2.1. Expression Levels of m⁶A-RMRs in Patients with Various Inflammatory Disorders and Tumors

We collected 23 microarray datasets of acute inflammation, metabolic and cardiovascular diseases, autoimmune diseases, and organ failures (Table 1); one microarray of Middle-East respiratory syndrome (MERS) coronavirus-infected human microvascular endothelial cells; one microarray dataset (nine comparisons) of subacute respiratory syndrome coronavirus- (SARS-CoV-) infected human airway epithelial cells; one microarray dataset (23 comparisons) of influenza virus-infected lung epithelial cells (Table 2); and eight microarray datasets of endothelial cells (Table 3) from National Institutes of Health- (NIH-) National Center for Biotechnology Information- (NCBI-) Gene Expression Omnibus (GEO) databases (https://www.ncbi.nlm.nih.gov/gds/). These datasets were analyzed with GEO2R (https://www.ncbi.nlm.nih.gov/geo/geo2r/). Some datasets were overlapped with our previous studies [44]. In addition, the Oncomine database (https://www.oncomine.org) was used to analyze the gene expression profile from 19 tumors [45], with threshold parameters of fold change > 2, p < 0.05, and gene rank in the top 10%. Because these microarray studies employed diverse cell types, we were unable to compare the effects of illness circumstances on m⁶A-RMR regulation in the same cell types. It is worth noting that our strategy was well justified. For example, we and others frequently investigated gene expression in nonideal heterogeneous peripheral blood mononuclear cell populations (PBMCs) in pathophysiological conditions, which are made up of a variety of cell types (also see Discussion).

2.2. Expression Profile of m⁶A-RMRs in Single-Cell RNA Sequencing (scRNA-Seq) Datasets from Studies of Sepsis, Atherosclerosis, Tumors, and Endothelial Cell

Five scRNA-Seq datasets were collected from the Single Cell Portal database (https://singlecell.broadinstitute.org/single_cell), including one study about sepsis, one study about atherosclerosis, two studies about tumors (astrocytoma and melanoma), and one study about endothelial cell populations (Supplementary Table 1). The expressions of 29 m⁶A-RMRs were online analyzed.

2.3. Expression Regulation Analysis of m⁶A-RMRs from Deficiency of Folate Cycle and Metabolism-Related Enzymes, m⁶A-RMRs, H3K4 Methylase, DNA Methyltransferase, Regulatory T Cells’ Signature Genes, Proinflammatory Cytokines, Oncogene, Tumor Suppressors, and Immune Checkpoint Receptors

The 68 microarray datasets in the NIH-NCBI-GeoDataset database (https://www.ncbi.nlm.nih.gov/gds/) were collected in analyzing the regulatory mechanisms of m⁶A-RMRs (Supplementary Table 2). There are six microarrays about deficiencies of the folate cycle and metabolism-related enzymes and m⁶A-RMRs, four datasets of deficiencies of H3K4 methylase, six datasets of deficiencies of DNA methyltransferase, three datasets of Treg versus conventional T cells, 11 datasets of deficiencies of Treg cells’ signature genes, one dataset of macrophage polarization, eight datasets of deficiencies of proinflammatory cytokines, nine datasets of deficiencies of oncogene, ten datasets of deficiencies of tumor suppressors, and ten datasets of deficiencies of immune checkpoint receptors.

2.4. Statistical Analysis of Microarray Data

The expression changes of m⁶A-RMRs are not big enough in diseases; we may miss some important information if the expression change fold thresholds are set high. Thus, the expression changes were listed in the results with p < 0.05 (statistical significance), m⁶A-RMRs with expression changes more than 1-fold (red) were defined as the upregulated genes, while genes with expression decreased more than 1-fold (blue) were defined as downregulated genes. The missing value was marked with “NA,” and “NA” was excluded in the total count number.

3.1. A Few m⁶A-RNA Methylation Regulators (m⁶A-RMRs) Are Upregulated, and Most m⁶A-RMRs Are Downregulated in Sepsis, Sepsis plus Acute Respiratory Distress Syndrome, Sepsis plus Shock, and Trauma, and Upregulated Two RNA Methyltransferases WTAP and PCIF1 and Three RNA Binding Proteins YTHDF3, IGF2BP2, and IGF2BP3 May Promote Acute Inflammations

We hypothesized that pathological conditions significantly modulate the expressions of m⁶A-RMRs in cell type-specific and disease-specific manners. To test this hypothesis, we collected 29 m⁶A-RMRs in five groups from PubMed database (https://pubmed.ncbi.nlm.nih.gov/) (Figure 2 and Supplementary Table 3). The 29 m⁶A-RMRs included (1) 10 m⁶A-RNA methylation writers (methyltransferases) including METTL14, METTL3, WTAP, KIAA1429, ZC3H13,CBLL1, RBM15, METTL16,RBM15B, and PCIF1; (2) two m⁶A-RNA demethylases, FTO and ALKBH5; (3) 11 RNA binding proteins (readers) including YTHDC1, YTHDF1, YTHDF2, YTHDF3, YTHDC2, HNRNPA2B1, EIF3A, IGF2BP1, IGF2BP2, IGF2BP3, and FMR1; (4) three m⁶A-dependent RNA binding proteins including HNRNPC, RMMX, and PRRC2A; and (5) three m⁶A-repelled RNA binding proteins including ELAVL1, G3BP1, and G3BP2. As shown in Figure 3(a), one to four out of 28 m⁶A-RMRs were upregulated; and 14 to 15 out of 28 m⁶A-RMRs were downregulated in 0 day and 7 days in patients with sepsis and sepsis plus acute respiratory distress syndrome (ARDS) [46], respectively. In the second datasets with sepsis and sepsis plus shock, two to four out of 26 m⁶A-RMRs were upregulated; and 10 to 17 out of 26 m⁶A-RMRs were downregulated in 1 day and 3 days in patients with sepsis and sepsis plus shock [47, 48], respectively. In the third datasets with leukocytes, monocytes, and T cells from trauma, zero to one out of 22 m⁶A-RMRs was upregulated; and 5 to 11 out of 22 m⁶A-RMRs were downregulated in leukocytes, monocytes, and T cells from trauma [49], respectively.

We then analyzed the shared and disease-specific m⁶A-RMRs using Venn diagram analysis. As shown in Figure 3(b), four diseases including sepsis at 0 day and 7 days and sepsis plus ARDS at 0 day and 7 days shared one RNA methyltransferase WTAP. In the second sepsis datasets, four diseases including sepsis at 1 day and 3 days and sepsis plus shock at 0 day and 7 days shared one RNA methyltransferase PCIF1 and one RNA methylation reader IGF2BP3. In the trauma datasets, three immune cell types, leukocytes, monocytes, and T cells did not share any upregulated m⁶A-RMRs. When comparing three groups of acute inflammations, sepsis plus ARDS shared three upregulated m⁶A-RMRs such as writers WTAP, PCIF1, and reader YTHDF3 with sepsis plus shock; sepsis plus ARDS shared upregulation of one RNA binding protein IGF2BP2 with trauma, and sepsis plus shock shared one RNA binding protein IGF2BP3 with trauma.

We noticed that in the trauma datasets (Figure 3(b)), three immune cell types, leukocytes, monocytes, and T cells did not share any upregulated RNA methylation regulators, suggesting that upregulation of m⁶A-RMRs is in a cell-specific manner. To further examine this issue, we collected single-cell RNA-sequencing (scRNA-Seq) datasets in sepsis [57] in a comprehensive single-cell sequencing database (https://singlecell.broadinstitute.org/single_cell). As shown in Figure 3(c), the 11 out of 29 m⁶A-RMRs (38%) including WTAP, ZC3H13, YTHDC1, YTHDF2, HNRNPA2B1, EIF3A, HNRNPC, RBMX, ELAVL1, G3BP1, and G3BP2 were expressed differentially in 13 different immune cell types (natural killer cell (NK), granulomyeloid progenitor (GMP), megakaryocyte-erythrocyte progenitor (MEP), common lymphoid progenitor (CLP), myeloblasts (MYL), T cells, dendritic cell (DC), common myeloid progenitor (CMP), monocyte (mono), hematopoietic stem cell/multipotent progenitor (HSC/MPP), B cells, CD14⁺RETN⁺ALOX5AP⁺IL1R2⁺ monocyte (iMS1), and red blood cells (RBC)) in bacterial sepsis. The expressions of most m⁶A-RMRs except HNRNPA2B1 and HNRNPC were lower in monocyte subsets than those in other peripheral mononuclear cell (PBMC) types (Figure 3(d)).

These results have demonstrated that first, a few m⁶A-RMRs are upregulated and most m⁶A-RMRs are downregulated in sepsis, sepsis plus acute respiratory distress syndrome, sepsis plus shock, and trauma; second, two RNA methyltransferases WTAP and PCIF1 and three RNA binding proteins such as YTHDF3, IGF2BP2, and IGF2BP3 are commonly upregulated in sepsis, sepsis plus ARDS, sepsis plus shock, and trauma, suggesting that these five m⁶A-RMRs are the emergency m⁶A-RMRs for promoting acute inflammatory diseases; third, nine m⁶A-RMRs including METTL3, RBM15, FTO, YTHDC2, HNRNPA2B1, EIF3A, HNRNPC, G3BP1, and CBLL1 are commonly downregulated in sepsis, sepsis plus ARDS, sepsis plus shock, and trauma, suggesting that those m⁶A-RMRs play more important roles in maintaining homeostasis and suppressing inflammation than emergency roles for acute inflammations; and fourth, the expressions of 38% m⁶A-RMRs in immune cell types in response to sepsis stimulation are different.

3.2. Type 2 Diabetes Has More Modulation of m⁶A-RMR Expressions Than Atherogenic Diseases and Obesity; Nearly Half of 29 m⁶A-RMRs Are Downregulated as Atherosclerosis Progression Compared with That of Atherosclerosis Regression

We hypothesized that major metabolic cardiovascular disease groups such as obesity [58, 59], type 2 diabetes, and atherogenic diseases differentially modulate the expressions of m⁶A-RMRs. To test this hypothesis, we collected five datasets of obesity including obese, metabolically unhealthy obesity (MUO), metabolically healthy obesity (MHO), obese with insulin resistance (ob IR), and obese with insulin sensitivity (ob IS); four datasets of type 2 diabetes; and five datasets of atherosclerosis, familial hypercholesterolemia (FHC) plus atherosclerosis, and familial combined hyperlipidemia (FCH) from the NCBI-GeoDataset (https://www.ncbi.nlm.nih.gov/gds/). As shown in Figure 4(a), in five diseases in the obesity group, zero to three m⁶A-RMRs were upregulated and zero to 10 m⁶A-RMRs were downregulated. However, in the four type 2 diabetes datasets, zero to 11 m⁶A-RMRs were upregulated and one to seven m⁶A-RMRs were downregulated. The five datasets of atherosclerotic diseases had modulations of m⁶A-RMRs higher than those in the obesity group but lower than those of type 2 diabetes group: zero to seven m⁶A-RMRs were upregulated, and zero to seven m⁶A-RMRs were downregulated.

We also analyzed the shared and disease-specific m⁶A-RMRs using Venn diagram analysis (Figure 4(b)). In six m⁶A-RMRs upregulated in the obesity group, two regulators IGF2BP3 (RNA methyltransferase) and G3BP1 (m⁶A repelled RNA binding protein) were shared by obesity IS and obesity IR. Among the 15 m⁶A-RMRs upregulated in four different tissues in type 2 diabetes, one regulator HNRNPA2B1 (reader) was shared by liver, subcutaneous adipose, and visceral adipose, and one regulator G3BP1 was shared by liver and visceral adipose. In 10 m⁶A-RMRs upregulated in atherosclerotic diseases, one regulator WTAP (RNA methyltransferase) was shared by atherosclerosis (athero) carotid artery and atheromacrophages, and two regulators PCIF1 (RNA methyltransferase) and PRRC2A (m⁶A dependent RNA binding protein) were shared by atheromacrophages and FHC and atheromonocytes. Among 21 m⁶A-RMRs upregulated in three major metabolic cardiovascular disease groups, one regulator WTAP (RNA methyltransferase) was shared by three major groups; one regulator IGF2BP3 was shared by obesity and athero; three regulators FTO (demethylase), YTHDF2 (RNA binding protein), and G3BP1 were shared by obesity and type 2 diabetes; and four regulators such as PCIF1 (RNA methyltransferase), PRRC2A (m⁶A-dependent RNA binding protein), YTHDC2 (RNA binding protein), and HNRNPC (m⁶A-dependent RNA binding protein) were shared by type 2 diabetes and atherosclerotic diseases.

We then examined a hypothesis that atherosclerosis progression and regression differentially modulate the expressions of m⁶A-RMRs. We collected a dataset of scRNA-Seq in the scRNA-Seq database in the Broad Institute at MIT. As shown in Figure 4(c), the expressions of 14 out of 27 m⁶A-RMRs including Wtap, Pcif1, Alkbh5, Ythdc1, Ythdf1, Ythdf2, Ythdf3, Hnrnpa2b1, Eif3a, Fmr1, Hnrnpc, Prrc2a, G3bp1, and G3bp2 were decreased in the progressive atherosclerosis compared with those in the regressive atherosclerosis. The expression of Elavl1 was increased in the progressive atherosclerosis compared with that in the regressive atherosclerosis. Of note, the expressions of Virma and Igf2bp1 were not found. These results have illustrated that (1) atherosclerotic macrophages have higher upregulation of m⁶A-RMRs than atherosclerotic carotid artery, and FHC and atherosclerotic monocytes have more modulation of m⁶A-RMRs than FHC and atherosclerotic T cells; (2) type 2 diabetes adipose tissues have the highest modulation of m⁶A-RMRs (11 upregulated and three downregulated) among the 14 metabolic disease datasets; (3) obesity diseases have more downregulation than upregulation of m⁶A-RMRs, and (4) nearly half of 29 m⁶A-RMRs are downregulated as atherosclerosis progression compared with that of atherosclerosis regression.

Additionally, for further understanding the m⁶A-RMRs’s modulation in cardiovascular diseases, our study of m⁶A-RMRs in atherosclerosis and other three studies of m⁶A-RMRs in cardiovascular disease (CVDs) [60–62] were compared. From Table 4, METTL14 is upregulated in our study and other two studies; FTO is downregulated in our study and other two studies. At least, these data revealed that in atherosclerosis and other CVDs, m⁶A writer METTL14 is upregulated and eraser FTO is downregulated.

3.3. Inflammatory Bowel Diseases and Rheumatoid Arthritis Modulate m⁶A-RMRs More Than Autoimmune Lupus and Psoriasis; and Some Autoimmune Diseases Share Five Upregulated m⁶A-RMRs such as PCIF1, G3BP2, G3BP1, WTAP, and FTO

We hypothesized that major autoimmune diseases including rheumatoid arthritis (RA), psoriasis (PS), acute cutaneous lupus (ACLE), chronic cutaneous lupus (CCLE), subacute cutaneous lupus (SCLE), Crohn’s disease (CD), and ulcerative colitis (UC) differentially modulate the expressions of m⁶A-RMRs. To examine this hypothesis, we collected four microarray datasets (eight comparisons) from the NCBI-GeoDatasets (https://www.ncbi.nlm.nih.gov/gds/). In a RA dataset, four out of 26 m⁶A-RMRs were upregulated; and 10 out of 26 m⁶A-RMRs were downregulated. In four lupus datasets, two to six out of 28 m⁶A-RMRs were upregulated; and one to four out of 28 m⁶A-RMRs were downregulated. In three colitis datasets, five to eight m⁶A-RMRs were upregulated; and eight to nine m⁶A-RMRs were downregulated (Figure 5).

We also analyzed the shared and disease-specific m⁶A-RMRs using Venn diagram analysis. In seven m⁶A-RMRs upregulated in the autoimmune skin disease group, one m⁶A-RMR WTAP was shared by four autoimmune skin diseases; one m⁶A-RMR G3BP1 was shared by three autoimmune skin diseases including ACLE, CCLE, and psoriasis; one m⁶A-RMR PRRC2A was shared by three autoimmune skin diseases including CCLE, psoriasis, and SCLE; and one m⁶A-RMR G3BP2 was shared by two autoimmune skin diseases including CCLE and psoriasis. In 13 m⁶A-RMRs upregulated in the inflammatory bowel disease group, one m⁶A-RMR ELAVL1 was shared by UC sigmoid colon or rectum and UC PBMCs; one m⁶A-RMR WTAP was shared by UC sigmoid colon or rectum and CD PBMCs; and four m⁶A-RMRs including ZC3H13, RBM15B, IGF2BP2, and IGF2BP3 were shared by CD PBMCs and UC PBMCs. In 19 m⁶A-RMRs upregulated in all eight autoimmune diseases, one m⁶A-RMR PCIF1 was shared by RA and inflammatory bowel disease; one m⁶A-RMR G3BP2 was shared by RA and autoimmune skin diseases; and three m⁶A-RMRs including G3BP1, WTAP, and FTO were shared by autoimmune skin diseases and inflammatory bowel diseases. Taken together, our results have shown that first, inflammatory bowel diseases and RA modulate m⁶A-RMRs more than autoimmune lupus and psoriasis; second, autoimmune skin diseases share four upregulated m⁶A-RMRs including G3BP1, G3BP2, PRRC2A, and WTAP; third, inflammatory bowel diseases share six upregulated m⁶A-RMRs including ELAVL1, WTAP, ZC3H13, RBM15B, IGF2BP2, and IGF2BP3; and fourth, autoimmune diseases have no any commonly shared m⁶A-RMRs but share five upregulated m⁶A-RMRs between groups including PCIF1, G3BP2, G3BP1, WTAP, and FTO.

3.4. Some Organ Failures Share Eight Upregulated m⁶A-RMRs between Groups including Four Writers WTAP, PCIF1, RBM15, and RBM15B; Two m⁶A-Dependent RNA Binding Proteins PRRC2A and HNRNPC; and Two m⁶A-Repelled RNA Binding Proteins IGF2BP2 and IGF2BP3; End-Stage Renal Failure (ESRF) Downregulates 85% of 26 m⁶A-RMRs, and Upregulation of 45% m⁶A-RMRs in Hemodialysis from ESRF (15%) May Be Associated with Clinical Benefits

We hypothesized that major organ failures including heart failure, hepatitis B virus-associated acute liver failure (HBV-ALF), end-stage renal failure (ESRF), and hemodialysis modulate the expressions of m⁶A-RMRs. We collected four microarray datasets from the NCBI-GeoDatasets (https://www.ncbi.nlm.nih.gov/gds/) to examine this hypothesis. As shown in Figure 6, in the heart failure dataset, six out of 26 m⁶A-RMRs were upregulated; and seven out of 26 m⁶A-RMRs were downregulated. In the HBV-ALF dataset, eight out of 26 m⁶A-RMRs were upregulated; and 17 out of 26 m⁶A-RMRs were downregulated. In the ESRF dataset, four out of 26 m⁶A-RMRs were upregulated; and 22 out of 26 m⁶A-RMRs were downregulated. In the hemodialysis dataset, 10 out of 22 m⁶A-RMRs were upregulated; and five out of 22 m⁶A-RMRs were downregulated. We also analyzed the shared and disease-specific m⁶A-RMRs using Venn diagram analysis. In 20 m⁶A-RMRs upregulated in four organ failure groups, one upregulated m⁶A-RMR (RBM15B, RNA methyltransferase) was shared by HF and HBV-ALF; one upregulated m⁶A-RMR (PRRC2A) was shared by HF and ESRF; two upregulated m⁶A-RMRs (IGF2BP2 and IGF2BP3, RNA binding proteins) were shared by HBV-ALF and ESRF; and four m⁶A-RMRs including WTAP (RNA methyltransferase), RBM15 (RNA methyltransferase), PCIF1 (RNA methyltransferase), and HNRNPC (m⁶A-dependent RNA binding protein) were shared by HBV-ALF and hemodialysis.

These results demonstrated that first, four major organ failures have no commonly shared upregulated m⁶A-RMRs but share eight m⁶A-RMRs between groups including RBM15B, IGF2BP2, IGF2BP3, PRRC2A, WTAP, RBM15, PCIF1, and HNRNPC; second, hemodialysis is the only organ failure that upregulates more m⁶A-RMRs than downregulates m⁶A-RMRs; other organ failures have less upregulation of m⁶A-RMRs than downregulation of m⁶A-RMRs; third, surprisingly, ESRF and hemodialysis have no any shared upregulated m⁶A-RMRs, suggesting that the clinical improvements of hemodialysis from ESRF are benefited from upregulation of m⁶A-RMRs by hemodialysis; and fourth, ESRF modulates 100% of 26 m⁶A-RMRs, which is the highest modulation rate found among all the diseases examined.

3.5. MERS-CoV Infections at 36 Hours in Human Microvascular Endothelial Cells Modulate the Highest Numbers of m⁶A-RMRs (Upregulated 20 and Downregulated Four) among Three Types of Viral Infections (MERS-CoV, SARS-CoV, and Influenza Virus); and the Differences in Modulating the Expressions of m⁶A-RMRs May Be Related to the Virulence of Virus and Virus Replication

We hypothesized that viral infections with Middle East respiratory syndrome (MERS) coronavirus infection [63], severe acute respiratory syndrome (SARS) coronavirus (SARS-CoV) [64], and influenza virus infections [65] modulate the expressions of m⁶A-RMRs. To test this hypothesis, we collected three groups of microarray datasets from the NCBI-GeoDatasets (https://www.ncbi.nlm.nih.gov/gds/). We found that MERS-CoV infection of human microvascular endothelial cells at the time course of 0, 12, 24, 36, and 48 hours (h) [63] after infection resulted in upregulation of two to 20 out of 28 m⁶A-RMRs and downregulation of one to 10 out of 28 m⁶A-RMRs (Figure 7). In addition, SARS-CoV infection of human airway epithelial cells at the time course of 0, 24, 36, 48, 60, 72, 84, and 96 hours (h) [64] after infection resulted in upregulation of two to nine out of 27 m⁶A-RMRs and downregulation of zero to eight out of 27 m⁶A-RMRs. H1N1 influenza virus infection of human airway epithelial cells at the time course of 0, 6, 12, 18, 24, 36, and 48 hours after infection resulted in upregulation of one to eight out of 27 m⁶A-RMRs and downregulation of five to 11 out of 27 m⁶A-RMRs. Moreover, infections of Calu-3 cells (a non-small-cell lung cancer cell line) with four different strains of influenza viruses (H7N9, H7N7, H5N1, and H3N2) [65] resulted in upregulation of four to 16, one to nine, two to 12, and one to eight out of 28 m⁶A-RMRs, respectively, and downregulation of three to 11, four to 17, three to 14, and four to 16 out of 28 m⁶A-RMRs, respectively. Taken together, these results have demonstrated that (1) MERS-CoV infections in human microvascular endothelial cells modulate the highest numbers of m⁶A-RMRs (upregulated 20 and downregulated four at the time course of 36 hours) among the three types of viral infections examined; (2) similar to MERS-CoV infection, SARS-CoV infections in human airway epithelial cells upregulated more m⁶A-RMRs than downregulated but significantly less than that in MERS-CoV infections; (3) influenza virus (H1N1) infection in human airway epithelial cells downregulated more m⁶A-RMRs than upregulated; and (4) novel avian-origin influenza virus (IAV) H7N9 infections at 24-hour postinfection upregulated 16 and downregulated six out of 28 m⁶A-RMRs, which are significantly different from that of upregulation of nine, 12, and eight and downregulation of 17, 14, and 13 out of 28 m⁶A-RMRs by highly pathogenic avian-origin influenza viruses H7N7, H5N1, and human seasonal H3N2 IAV, respectively. The differences in modulating the expressions of m⁶A-RMRs may be related to the virulence of the virus, virus replication, and other aspects [65].

3.6. Proinflammatory Lipid oxPAPC and NOTCH1 Knockdown Modulates m⁶A-RMRs More Than Other DAMP Stimulation of ECs including LPS, oxLDL, and IFNs; and Two m⁶A-RMRs Such as Hnrnpa2b1 and Eif3a Were Differentially Expressed in Three Aortic EC Clusters

We hypothesized that pathogen-associated molecular patterns (PAMPs)/danger-associated molecular patterns (DAMPs) modulate the expressions of m⁶A-RMRs in various endothelial cells. To examine this hypothesis, we collected eight microarray datasets from various endothelial cells stimulated by PAMPs/DAMPs. As shown in Figure 8(a), influenza virus infection of human umbilical vein endothelial cells (HUVEC) resulted in upregulation of nine out of 26 m⁶A-RMRs (with the highest upregulation of writer KIAA1429 for 10 folds) and downregulation of 16 out of 26 m⁶A-RMRs. By comparison, Kaposi’s sarcoma-associated herpesvirus (KSHV) infection of human dermal endothelial cells modulated m⁶A-RMR expressions significantly less than that of influenza virus infection, upregulating zero and downregulating two out of 22 m⁶A-RMRs. Toll-like receptor 4 (TLR4) ligand lipopolysaccharide (LPS) stimulation of human lung microvascular endothelial cells for 4, 8, and 24 hours resulted in upregulation of two to three m⁶A-RMRs (with the high upregulation of writer WTAP) and downregulation of one to two m⁶A-RMRs, respectively. In addition, cytokine interferon-α (IFNα), IFNβ, and IFNγ treatments of HUVEC led to upregulation of zero to one and downregulation of one to five out of 22 m⁶A-RMRs, respectively. The knockdown (KD) of proinflammatory NOTCH1 signaling and NOTCH1 KD plus proinflammatory cytokine interleukin-1β (IL-1β) in HUVEC led to upregulation of seven and five out of 25 m⁶A-RMRs and downregulation of five and six out of 25 m⁶A-RMRs. The differences of m⁶A-RMR expression modulation between NOTCH1 KD and NOTCH1 KD plus IL-1β may contribute to the inflammatory status of treated HUVEC [66]. In contrast, another report found that NOTCH1 is antiatherogenic; and proinflammatory lipids oxidized 1-palmitoyl-2-arachidonoyl-sn-glycero-3-phosphocholine (Ox-PAPC) decrease NOTCH1 expression in human aortic endothelial cells (HAEC) [67]. In this experimental setting, NOTCH1 KD in HAEC resulted in upregulation of five and downregulation of 10 out of 25 m⁶A-RMRs; proinflammatory lipid oxPAPC stimulation in HAEC led to upregulation of 10 and downregulation of seven out of 25 m⁶A-RMRs. It was reported that oscillatory shear (OS) present on the fibrosa stimulates fibrosa human aortic valve endothelial cells (HAVEC), which may contribute to aortic valve disease [68]. In this experimental setting, oscillatory shear on fibrosa HAVEC and ventricularis HAVEC resulted in upregulation of four and six and downregulation of two and three out of 28 m⁶A-RMRs, respectively. Moreover, mouse aortic endothelial cells (MAEC) isolated from atherogenic apolipoprotein E-deficient (ApoE^-/-) mice [3], TLR4 ligand LPS-treated MAEC, another TLR4 ligand oxidized low-density lipoprotein- (oxLDL-) stimulated MAEC [69, 70], and oxPAPC-stimulated MAEC [71] upregulated two, two, five, and eight out of 21 m⁶A-RMRs and downregulated five, six, three, and five out of 21 m⁶A-RMRs, respectively.

We then examined the expressions of m⁶A-RMRs in three mouse aortic endothelial cell (EC) clusters identified recently with single-cell RNA sequencing (scRNA-Seq) [72]. As shown in Figure 8(b), 12 out of 26 m⁶A-RMRs (46%) including Wtap, Zc3h13, Pcif1, FTO, Alkbh5, Ythdc1, YthDf2, Ythdf3, Fmr1, Prrc2a, Elavl1, and G3bp1 were in medium expression levels in almost all aortic cell populations of normal mouse aortas; and two m⁶A-RMRs such as Hnrnpa2b1 and Eif3a were differentially expressed in three aortic EC clusters, which were Cytl1⁺Gkn3⁺ endothelial cell (EC) cluster 1, Fabp4⁺Gphbp1⁺Rgcc⁺ EC cluster 2, and Ccl21a⁺Lrg1⁺ EC cluster 3 [72]. Of note, the expressions of Mettl14, Igf2bp1, and Igf2bp2 were not found.

These results conclude that (1) influenza virus infection of HUVEC results in expression modulation of most m⁶A-RMRs (25 out of 26), which were similar to that found in influenza virus infections of human airway epithelial cells and Calu-3 lung cells for 24 hours in Figure 7; (2) proinflammatory lipids oxPAPC and NOTCH1 knockdown modulate m⁶A-RMRs more than other DAMP stimulation of ECs including LPS, oxLDL, and IFNs; (3) ten m⁶A methyltransferases were not significantly modulated in MAEC from ApoE KO aorta with or without further stimulations of LPS, oxLDL, and oxPAPC, respectively (Figure 8, GSE39264); and (4) 12 out of 26 m⁶A-RMRs (46%) were in medium expression levels in almost all aortic cell populations of normal mouse aortas; and two m⁶A-RMRs including Hnrnpa2b1 and Eif3a were differentially expressed in three aortic EC clusters.

3.7. Upregulated m⁶A-RMRs Were More Than Downregulated m⁶A-RMRs in Various Cancer Types; Head and Neck Cancer, Cervical Cancer, Brain and CNS Cancer, Other Cancer, and Kidney Cancer Upregulated ≥10 m⁶A-RMRs; and IGF2BP3, G3BP1, IGF2BP2, and HNRNPC Were Upregulated in ≥10 Cancer Types

We hypothesized that the expressions of m⁶A-RMRs are modulated in various cancer cell populations. To test this hypothesis, we collected 19 cancer datasets in a comprehensive cancer gene expression database Oncomine (https://www.oncomine.org) [45] including brain and CNS cancer, head and neck cancer, esophageal cancer, gastric cancer, bladder cancer, breast cancer, cervical cancer, ovarian cancer, colorectal cancer, kidney cancer, liver cancer, lung cancer, pancreatic cancer, prostate cancer, leukemia, lymphoma, melanoma, myeloma, sarcoma, and other cancer (Figure 9(a)). As shown in Figure 9, the numbers in red indicated the numbers of studies with upregulated m⁶A-RMRs; and the numbers in blue indicated the numbers of studies with downregulated m⁶A-RMRs. The results showed that upregulated m⁶A-RMRs were more than downregulated m⁶A-RMRs in various cancer types (number of red cells is more than that of blue cells). Of note, the “Significant Unique Analyses” indicated the numbers of studies in which the analyzed genes were significantly upregulated (red) or downregulated (blue). The “Total Unique Analysis” indicated the numbers of studies in which the analyzed genes were included (p < 0.05, fold change > 2). Three m⁶A-RMRs such as METTL16, YTHDC1, and PRRC2A were not found.

We then determined the top cancer types that upregulated or downregulated m⁶A-RMRs the most. As shown in Figures 9(b) and 9(c), head and neck cancer, cervical cancer, brain and CNS cancer, other cancer, and kidney cancer upregulated ≥10 m⁶A-RMRs. In contrast, breast cancer, ovarian cancer, kidney cancer, esophageal cancer, leukemia, and lymphoma downregulated ≥four m⁶A-RMRs. In addition, lymphoma, leukemia, other cancer, sarcoma, and brain and CNS cancer modulated ≥five m⁶A-RMRs in uncertain manners. We also determine the top m⁶A-RMRs that upregulated or downregulated in cancer types the most. As shown in Figures 9(d) and 9(e), IGF2BP3, G3BP1, IGF2BP2, and HNRNPC were upregulated in ≥10 cancer types. In contrast, ZC3H13, IGF2BP1, WTAP, FTO, and YTHDC2 were downregulated in ≥four cancer types. In addition, G3BP2, WTAP, PCIF1, HNRNPA2B1, EIF3A, and IGF2BP2 were modulated in ≥three cancer types in uncertain manners.

The results indicated that most m⁶A-RMRs were downregulated in acute inflammatory diseases, while in cancer, the upregulated m⁶A-RMRs were more than downregulated m6A-RMRs. This situation is because acute inflammation is a part of an innate immune system reaction that can be produced by various factors, such as signaling pathways of the receptors for danger-associated molecule patterns (DAMPs/PAMPs) derived from pathogens, viruses, bacteria, and toxic substances as well as cytokine signals and stress hormone signal. These factors may activate the interactions and cross-talks among the receptors of cytokines, viruses, DAMPs, or PAMPs and promote the migration of macrophages or neutrophils to the area of inflammation [73]. However, during cancer occurs, tumor cells will generate tumor antigens that activate adaptive immune responses and induce T cells and B cells. T cell activation could cause a variety of immune signaling, including T cell antigen receptor signaling, costimulation signaling, immune checkpoint coinhibition signaling, and cytokine signaling. The composition of signaling receptors and pathways is different between acute inflammation and cancer. Therefore, RNA methylation and its transcription, splicing, and mRNA stability are also different under these two situations.

We further hypothesized that m⁶A-RMRs were differentially expressed as different cancer progression. To examine this hypothesis, the expression matrix GSE114783 was downloaded from the NIH-NCBI-GeoDataset database (https://www.ncbi.nlm.nih.gov/gds); and the expressions of m⁶A-RMRs were screened. The heat map was generated from Cluster web tool [45]. As shown in Supplementary figure 1A, the results showed that the expressions of m⁶A-RMRs were different among healthy controls, hepatitis B virus carriers, and the patients with liver cirrhosis and hepatocellular carcinoma, especially genes in the red box; compared with healthy controls, there were more upregulated m⁶A-RMRs in patients with chronic hepatitis B (CHB) group than those in healthy controls. Potentially, due to the tumor heterogeneity or the small sample sizes, there were no differentially expressed m⁶A-RMRs in the hepatocellular carcinoma group compared with healthy controls. In addition, as shown in Supplementary figure 1B, using the same method as Supplementary figure 1A, the expressions of m⁶A-RMRs were analyzed in human preinvasive and invasive cervical squamous cell carcinomas and normal cervical epithelia. The differential expressions of m⁶A-RMRs were shown in the red box. The 22 m⁶A-RMRs were examined in these three groups. The results showed that the expressions of WTAP were downregulated; and the expressions of four m⁶A-RMRs such as HNRNPA2B1, IGF2BP2, FMR1, and HNRNPC were upregulated in patients with preinvasive and invasive cervical squamous cell carcinomas compared with normal controls. Moreover, as shown in Supplementary figure 1C, the expressions of m⁶A-RMRs were analyzed in prostate cancer from benign prostatic hyperplasia to metastatic prostate cancer. The expressions of RBMX were gradually increased from prostatic intraepithelial neoplasia to hormone-naïve prostate cancer compared with normal samples. There were more obvious differential expressions of m⁶A-RMRs between adjacent prostate cancer samples and prostate carcinoma samples. The expressions of two m⁶A-RMRs such as EIF3A and RBMX were upregulated in prostate carcinoma samples compared with normal prostate epithelium-adjacent samples.

Finally, we examined the expressions of 29 m⁶A-RMRs [74] in single-cell RNA-Seq data. As shown in Supplementary figure 2A, in a single-cell RNA-seq analysis of astrocytoma dataset [75], 23 out of 28 m⁶A-RMRs (82%) (except ALKBH5, IGF2BP1, IGF2BP2, IGF2BP3, and G3BP2) were in the medium to high expression levels in malignant cells of astrocytoma. In addition, as shown in Supplementary figure 2B, in a melanoma intratumor heterogeneity dataset [76], 24 out of 29 m⁶A-RMRs (83%) (except RBM15B, IGF2BP1, IGF2BP3, HNRNPA2B1, and HNRNPC) were in the medium to high expression levels in cancer-associated fibroblasts (CAF) of melanoma.

Taken together, these results have demonstrated that first, upregulated m⁶A-RMRs were more than downregulated m⁶A-RMRs in various cancer types; second, head and neck cancer, cervical cancer, brain and CNS cancer, other cancer, and kidney cancer upregulated ≥10 m⁶A-RMRs; third, IGF2BP3, G3BP1, IGF2BP2, and HNRNPC were upregulated in ≥10 cancer types; fourth, four m⁶A-RMRs such as HNRNPA2B1, IGF2BP2, FMR1, and HNRNPC were upregulated in patients with preinvasive and invasive cervical squamous cell carcinomas; and the expressions of two m⁶A-RMRs such as EIF3A and RBMX were upregulated in prostate carcinoma samples; and fifth, 82-83% of 29 m⁶A-RMRs were in the medium to high expression levels in astrocytoma and melanoma.

3.8. M1 Macrophage Polarization Upregulates Seven m⁶A-RMRs including METTL14, WTAP, PCIF1, HNRNPA2B1, IGF2BP2, FMR1, and G3BP1, among Which WTAP and IGF2BP2 May Not Only Promote M1 Polarization but Also Inhibit M2 Polarization

We hypothesized that macrophage polarization from M0 into type 1 proinflammatory macrophages (M1) or M2 anti-inflammatory macrophages modulate the expressions of m⁶A-RMRs. To examine this hypothesis, we collected a microarray dataset GSE85346 from the NCBI-GeoDataset database. As shown in Figure 10, M1 macrophage polarization upregulated seven m⁶A-RMRs including METTL14, WTAP, PCIF1, HNRNPA2B1, IGF2BP2, FMR1, and G3BP1 whereas M2 polarization into M2a, M2b, or M2c downregulated three m⁶A-RMRs including WTAP, IGF2BP2, and IGF2BP3. These results suggest that these M1 polarization upregulated seven m⁶A-RMRs may promote proinflammatory M1 polarization. In contrast, two M1 polarization-upregulated and M2 polarization-downregulated m⁶A-RMRs such as WTAP and IGF2BP2 may not only promote M1 polarization but also inhibit M2 polarization.

Macrophages are diverse immune cells polarized by numerous stimuli, resulting in a wide range of traits and functions. The polarized process from M0 to M1 is induced by the stimulations of TLR ligands or Th1 cytokines, like TNF-a, IFN-γ, and CSF2 signaling. Under inflammation, these macrophage polarization signals are specific and robust [77], whereas, during tumorigenesis, so many factors can be involved. Moreover, different cancers are related to different cell types. A great amount of different signaling cross-talks may cause chaos signaling [78]. Consequently, the signaling that modulates the upregulation of WTAP could be diffused by the chaos cross-talking signaling.

3.9. The 86% of m⁶A-RMRs Were Differentially Expressed in the Six Spleen Treg Clusters; and 8 Out of 12 Treg Signature Genes including FOXO1, HDAC9, Dicer, BLIMP1, GATA3, EP300, BCL6, and PPAR Significantly Regulated the Expressions of m⁶A-RMRs

Our and others’ reports demonstrated that CD4⁺Foxp3⁺ regulatory T cells play significant roles in suppressing immune responses, inflammations, atherosclerosis [13, 25–27, 70, 79–83], and antitumor immune reactions [84] and promoting tissue regeneration. We hypothesized that m⁶A-RMRs are modulated in Treg differentiation and Treg responses to tissue microenvironment. As shown in Figure 11(a), Treg upregulated six m⁶A-RMRs such as ZC3H13, YTHDC1, YTHDC2, IGF2BP1, IGF2BP3, and HNRNPC in three Treg datasets compared to that of CD4⁺Foxp3^- effector T cells.

In addition, a recent report with single-cell RNA sequencing (scRNA-Seq) characterizes spleen Treg into six clusters including S100a4highS100a6high cluster 1 (activated), Itgb1high cluster 2 (activated), Dusp2highNr4a1highFoxp3highIL2rahigh cluster 3 (activated), Ikzf2highFoxp3high cluster 4 (resting), Bach2high cluster 5 (resting), and Satb1highSellhigh cluster 6 (resting) [85]. We hypothesized that the m⁶A-RMRs are differentially expressed in these six Treg clusters. As shown in Figure 11(b), four m⁶A-RMRs including PCIF1, METTL3, HNRNPA2B1, and EIF3A were upregulated in the cluster 1 Treg; two m⁶A-RMRs including RBM15 and YTHDC2 were upregulated in the cluster 2 Treg; seven m⁶A-RMRs including ALKBH5, YTHDF3, RBM15B, HNRNPC, ZC3H13, FTO, Wand TAP were upregulated in cluster 3 Treg; four m⁶A-RMRs including IGF2BP2, RBMX, G3BP2, and PRRC2A were upregulated in cluster 4 Treg; three m⁶A-RMRs including METTL14, YTHDF2, and IGF2BP3 were upregulated in cluster 5 Treg; and five m⁶A-RMRs including G3BP1, ELAVL1, IGF2BP1, FMR1, and METTL16 were upregulated in cluster 6 Treg. Therefore, 25 out of 29 m⁶A-RMRs (86%) were differentially expressed in the six spleen Treg clusters.

Moreover, we hypothesized that Treg signature genes play significant roles in regulating the expressions of m⁶A-RMRs. To examine this hypothesis, we collected the microarray datasets of 12 Treg signature gene deficiencies. As shown in Figure 11(c), the deficiencies of type 2 T helper (Th2) transcription factor (TF) GATA binding protein 3 (GATA3), follicular Th cell (Tfh) TF B cell lymphoma 6 protein (BCL6), histone deacetylase 6 (HDAC6), HDAC9, microRNA maturation enzyme DICER, interleukin-2 receptor b chain (IL2rb), B-lymphocyte-induced maturation protein 1 (BLIMP1, PR domain Zinc finger protein 1), peroxisome proliferator activated receptor gamma (PPARg) (visceral adipose tissue, VAT), PPARg (lymph nodes, LN), IKAROS family zinc finger 4 (EOS), Foxhead box 1 (FOXO1), and histone acetyltransferase P300 (EP300) modulated and upregulated 1, 6, 0, 3, 5, 0, 6, 3, 1, 1, 6, and 5 m⁶A-RMRs, respectively, and downregulated 6, 0, 2, 5, 3, 1, 3, 2, 0, 0, 11, and 2 m⁶A-RMRs, respectively. The 12 Treg signature genes were ranked based on the total modulation numbers when they were deficient, FOXO1 (17) > HDAC9 (8) = Dicer (8) = BLIMP1 (8) > GATA3 (7) = EP300 (7) > BCL6 (6) > PPARg (VAT) (5) > HDAC6 (2) > IL − 2RB (1) = PPARG (LN) (1) = EOS (1). It has been reported that mechanistic target of rapamycin kinase (mTOR)/interferon regulatory factor 4 (IRF4)/GATA3 promotes IL-4⁺ Th2-like Treg; phosphatidylinositol-4,5-bisphosphate 3-kinase (PI3K)/AKT serine/threonine kinase 1 (Akt)/Foxo, Th1 TF T-box transcription factor 21 (T-bet), hypoxia inducible factor 1 subunit alpha (HIF-1a), glycolysis promote interferon-gamma (IFNg)⁺ Th1-like Treg, microbiota/MAF BZIP transcription factor (c-Maf)/signal transducer and activator of transcription 3 (Stat3)/RAR-related orphan receptor C (RORgt), aryl hydrocarbon receptor (AhR) promote IL-17+ Th17 like Treg, and mTOR/Stat3/hepatocyte nuclear factor 1-alpha (TCF1) promote Bcl6+ follicular Treg (Tfr).

Our results have shown that (1) Treg upregulated six m⁶A-RMRs including ZC3H13, YTHDC1, YTHDC2, IGF2BP1, and HNRNPC; (2) 25 out of 29 m⁶A-RMRs (86%) were differentially expressed in the six spleen Treg clusters, suggesting that m⁶A-RMRs play significant roles in reprogramming Treg activation status (clusters 1-3) and resting status (clusters 4-6); and (3) the transcription data from 12 Treg signature gene deficiencies suggest that eight out of 12 Treg signature genes including FOXO1, HDAC9, Dicer, BLIMP1, GATA3, EP300, BCL6, and PPARγ regulate the expressions of m⁶A-RMRs significantly; and m⁶A-RMRs play important roles in reshaping Treg stability and plasticity.

3.10. Immune Checkpoint Receptors TIM3, TIGIT, PD-L2, and CTLA4 Play Significant Roles in Modulating the Expressions of m⁶A-RMRs; and Inhibition of Costimulation Receptor CD40 with Anti-CD40 Antibody Significantly Upregulates the Expressions of m⁶A-RMRs

We recently reported that cosignaling receptors localized on cell membrane including costimulation receptors and immune checkpoint receptors (coinhibition receptors) serve as a prototypic cell-cell contact interaction and regulate T cell plasticity, immune tolerance, cellular physiology, and sterile inflammatory disorders [86]. We hypothesized that cosignaling receptors regulate the expressions of m⁶A-RMRs. We collected 10 cosignaling receptor deficiency/suppression microarray datasets from the NIH-NCBI GeoDataset database. As shown in Figure 12, cytotoxic T-lymphocyte-associated protein 4 (CTLA4) KO and anti-CTLA4 upregulated 2-3 m⁶A-RMRs and downregulated one m⁶A-RMR. CD80⁺ programmed cell death 1 ligand 2 (PD-L2)+ (memory B cells) [87] upregulated three and downregulated five m⁶A-RMRs. PD-L1 and anti-PD-L1 upregulated one m⁶A-RMRs and downregulated zero m⁶A-RMRs. T cell immunoglobulin and mucin domain-containing protein 3 (TIM-3) KO, CD3⁺CD8⁺TIM-3^- cells [88], and CD8⁺PD1⁺TIM3^- cells [89] upregulated 0-5 m⁶A-RMRs and downregulated 2-11 m⁶A-RMRs. T cell immunoreceptor with Ig and ITIM domain- (TIGIT-) Treg upregulated 3 m⁶A-RMRs and downregulated 13 m⁶A-RMRs. In addition to the modulation of m⁶A-RMR expressions by immune checkpoint receptors, the modulation with anti-CD40 (tumor necrosis factor receptor superfamily member 5), a costimulation receptor, resulted in upregulation of 5-12 m⁶A-RMRs and downregulation of 1-2 m⁶A-RMRs. Taken together, these results have demonstrated that immune checkpoint receptors TIM3, TIGIT, PD-L2, and CTLA4 play significant roles in modulating the expressions of m⁶A-RMRs; and inhibition of costimulation receptor CD40 with anti-CD40 antibody significantly upregulates the expressions of m⁶A-RMRs, which is associated with suppression of autoimmune responses by anti-CD40 antibody therapy [90].

3.11. Proinflammatory Cytokine Signaling Pathways Significantly Modulate the Expressions of m⁶A-RMRs, and Suppression of Proinflammatory Cytokine Pathways Upregulate More Than Downregulate the Expressions of m⁶A-RMRs

We recently reported that cytokines including IL-17 [91–93], IL-2 [26], and IL-35 [15, 94, 95] play significant roles in inflammation regulation, Treg maintenance, and various pathologies. We hypothesized that signal pathways of cytokines and cytokine receptors regulate the expressions of m⁶A-RMRs. We collected eight cytokine and cytokine receptor microarray datasets from the NIH-NCBI GeoDataset database. As shown in Figure 13, deficiencies of tumor necrosis factor-α (TNFα, TNF receptor 1-2 (TNFR1-2), interferon-γ (IFNγ), IFNγ receptor-1 (IFNγR1), IL6, IL-17 receptor (IL17R), IL18, and IL18 upregulated 3, 7, 1, 7, 2, 3, 5, and 2 m⁶A-RMRs, respectively, and downregulated 0, 3, 0, 1, 2, 2, 1, and 3 m⁶A-RMRs, respectively. These results have demonstrated that (1) proinflammatory cytokine signaling pathways significantly modulate the expressions of m⁶A-RMRs, and (2) suppression of proinflammatory cytokine pathways upregulates more than downregulates the expressions of m⁶A-RMRs.

3.12. NF-κB Components and JAK/STAT Signaling (Except STAT1) Play Important Roles in Upregulating More Than Downregulating m⁶A-RMRs; and Tumor Suppressors TP53, PTEN, and APC Play Significant Roles in Downregulating More Than Upregulating m⁶A-RMRs

We and others reported that proinflammatory/immune regulatory transcription factors (TFs), Janus kinase (JAK)/signal transducer and activator of transcription (STAT) [94], nuclear factor kappa B (NF-κB) [96], tumor protein P53 (TP53), phosphatase and tensin homolog (PTEN), and adenomatous polyposis coli (APC regulator of WNT signaling pathway, APC) [97] play significant roles in regulating inflammation, Treg maintenance, cytokine responses, tumorigenesis, and development. We hypothesized that proinflammatory and immune regulatory TFs, NF-κB, and JAK/STAT pathways and three tumor suppressors TP53, PTEN, and APC pathways regulate the expressions of m⁶A-RMRs. We collected five NF-κB subunit deficiency datasets (seven comparisons), one JAK2 deficiency dataset, one STAT1 deficiency dataset, two STAT3 deficiency datasets, six TP53 deficiency datasets (eight comparisons), five PTEN datasets, and one APC dataset from the NIH-NCBI GeoDataset database. As shown in Figure 14, deficiencies of IKK2 (cell line), IKK2 (tumor), IKK complex, IKK complex, IKK2, RELA, and Rela upregulated 4, 2, 1, 1, 5, 3, and 2 m⁶A-RMRs, respectively, and downregulated 7, 3, 7, 6, 6, 13, and 13, respectively. Of note, deficiencies of NF-κB signaling components resulted in more downregulation than upregulation of m⁶A-RMRs. Deficiencies of JAK2, STAT1, STAT3, and STA3 resulted in upregulating 1, 10, 2, and 4 m⁶A-RMR, respectively, and downregulating 5, 2, 4, and 8 m⁶A-RMRs, respectively. STAT1 deficiency upregulated more than downregulated m⁶A-RMRs, but deficiencies of JAK2 and STAT3 downregulated more than upregulated m⁶A-RMRs. Deficiencies of TP53 in eight datasets upregulated 8, 9, 9, 15, 4, 10, 3, and 4 m⁶A-RMRs, respectively, and downregulated 7, 5, 0, 0, 5, 7, 3, and 1 m⁶A-RMRs, respectively. Of note, six out of 8 TP53-deficient datasets had more upregulation than downregulation of m⁶A-RMR, suggesting that tumor progression triggered by TP53 deficiencies may require more upregulation than downregulation of m⁶A-RMRs. Deficiencies of PTEN in five datasets upregulated 10, 2, 5, 4, and 4 m⁶A-RMRs, respectively, and downregulated 0, 14, 2, 2, and 11 m⁶A-RMRs, respectively. Of note, three out of five PTEN deficient datasets had more upregulation than downregulation of m⁶A-RMR, suggesting that tumor progression triggered by PTEN deficiencies may require more upregulation than downregulation of m⁶A-RMRs. Deficiencies of APC upregulated 14 and downregulated 4 m⁶A-RMRs, respectively. These results have exhibited that (1) NF-κB signaling components play important roles in upregulating more than downregulating m⁶A-RMRs; (2) JAK/STAT signaling pathways play important roles in upregulating more than downregulating m⁶A-RMRs except STAT1; and (3) tumor suppressors TP53, PTEN, and APC play significant roles in downregulating more than upregulating m⁶A-RMRs, suggesting that upregulation of m⁶A-RMRs may favor more than inhibit tumor progression.

3.13. Methionine-Homocysteine Cycle Enzyme Mthfd1 Downregulates More Than Upregulates m⁶A-RMRs; One m⁶A Writer RBM15 and One m⁶A Eraser FTO Significantly Modulate the Expressions of m⁶A-RMRs; and H3K4-Specific Methyltransferase MLL1 and DNA Methyltransferase, DNMT1, Significantly Regulate the Expressions of m⁶A-RMRs

We reported that hyperhomocysteinemia promotes differentiation of the inflammatory Ly6C⁺ monocyte [5, 98, 99], decreased high-density lipoprotein (HDL) [100, 101], endothelial cell activation, and programmed cell death [6] and atherosclerosis [102]. We hypothesized that m⁶A-RMRs, histone lysine methylases, DNA methyltransferases, and methyl donating homocysteine-methionine metabolism cycle play significant roles in regulating the expressions of m⁶A-RMRs as a feedback mechanism. To examine this hypothesis, we collected numerous microarray datasets including a methionine diet-fed model, Mthfd1 KO, Mthfr KO, eight histone 3 lysine 4 (H3K4) methylase KOs, and eight DNA methyltransferase KOs from the NIH-NCBI-GeoDataset database. As shown in Figure 15(a), the methionine diet-fed model upregulated three m⁶A-RMRs and downregulated one m⁶A-RMR; deficiencies of methylenetetrahydrofolate dehydrogenase (Mthfd1) and methylenetetrahydrofolate reductase (Mthfr) resulted in upregulation of one and zero m⁶A-RMR, respectively, and downregulation of two and one m⁶A-RMRs, respectively. Deficiencies of m⁶A-RMRs RBM15 and FTO upregulated 1 to 11 m⁶A-RMRs and downregulated 3-5 m⁶A-RMRs. As shown in Figure 15(b), deficiencies of histone lysine 4 methylase in eight datasets upregulated one to 13 m⁶A-RMRs and downregulated zero to 11 m⁶A-RMRs. Among the eight comparison datasets, three datasets of deficiencies of H3K4-specific methyltransferase MLL1 [103] modulated m⁶A-RMRs the most, with 10 out of 29, 20 out of 29, and 16 out of 29 m⁶A-RMRs, respectively. Moreover, deficiencies of DNA methyltransferase 1 (DNMT1) in GSE54841, GSE27434, GSE44277, and GSE86147; DNMT3A at GSE54841 and GSE42304; and DNMT3B at GSE54841 and GSE75401 upregulated 0, 1, 0, 3, 0, 1, 0, and 0 m⁶A-RMRs, respectively, and downregulated 2, 8, 12, 15, 1, 3, 1, and 1 m⁶A-RMRs, respectively. Among the eight datasets of DNA methyltransferase deficiencies, DNMT1 deficiencies modulated the most with 9, 12, and 18 m⁶A-RMRs except GSE54841 (Figure 15(c)). Taken together, these results have demonstrated that (1) methionine diet-fed model and methionine-homocysteine cycle enzyme Mthfr do not significantly regulate the expressions of m⁶A-RMRs but methionine-homocysteine cycle enzyme Mthfd1 downregulates more than upregulates m⁶A-RMRs; (2) one m⁶A writer RBM15 and one eraser FTO significantly modulate the expressions of m⁶A-RMRs, suggesting a feedback mechanism for regulation of m⁶A-RMR expressions; (3) H3K4-specific methyltransferase MLL1 significantly regulates the expressions of m⁶A-RMRs; and (4) DNA methyltransferase, DNMT1, modulates the expressions of m⁶A-RMRs.

3.14. The 18 Out of 165 ROS Regulators (11%) Were Modulated by Four Human m⁶A Writers KIAA1429, METTL14, METTL3, and WTAP, and Gene KO Data Showed That 40 Out of 165 ROS Regulators (24%) Were Modulated by m⁶A Eraser FTO and Two m⁶A Writers METTL3 and WTAP

Our and others’ reports showed that reactive oxygen species (ROS) play significant roles not only in mediating endothelial cell activation and vascular inflammation [104] but also in sensing metabolic homeostasis and stress in organelle metabolic process as an integrated system. An important question remained whether m⁶A-RMRs regulate ROS regulators. We hypothesized that m⁶A-RMRs regulate the expressions of ROS regulators. To examine this hypothesis, the ROS regulators were the targets of RNA methylation regulators in human cell studies as shown in Table 5. The TREW data (Target of m⁶A Readers, Erasers, and Writers) was download from Met-DB v2.0 (MeT-DB V2.0: the MethylTranscriptome DataBase Version 2.0 http://180.208.58.19/metdb_v2/html/index.php) [105]. The 165 ROS regulators were examined as we reported. The result showed that 18 out of 165 ROS regulators (11%) (F2RL1, PDK4, TIGAR, BCL2, SESN2, GNAI2, DDIT4, SH3PXD2A, FOXM1, AATF, TGFB1, TSPO, G6PD, GNAI3, and CYP1B1) were modulated by four m⁶A writers KIAA1429, METTL14, METTL3, and WTAP (several ROS regulators were modulated in more than one position in chromosome or by more than one RNA methylation regulators) (Table 5). The deficiencies of four m⁶A writers (KIAA1429, METTL14, METTL3, and WTAP) downregulated the m⁶A modification of ROS regulators. As shown in Supplementary Table 4, the expressions of ROS regulators were also the targets of RNA methylation regulators in mouse cell studies. TREW (Target of m⁶A Readers, Erasers and Writers) was download from Met-DB v2.0 (MeT-DB V2.0: the MethylTranscriptome DataBase Version 2.0 http://180.208.58.19/metdb_v2/html/index.php) [105]. The 165 ROS regulators were examined; and the result showed that 40 out of 165 ROS regulators (24%) were modulated by m⁶A eraser FTO and two m⁶A writers METTL3 AND WTAP. The deficiency of m⁶A eraser FTO upregulated the m⁶A modification of ROS regulators, and the deficiencies of two m⁶A writers METTL3 or WTAP downregulated the m⁶A modification of ROS regulators.