Abstract

   Colonic mucosal barrier dysfunction is one of the major causes of
   inflammatory bowel disease (IBD). However, the mechanisms underlying
   mucosal barrier dysfunction are poorly understood. N^6-methyladenosine
   (m^6A) mRNA modification is an important modulator of
   epitranscriptional regulation of gene expression, participating in
   multiple physiological and pathological processes. However, the
   function of m^6A modification in colonic epithelial cells and stem
   cells is unknown. Here, we show that m^6A modification is essential for
   maintaining the homeostatic self-renewal in colonic stem cells.
   Specific deletion of the methyltransferase 14 (Mettl14) gene in mouse
   colon resulted in colonic stem cell apoptosis, causing mucosal barrier
   dysfunction and severe colitis. Mechanistically, we revealed that
   Mettl14 restricted colonic epithelial cell death by regulating the
   stability of Nfkbia mRNA and modulating the NF-κB pathway. Our results
   identified a previously unidentified role for m^6A modification in
   colonic epithelial cells and stem cells, suggesting that m^6A
   modification may be a potential therapeutic target for IBD.
     __________________________________________________________________

   m^6A RNA modification plays important roles in the homeostatic
   self-renewal of colonic stem cells and epithelial cells.

INTRODUCTION

   The intestinal tract is the most rapidly self-renewing tissue in adult
   mammals. It functions both in nutrient absorption and as a barrier that
   protects against environmental insults. The single-layered, repetitive
   crypt-villus unit ([54]1) of the intestinal epithelium consists of
   constantly proliferating stem cells and postmitotic cells and is
   critical for the maintenance of intestinal function. At the base of the
   crypt, continuously proliferating stem cells divide into
   transit-amplifying (TA) cells. TA cells, after dividing four to five
   times ([55]2), differentiate into postmitotic absorptive and secretory
   epithelial cells. Absorptive enterocytes, also called colonocytes in
   the colon, characterized by a brush border, are responsible for
   nutrient uptake. Goblet cells, Paneth cells, enteroendocrine cells, and
   tuft cells belong to the secretory cell class. Goblet cells secrete
   mucins, which are highly glycosylated proteins that protect against
   mechanical stress ([56]3). Paneth cells reside only in the base of the
   crypt in the small intestine and form a niche essential for the
   maintenance of stem cells ([57]4), while in the colon deep crypt
   secretory (DCS) cells, which express regenerating family member 4
   (REG4) ([58]5), serve the same role as Paneth cells in the small
   intestine. Enteroendocrine cells are scattered throughout the
   intestinal mucosa and critically regulate metabolism and digestion
   ([59]6). Tuft cells, a rare cell type ([60]7), are involved in immune
   regulation. The constant supply of all types of intestinal epithelial
   cells relies on a complex regulation of intestinal stem cell
   self-renewal and differentiation. Despite the robust regeneration of
   intestinal stem cells, intestinal homeostasis can be disrupted upon
   severe acute and chronic damage, which can lead to inflammatory bowel
   disease (IBD) ([61]8), such as Crohn’s disease and ulcerative colitis.
   The mechanism of the rapid and proper restoration of the integrity of
   the epithelial barrier is largely unknown.

   The nuclear factor κB (NF-κB) transcription factor complex includes
   five members, namely, RelA (p65), RelB, c-Rel, NF-κB1 (p50), and NF-κB2
   (p52). These proteins play important function in the immune response
   and cell survival by targeting proinflammatory genes, antiapoptotic
   proteins, and angiogenesis regulators. In resting cells, the p65-p50
   complex ([62]9) is inactivated by binding one of the inhibitory κB
   (IκB) proteins, such as IκBα (also called NFKBIA), IκBβ, or IκBε
   ([63]10). When stimulated by bacterial products, such as
   lipopolysaccharide or host cytokines like tumor necrosis factor α
   (TNFα), IκBs are phosphorylated by the IκB kinase (IKK) complex and
   degraded by the proteasome, leading to the release of p65 and p50.
   Nuclear translocation of p65 and p50 then triggers NF-κB–related gene
   transcription ([64]11). Depletion of NF-κB in immune cells leads to
   disruption of the immune response ([65]12), while inhibition of NF-κB
   in intestinal epithelial cells activates cell death pathways ([66]13).
   Furthermore, in intestinal epithelial cells, knockout (KO) of IKKβ
   failed to prevent dextran sulfate sodium (DSS)–induced inflammation and
   resulted in apoptosis, while in myeloid cells the loss of IKKβ
   decreased the expression of proinflammatory cytokines ([67]14–[68]16).
   In addition, mutation of the constitutively expressed form of IκB, also
   called NFKBIA, enhanced TNF-induced cell death in intestinal epithelial
   cells ([69]17). Thus, in the intestine, NF-κB signaling has a dual
   function related to both cell survival and inflammation.

   N^6-methyladenosine (m^6A) RNA modification is one of the most abundant
   and extensively distributed mRNA modifications ([70]18). It modulates
   RNA metabolism, including decay, translation, localization, and
   splicing ([71]19, [72]20). m^6A modification of mRNA is performed by a
   methyltransferase complex composed of methyltransferase-like 3
   (METTL3), METTL14, and Wilms’ tumor 1–associating protein (WTAP). m^6A
   modification is erased by two demethylases, which are fat mass- and
   obesity-associated protein (FTO) ([73]21) and α-ketoglutarate–dependent
   dioxygenase AlkB homolog 5 (ALKBH5) ([74]22). In the methyltransferase
   complex, METTL3 is the only catalytically active subunit, METTL14 is
   essential for substrate recognition, and WTAP interacts with METTL3 and
   METTL14 to regulate m^6A deposition ([75]23, [76]24). Emergent evidence
   has revealed that m^6A modification is indispensable for the
   development of embryonic stem cells ([77]25), hematopoietic stem cells
   ([78]26), T cells ([79]27, [80]28), and neural cells ([81]29).
   Moreover, dysfunctional m^6A modification is involved in the
   development of human acute myeloid leukemia ([82]30) and other cancers
   ([83]31). However, how m^6A modification regulates the function of
   colonic epithelial cells is still unclear. Although previous work
   demonstrated that an m^6A reader protein, YTH domain-containing family
   protein 1 (YTHDF1), drives Wnt-related intestinal regeneration and
   tumorigenesis; however, YTHDF1 is dispensable for normal intestinal
   development in mice and YTHDF1 depletion does not affect body weight
   and crypt-villus architecture ([84]32). Other reports published
   recently showed the promotive role of METTL3 in colorectal cancer (CRC)
   ([85]33, [86]34); however, the significance of m^6A modification in
   colonic epithelial cells under physiological conditions remains to be
   determined.

   Here, we report that, in the steady state, m^6A modification is
   indispensable for the maintenance of colonic epithelial homeostasis.
   Mice with colon-specific KO of Mettl14 showed disturbed integrity of
   the colonic epithelial barrier, resulting in severe colitis. The mRNA
   of Nfkbia, the NF-κB inhibitor, is the direct target of m^6A
   modification. Thus, depletion of m^6A modification led to enhanced
   stability of Nfkbia mRNA and accumulation of NFKBIA protein, which
   down-regulated the NF-κB–mediated antiapoptotic pathway and promoted
   colonic cell death.

RESULTS

Deletion of Mettl14 in colonic epithelial cells leads to a wasting phenotype
and spontaneous colitis

   We first investigated the dynamic changes in m^6A levels in colonic
   epithelial cells during mouse development. The global m^6A level in
   intestinal mRNAs was lower in adult 8-week-old mice than at 2 weeks
   after birth (fig. S1A), implying an important function of m^6A
   modification in the development of colonic epithelial cells. To
   investigate the genetic requirement for Mettl14 in colonic epithelial
   cells, we conditionally deleted Mettl14 by generating Villin-cre^+;
   Mettl14^f/f (hereafter called Mettl14 KO) mice (fig. S1B), in which the
   Vil1 gene promoter–driven Cre recombinase would specifically delete the
   floxed Mettl14 gene in all intestinal epithelial cells. The Mettl14 KO
   efficiency was confirmed by genotyping, quantitative reverse
   transcription polymerase chain reaction (qRT-PCR) (fig. S1C), and
   immunohistochemistry (fig. S1, D and E). Consequently, the total m^6A
   level in mRNAs in the Mettl14 KO group was decreased in comparison with
   Mettl14^f/f littermates [hereafter called Mettl14 wild type (WT)], as
   verified by m^6A dot blot (fig. S1F).

   At 2 weeks of age, Mettl14 KO mice were distinguishable from Mettl14 WT
   due to their notable growth retardation ([87]Fig. 1A). From 2 to 8
   weeks, Mettl14 KO mice demonstrated lower weight gains ([88]Fig. 1B)
   than Mettl14 WT mice. This difference was associated with a shorter
   colon in the Mettl14 KO mice ([89]Fig. 1C). Furthermore, Mettl14 KO
   mice developed diarrhea as early as at 2 weeks of age ([90]Fig. 1D).
   The decrease in the global level of m^6A mRNA modification in colonic
   epithelial cells from infancy to adulthood suggests that m^6A
   modification deficiency has a significant impact on the developing
   colon. Furthermore, mouse survival was monitored. We observed a
   significantly lower survival in Mettl14 KO mice, and the mortality
   reached its peak in mice around 3 or 4 weeks old, ranging from 42.86 to
   64.29% ([91]Fig. 1E). However, no mice died after the age of 4 weeks,
   consistent with the data showing that Mettl14 KO mice have the greatest
   suppression of weight gain at around 3 or 4 weeks of age. Detailed
   autopsy of the Mettl14 KO mice revealed a shorter colon that was filled
   with bloody feces ([92]Fig. 1F). To further explore the impact of
   Mettl14 depletion on the changes in the morphology of intestinal
   epithelial cells, hematoxylin and eosin (H&E) and Muc2 staining was
   performed in mice of different ages. Aggravated deterioration of the
   colon was detected from 2 to 8 weeks of age, which was characterized by
   the loss of goblet cells, crypt architectural distortion, inflammatory
   cell infiltration, and epithelial ulceration ([93]Fig. 1, G to J).
   However, we found no significant changes in the length of small
   intestine, the length of villus, and the integrity of crypts, nor with
   the relative expression of specific markers of different cell subsets
   between the two groups (fig. S2, A to E). These results reveal that
   METTL14-mediated m^6A modification plays an important role in the
   development of colonic epithelial cells, and depletion of Mettl14
   triggers a wasting phenotype in the infant stage followed by
   spontaneous colitis.

Fig. 1. Mettl14 depletion in intestinal epithelial cells triggers a wasting
phenotype and spontaneous colitis.

   [94]Fig. 1.
   [95]Open in a new tab

   (A) Representative image of Mettl14 WT and Mettl14 KO mice at 2 weeks
   of age. (B and C) Body weight [(B), n = 5 to 8 per group] and colon
   length [(C), n = 6 to 10 per group] of Mettl14 WT and Mettl14 KO mice
   from 2 to 8 weeks. (D) Fecal water content of Mettl14 WT and Mettl14 KO
   mice at 2 weeks of age (n = 6 per group). (E) Survival rate of Mettl14
   WT (n = 12) and Mettl14 KO (n = 14) mice was observed from 2 to 8
   weeks, and the corresponding survival curves were plotted. (F)
   Representative image of autopsies from Mettl14 WT and Mettl14 KO colon.
   (G) Representative H&E staining of proximal and distal colon sections
   from Mettl14 WT and Mettl14 KO mice at 2, 4, and 8 weeks of age (scale
   bars, 100 μm). (H) Representative Muc2 staining of proximal and distal
   colon sections from Mettl14 WT and Mettl14 KO mice at 8 weeks of age
   (scale bars, 70 μm). (I and J) Quantitative analysis of H&E score [(I),
   n = 4 to 7 per group] and the number of goblet cells [(J), n = 10 per
   group]. Data are presented as means ± SD. Two-sided Student’s t test (B
   to D, I, and J) and log-rank (Mantel-Cox) test (E) were performed (*P <
   0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001).

Mettl14 depletion leads to colonic epithelial cell death

   To address how Mettl14 deficiency affects the function of different
   types of colonic epithelial cells, we first examined the expression of
   genes encoding specific markers of total colonic epithelial cells, stem
   cells, goblet cells, and colonocytes ([96]Fig. 2A). All these marker
   genes were down-regulated in Mettl14 KO mice compared to their WT
   littermates, suggesting that the KO of Mettl14 affects all types of
   colonic epithelial cells in the 2-week-old mice. Mucus production by
   goblet cells serves as an indispensable barrier controlling the
   microbiome in the colon ([97]3), which could be specifically stained by
   Alcian blue–periodic acid–Schiff (AB-PAS) and Muc2. AB-PAS and Muc2
   staining also showed that Mettl14 KO mice exhibited a smaller number of
   goblet cells than the WT controls ([98]Fig. 2, B and C). To determine
   whether the loss of colonic epithelial cells was due to decreased cell
   survival, we performed the terminal deoxynucleotidyl
   transferase–mediated deoxyuridine triphosphate nick end labeling
   (TUNEL) cell death assay on colon sections and found higher numbers of
   TUNEL-positive cells in the Mettl14 KO group than in the Mettl14 WT
   group ([99]Fig. 2, D and E). Furthermore, the proportion of cleaved
   caspase-3–positive cells in the colon was higher in the Mettl14 KO
   group than in the Mettl14 WT group ([100]Fig. 2, F and G).
   Subsequently, we performed Ki67 staining to further evaluate whether
   Mettl14 deficiency could also affect cell proliferation. Although no
   significant difference was found in cell proliferation between the WT
   and KO groups ([101]Fig. 2, H and I), the cell apoptosis versus
   proliferation ratio was much higher in Mettl14 KO mice compared to the
   WT controls ([102]Fig. 2J). The expression of both CD3 and CD11B in the
   Mettl14 KO group was comparable to that in the Mettl14 WT group,
   excluding the presence of ongoing inflammation in the 2-week-old mice
   (fig. S3, A to D). Together, these data document that m^6A modification
   and Mettl14 are critical for the maintenance of colonic epithelial cell
   homeostasis by restricting cell death.

Fig. 2. Mettl14 depletion leads to colonic epithelial cell death.

   [103]Fig. 2.
   [104]Open in a new tab

   (A) qRT-PCR analysis of gene expression in Mettl14 WT and Mettl14 KO
   mice at 2 weeks of age. Vil1, intestinal epithelial cell marker; Lgr5,
   intestinal stem cell marker; Gob5, goblet cell marker; Alpi, colonocyte
   marker (n = 4 to 5 per group). (B) Left: Representative AB-PAS–stained
   proximal and distal colon sections of Mettl14 WT and Mettl14 KO mice at
   2 weeks of age (scale bars, 100 μm). Right: Representative Muc2
   staining of proximal and distal colon sections from Mettl14 WT and
   Mettl14 KO mice at 2 weeks of age (scale bars, 70 μm). (C) Quantitative
   analysis of AB-PAS and Muc2 staining (n = 4 per group). (D and E)
   Representative TUNEL staining of proximal and distal colon sections
   from Mettl14 WT and Mettl14 KO mice at 2 weeks of age [(D), scale bar,
   100 μm] and quantitative analysis [(E), n = 7 per group].
   TUNEL-positive cell (arrows). (F and G) Representative cleaved
   caspase-3–stained proximal and distal colon sections of Mettl14 WT and
   Mettl14 KO mice at 2 weeks of age [(F); scale bar, 100 μm] and
   quantitative analysis [(G), n = 7 per group]. Cleaved
   caspase-3–positive cell (arrows). (H and I) Representative Ki67
   staining of distal colon sections from Mettl14 WT and Mettl14 KO mice
   at 2 weeks of age [(H); scale bars, 100 μm] and quantitative analysis
   [(I), n = 5 per group]. (J) Quantitative analysis of cell apoptosis
   versus cell proliferation ratio (n = 7 per group). Data are presented
   as means ± SD. Two-sided Student’s t test (A, C, E, G, I, and J) was
   performed (ns, no significance; *P < 0.05; **P < 0.01; ***P < 0.001).

Single-cell RNA sequencing reveals colonic stem cell dysfunction in Mettl14
KO mice

   To investigate the heterogeneity of colonic epithelial cells and
   identify the impact of m^6A deficiency on distinct types of epithelial
   cells, we performed single-cell RNA sequencing (scRNA-seq) analysis.
   Through sequencing 7863 epithelial cells, we identified a total of
   seven clusters of colonic epithelial cell types using t-distributed
   stochastic neighborhood embedding (t-SNE) ([105]Fig. 3A) and labeled
   each cluster by the expression of known marker genes (fig. S4, A to H),
   including stem cells, TA cells, DCS cells, tuft cells, colonocytes,
   enteroendocrine cells, and goblet cells. Subsequently, we defined a
   gene expression signature for each cell type using our scRNA-seq data,
   highlighting known markers ([106]Fig. 3B). By splitting these cells
   into Mettl14 WT and Mettl14 KO groups (fig. S5, A and B), we observed a
   difference in cellular composition between the two groups ([107]Fig.
   3C). In the Mettl14 KO group, the proportion of stem cells, goblet
   cells, and colonocytes was much decreased, while the population of TA
   cells was expanded.

Fig. 3. scRNA-seq reveals dysfunction of colonic epithelial cells in Mettl14
KO mice.

   [108]Fig. 3.
   [109]Open in a new tab

   (A) t-SNE plot of 7863 epithelial cells from Mettl14 WT and Mettl14 KO
   mouse colons of 2 weeks of age showing seven major clusters. (B) Gene
   expression heatmap of seven clusters. Rows represent signature genes,
   and columns represent different clusters. (C) Proportion of each
   epithelial cell cluster among the total epithelial cells. (D) Violin
   plots showing the entire range of Mettl3 and Mettl14 gene expression
   levels per single cell in each cluster after imputation. (E) Bubble
   heatmap showing expression of selected marker genes for each cluster.
   (F) Representative bright-field images of colonic organoids grown from
   Mettl14 WT and Mettl14 KO mice. (G) Organoid number per 100 crypts per
   well in 3D culture medium (n = 20 for each group). (H) Statistical
   analysis of relative levels of Mettl14 in surviving organoids from
   Mettl14 WT and Mettl14 KO mice (n = 4 per group). Data are presented as
   means ± SD. Two-sided Student’s t test (G and H) was performed (***P <
   0.001).

   In the colon, Reg4^+ DCS cells, instead of Paneth cells, are
   intermingled with Lgr5^+ stem cells at the crypt base and act as a stem
   cell niche in colonic crypts ([110]5). It has been reported that after
   treatment with DSS, cells expressing high levels of Kit ([111]35) and
   Reg4 convert to stem-like cells in the colon ([112]36). These findings
   imply that Reg^+ DCS cells can assume the properties of stem cells
   after stem cell loss. Consistently, the DCS cell population, marked by
   Kit and Reg4 (fig. S4, C and D), was also increased in the Mettl14 KO
   group ([113]Fig. 3C). In addition, we checked the expression levels of
   Mettl14 and Mettl3 in all the clusters and found relatively higher
   expression of both Mettl14 and Mettl3 in stem cells and TA cells
   ([114]Fig. 3D), implying that m^6A modification mainly affects the
   function of stem cells and progenitor cells. Furthermore, bubble
   heatmap analysis demonstrated a difference in the expression of marker
   genes between the Mettl14 WT and Mettl14 KO groups ([115]Fig. 3E): The
   mRNA levels of stem cell marker (Lgr5), goblet cell transcription
   factor (Spdef), enteroendocrine cell marker (Chga), and colonocyte
   marker (Alpi) were all decreased after Mettl14 depletion, which was
   consistent with the qPCR data ([116]Fig. 2A).

   As all the mature epithelial cells are derived from stem cells, we
   postulated that m^6A modification depletion mainly disrupts the
   function of the epithelial stem cells and, as a result, affects the
   development of all the matured epithelial cells. Therefore, to directly
   dissect the impact of Mettl14 depletion on the stem cell function, we
   isolated colonic crypts from both Mettl14 WT and Mettl14 KO mice and
   cultured them in a three-dimensional (3D) organoid growth medium. We
   observed that the organoid-forming efficiency of the crypts from the
   Mettl14 KO group was lower than from the Mettl14 WT group ([117]Fig. 3,
   F and G), consistent with the scRNA-seq data presented above. Then, we
   further examined the expression level of Mettl14 in organoids and found
   that Mettl14 was not completely depleted in the surviving organoids in
   the KO groups ([118]Fig. 3H), suggesting that Mettl14-deficient stem
   cells were unable to survive in organoid culture. Together, these data
   suggest that Mettl14 deficiency leads to the loss of colonic stem
   cells, inhibiting the formation and function of the mature colonic
   epithelial cells.

Mettl14 deficiency in Lgr5^+ stem cells promotes cell apoptosis and
DSS-induced colitis

   Because Lgr5 is a marker of intestinal epithelial stem cells and Lgr5
   mRNA was down-regulated, we investigated whether the m^6A machinery
   regulates colonic epithelial cell function by directly targeting Lgr5
   mRNA for m^6A-mediated degradation. However, we did not find Lgr5 mRNA
   in multiple MeRIP-seq (m^6A methylated RNA immunoprecipitation
   sequencing) databases as 3′ untranslated region (UTR) of Lgr5 mRNA does
   not have the typical m^6A modification consensus motifs ([119]37).

   To investigate the role of METTL14 in Lgr5^+ stem cells in vivo, we
   generated a mouse model with a conditional Mettl14 KO in Lgr5^+ stem
   cells by crossing Lgr5-EGFP-IRES-creERT2 (Lgr5-creERT) mice with floxed
   alleles of Mettl14 (hereafter called Mettl14^f/f Lgr5^creERT mice).
   When these engineered mice were treated with tamoxifen, the Lgr5
   promoter–driven Cre recombinase specifically deleted the floxed Mettl14
   gene in Lgr5-expressing intestinal stem cells and their derived
   epithelial cell progeny. We first determined whether Mettl14 deficiency
   in stem cells affected the survival of intestinal stem cells. After 7
   days of tamoxifen administration, more cleaved caspase-3–positive cells
   were detected in Mettl14^f/f Lgr5^creERT mice than in Mettl14^f/f mice
   ([120]Fig. 4, A and B). It is well documented that rapidly
   proliferating and regenerating intestinal stem cells ensure intestinal
   epithelial homeostasis during injury repair. To gain a better
   functional insight into the effect of Mettl14 depletion on Lgr5^+ stem
   cells, we treated the mice with DSS to accelerate the deterioration of
   intestinal epithelial cells. The Mettl14^f/f Lgr5^creERT mice developed
   more severe colitis than their littermates, as judged by weight loss
   ([121]Fig. 4C), colon length ([122]Fig. 4, D and E), and colitis
   severity score defined by endoscopy ([123]Fig. 4F). These results
   demonstrate that Mettl14 depletion in Lgr5^+ stem cells results in
   increased stem cell apoptosis and increased severity of DSS-induced
   colitis.

Fig. 4. Mettl14 deficiency in Lgr5^+ stem cells promotes cell apoptosis and
DSS-induced colitis.

   [124]Fig. 4.
   [125]Open in a new tab

   (A and B) Mettl14^f/f Lgr5^creERT and Mettl14^f/f littermates were
   treated with tamoxifen daily for 7 days and then sacrificed at day 10
   for cleaved caspase-3 staining. Representative cleaved caspase-3
   staining of distal colon sections from Mettl14^f/f Lgr5^creERT and
   Mettl14^f/f mice [(A); scale bars, 70 μm] and quantitative analysis
   [(B), n = 10 per group]. (C to F) To induce colitis, after treatment
   with tamoxifen daily for 7 days and resting for 7 days, mice were
   treated with 2% DSS water for a further 7 days. Weight loss of
   Mettl14^f/f Lgr5^creERT and Mettl14^f/f mice [(C), n = 8 per group].
   Representative photo of colon (D) and colon length measurements from
   Mettl14^f/f Lgr5^creERT and Mettl14^f/f mice on day 9 [(E), n = 8 per
   group]. Colonoscopy severity score analysis on day 7 [(F), n = 8 per
   group]. Data are presented as means ± SD. Two-sided Student’s t test
   (B) and two-tailed Mann-Whitney U test (C, E, and F) were performed (*P
   < 0.05; **P < 0.01; ****P < 0.0001).

Mettl14 regulates colonic epithelial cell homeostasis by restricting
TNF-induced cell death

   To comprehensively understand the molecular pathways by which
   Mettl14-mediated m^6A modification regulates the homeostasis of colonic
   epithelial cells, we conducted a bulk RNA-seq analysis and compared the
   gene expression of the colonic epithelial cells purified from Mettl14
   KO mice and their WT littermates. The loss of Mettl14 in epithelial
   cells resulted in an up-regulation of 1014 genes and down-regulation of
   502 genes ([126]Fig. 5, A and B). Kyoto Encyclopedia of Genes and
   Genomes (KEGG) pathway enrichment analysis showed that the
   differentially expressed genes were significantly enriched for pathways
   involved in TNF signaling and apoptosis ([127]Fig. 5C). To reveal the
   major signaling pathway affected in Mettl14 KO mice in an unbiased way,
   the bulk RNA-seq data were subjected to gene set enrichment analysis
   (GSEA). We found that the Mettl14 KO group was enriched in signaling
   pathways related to TNF-induced cell death via NF-κB inhibition
   ([128]Fig. 5D). It has been mentioned above that inhibition of NF-κB
   restricts proinflammatory signaling in myeloid cells, while the
   inhibition of NF-κB favors TNF-induced cell death in intestinal
   epithelial cells ([129]12, [130]13, [131]38). Next, we determined the
   expression levels of key genes involved in the NF-κB–mediated cell
   death pathway. Using qRT-PCR analysis, we found that Tnf, Nfkb2, and
   Nfkbia mRNA levels were markedly up-regulated in colonic epithelial
   cells in the Mettl14 KO group, while the expression of the other genes
   related to TNF signaling did not change ([132]Fig. 5E). Together, these
   data reveal that Mettl14 suppresses TNF-induced cell death via the
   NF-κB–mediated antiapoptotic pathway.

Fig. 5. m^6A modification regulates colonic epithelial cell homeostasis by
restricting TNF-induced cell death.

   [133]Fig. 5.
   [134]Open in a new tab

   (A) Volcano plot of comparative RNA-seq data between 2-week-old Mettl14
   WT and Mettl14 KO mice. The x axis specifies the log[2] fold changes
   (FC), and the y axis specifies the −log[10] P value (Mettl14 KO
   compared to Mettl14 WT). Blue dots represent down-regulated genes
   [log[2](FC) ≤ −1 and P ≤ 0.05], and red dots represent up-regulated
   genes [log[2](FC) ≥ 1 and P ≤ 0.05]. (B) Number of genes that increased
   significantly (red) and decreased (blue) in Mettl14 KO versus Mettl14
   WT colonic epithelial cells. (C) KEGG enrichment analysis of
   differentially expressed genes. (D) GSEA plot determined by RNA-seq
   profiling. Analysis completed on genes ranked by log[10] false
   discovery rate (FDR) and FC sign, with enrichment determined after 1000
   permutations. The statistics were computed using GSEA and controlled
   for multiple comparisons by FDR. NES, normalized enrichment score. (E)
   qRT-PCR was used to verify the differentially expressed genes related
   to the TNF signaling pathway (n = 6 to 9 per group). Data are presented
   as means ± SD. Two-sided Student’s t test (E) was performed (**P <
   0.01; ***P < 0.001).

Nfkbia is the direct target of m^6A modification in colonic epithelial cells

   To identify the direct target gene of m^6A modification in colonic
   epithelial cells, we cross-referenced up-regulated genes from the bulk
   RNA-seq to four reported MeRIP-seq datasets obtained from the human
   gastrointestinal tract and cell lines ([135]37). This analysis
   identified 366 overlapping genes in all datasets ([136]Fig. 6A); among
   them, Nfkbia was the only target of m^6A modification that was involved
   in NF-κB pathway. We confirmed the MeRIP-seq data using m^6A RNA
   immunoprecipitation qRT-PCR and found enrichment of m^6A binding to
   Nfkbia mRNA 3′UTR ([137]Fig. 6B). It is well known that m^6A
   modification enhances target mRNA degradation by addition of m^6A
   modification to the 3′UTR region of the gene ([138]39). Depletion of
   m^6A modification resulted in the accumulation of Nfkbia mRNA due to a
   lower rate of degradation than WT littermates, as confirmed by the RNA
   decay assay in which actinomycin D was used to block the mRNA
   transcription ([139]Fig. 6C) ([140]27). Furthermore, we found that
   Nfkbia has a broad distribution and relatively higher expression in
   stem cells and TA cells by analyzing our scRNA-seq data ([141]Fig. 6D).
   Meanwhile, the Nfkbia mRNA level was up-regulated mostly in Mettl14 KO
   stem cells and TA cells ([142]Fig. 6E). In addition, the up-regulation
   of Nfkbia mRNA expression resulted in a higher level of NFKBIA protein
   ([143]Fig. 6F). It is known that increased levels of NFKBIA protein
   inhibit the canonical activity of NF-κB and promote TNF-stimulated cell
   death ([144]17). Consistently, the expression of the proapoptotic
   protein Bax was increased, and expression of the antiapoptotic protein
   Bcl2 was decreased in Mettl14 KO mice ([145]Fig. 6F).

Fig. 6. Nfkbia is the direct target of m^6A modification in colonic
epithelial cells.

   [146]Fig. 6.
   [147]Open in a new tab

   (A) UpSet plot showed genes overlapped between the RNA-seq data and
   m^6A RIP sequencing datasets. DEGs, differentially expressed genes. (B)
   mRNA isolated from colonic epithelial cells was immunoprecipitated (IP)
   with an anti-m^6A antibody, and qRT-PCR was used to assess the mRNA
   level of Nfkbia (n = 5 per group). (C) Colonic epithelial cells were
   treated with actinomycin D for 4 hours, and Nfkbia mRNA was normalized
   for 0 hours (n = 6 per group). (D) Violin plot showing the distribution
   of Nfkbia per single cell in each cluster. (E) Bubble heatmap showing
   expression of TNF signaling pathway–related genes for each cluster. (F)
   Representative immunoblot for NFKBIA, Bcl2, and Bax in colonic
   epithelial cells from 2-week-old mice; glyceraldehyde-3-phosphate
   dehydrogenase (GAPDH) was used as a loading control. (G) Crypts
   isolated from Mettl14 WT and Mettl14 KO mice were grown in 3D organoid
   culture medium with or without RIPK1 inhibitor (GSK’963, 1 μM) (n = 5
   per group). Data are presented as means ± SD. Two-sided Student’s t
   test (B, C, and G) was performed (**P < 0.01).

   Given that TNF-induced cell death is dependent on the activity of
   receptor-interacting serine/threonine kinase 1 (RIPK1) ([148]40), we
   investigated whether inhibition of RIPK1 activity can rescue the
   phenotype. Crypts isolated from Mettl14 KO and Mettl14 WT mice were
   incubated in a 3D organoid culture medium with or without the RIPK1
   inhibitor GSK’963. We found that organoid-forming efficiency in the
   Mettl14 KO group with GSK’963 was increased compared to the Mettl14 KO
   group, while GSK’963 had no impact on the organoid formation in the
   Mettl14 WT group ([149]Fig. 6G), confirming that the Mettl14 deficiency
   that led to colonic epithelial cell death was dependent on RIPK1
   activity. Although Wnt signaling is indispensable for the
   differentiation and proliferation of intestinal stem cells, no
   differences in Wnt signaling were observed between Mettl14 WT and
   Mettl14 KO mice (fig. S6). Together, we conclude that m^6A modification
   sustains colonic epithelial cell survival by regulating the stability
   of Nfkbia mRNA and, therefore, modulates TNF-induced cell death via
   NF-κB activity.

DISCUSSION

   Our present study demonstrated that METTL14 is essential for
   maintaining colonic epithelial homeostasis by controlling epithelial
   cell death at steady state. Mettl14 KO–mediated loss of m^6A
   modification led to the accumulation of Nfkbia mRNA and protein in
   intestinal epithelial cells, promoting cell death through the
   inhibition of NF-κB signaling, thus triggering the loss of colonic stem
   cells and resultant colitis. This is a comprehensive study focusing on
   the physiological function of METTL14 and m^6A modification in colonic
   epithelial cells and stem cells.

   Although a recent study reported that the modulation of translation by
   the m^6A reader protein YTHDF1 affects intestinal tumorigenesis by
   targeting Wnt signaling ([150]32), our work has delineated the novel
   function of the m^6A “writer” METTL14 in intestinal epithelial cells
   under physiological conditions. Genetic ablation of the m^6A writer
   protein METTL14 in mice resulted in a wasting phenotype and impairment
   of epithelial development at 2 weeks of age and spontaneous colitis in
   adults. Ythdf1 KO mice did not exhibit an intestinal phenotype under
   normal conditions, while Mettl14 KO mice spontaneously developed severe
   colon damage and had a lower survival rate than Mettl14 WT mice. One
   possible explanation is that the METTL14-associated m^6A writer complex
   adds m^6A modification to a number of different target mRNAs and thus
   affects RNA metabolism, including decay, splicing, translocation, and
   translation, while Ythdf1 recognizes only a few of the m^6A-marked RNAs
   and mostly regulates the efficiency of the target RNA translation
   ([151]41). To investigate whether Mettl14 deficiency regulates
   intestinal epithelial cell differentiation, proliferation, and
   apoptosis, subsequently inducing the wasting phenotype and spontaneous
   colitis, we performed histological staining, qRT-PCR, scRNA-seq, and
   organoid culture analyses. However, we did not observe significant
   changes in the length of the small intestine, the integrity of the
   crypts, and the length of the villus, nor in the relative expression of
   specific markers of the different cell subsets in the small intestine,
   suggesting that the small intestine is not responsible for the marked
   wasting phenotype of the Mettl14 KO. The Mettl14 KO mice had a much
   higher colonic apoptotic rate and apoptosis-to-proliferation ratio than
   the WT mice, while no difference was seen in cell proliferation. In
   addition, Mettl14-deficient colonic crypts hardly survived in organoid
   culture, and a significant loss of stem cells was found in the Mettl14
   KO mice compared with the WT group. Our scRNA-seq data show that
   Mettl14 is highly expressed in stem cells and TA cells compared with
   other subsets in mice. Human Atlas studies indicate that METTL14 is
   expressed at much higher levels in differentiated cell enterocytes and
   secretory cells compared to stem cells and TA cells. While this seems
   inconsistent to Human Atlas studies, our data are based on objective
   scRNA-seq of individual mice. This could be due to species differences
   between humans and mice, or due to the sampling timing, as we used
   2-week-old pups for scRNA-seq, while the Human Atlas studies are mostly
   based on adult samples. Furthermore, Mettl14 depletion in Lgr5^+ stem
   cells was more likely to promote a steady rate of apoptosis and
   DSS-induced colitis. Together, these data provide strong evidence that
   the colonic phenotype in Mettl14 KO mice is mainly due to defects in
   the colonic stem cells and increased intestinal epithelial cell
   apoptosis.

   We then explored the mechanism underlying the effects of the
   Mettl14-mediated m^6A modification in colonic epithelial cell
   homeostasis and apoptosis. We found that Mettl14 deficiency in
   intestinal epithelial cells promotes TNF-induced cell death by
   suppressing NF-κB–mediated antiapoptotic pathway. It has been shown
   that TNF binds to TNF receptor 1 (TNFR1) and forms the TNFR1-associated
   signaling complex by recruiting cytoplasmic adaptor molecules, such as
   TNF receptor–associated death domain (TRADD), TNF receptor–associated
   factor 2 (TRAF2), RIPK1, and cellular inhibitors of apoptosis 1 and 2
   (c-IAP1/2). The TNFR1-associated signaling complex rapidly activates
   IKK and degrades NFKBIA, and thus induces the activation of NF-κB
   signaling, which promotes survival and the inflammatory response.
   Inhibition of NF-κB by overexpression of nondegradable NFKBIA favors
   RIPK1-dependent cell death ([152]17, [153]40). Similarly, our data show
   a much higher level of NFKBIA in Mettl14 KO mice and that Mettl14
   deficiency led to colonic epithelial cell death in a manner dependent
   on RIPK1 activity.

   Several studies on METTL14 suggest that its role in CRC is as an
   antitumor protein ([154]42, [155]43), implying that METTL14 could
   inhibit intestinal epithelial cell proliferation, while our data mainly
   attributed METTL14 function to maintain intestinal epithelial cell
   survival. There could be several possible reasons: (i) The other
   studies mainly draw their conclusion based on in vitro studies with CRC
   cell lines, which were well known to contain a lot of gene mutations,
   such as APC, KRAS, BRAF, and TP53, that might interfere or synergize
   with METTL14 and led to different conclusions, and (ii) m^6A is
   involved in almost every process of cell development, proliferation,
   apoptosis, and carcinogenesis. METTL14 may play completely different
   roles by different mechanisms in maintaining the homeostasis of colonic
   epithelial cells and in the occurrence and development of CRC. It is
   well known that Wnt signaling modulates the self-renewal and
   differentiation of intestinal stem cells ([156]44). Han et al.
   ([157]32) found that Ythdf1 amplified Wnt signaling by targeting mRNA
   translation of Tcf7l2 in colon tumors. However, Wnt signaling was not
   perturbed in Mettl14 KO colonic epithelial cells (fig. S6), further
   indicating that Mettl14 does not affect the differentiation and
   proliferation of stem cells. Another report showed that METTL3
   facilitated the development of CRC by activating the m^6A-GLUT1-mTORC1
   axis ([158]34), and transgenic Mettl3^+/− mice had fewer tumors after
   azoxymethane/DSS treatment compared to their WT littermates. Our work
   focused on the specific function of METTL14 in colonic epithelial cells
   under normal conditions by analyzing the phenotype of Mettl14 KO mice.
   Moreover, m^6A modification regulates chromosome-associated regulatory
   RNAs, modulating both the chromatin status and transcription ([159]45);
   this interaction adds further complexity to the regulatory network
   downstream of m^6A. Thus, further study will be informative in
   understanding the molecular mechanism by which different modulators of
   m^6A mRNA modification regulate the function and development of
   intestinal epithelial cells.

   Collectively, our work demonstrates the critical role of Mettl14 in
   maintaining the homeostasis of colonic epithelial cells. Deficiency of
   METTL14 and m^6A modification in mice lead to severe colitis, though
   inhibiting the NF-κB–mediated antiapoptotic pathway, suggesting a
   potential therapeutic strategy for colitis intervention by modulating
   METTL14 and the NF-κB signaling machinery.

MATERIALS AND METHODS

Mice

   Mettl14^f/f mice were generated as previously described with
   CRISPR-Cas9 technology by insertion of two loxp sites into Mettl14
   genome loci ([160]27). Mettl14^f/f mice without Villin-Cre were used as
   WT controls (Mettl14 WT) for Mettl14 KO mice. Mettl14^f/f mice were
   crossed with Lgr5-eGFP-IRES-creERT2 (Lgr5-Cre) mice to generate Mettl14
   depletion in Lgr5^+ stem cells.

   Villin-Cre (B6.Cg strain; stock no. 004586) and Lgr5-eGFP-IRES-CreERT2
   (B6.129P2 strain; stock no. 002120) mice were obtained from The Jackson
   Laboratory and had been backcrossed to C57BL/6N mice (Charles River
   Laboratories) for more than 10 generations. All of the KO and WT mice
   were sex- and age-matched littermates and were cohoused for any
   experiments described. All the mice were maintained under specific
   pathogen–free conditions and used in accordance with the animal
   experimental guidelines set by the Institutional Animal Care and Use
   Committee (IACUC). This study has been approved by the IACUC of the
   Shanghai Jiao Tong University School of Medicine (protocol no.
   A-2015-016) and the Shanghai Jiao Tong University (protocol no.
   2017043).

Isolation of mouse intestinal epithelial cells

   For the isolation of colonic epithelial cells, the colon tissue was cut
   into 2-mm segments and incubated in RPMI 1640 medium containing 2.5 mM
   EDTA and 2% fetal serum albumin for 30 min at 4°C on a rocker.
   Epithelial cells were released by vigorous shaking and passed through a
   70-μm strainer and then washed with cold Dulbecco’s phosphate-buffered
   saline (DPBS) containing 2% fetal serum albumin.

Mouse colonic organoid culture

   IntestiCult Organoid Growth Medium (Mouse) (STEMCELL Technologies,
   catalog no. 06005) was used for the establishment and maintenance of
   mouse colonic organoids. Briefly, the mice were sacrificed and the
   colon was harvested from each mouse. The colon was flushed gently with
   cold PBS containing 1% fetal serum albumin. Small scissors were used to
   make a longitudinal incision along the entire length of the colon, and
   then the colon was cut into 2-mm pieces. A 10-ml serological pipette
   was used to wash the colonic pieces by pipetting up and down three
   times. The supernatant was removed. This wash procedure was repeated 15
   to 20 times or until supernatant was clear. Then, the colonic pieces
   were suspended in 25-ml Gentle Cell Dissociation Reagent (STEMCELL
   Technologies, catalog no. 07174) in a shaker at 20 rpm for 20 min at
   room temperature. The colonic pieces were resuspended in 10 ml of cold
   PBS containing 0.1% bovine serum albumin and pipetted up and down three
   times, and the supernatant was passed through a 100-μm strainer into a
   50-ml conical tube. This procedure was repeated six times, and the
   crypts were collected. The quality of the suspensions was assessed, and
   the crypt number was counted with an inverted microscope. Crypts were
   used for organoid development in domes made by Matrigel (Corning,
   catalog no. 356231) and IntestiCult Organoid Growth Medium (1:1).
   Suspension (50 μl) for each dome was pipetted into a 24-well plate.
   Next, the plate was incubated at 37°C for 10 to 15 min until the
   Matrigel was solidified. Last, 750 μl of complete IntestiCult Organoid
   Growth Medium was added to each well and incubated at 37°C and 5%
   CO[2].

RNA isolation and qRT-PCR

   RNA was extracted using the TRIzol reagent (Invitrogen) and further
   purified using the RNase-Free DNase Set (QIAGEN, catalog no. 79256).
   The Maxima H Minus Reverse Transcriptase Kit (Thermo Fisher Scientific,
   catalog no. EP0753) was used for complementary DNA (cDNA) synthesis.
   The iTaq Universal SYBR Green Supermix was used for real-time PCR, and
   L32 or Hprt mRNA level was used as internal control to calculate mRNA
   relative abundance. The primer sequences used for qPCR were as follows:

   Mettl14: 5′-CTCCAAACTCAAAACGGAAGTGT-3′ (forward) and
   5′-ATGGGGATTTAAGCTCTGCGT-3′ (reverse); Vil1:
   5′-GCTTGCCACAACTTCCTAAGAT-3′ (forward) and 5′-TCAGTTTAGTCATGGTGGACGA-3′
   (reverse); Lgr5: 5′-CCTACTCGAAGACTTACCCAGT-3′ (forward) and
   5′-GCATTGGGGTGAATGATAGCA-3′ (reverse); Gob5: 5′-GGAAGGCAAAGCCTGAATAT-3′
   (forward) and 5′-GGCTCATCATTGCCTAGAGG-3′ (reverse); Alpi:
   5′-AGGACATCGCCACTCAACTC-3′ (forward) and 5′-GGTTCCAGACTGGTTACTGTCA-3′
   (reverse); Chga: 5′-CGATCCAGAAAGATGATGGTC-3′ (forward) and
   5′-CGGAAGCCTCTGTCTTTCC-3′ (reverse); Tnf: 5′-CCCTCACACTCAGATCATCTTCT-3′
   (forward) and 5′-GCTACGACGTGGGCTACAG-3′ (reverse); Tnfaip2:
   5′-AGGAGGAGTCTGCGAAGAAGA-3′ (forward) and 5′-GGCAGTGGACCATCTAACTCG-3′
   (reverse); Tnfaip8l1: 5′-GTTTGTGGACAATACCAGCAGT-3′ (forward) and
   5′-GTTCTTCACTACCCTCTGTGC-3′ (reverse); Tnfrsf11a:
   5′-GGACGGTGTTGCAGCAGAT-3′ (forward) and 5′-GCAGTCTGAGTTCCAGTGGTA-3′
   (reverse); Nfkb1: 5′-GGAGGCATGTTCGGTAGTGG-3′ (forward) and
   5′-CCCTGCGTTGGATTTCGTG-3′ (reverse); Nfkb2: 5′-AGTGTGCGCTGTGTCTGTAG-3′
   (forward) and 5′-GTTCTTCTTGGTTACATGCAGGA-3′ (reverse); Map3k14:
   5′-TGTGGGAAGTGGGAGATCCTA-3′ (forward) and 5′-GGCTGAACTCTTGGCTATTCTCA-3′
   (reverse); Nfkbia: 5′-TGAAGGACGAGGAGTACGAGC-3′ (forward) and
   5′-TTCGTGGATGATTGCCAAGTG-3′ (reverse); Fadd: 5′-GCGCCGACACGATCTACTG-3′
   (forward) and 5′-TTACCCGCTCACTCAGACTTC-3′ (reverse); Traf6:
   5′-AAAGCGAGAGATTCTTTCCCTG-3′ (forward) and 5′-ACTGGGGACAATTCACTAGAGC-3′
   (reverse); Nfkbia for RIP: 5′-GCCAGCGTCTGACCATTATAAG-3′ (forward) and
   5′-ACTCTATTTGTACAAATATACAAGTCCAC-3′ (reverse); Ascl2:
   5′-AAGCACACCTTGACTGGTACG-3′ (forward) and 5′-AAGTGGACGTTTGCACCTTCA-3′
   (reverse); Tcf7l2: 5′-TCATCACGTACAGCAATGAACA-3′ (forward) and
   5′-CGACAGCGGGTAATATGGAGAG-3′ (reverse); Ephb2:
   5′-GCGGCTACGACGAGAACAT-3′ (forward) and 5′-GGCTAAGTCAAAATCAGCCTCA-3′
   (reverse); Axin2: 5′-ATGAGTAGCGCCGTGTTAGTG-3′ (forward) and
   5′-GGGCATAGGTTTGGTGGACT-3′ (reverse); Hprt: 5′-TCAGTCAACGGGGGACATAAA-3′
   (forward) and 5′-GGGGCTGTACTGCTTAACCAG-3′ (reverse); L32:
   5′-AAAAACAGACGCACCATCGAA-3′ (forward) and 5′-TTCAGGTGACCACATTCAGGG-3′
   (reverse).

Diarrhea analysis

   We calculated the proportion of water in feces to determine the
   diarrhea situation of 2-week-old mice. Briefly, sterile 1.5-ml tubes
   were weighed, and stools were collected by these tubes and weighed.
   Then, the lid of the tubes was opened and the tubes were put on a
   constant temperature flat bed at 80°C overnight. The tubes were weighed
   again, and the reduced weight is water.

Immunohistochemical and histological analysis

   Mouse intestinal tissue segments were harvested and fixed in 10%
   formalin solution overnight at room temperature. Postfixation samples
   were moved to 70% ethanol and then sent to the Histology Core of the
   Shanghai Institute of Immunology for processing. Histological grading
   was performed by two blinded investigators as described previously
   ([161]46). Briefly, the amount and depth of inflammation, crypt damage,
   or regeneration were used to grade the histological score. For
   immunohistochemical staining, the sections were deparaffinized and
   washed in PBS. Antigen retrieval was performed by heating the sections
   in 10 mM sodium citrate buffer (pH 6.0). The sections were washed in
   PBS after cooling, incubated in 3% hydrogen peroxide for 10 min at room
   temperature, and then washed again in PBS. The sections were blocked in
   PBS containing 1% bovine serum albumin for 1 hour at room temperature
   and stained in blocking buffer containing primary antibodies overnight
   at 4°C. On the following day, the sections were warmed to room
   temperature for 1 hour and stained with a horseradish peroxidase
   (HRP)–polymer complex for 20 min, followed by incubation with secondary
   antibody for 20 min. The sections were washed three times in PBS,
   developed with 3,3’-diaminobenzidine (DAB) reagent (Peroxidase
   Substrate Kit, ZSGB-BIO, catalog no. ZLI-9018), and counterstained with
   hematoxylin. The sections were washed with tap water and then
   subsequently washed with increasing ethanol concentrations for
   dehydration. Once mounted and air-dried, the slices were viewed under
   Digital Whole Slide Scanner (Hamamatsu, NanoZoomer S360) and analyzed
   with NDP.view2 software. Tissue sections were assessed for pathology as
   previously described. The list of antibodies used in the
   immunohistochemical and histological analyses is shown in table S1.

TUNEL staining

   For TUNEL staining, colon sections were treated with a TUNEL kit
   (Sigma-Aldrich, St. Louis, MO, USA) according to the manufacturer’s
   instructions. TUNEL-positive cells were observed under a confocal
   microscope (FLUOVIEW FV3000, Olympus) and analyzed with Imaris
   software.

Immunofluorescence

   Formalin-fixed paraffin-embedded tissue sections were deparaffinized,
   and antigen retrieval was performed by heating the sections in 10 mM
   sodium citrate buffer (pH 6.0). The sections were washed with PBS three
   times and then rinsed with PBS containing 0.5% Triton X-100 and 0.05%
   Tween 20 for 10 min. The sections were added in blocking buffer (PBS
   and 0.025% Triton X-100, 0.05% Tween 20, and 5% bovine serum albumin)
   for 30 min at room temperature. Primary antibodies were incubated
   overnight at 4°C, and the secondary antibodies were incubated for 1
   hour at room temperature on the following day. The sections were
   mounted with Fluoroshield histology medium containing
   4′,6-diamidino-2-phenylindole (DAPI) (Sigma-Aldrich, F6057). The slices
   were viewed under a confocal microscope (FLUOVIEW FV3000, Olympus) and
   analyzed with Imaris software. The list of antibodies used in the
   immunofluorescence analysis is shown in table S1.

DSS-induced colitis analysis

   Mettl14^f/f Lgr5^creERT and Mettl14^f/f WT control mice were treated
   with tamoxifen for 7 days and then rested for 7 days, and 2% DSS water
   was added. The weight of mice was observed at the same time point every
   day. On days 6 and 7, the severity of DSS-induced colitis was detected
   by endoscopy analysis. Stool, blood vessels, particle size, and
   translucency were used to evaluate the colitis score, and detailed
   scoring criteria were performed as described previously ([162]47). The
   mice were euthanized on day 9, and the colon lengths were detected.

Tamoxifen treatment

   For Mettl14^f/f Lgr5^creERT mice, tamoxifen-inducible, Cre-mediated
   recombination will result in deletion of the floxed sequences in the
   Lgr5-expressing cells of the offspring. Tamoxifen was added to corn oil
   (20 mg/ml) and shaken overnight at 37°C protected from light. Mice were
   administered tamoxifen (75 mg/kg body weight) via intraperitoneal
   injection daily for 7 days and then rested for 3 days for cell
   apoptosis analysis or 7 days before DSS-induced colitis.

Western blot

   Total protein of colonic epithelial cells was extracted with
   radioimmunoprecipitation assay lysis buffer supplemented with protease
   inhibitors (Thermo Fisher Scientific). Antibodies against METTL14
   (Sigma-Aldrich, catalog no. HPA038002), NFKBIA (Cell Signaling
   Technology, catalog no. 9242s), Bax (Cell Signaling Technology, catalog
   no. 2772), Bcl2 (ProteinTech, catalog no. 12789), and
   glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (Cell Signaling
   Technology, catalog no. 2118) were used at a 1:1000 dilution in 5%
   nonfat milk buffer at 4°C overnight. After that, an HRP-conjugated
   secondary antibody (Cell Signaling Technology, catalog no. 7074) was
   incubated at room temperature for 1 hour. Signals were detected with an
   Immobilon Western HRP substrate (Millipore, WBKLS0500) and visualized
   using Amersham Imager 600 System (GE Healthcare Bio-Sciences) and
   quantified by gel analysis using ImageJ software (National Institutes
   of Health, USA).

m^6A-RIP-qPCR

   m^6A-RIP-qPCR analysis was conducted as described previously ([163]48).
   Briefly, total RNA was isolated from colonic epithelial cells.
   Polyadenylated RNA was further enriched from total RNA by using the
   Dynabeads mRNA Purification Kit (Invitrogen) and incubated with
   anti-m^6A antibody (Synaptic System, catalog no. 202003) or rabbit
   immunoglobulin G (IgG). The mixture was then immunoprecipitated by
   incubation with 50 μl of protein A beads (Sigma-Aldrich, P9424). After
   being washed three times, bound RNA was eluted from the beads with m^6A
   (0.5 mg/ml; Berry & Associates, PR3732) and then extracted by TRIzol.
   qRT-PCR was conducted, and the expression of Nfkbia was normalized to
   the input sample. Rps21 was chosen as the m^6A-negative control, and
   ptpn4 was chosen as the m^6A-positive control ([164]48).

RNA decay analysis

   Colonic epithelial cells from 2-week-old mice were plated on 96-well
   plates with 1 × 10^5 cells per well. After treatment with actinomycin D
   (5 μg/ml) (Sigma-Aldrich, catalog no. A1410) for 4 hours, cells were
   collected and subjected to RNA extraction. Total RNA was isolated, and
   qRT-PCR was conducted for mRNA levels as described above.

m^6A dot-blot assay

   mRNA was purified by a Dynabeads mRNA purification kit (Ambion, 61006)
   and quantified using ultraviolet spectrophotometry. The m^6A dot blot
   was performed on Bio-Dot Apparatus (Bio-Rad, no. 170-6545). In brief,
   the primary rabbit anti-m^6A antibody (Synaptic Systems, 202003) was
   applied to the charged nylon-based membrane containing mRNA samples.
   HRP-conjugated goat anti-rabbit IgG (DakoCytomation, p0448) was added
   to the blots and then developed with enhanced chemiluminescence (GE
   Healthcare, RPN2232). The signal density of the dot-blot experiment is
   quantified with Gel-Pro analyzer software (Media Cybernetics) in all
   experiments.

RNA sequencing

   The colonic epithelial cells were isolated from 2-week-old mice as
   described above. RNA was extracted using the TRIzol reagent
   (Invitrogen) according to the manufacturer’s instructions. Libraries
   were constructed and sequenced using the BGISEQ-500 sequencer. Raw
   sequencing reads were cleaned by removing adaptor sequences, reads
   containing poly-N sequences, and low-quality reads, and clean reads
   were then mapped to GenBank to identify known mouse mRNA. RNA-seq and
   analysis were conducted by OE Biotech Co. Ltd.

scRNA-seq and data analysis

   Colonic epithelial cells were stained with an EpCAM and live/dead cell
   kit (Thermo Fisher Scientific, catalog no. [165]L34965) on ice in the
   dark for 30 min. The EpCAM^+ live cells were sorted by flow cytometry
   (Beckman Coulter MoFlo Astrios EQ), and cell number was counted. The
   cells were diluted to a final concentration range of 250 to 400
   cells/μl for single-cell sequencing according to the manufacturer’s
   instructions (10x Genomics kit, chromium single cell 3′ reagent kits
   V2) ([166]49). For analysis of single-cell data, a total of 7863 cells
   after filtering and removing immune clusters were used. The raw
   sequencing reads from 10x Genomics 3-prime sequencing were processed
   with Cell Ranger ([167]50) to generate gene–cell count matrices. This
   matrix was filtered, retaining cells with more than 500 genes and less
   than 20% mitochondria transcripts; genes expressed more than five
   cells. Datasets from Mettl14 WT and Mettl14 KO samples were integrated
   with Seurat using canonical correlation analysis (CCA) with Seurat
   ([168]51). The top 15 CCA components were aligned to remove batch
   effect. Seurat was also used for data normalization with log (counts
   per 10,000 + 1), scaling, dimension reduction, clustering, and marker
   gene identification (Wilcoxon test by default). t-SNE was used for 2D
   visualization. We removed clusters of immune cells based on the
   expression of gene Ptprc (CD45) and retained only nonimmune cells.
   Clusters of the same cell types were merged together. To better inspect
   the gene expression distribution of modulators of m^6A modification,
   which have relatively low expression and tend to have more dropout
   ([169]52), we used ALRA ([170]53) to impute the merged expression
   matrix from WT and KO samples in [171]Fig. 4D.

Statistical analysis

   Statistical analysis was performed with GraphPad Prism 8.01. Unpaired
   Student’s t test and two-tailed Mann-Whitney U test were used for
   measurement data of two-group analysis. One-way analysis of variance
   (ANOVA) was used for measurement data of more than two groups. All
   general statistical analysis was calculated with a confidence interval
   of 95%. P values ≤ 0.05 were considered as statistically significant
   (*P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001). Survival curves
   were compared by the log-rank test. Data are represented as means ± SD
   as indicated in the figures. All authors had access to the data and
   reviewed and approved the final manuscript.

Acknowledgments