucose, etc.) (Galon and Bruni, 2019). Both cold and altered tumors are derived from tumor-cell-intrinsic immunosuppression and impede effective antitumor immunity. Thus, in order for immunotherapies to have more impact, cold/altered tumors must be turned into hot tumors (Galon and Bruni, 2019; Ochoa et al., 2020).

Based on the classification into hot, altered, and cold tumors, researchers have explored different strategies to turn cold tumors into hot tumors. For example, the colony-stimulating factor-1 receptor (CSF-1R) is an attractive combination immunotherapeutic agent for tumor treatment by targeting TAMs (Mok et al., 2014; Cannarile et al., 2017; Razak et al., 2020). Furthermore, combined intratumoral interleukin (IL)-12 application with CTLA-4 was shown to lead to glioblastoma eradication through the elevation of CD4⁺ T-cell counts and Treg attenuation (Vom et al., 2013). As our understanding of cold and hot tumors expanded, strategies to turn cold tumors into hot tumors have been reported including creating local inflammation in the TME, increasing the levels of TILs, and decreasing levels of immunosuppressive cells by neutralizing immunosuppressive factors, targeting cellular metabolic and epigenetic reprogramming, normalizing tumor vasculature, and targeting tumor-cell-intrinsic oncogene pathways (Duan et al., 2020; Ochoa et al., 2020). An overview of the characteristics of hot, altered, and cold tumors as well as the strategies to turn cold tumors into hot ones is presented in Figure 1.

Epigenetic events are implicated in almost all major bioprocesses. These epigenetic events, which consist of DNA methylation, histone modification, and RNA-mediated processes, are reversible and dynamic chemical modifications (Ling and Ronn, 2019). These modifications are cooperatively interpreted by a multitude of guiding enzymes that can be classified into “writer,” “eraser,” and “reader” proteins. Disruption of any of these proteins contributes to disease development, including cancer (Dawson, 2017). DNA methylation and histone modification are essential for controlling chromatin remodeling and gene expression epigenetically. Nevertheless, the field of RNA-mediated processes has not moved forward very much (Deng et al., 2018; Jung et al., 2020). There is still a lot to uncover in terms of RNA-mediated processes, their regulation, and effects, etc., but more than 160 chemical RNA modifications have been identified since the 1950s, advancing our understanding of the biogenesis and function of RNA (Saletore et al., 2012). m⁶A, the methylation of adenosine (A) at the N⁶ position, was the first identified RNA modification and has been defined as the most widespread internal chemical modification in eukaryotic mRNA. Furthermore, m⁶A has also been identified in non-coding RNAs, such as ribosomal (rRNAs), small nuclear (snRNAs), small nucleolar (snoRNAs), micro- (microRNAs), long non-coding (lncRNAs), and circular (circRNAs) RNAs (Dominissini et al., 2012). Next-generation sequencing (NGS) studies have shown that m⁶A RNA modification sites in mRNA, microRNAs, lncRNAs, and circRNAs are non-randomly distributed but have the DRACH consensus sequence (D = G/A/U; R = G/A; H = A/C/U; G/C/U: guanosine/cytidine/uridine) and are highly enriched in the coding sequence, 3′-untranslated region (3′-UTR), and around stop codons (Meyer et al., 2012). Notably, the development of NGS-based approaches for m⁶A sequencing promises to delineate the landscape of the m⁶A epitranscriptome in various cellular contexts (Garcia-Campos et al., 2019; Huang et al., 2020b). In line with DNA methylation and histone modification, m⁶A RNA modification is a reversible and dynamic process that can be installed, removed, and recognized by its writers, erasers, and readers, respectively (Wang et al., 2020d; Figure 2).

m⁶A writers install m⁶A through a methyltransferase complex (MTC) composed of several components. Methyltransferase-like 3 (METTL3), METTL14, and Wilms’ tumor 1-associated protein (WTAP) are core components of the m⁶A MTC (Bokar et al., 1997). METTL3 is the only catalytic subunit, which installs m⁶A by binding to the methyl donor, S-adenosylmethionine (SAM), and transferring the methyl groups to adenine in the RNA molecule, producing S-adenosyl homocysteine (SAH). METTL3 and METTL14 are co-localized in nuclear speckles and form METTL3-METTL14 heterodimer complexes in a 1:1 ratio. METTL14 also contains the catalytic donor; however, METTL14 itself is not a catalytic subunit but maintains METTL3 conformation and identifies catalytic substrates (Wang P. et al., 2016; Wang X. et al., 2016). Moreover, METTL14 cooperates with the histone mark, histone H3 lysine 36 trimethylation (H3K36me3), to carry out m⁶A RNA methylation, suggesting a co-transcriptional mechanism underlying histone modification and RNA methylation in mammalian transcriptomes (Huang et al., 2019). WTAP does not have catalytic function but facilitates m⁶A deposition through recruitment of METTL3-METTL14 heterodimer complexes as well as localization to nuclear speckles (Ping et al., 2014). RNA-binding motif protein 15 (RBM15) and RBM15B, which have no catalytic function, interacts with METTL3 and WTAP and assists these two core components to reach their target RNA sites for m⁶A RNA modification in nuclear speckles (Knuckles et al., 2018). Zinc finger CCCH-type containing 13 (ZC3H13) controls the MTC by binding to WTAP and is required for the nuclear localization of the ZC3H13-WTAP-Virilizer-Hakai complex, which is essential for facilitating m⁶A methylation and mouse embryonic stem cell pluripotency (Wen et al., 2018). Vir-like m⁶A methyltransferase associated (VIRMA), also called KIAA1429, mediates preferential m6A mRNA methylation in the 3′-UTR and near stop codon (Yue et al., 2018). Furthermore, the MTC contains other components, such as METTL16 and METTL5. METTL16 has been suggested to function alone in catalyzing m⁶A modification on the U6 snRNA (Warda et al., 2017), whereas METTL5 acts as an independent RNA methyltransferase and is required for 18S rRNA m⁶A modification (Leismann et al., 2020). Moreover, zinc finger CCHC-type containing 4 (ZCCHC4) was identified as an RNA methyltransferase in 2019 and is essential for the independent methylation of 28S rRNA (Ma et al., 2019).

m⁶A RNA modification can be removed by a handful of specific demethylases known as erasers. The fat mass and obesity-associated protein (FTO) was identified as the first m⁶A demethylase in 2011 (Jia et al., 2011). FTO is an α-ketoglutarate (αKG)-dependent demethylase located in both the cell nucleus and cytoplasm (Gulati et al., 2014). FTO first oxidizes m⁶A to form N⁶-hydroxymethyladenosine (hm⁶A). Then, hm⁶A is converted to N⁶-formyladenosine (f⁶A). Lastly, f⁶A is converted to adenosine, thus removing the m⁶A RNA modification in the nucleus (Wang et al., 2020d). Furthermore, FTO also demethylates N⁶,2′-O-dimethyladenosine (m⁶A_m) in snRNA and N¹-methyladenosine (m¹A) in tRNA in the nucleus (Wei et al., 2018). It is worth mentioning that FTO can mediate mRNA and cap m⁶A_m demethylation as well as tRNA m¹A demethylation in the cytoplasm (Wei et al., 2018). Moreover, ALKB homolog 5 (ALKBH5) is another vital m⁶A eraser, which is Fe(II) dependent, locates in the nucleus, and seems to be an m⁶A-specific demethylase involved in m⁶A RNA modification (Zheng et al., 2013). Moreover, Ueda et al. (2017) recently identified ALKBH3, an m⁶A eraser suggested to be present in both, in the cytoplasm and nucleus, promoting the demethylation of target mammalian tRNA.

The reversible processes of m⁶A RNA installation and removal occur through the alteration of the RNA structure. RNA-mediated biological functions are also regulated by m⁶A-binding proteins, which are called m⁶A readers (Li A. et al., 2017). On the one hand, cytoplasmic mRNA is decoded in the ribosome to produce a protein. On the other hand, messenger ribonucleoprotein (mRNP) foci are essential for the storage or degradation of cytoplasmic RNA. The YT521-B homology (YTH) domain-containing proteins (YTHDFs) and insulin-like growth factor-2 mRNA-binding proteins (IGF2BPs) play crucial roles in RNA-mediated biological functions by binding to m⁶A domains in the cytoplasm. YTHDFs include YTHDF1, YTHDF2, and YTHDF3. YTHDF1 selectively binds to m⁶A and recruits translation initiation factors, including the eukaryotic translation initiation factors (eIFs) 3/4E/4G, poly(A) binding protein (PABP), and 40S ribosomal subunit, to magnify RNA translation (Wang et al., 2015). The first identified m⁶A reader was YTHDF2, which recognizes m⁶A-modified RNA degradation sites via its C-terminal region and recruits the carbon catabolite repressor 4-negative on TATA (CCR4-NOT) deadenylase complex through its N-terminal region (Du et al., 2016; Zhang C. et al., 2020). YTHDF3 has overlapping roles in RNA fate through augmenting RNA translation in cooperation with YTHDF1 and promoting RNA degradation via synergy with YTHDF2 (Li A. et al., 2017; Shi et al., 2017). Cytoplasmic IGF2BPs, including IGF2BP1, IGF2BP2, and IGF2BP3, bind directly to m⁶A-modified RNA through its K homology domains and promote the stability and translation of RNA (Kataoka, 2019). Cytoplasmic YTH domain-containing protein 2 (YTHDC2) is another m⁶A reader that can recognize m⁶A and bind to meiosis-specific coiled-coil domain (MEIOC) and 5′-3′exoribonuclease 1, further increasing m⁶A-modified RNA translation (Hsu et al., 2017). Notably, m⁶A readers can also bind m⁶A in the nucleus. For example, YTHDC1 promotes exon inclusion in RNA by amplifying serine- and arginine-rich splicing factor 3 (SRSF3) or blocking serine- and arginine-rich splicing factor 10 (SRSF10) in the nucleus (Xiao et al., 2016). Furthermore, YTHDC1 plays a role in facilitating m⁶A-methylated RNA export from the nucleus to the cytoplasm (Roundtree et al., 2017). Additionally, heterogeneous nuclear ribonucleoproteins (hnRNPs), including HNRNPA2B1, HNRNPC, and HNRNPG, recognize m⁶A and act as “m⁶A switches” that accelerate RNA and primary microRNA processing by changing the RNA structure (Alarcon et al., 2015; Liu et al., 2015; Zhou et al., 2019).

In summary, studies have shown that m⁶A RNA modifications are implicated in a wide range of biological processes. Nevertheless, structural and biochemical data on m⁶A writers, erasers, and readers need to be further verified, and the detailed mechanisms regulated by these proteins remain undetermined. It is reasonable to believe that there are more m⁶A writer, eraser, and reader components, and that the mechanism underlying these protein-mediated RNA modifications will be elucidated with the development of quantification and sequencing methodologies (Bodi and Fray, 2017; Chen et al., 2019). A summary of the currently known m⁶A writers, erasers, and readers is presented in Table 1.

In colorectal cancers with low mutational burden, which are resistant to immunotherapy, depletion of METTL3 and METTL14 increases the expression of CD8⁺ T cells and the secretion of IFN-γ via the m⁶A reader YTHDF2 (Wang et al., 2020b). Another study showed that tumors with decreased levels of METTL3 have increased DC infiltration, MHC expression, and levels of costimulatory and adhesion molecules in the TME (Shen et al., 2021a). On the contrary, loss of METTL3 has also been shown to promote tumor growth and metastasis. For example, METTL3-deficient mice show increased immunosuppressive cell (TAMs, Tregs) infiltration into tumors (Yin et al., 2021). Yao et al. (2021) showed that METTL3 is responsible for the expression of T follicular helper cells, which are specialized effector CD4⁺ T cells. Loss of METTL3 results in inactivation of T follicular helper cell differentiation by promoting the decay of T follicular helper cell signature genes, including Tcf7 transcripts. Using CRISPR-Cas9 screening, Tong et al. (2021) demonstrated that loss of METTL3 leads to the removal of m⁶A RNA modification on Irakm IL-1 receptor-associated kinase 3 (IRAK3) mRNA, slowing down its degradation and ultimately attenuating toll-like receptor 4 (TLR4) signaling-mediated macrophage activation. Particularly, the authors suggested that METTL3 augments the tumoricidal ability of macrophages by promoting the polarization bias of TAMs toward the M1 macrophage phenotype and rescuing infiltrating CD4⁺ and CD8⁺ T cells (Tong et al., 2021). Recently, mechanistic investigations found a positive role of ALKBH5 in Tregs and MDSCs by targeting Mct4/Slc16a3. Notably, low levels of ALKBH5 in clinical settings are correlated with low Treg cell numbers (Li et al., 2020c). However, another study by Tang et al. (2020) showed that deletion of ALKBH5 decreases the infiltration of CD8⁺ T cells in pancreatic adenocarcinoma.

Lysosomal proteases are responsible for antigen degradation in DCs (Cebrian et al., 2011). In a study by Han D. et al. (2019), YTHDF1 was shown to have a negative correlation with CD8⁺ T-cell infiltration in colon cancer patients. Mechanistically, YTHDF1 in DCs can recognize lysosomal proteases, leading to the inactivation of cross-presentation. Loss of YTHDF1 promotes DC-mediated cross-presentation of tumor antigens and cross-priming of CD8⁺ T cells in vitro and in vivo (Han D. et al., 2019). Additionally, other m⁶A RNA modification regulators have also been found to have a close relationship with immune cells in tumors. For instance, the expression of METTL14 and ZC3H13 is positively correlated with infiltrating levels of CD4⁺ T cells, CD8⁺ T cells, and DCs, but negatively correlated with those of Tregs in breast cancer (Gong et al., 2020). In head and neck squamous cell carcinoma, low expression of YTHDC2 is positively correlated with the low levels of B cells, CD8⁺ T cells, CD4⁺ T cells, neutrophils, and infiltrating DCs (Li et al., 2020d). IFN-γ is the main proinflammatory cytokine produced by cytotoxic T cells, enhancing antigen presentation to cytotoxic T cells by facilitating MHC I and immunoproteasome expression in tumor cells (Cheon et al., 2014). YTHDF2 is responsible for RNA-binding motif 4 (RBM4)-mediated suppression of IFN-γ-induced M1 macrophage polarization and glycolysis (Huangfu et al., 2020). In a recent study by Shen et al. (2021a), downregulation of METTL3 was shown to contribute to increasing the levels of MHC molecules (Shen et al., 2021a). More recently, the levels of YTHDC2, HNRNPC, and VIRMA were suggested to be negatively correlated, whereas WTAP was positively correlated, with MHC molecules in endometrial cancer (Zhao et al., 2021). A comprehensive study showed that a low risk score of m⁶A signature is significantly correlated with a high expression of immune cell checkpoint molecules, such as PD-1, PD-L1, and CTLA-4 (Mo et al., 2020). Nevertheless, the mechanisms whereby m⁶A RNA modification regulators exert their action in immune cells of the TME remain unclear.

Several studies have shown that METTL3 acts as an oncogenic regulator by activating tumor-cell-intrinsic pathways in tumors. For example, in hepatoblastoma, upregulation of METTL3 promotes the proliferation, migration, and invasion of hepatoblastoma cells by activating the Wnt/β-catenin signaling pathway (Liu et al., 2019; Cui et al., 2020). In colorectal cancer, METTL3 promotes tumor metastasis, stemness, and chemoresistance through activation of MAPK and Wnt/β-catenin signaling (Peng et al., 2019; Liu et al., 2021b). Furthermore, METTL3 facilitates the proliferation and invasion of esophageal cancer cells via activation of Wnt/β-catenin and AKT signaling (Hou et al., 2020). In contrast, Yin et al. (2021) recently showed that ablation of METTL3 orchestrates tumor growth and metastasis by facilitating ERK-NF-κB/STAT3 signaling. Liu et al. (2018) showed that METTL14 mutation and loss of METTL3 expression contribute to increased proliferation and tumorigenicity of endometrial cancer cells by activating AKT signaling. Moreover, METTL3 knockdown in a multiplicity of tumor cell lines leads to the activation of PI3K/AKT/mTOR signaling (Zhao et al., 2020). Wang Y. et al. (2021) indicated that METTL14 may be a favorable prognostic factor for clear cell renal cell carcinoma (ccRCC). Mechanistically, loss of METTL14 increases gastric cancer cell proliferation and invasiveness by promoting the activation of Wnt and PI3K/AKT signaling. In contrast, knockdown of FTO restricts the activation of Wnt and PI3K/AKT signaling (Zhang et al., 2019). Recently, Liu et al. (2021a) showed that METTL3 and METTL14 are required for senescence-associated secretory phenotype (SASP)-mediated tumor-promoting and immune-surveillance functions of senescent cells through the activation of NF-κB signaling. Frizzled proteins are key Wnt receptors whose activation contributes to the stabilization of cytoplasmic β-catenin (MacDonald and He, 2012). The activity of FTO and ALKBH5 lead to PARP inhibitor resistance in BRCA-deficient epithelial ovarian cancer (EOC) cells by upregulating the Wnt/β-catenin pathway through stabilization of Frizzled 10 protein (Fukumoto et al., 2019). YTHDF1 has been shown to promote stemness, tumor cell proliferation, and metastasis by activating the Wnt/β-catenin pathway through the stabilization of Frizzled 5 and 7 (Bai et al., 2019; Han et al., 2020; Liu et al., 2020; Pi et al., 2021).

As mentioned earlier, altered tumors are characterized by the presence of tumor angiogenesis. METTL3 has been shown to facilitate miR-143-3p biogenesis, promoting the brain metastasis in lung cancer patient samples through the miR-143-3p/Vasohibin/VEGFA axis (Wang et al., 2019). In line therewith, Wang G. et al. (2021) showed that METTL3 is responsible for the activation of tyrosine kinase endothelial (TEK)-VEGFA-mediated tumor progression and angiogenesis in bladder cancer. In colon cancer, the m⁶A RNA modification reader, IGF2BP3, can bind to the VEGF mRNA to promote its expression and stability. Thus, loss of IGF2BP3 restricts angiogenesis by inhibiting VEGF (Yang et al., 2020). Upregulation of TGF-β in the TME also contributes to altered tumor formation by suppressing T-cell proliferation and stimulating Treg development (Chen and Ten Dijke, 2016). In TGF-β-induced epithelial-mesenchymal transition (EMT) of lung cancer cell lines, the level of METTL3 was found to be upregulated. Loss of METTL3 attenuates TGF-β-induced morphological conversion of lung cancer cells, their cell migration potential, and EMT progression (Wanna-Udom et al., 2020). Mechanistic investigations found that METTL3 increases TGF-β1 mRNA decay and impairs TGF-β1 translation progress. Furthermore, ablation of METTL3 disrupts the autocrine action of TGF-β1 by interrupting TGF-β1 dimer formation and TGF-β1-induced EMT in cancer cells (Li et al., 2020a). Importantly, the level of VEGFA and content of TGF-β1 in the TME are decreased in ALKBH5-deficient melanoma cells (Li et al., 2020c).

Recently, m⁶A RNA modification was recognized to be responsible for metabolic competition-mediated tumorigenesis. Cancer cells with metabolic competition contribute to tumorigenesis through inhibiting T-cell responses and increasing T-cell depletion (Kedia-Mehta and Finlay, 2019). Upregulation of ALKBH5 was shown to contribute to breast cancer initiation by attenuating NANOG mRNA methylation and thereby increasing NANOG expression under hypoxia (Zhang et al., 2016). FTO was found upregulated in tumor suppressor von Hippel-Lindau (VHL)-deficient ccRCC. Mechanistically, FTO increases metabolic reprogramming and survival of VHL-deficient ccRCC cells by targeting SLC1A5 in a hypoxia-inducible factor (HIF)-independent way (Xiao et al., 2020). Furthermore, Niu et al. (2021) recently showed that the posttranscriptional regulation of the abnormal expression of aldolase A (ALDOA) under hypoxia was positively modulated by FTO-mediated m⁶A RNA modification in a YTHDF2-dependent manner in liver cancer cells, and hypoxia-mediated high level of ALDOA contributed to liver cancer development by promoting glycolysis metabolism and its terminal product lactate expression. Additionally, FTO promotes tumor cell glycolysis by activating PI3K/AKT signaling or in a YTHDF2-dependent manner (Liu et al., 2017; Liu et al., 2021d). In addition, upregulation of METTL3 in gastric cancer promotes tumor angiogenesis and glycolysis by promoting IGF2BP3-dependent hepatoma-derived growth factor (HDGF) mRNA stability, which is essential for increasing in glycolysis by activating GLUT4 and ENO2 in gastric cancer cells (Wang et al., 2020c).

Tumors with high mutational burden carry neoantigens that are sensitive to immune cells and immune checkpoint blockade (Samstein et al., 2019). Recently, numerous systematic and comprehensive studies have suggested a close relationship between m⁶A RNA modification and mutational burden. m⁶A RNA modification patterns are quantified as m⁶Ascore by a specific procedure (Zeng et al., 2019). Zhang B. et al. (2020) comprehensively investigated the m⁶A RNA modification patterns of 1,938 gastric cancer samples based on 21 m⁶A regulators and systematically analyzed the correlation between the m⁶Ascore and TME cell-infiltrating characteristics. They found that a low m⁶Ascore is markedly correlated with increased mutational burden and activation of immunity and correlated with increased neoantigen load and enhanced response to anti-PD-1/L1 treatment (Zhang B. et al., 2020). Another study showed a wide range of FTO, RBM15, and YTHDF1 inter-group expression differences between high- and low- tumor mutational burden cancer tissues (Liu et al., 2021c). Consistently, a recent study indicated that there is a positive correlation between the m⁶A signature and tumor mutational burden scores in 16 cancer types (Shen et al., 2021b). Furthermore, in colon cancer patients, a low m⁶Ascore is associated with high tumor mutational burden, PD-L1 expression, and mutation rates in significantly mutated genes (Chong et al., 2021). It is also noteworthy that colorectal cancers with low mutational burden were suggested to be resistant to anti-PD-1 immunotherapy through the inhibition of IFN-γ-mediated CD8⁺ T-cell secretion by METTL3 and METTL14 (Wang et al., 2020b). Nevertheless, the mechanisms whereby m⁶A RNA modification regulates the tumor mutational burden require further investigation.

Collectively, the mechanisms underlying m⁶A RNA modification-mediated cold tumor formation include immune cell regulation in the TME, targeting of tumor-cell-intrinsic pathways, facilitation of soluble inhibitory mediators in the TME, increase of metabolic competition in the TME, and effect on tumor mutational burden. Notably, several specific m⁶A regulators play dual roles in cold tumor formation, such as METTL3, METTL14, and YTHDF1, suggesting the exact role m⁶A RNA modification-mediated cold tumor formation is tumor-type dependent. Furthermore, the abnormal expression of m⁶A regulators contribute to cold tumor formation is not through one mechanism alone, they always play roles in cold tumor formation by several mechanisms. For example, METTL3 is involved in cold tumor formation via regulating immune cell expression, targeting of tumor-cell-intrinsic pathways, facilitating soluble inhibitory mediators, increasing metabolic competition, and affecting tumor mutational burden together, which indicated the extensive role of m⁶A RNA modification in cold tumor formation. In addition, some different m⁶A regulators are implicated in cold tumor formation by the same mechanism, such as METTL3 and ALKBH5, they both lead to cold tumor formation through VEGFA expression, indicating they may play a role in the cold tumor formation synergistically, which needs to be validated in the future. The studies involved in m⁶A RNA modification in cold tumor are just getting started; the related mechanism is still unclear and needs to be illustrated in the future. An overview of the uncovered mechanisms till now is presented in Figure 3.