Mettl3-mediated mRNA m6A modification controls postnatal liver development by modulating the transcription factor Hnf4a

Hepatic specification and functional maturation are tightly controlled throughout development. N6-methyladenosine (m6A) is the most abundant RNA modification of eukaryotic mRNAs and is involved in various physiological and pathological processes. However, the function of m6A in liver development remains elusive. Here we dissect the role of Mettl3-mediated m6A modification in postnatal liver development and homeostasis. Knocking out Mettl3 perinatally with Alb-Cre (Mettl3 cKO) induces apoptosis and steatosis of hepatocytes, results in severe liver injury, and finally leads to postnatal lethality within 7 weeks. m6A-RIP sequencing and RNA-sequencing reveal that mRNAs of a series of crucial liver-enriched transcription factors are modified by m6A, including Hnf4a, a master regulator for hepatic parenchymal formation. Deleting Mettl3 reduces m6A modification on Hnf4a, decreases its transcript stability in an Igf2bp1-dependent manner, and down-regulates Hnf4a expression, while overexpressing Hnf4a with AAV8 alleviates the liver injury and prolongs the lifespan of Mettl3 cKO mice. However, knocking out Mettl3 in adults using Alb-CreERT2 does not affect liver homeostasis. Our study identifies a dynamic role of Mettl3-mediated RNA m6A modification in liver development.

The liver is the primary organ responsible for metabolism, lipid transportation, drug detoxification, and hormone secretion 1 . Hepatic specification and the dramatic functional transition from haematopoiesis to metabolism of the liver are tightly controlled by the intricate crosstalk among extracellular signals, transcription factors, and epigenetic regulators. Previous reports showed that coordination of liverenriched transcription factors, histone modifications, and DNA methylation orchestrates the hepatic differentiation program during development and maintains liver homeostasis in adults 2,3 .
In this work, we generate hepatic-specific Mettl3 knockout (Mettl3 cKO) mice by crossing Mettl3 flox/flox mice with Albumin (Alb)enhancer/promoter driven-Cre transgenic mice to investigate the role of m 6 A modification in liver development. Hepatic perinatal loss of Mettl3 causes severe liver damage, including steatosis, apoptosis, and fibrosis, and finally results in lethality within 7 weeks. Using m 6 A-RNA immunoprecipitation (m 6 A-RIP) sequencing and RNA-sequencing, we identify that crucial liver-enriched transcription factors, including Hnf4a, are modified by m 6 A in liver development. Loss of Mettl3 induces depletion of m 6 A on Hnf4a transcripts, decreases its transcript stability in an Igf2bp1dependent manner, and down-regulates Hnf4a expression, while overexpressing Hnf4a with AAV8 alleviates the liver injury and prolongs lifespan of Mettl3 cKO mice. However, deletion of Mettl3 in adult mouse livers using Albumin-enhancer/promoter-driven Cre ERT2 shows minimal effects on liver homeostasis. In conclusion, we elucidate a dynamic role of Mettl3-mediated RNA m 6 A modification during mouse postnatal liver development and decipher a novel function of epitranscriptomic control of liver organogenesis.

Generation of hepatic specific Mettl3 knockout mice
To study the role of m 6 A modification in liver development, we first tested the expression level of critical subunits of the m 6 A methyltransferase complex, Mettl3 and Mettl14 7 . Both components showed shallow protein levels in mouse neonates (within one day after birth), increased gradually, and peaked at 2-3 weeks, and then decreased from 4 weeks onwards ( Supplementary Fig. 1a). A similar trend was observed in human livers with high expression in children and a subsequent decline with age ( Supplementary Fig. 1b). These results indicated that m 6 A is dynamically regulated in postnatal liver development. Global knockout of either Mettl3 or Mettl14 results in embryonic lethality caused by gastrulation defects [19][20][21] . Thus, to study the role of m 6 A modification, we generated mice with hepatic specific knockout of the catalytic subunit of the m 6 A methyltransferase complex, Mettl3, by crossing Mettl3 flox/flox mice (with loxP sites flanking exons 2 and 4) with Alb-enhancer/promoter-driven Cre transgenic mice ( Supplementary Fig. 1c, d). The specific knockout of Mettl3 in the liver was confirmed by genomic PCR, quantitative real-time PCR (RT-qPCR), western blot, and immunochemistry ( Fig. 1a-d and Supplementary Fig. 1e-j). Genomic PCR and RT-qPCR showed that efficient knockout of Mettl3 started from day 1 after birth ( Fig. 1b and Supplementary Fig. 1k), along with Cre expression (Supplementary Fig. 1i). As expected, livers from Mettl3 cKO mice showed a significant decrease in mRNA m 6 A levels compared to control mice ( Supplementary Fig. 1l). In addition, we also observed that knocking out Mettl3 led to disruption of Mettl14 ( Supplementary Fig. 1m), which is in accordance with previous reports 19 .

Hepatic Mettl3 knockout results in postnatal lethality
Hepatic Mettl3 knockout mice were born at almost expected Mendelian frequencies (Fig. 1e), excluding the possibility of prenatal lethality. However, both male and female knockout mice were smaller in body size than their age and sex-matched wild-type (WT) control (Control, also hereafter in similar experiments) littermates (Fig. 1f). This difference appeared at 2 weeks after birth and gradually became more pronounced at 4 and 5 weeks ( Fig. 1g and Supplementary Fig. 1n, o). Moreover, all the Mettl3 cKO mice died within 7 weeks after birth, while heterozygous knockout individuals were fertile and survived for over 12 months without discernible defects in development (Fig. 1h), indicating that one allele of Mettl3 was sufficient to maintain normal development and function of mouse livers. These results demonstrate that Mettl3 is critical for postnatal liver development, especially during the highly proliferative stages from 0 to 4 weeks after birth.

Hepatic Mettl3 deletion causes liver injury
To delineate the exact role of Mettl3 in liver organogenesis, we dissected the livers of Control and Mettl3 cKO mice at different time points after birth. Grossly, an obvious mottled appearance, which is an indication of lipid deposition, was observed in the livers of Mettl3 cKO mice 2 weeks after birth (Fig. 2a). The livers of 4-week-old Mettl3 cKO mice were yellow, smaller, and stiffer than the WT Control. These differences became more significant at 5 weeks (Fig. 2a). The liver weight of Mettl3 cKO mice decreased since 3 weeks after birth (Fig. 2b), while the liver weight to body weight ratio was slightly increased ( Supplementary  Fig. 2a). Serum indicators of liver function showed that Mettl3 deficiency caused defects in metabolism, detoxification, protein synthesis, and secretion functions in the liver from 1-2 weeks after birth, indicating progressive liver damage (Fig. 2c-j and Supplementary Fig. 2b).
Mettl3 deletion results in apoptosis, steatosis, fibrosis, and activation of hepatic progenitors Although histologic analysis showed no apparent differences between the Control and Mettl3 cKO mice at 1 day and 1 week after birth ( Supplementary Fig. 3a, b), marked pathological lesions were observed in Mettl3 cKO mice since 2 weeks old (Fig. 3a). Increased lipid droplet deposition in Mettl3 cKO livers was observed 2 weeks after birth, confirmed by BODIPY staining and Oil Red O staining for both frozen liver tissues ( Fig. 3b and Supplementary Fig. 3c, d) and primary hepatocytes ( Supplementary Fig. 3e). We also observed enlarged cell size, enlarged nucleus, and increased apoptosis of hepatocytes in Mettl3 cKO mice starting from 2 weeks and expansion of ductular cells at 3 weeks after birth (Fig. 3a-d and Supplementary Fig. 3f, g). In addition, prominent fibrosis in Mettl3 cKO livers was seen at 4 weeks and became more pronounced at 5 weeks (Fig. 4a-c and Supplementary Fig. 4a, b). We did not observe any abnormalities of heterozygous cKO livers by histological analysis (Supplementary Fig. 4c). Consistent with the expansion of ductular cells, we detected marked increases of Sox9, CK19, and Ki67 positive cells in Mettl3 cKO livers ( Fig. 4d and Supplementary Fig. 4d). Meanwhile, RT-qPCR of liver tissues collected at different time points showed that hepatocyte markers (Albumin (Alb)) decreased, while hepatic progenitor markers (Afp, Krt7, Krt19, Epcam, and Sox9) and fibrosis markers (Col1a1, Acta2, and Pdgfrb) increased in Mettl3 cKO individuals (Fig. 4e). These changes were confirmed by western blot (Fig. 4f). The above results demonstrate that Mettl3 deletion in hepatocytes perinatally leads to hepatocyte injury, activation of progenitor cells, and fibrosis.
Transcriptome-wide m 6 A-RIP sequencing to identify potential targets of Mettl3 As Mettl3 is the key catalytic subunit of the m 6 A methyltransferase machinery, to gain a comprehensive insight into the molecular mechanisms underlying Mettl3 regulating postnatal liver development, we first quantified the m 6 A levels on mRNAs of the liver tissues from different developmental stages using LC-MS/MS. Global mRNA m 6 A levels of postnatal livers increased after birth, peaked at 2 weeks, and decreased then ( Supplementary Fig. 5a). Next, we profiled the genome-wide m 6 A methylation distribution using m 6 A-RIP sequencing of RNAs from five developmental time points (1 day and 1, 2, 4, and 8 weeks after birth) of mouse liver tissues. The distribution of m 6 A modifications was dynamically regulated during different stages of postnatal liver development ( Fig. 5a and Supplementary Fig. 5a-d). We identified 15139, 12483, 12615, 10561, and 6806 m 6 A peaks, corresponding to 6330, 5522, 5515, 4856, and 3445 genes from the above five groups, respectively (Supplementary Fig. 5b and Supplementary Dataset 1). Global m 6 A peak enrichment peaked 2 weeks after birth, along with patterns of bulk m 6 A levels during mouse liver development ( Supplementary Fig. 5a, d). In line with previous reports, m 6 A peaks were significantly enriched in the vicinity of the stop codon (Fig. 5a, b), and the consensus motif "GGAC" was most commonly enriched in peaks from all samples ( Supplementary Fig. 5e). Interestingly, "liver development" was one of the most significantly enriched terms of m 6 A modified genes by Gene Ontology (GO) analysis at all time points ( Fig. 5c and Supplementary Dataset 2). mRNAs of several key liverenriched transcription factors were highly methylated by m 6 A in liver tissues, including Hnf4a, Hnf1a, Ppara, and Cebpa ( Fig. 5d and Supplementary Fig. 5f). These factors play essential roles in both liver development in vivo and hepatocyte differentiation in vitro 2 . Next, we used gene-specific m 6 A-RIP-qPCR assays to confirm the authentic deposition of RNA m 6 A modifications by Mettl3 on liver-enriched factors using liver tissues from Control and Mettl3 cKO mice. We observed a significant decrease of m 6 A deposition on Hnf1a, Hnf4a, Ppara, Cebpa et al. in Mettl3 cKO mice at both 2 weeks and 4 weeks (Fig. 5e). These results indicate that Mettl3-mediated m 6 A modification is dynamically regulated during liver development and modifies crucial transcription factors controlling liver specification and function. m 6 A regulates pathways of liver development and metabolism by controlling mRNA stability of the core transcription factor Hnf4a To gain further insights into the mechanism of Mettl3 regulating liver development, we conducted RNA-sequencing for liver tissues from Control and Mettl3 cKO mice at 1 day, 1 week, 2 weeks, and 4 weeks after birth. There were much more differentially regulated genes (DEGs) between Control and Mettl3 cKO mice at later time points ( Supplementary Fig. 6a, Supplementary Dataset 3), which is consistent with our observations that Mettl3 cKO mice showed progressive severe liver damage 2 weeks after birth onward (Fig. 3a). Gene set enrichment analysis (GSEA) showed that targets of Hnf4a and Hnf1a were significantly repressed in Mettl3 cKO livers even at 1 day postnatally ( Fig. 6a and Supplementary Fig. 6b). Dual-luciferase reporter assay and mutagenesis assay ( Fig. 6b and Supplementary Fig. 6c, d) showed that co-transfection with WT, but not catalytic mutant Mettl3 22-24 , significantly promoted luciferase activity in reporters carrying WT Hnf4a and Hnf1a fragments, while such increases were abolished when the m 6 A consensus motifs were mutated, confirming that the regulation of Hnf4a and Hnf1a by Mettl3 was indeed relying on m 6 A methylation of their transcripts. Both RT-qPCR and western blot confirmed that Hnf4a was downregulated in Mettl3 cKO livers at different time points postnatally ( Fig. 6c-e). Although the RNA level of Hnf1a was downregulated at all time points ( Supplementary Fig. 6e), we observed a dramatic decrease of Hnf1a protein with age and only observed a difference between Control and Mettl3 cKO mouse livers 1 week after birth ( Supplementary Fig. 6f), indicating a less essential role of Hnf1a in the maturation of hepatocytes, which is consistent with previous studies 25 . Since Hnf4a is a master transcription factor required for liver development in both foetuses and adults and controls most aspects of mature hepatocyte function 26,27 , we mainly focused on Hnf4a for further studies. RNA-sequencing data showed that along with the downregulation of Hnf4a, most Hnf4a target genes, such as Apoa2, Apoc3, Cyp8b1, and Mttp, were repressed in Mettl3 cKO individuals (Supplementary Fig. 6g, Supplementary Dataset 3), which was validated by RT-qPCR ( Supplementary Fig. 6h). We also noticed that Smad signaling, the central mediator of fibrosis 28 , was significantly enriched in Mettl3 cKO mouse liver tissues at 4 weeks ( Supplementary Fig. 7a), supporting the phenomenon that massive liver fibrosis was induced in Mettl3 cKO animals (Fig. 4). These results indicate that Mettl3-mediated m 6 A controls the expression of crucial liver developmental genes during liver development. m 6 A modification is involved in various aspects of RNA metabolism, including transcription, splicing, nuclear transportation, stability, and translation. Because we observed decreased expression of Hnf4a at both mRNA and protein levels, we determined the alternative splicing, nucleus-cytoplasm transportation, and mRNA stability of Hnf4a mRNA. Alternative splicing analysis showed no differences on Hnf4a transcripts in RNA-sequencing data from Control and Mettl3 cKO livers (Supplementary Dataset 4). The distribution of Hnf4a mRNA in nuclear and cytoplasm was also not affected by Mettl3 knockout (Supplementary Fig. 7b-d). Only mRNA stability showed significant changes in primary hepatocytes and the HepG2 cells with Mettl3 inhibition These results demonstrated that Mettl3 deficiency downregulated Hnf4a expression by reducing the half-life of Hnf4a mRNA. m 6 A modification controls RNA fate mainly through "reader" proteins recognizing and binding to m 6 A-containing transcripts. Among identified m 6 A "readers", insulin-like growth factor 2 mRNAbinding proteins (IGF2BPs, including IGF2BP1/2/3) are known to promote the stability of their target mRNAs 17 . To further delineate the mechanism of m 6 A controlling Hnf4a expression, we checked previous publications and found that deletion of IGF2BP1 leads to destabilization of HNF4A mRNA in HepG2 cells while interfering with the other two members did not affect HNF4A mRNA degradation (Supplementary Fig. 7j) 17 , indicating that IGF2BP1 may directly recognize m 6 A on HNF4A mRNA and maintain its levels in the liver context. Thus, we tested Igf2bp1 binding on Hnf4a mRNA with RIP experiments. The results showed that Igf2bp1 could efficiently bind to Hnf4a transcripts in mouse livers of both 2 weeks and 4 weeks, and the enrichment significantly decreased after Mettl3 knockout (Fig. 6h). Accordingly, knocking down IGF2BP1 with small Hairpin RNA (shRNA) in HepG2 cells also decreased HNF4A mRNA half-life, similar to Mettl3 disruption ( Fig. 6i and Supplementary Fig. 7k, l). These data demonstrated that Mettl3-mediated m 6 A controls the expression of Hnf4a by regulating its mRNA stability in an IGF2BP1-dependent manner.

Hepatic Hnf4a overexpression alleviated liver injury caused by Mettl3 knockout
To further strengthen our conclusion that Hnf4a is the primary mediator of Mettl3 function in liver development, we conducted rescue experiments using AAV serotype 8 (AAV8) to express Hnf4a under the control of a liver-specific promoter (thyroxine-binding globulin, TBG) (AAV8-TBG-Hnf4a) on Mettl3 cKO mice (Fig. 6j). Injection of AAV8-TBG-Hnf4a by superficial temporal vein on day two after birth successfully overexpressed Hnf4a in the liver ( Supplementary Fig. 7m) and alleviated liver damage caused by hepatic Mettl3 knockout compared to AAV8-Ctrl at two weeks, evidenced by an increased number of Mettl3 cKO mice at different time points postnatally via RT-qPCR (n = 3 for 1 day Control and 3 weeks cKO group; n = 5 for 3 weeks Control group; n = 4 for other groups). c Western blot for Mettl3 in Control and Mettl3 cKO mouse liver tissues at two weeks after birth (6 experiments were repeated independently with similar results). Gapdh was used as a loading control (also hereafter in similar experiments). d Immunohistochemistry staining of Mettl3 in 4 weeks old Control and Mettl3 cKO mouse livers (6 experiments were repeated independently with similar results). Scale bar = 50 μm. e The number of offspring with different genotypes from intercrossing Mettl3 flox/flox and Mettl3 flox/-/Alb-Cre mice. f Representative appearance of sex-matched Control mice and Mettl3 cKO littermates at 4 weeks after birth. g Body weight of male Control and Mettl3 cKO littermates at different time points after birth (n = 4 for 1 week cKO group; n = 6 for 4 weeks and 5 weeks Control groups; n = 9 for 3 weeks Control group; n = 12 for 3 weeks cKO group; n = 13 for 2 weeks Control group; n = 15 for day 1 and 1 week Control group; n = 7 for other groups). h Survival curves of Control, Mettl3 cKO, and Mettl3 heterozygous (Mettl3 flox/-/Alb-Cre) littermates (n = 25 for each group). Data in b and g were shown as mean ± SEM with the indicated significance (*P < 0.05, **P < 0.01, ***P < 0.001; two-tailed student's t-test). Data in (h) were analyzed by Log-rank (Mantel-Cox) test with the indicated significance (***P < 0.001). Source data are provided as a Source Data file.   week cKO group and 5 weeks groups; n = 5 for 4 weeks groups; n = 6 for 2 weeks cKO group; n = 7 for day 1 cKO group; n = 8 for 1 week Control group; n = 9 for 3 weeks Control group; n = 10 for 2 weeks Control group; n = 12 for 3 weeks cKO group; n = 15 for day 1 Control group). c-j Serum levels of ALT c, AST d, Albumin e, ALP f, Total bile acid g, Total bilirubin h, Direct bilirubin i, and Cholesterol j of Control and Mettl3 cKO mice at different time points postnatally (n = 2 for day 1 cKO group; n = 3 for day 1 Control group; n = 4 for 1 week cKO and 5 weeks groups; n = 5 for 4 weeks groups; n = 6 for 1 week Control group, 2 weeks groups, and 3 weeks Control group; n = 7 for 3 weeks cKO group). Data in b-j were shown as mean ± SEM with the indicated significance (*P < 0.05, **P < 0.01, ***P < 0.001, two-tailed student's t-test). Source data are provided as a Source Data file.  Ki67 + proliferating hepatocytes and reduced hepatic steatosis (Fig. 6km). However, we did not see long-term benefits on mortality. This may attribute to the rapid dilution of AAV caused by the vigorous hepatocyte division within four weeks after birth 29 . Then we overexpressed Hnf4a by AAV-TBG-Hnf4a through tail vein injection at four-week-old Mettl3 cKO mice and found that Hnf4a overexpression significantly prolonged the life span of Mettl3 cKO mice (Fig. 6n). These results further demonstrated that Hnf4a is the primary factor mediating the function of Mettl3 in liver development.

Mettl3 is dispensable for homeostasis of adult liver
Given the lethality of Mettl3 cKO mice within 7 weeks, we generated conditional inducible Mettl3 knockout mice (Mettl3 icKO) by crossing Mettl3 flox/flox mice with Alb-Cre ERT2 mice. Mettl3 could be deleted in adult  mice by intraperitoneal (IP) injection of tamoxifen 30 . Genomic PCR, RT-qPCR, and western blot confirmed efficient and specific depletion of Mettl3 1 week after tamoxifen administration (Fig. 7a-e and Supplementary Fig. 8a). LC-MS/MS results also showed a significant decrease in bulk m 6 A modification of Mettl3 icKO mouse liver mRNAs (Fig. 7f). However, we did not observe any visible abnormalities in these Mettl3 icKO mice ( Fig. 7g and Supplementary Fig. 8b, c). Serological and histologic examinations showed minimal liver damage (Fig. 7h, i, and Supplementary Fig. 8d). Both mRNAs of Hnf4a and downstream targets of Hnf4a showed no differences between Control and Mettl3 icKO livers ( Supplementary Fig. 8e-k). These results indicate that although Mettl3 is essential for the early postnatal development of the liver, it is not crucial for the homeostasis of adult liver ( Supplementary Fig. 9).

Discussion
Previous reports had profiled transcriptome-wide m 6 A in porcine liver at three postnatal stages 31 and supplied a roadmap of m 6 A modification across human and mouse livers 32 , indicating dynamic changes of m 6 A modification during liver development. This study demonstrated a vital role of Mettl3-mediated m 6 A modification in mouse postnatal liver development. Using Mettl3 flox/flox /Alb-Cre mice, we found that Mettl3 cKO mice got steatosis at about 2 weeks after birth, liver fibrosis at 4-5 weeks, and finally died before 7 weeks due to severe liver injury. The liver experiences abrupt functional changes from intra-to the extra-uterine environment, corresponding to a functional shift from haematopoiesis to metabolism and immunity 33 . In general, fetal liver haematopoiesis is characterized by initiation (E11.5), peak (E14.5), recession (E15.5), and disappearance (3 days after birth), while the neonatal liver rapidly evolves into a vital organ for immunosurveillance and metabolism. Histological analysis showed that haematopoietic cells disappeared rapidly, and parenchymal cells occupied in hepatic constituent in the first week after birth, consistent with previous reports 34 . Mettl3 was barely expressed in neonatal mouse livers, while highly expressed at both fetal livers 35 and 1 week after birth at protein levels (even though the mRNA level only slightly changed during this process). The mechanism of how Mettl3 is tightly regulated during this perinatal period remains elusive and is worth considering. Given the functional transition of livers perinatally, the dynamic regulation and function of Mettl3 in this period might be involved in both haematopoietic and hepatic aspects. A series of studies have investigated the role of Mettl3-mediated m 6 A in haematopoietic system 36,37 , including early haematopoietic stem cell (HSC) development in the fetal liver 35 , revealing an essential role for m 6 A in both specification and homeostasis of haematopoietic system. However, the function of Mettl3 and m 6 A in hepatic lineage specification in prenatal development is currently not clear and worthy to be studied further.
Interestingly, the expression of Mettl3 and Mettl14 decreased synchronously in the liver after 4 weeks old, raising the possibility of a less critical role of m 6 A modification in adulthood and elder age. Indeed, we found that Mettl3 was not essential for liver homeostasis in adults. Previous studies also showed higher expression levels of the methyltransferase complex in early stages of differentiation or progenitors, such as in neurogenesis 38,39 and haematopoiesis 40 . The level of global m 6 A modification and methyltransferase expression is reduced in premature mesenchymal stem cells 41 , replicative senescent cells 42 , and peripheral blood mononuclear cells from old cohorts 43 . Aging is always accompanied by a progressive decline in regenerative capacity, especially in the liver 44 , and thus it will be intriguing to delineate whether the age-related decline of regeneration capacity is elicited by decreased m 6 A dynamics in older individuals. Although overexpressing RNA methyltransferases attenuates the senescence phenotype [41][42][43] , increased methyltransferase expression is related to aggravated liver metabolic disorders 45 and carcinogenesis progression 46 , including hepatocellular carcinoma 47 , indicating that fine-tuning regulation of m 6 A machinery is essential for physiological homeostasis.
As the largest digestive and metabolic organ in adults, the liver is responsible for transforming protein, glycogen, cholesterol, fatty acid, and many other complex molecules into elementary molecules. The proper development and function of the liver were maintained by a series of liver-enriched transcription factors and comprehensive regulatory networks among them 2,48 . Here we found that plenty of liverenriched transcription factor transcripts were modified by Mettl3mediated m 6 A modification, including Hnf1a, Hnf1β, Hnf4a, Ppara, Cebpa, Onecut1, Onecut2, Cited2, et al. Hnf4a seemed to be the most critical downstream mediator of Mettl3 in postnatal liver development. Furthermore, we observed a decreased Hnf4a at different time points on both mRNA and protein levels in Mettl3 cKO mouse livers compared with Control livers and significant dysregulation of its downstream targets.
Hnf4a is a master transcription factor required for mouse livers in both foetuses and adults 26 . It directly binds to almost half of the actively transcribed genes in adult livers and serves as a high-level transcription factor in hepatic transcriptional hierarchies 49,50 . Most of the abnormalities we found in Mettl3 cKO mice (including lipid deposition in the liver, increased bile acids in serum, liver injury, and lethality in young adults) phenocopied that of hepatic Hnf4a knockout mice 26 . However, Mettl3 cKO mice showed more severe liver injury than hepatic Hnf4a knockout mice. This might be explained by the fact that multiple targets were regulated by Mettl3-mediated m 6 A modification, and auto-regulatory and cross-regulatory circuits among them may further accelerate the collapse of liver development and function 49,51,52 . Among others, accumulating evidence showed profound crosstalk between m 6 A and histone/DNA epigenetic modifications 53 , which added another layer of complexity to explain the role of Mettl3 in liver development and function. Recent studies showed that m 6 A directly regulates heterochromatin organization 54,55 . Coordinated remodeling of heterochromatin is essential for liver development 56 , and disturbed heterochromatin landscape contributes to impaired hepatic function and tumorigenesis 57 . Therefore, it will be interesting to see whether an abnormal chromatin state accounts for Mettl3 cKO-induced liver injury. Besides, Mettl3 knockout may also lead to liver development defects by controlling other aspects of RNA metabolism, such as mRNA transportation, translation, and splicing of other regulators involved in liver development and biogenesis of miRNAs critical for hepatogenesis.
In summary, our study demonstrated a novel regulatory function of epitranscriptomics by Mettl3-mediated m 6 A modification on liver postnatal development and homeostasis, expanding our . c Quantification of the Masson's trichrome staining positive area (n = 8 for cKO group; n = 12 for Control group). d CK19 and Sox9 immunohistochemistry staining of liver sections from Control and Mettl3 cKO mice at 4 weeks after birth (6 experiments were repeated independently with similar results). Scale bar = 50 μm. e RT-qPCR analysis of hepatocyte markers, hepatic progenitor markers, and fibrosis markers for liver tissues from Control and Mettl3 cKO mice at different time points postnatally (n = 3 for 1 day Control group, 3 weeks cKO group, and 5 weeks groups; n = 5 for 3 weeks Control group; n = 4 for other groups). f Western blot for Albumin, Sox9, and αSMA of Control and Mettl3 cKO mouse liver tissues at 4 weeks after birth (3 experiments were repeated independently with similar results). Data in b, c, and e were shown as mean ± SEM with the indicated significance (*P < 0.05, **P < 0.01, ***P < 0.001; two-tailed student's t-test). Source data are provided as a Source Data file.
understanding of the regulatory network in mammalian liver development and function.       Table 1.

AAV virus preparation and in vivo transduction
Serotype 8 AAV (AAV8) was used in this study. Hnf4a coding sequencing was cloned into an AAV8 vector under the control of TBG promoter to achieve liver-specific overexpression of Hnf4a protein. The virus was packaged by Packgene Biotech Co., Ltd (Guangzhou, China) with a final titer larger than 2.0 × 10 13 viral genomes/mL. For AAV transduction in vivo, we used two strategies for rescue experiments. In the first strategy, each Mettl3 cKO mouse received 1.0 × 10 11 viral genomes of AAV8-Ctrl or AAV8-Hnf4a through the superficial temporal vein on day 2 after birth. Mice were sacrificed at 2 weeks old for further analysis. In another strategy, 1.0 × 10 11 viral genomes/mouse of AAV8-Ctrl or AAV8-Hnf4a was injected into Mettl3 cKO mice through the tail vein at 4 weeks old for further survival rate analysis.

Human specimens
Human liver tissues were obtained from donation after cardiac death (DCD) during liver transplantation in the Third Affiliated Hospital of Sun Yat-sen University. The study has been approved by the Medical Ethical Committees of the Third Affiliated Hospital of Sun Yat-sen University. The study design and conduct complied with all relevant regulations regarding the use of human study participants and was conducted following the criteria set by the Declaration of Helsinki.

Cell culture
HEK293T and HepG2 cells were obtained from American Type Culture Collection (ATCC) and maintained in DMEM-high glucose medium (Thermo Scientific, C11995500BT) supplemented with 10% fetal bovine serum (FBS) (PAN, P30-3302). Cells were incubated at 37°C in a humidified atmosphere of 5% CO 2 .

Primary hepatocyte isolation
Primary hepatocytes were isolated according to traditional two-step collagenase perfusion methods 59  . Cell pellets were collected by centrifugation at 50 g for 1 min at 4°C after three washes. Then cells were resuspended in Williams' Medium E (GIBCO, 12551032) supplemented with 10% FBS and 1% penicillin/streptomycin (KeyGEN Biotech, KGY0023) and seeded in Type I Collagen (Invitrogen, A048301) precoated culture plates and cultured at 37°C in 5% CO 2 incubator. The culture medium was changed at 2 hours of incubation. Cells were changed to serum-free medium 6 hours later and cultured overnight before use.

Western Blot
Tissue samples were lysed with RIPA buffer (50 mM Tris-HCl (PH 7.4), 150 mM NaCl, 0.1% SDS, 1% Triton X-100, 1% sodium deoxycholate, and 2 mM EDTA (PH 8.0)) containing protease inhibitor (Roche, 04693132001) and phosphatase inhibitor (Roche, 04906837001). For cells, samples were counted, washed twice with ice-cold PBS, and lysed the same as tissues. Then the lysates were separated with SDS-PAGE gels and transferred to nitrocellulose membranes. Membranes were blocked with TBS containing 5% (v/w) non-fat milk and 0.1% Tween-20 (Sigma-Aldrich, P1379) and incubated with primary and secondary antibodies sequentially. Protein bands were detected using Immobilon ECL Ultra Western HRP Substrate (Millipore, WBULS500) according to the manufacturer's instructions. Gapdh or β-actin was used as the loading control. Uncropped blots were supplied in the Source Data file. Antibodies used for western blot were listed in Supplementary Table 2.

RNA extraction and RT-qPCR
Total RNAs were extracted from tissues or cells using TRIzol (Ambion) according to the manufacturer's instructions and quantified by UV spectrophotometry. Reverse transcription was conducted using Pri-meScript™ RT Reagent Kit with gDNA Eraser (Perfect Real Time) (Takara, RR047B). RT-qPCR was then performed in triplicates on Light Cycler 480 II (Roche) using ChamQ Universal SYBR qPCR Master Mix (Vazyme, Q711-03). Gapdh was used as the internal control. The primers used for RT-qPCR were listed in Supplementary Table 1.

H&E, Masson's trichrome, and Immunohistochemistry staining
Liver samples were fixed in 4% paraformaldehyde and embedded with paraffin. Samples were sliced into 8 μm in thickness and then subjected to hematoxylin and eosin (H&E) staining or Masson's trichrome staining. For immunohistochemistry, sections were dewaxed, rehydrated, and then incubated in EDTA antigen retrieval buffer (ZSGB-BIO, ZLI-9072) for 5 min at 100°C. Slices were then incubated with 3% H 2 O 2 for 10 min, washed three times with PBS containing 0.02% Triton TM X-100 (Sigma-Aldrich, T8787), followed by incubation with primary antibodies overnight at 4°C. Horseradish peroxidase-conjugated antibody was used as the secondary antibody and incubated at 37°C for 1 h. The color was developed by incubation with Dako Real TM kit (Dako, K5007). Sections were counterstained with hematoxylin (Baso, BA4041) and checked under the microscope (Nikon). Primary antibodies used in the immunohistochemistry staining were listed in Supplementary Table 2. Quantification for Masson's trichrome staining positive area and αSMA immunohistochemistry staining positive area was conducted with 5-8 random fields (10*) each mouse using ImageJ software (version 1.8.0).

TUNEL assay
Liver tissues were fixed, embedded with OCT compound (Servicebio, G6059), and sliced into 8 μm in thickness, then permeated with PBS containing 0.25% Triton TM X-100 (Sigma-Aldrich, T8787) and stained with In Situ Cell Death Detection Kit (Roche, 11684795910) following the manufacturer's instructions. Hoechst 33342 (Beyotime, C1022) was used to counterstain nuclei. Sections were visualized under the confocal microscope (ZEISS, LSM 880), and images were analyzed by ZEN 2012 software. Quantification for TUNEL + cells/Hoechst 33342 + cells ratio and average nuclear diameter was conducted using ImageJ software (version 1.8.0).

PI staining
PI staining was performed on primary hepatocytes. The culture medium was removed, and cells were washed twice with pre-warmed PBS.  Igf2bp1 to Hnf4a in Control and Mettl3 cKO mouse liver tissues at 2 and 4 weeks after birth (n = 4 for each group). i RT-qPCR analysis of HNF4A mRNA levels in IGF2BP1 knockdown HepG2 cells at different time points after 5 μM actinomycin D treatment (n = 4 for each group). j Schematic diagram showing two rescue strategies with AAV8-Hnf4a. k Representative H&E staining and BODIPY staining photographs of 2-weeks-old liver sections from Mettl3 cKO mice intravenous injected with AAV8-Ctrl and AAV8-Hnf4a at day two after birth (6 experiments were repeated independently with similar results). Scale bar = 20 μm. l Statistical histogram of BODIPY staining (n = 5 for AAV8-Ctrl group; n = 7 for AAV8-Hnf4a group). m Statistical histogram of Ki67 immunohistochemical staining (n = 3 for AAV8-Ctrl group; n = 4 for AAV8-Hnf4a group). n Survival curves of Mettl3 cKO mice intravenous injected with AAV8-Ctrl and AAV8-Hnf4a (n = 8 for AAV8-Ctrl group; n = 9 for AAV8-Hnf4a group). Data in b, c, e-i, and l-m were shown as mean ± SEM with the indicated significance (*P < 0.05, **P < 0.01, ***P < 0.001; two-tailed student's ttest). Data in n were analyzed by Log-rank (Mantel-Cox) test with the indicated significance (***P < 0.001). Source data are provided as a Source Data file. RNAs were isolated and subjected to reverse transcription, and the mRNA levels of genes of interest were detected by RT-qPCR. For RNA stability RNA-sequencing, total RNAs were sent for RNA-sequencing.

RNA-sequencing and m 6 A-RIP sequencing
RNA-sequencing and m 6 A-RIP sequencing for liver tissues were conducted by Guangzhou Epibiotek Co., Ltd. Briefly, for liver tissue RNA-  ) was added throughout the entire process. Fragmented mRNAs were eluted by 100 μL 0.3 μg/μL Proteinase K (Thermo Scientific, AM2546) at 55°C for 1 h, and followed by phenol-chloroform extraction and ethanol precipitation purification procedure. The precipitated mRNAs were reverse transcribed, and enrichment was determined by RT-qPCR as mentioned above. RT-qPCR primers for m 6 A positive regions were marked as "positive-m 6 A-RT" (e.g., primers for two separate m 6 A positive regions in Ppara were listed as Pparapositive-m 6 A-1-RT and Ppara-positive-m 6 A-2-RT), and primers for m 6 A negative regions were marked as "negative-m 6 A-RT" (e.g., the primer for m 6 A negative regions in Ppara were listed as Ppara-negative-m 6 A-RT). Primers used for m 6 A-RIP-qPCR were listed in Supplementary  Table 1.

Dual-luciferase reporter assays
To construct the Mettl3 overexpression vector, the full-length coding sequences of mouse Mettl3 were amplified by PCR with Phanta Max DNA Polymerase (Vazyme, P505) using Mettl3-PKD-F and Mettl3-PKD-R primer, then cloned into a lentiviral vector PKD-EF1 using ClonExpress II One Step Cloning kit (Vazyme, C112). Mettl3-D395A-W398A catalytic mutant (DPPW/APPA) vector was constructed using the site-directed mutagenesis method (using PCR-based method with Mettl3 Fig. 7 | Mettl3 is dispensable for homeostasis of adult liver. a Schematic diagram of tamoxifen-induced liver-specific Mettl3 knockout (Mettl3 icKO) mouse generation. 4 weeks old Mettl3 flox/flox /Alb-Cre ERT2 mice were treated with tamoxifen (Tam) or olive oil (Oil) for 5 consecutive days and then labeled with 0 week. b Western blot for Mettl3 in indicated tissues from Control and Mettl3 icKO mice at 1 week after tamoxifen treatment (3 experiments were repeated independently with similar results). c Western blot for Mettl3 and Hnf4a in Control and Mettl3 icKO mouse liver tissues at different time points after tamoxifen treatment (3 experiments were repeated independently with similar results). d Densitometry analysis of western blot for Mettl3 from Control and Mettl3 icKO mouse liver tissues at different time points after tamoxifen treatment (n = 3 for each group). e RT-qPCR for Mettl3 expression in liver tissues at different time points after tamoxifen treatment (n = 3 for each group over 3 independent experiments). f LC-MS/MS analysis of m 6 A/A ration in Control and Mettl3 icKO mouse livers 2 weeks after tamoxifen treatment (n = 2 for each group). g Representative gross appearance of livers from Control and Mettl3 icKO mice at 1 week, 2 weeks, 4 weeks, and 6 weeks after tamoxifen treatment (3 experiments were repeated independently with similar results). h Serum levels of AST, ALP, and Cholesterol of Control and Mettl3 icKO mice at different time points after tamoxifen treatment (n = 3 for 6 weeks group; n = 6 for 4 weeks group; n = 10 for other groups). i Representative H&E staining photographs of Control and Mettl3 icKO mouse liver sections at indicated time points (3 experiments were repeated independently with similar results). Scar bar = 100 μm.
Data in d-f and h were shown as mean ± SEM with the indicated significance (*P < 0.05, **P < 0.01, ***P < 0.001; two-tailed student's t-test). Source data are provided as a Source Data file.
overexpression vector as template and Mettl3-AWWA-F and Mettl3-AWWA-R as primers). PKD-EF1 vector expressing EGFP served as a control in transfection. DNA fragments of Hnf1a and Hnf4a containing the WT m 6 A motifs and mutant motifs ( Supplementary Fig. 6c) were synthesized by Shanghai Generay Biotech Co., LTD., and then subcloned into pMIR-REPORT firefly luciferase reporter vector (Ambion, AM5795) between Mlu I and Sac I sites. All sequences were confirmed by Sanger sequencing. 50 ng Mettl3 vectors (Mettl3 overexpressing vector or control), 40 ng firefly luciferase reporter vectors with WT or mutated fragments, and 10 ng pRL Renilla Luciferase Control Reporter Vector (Promega, E2231) were co-transfected into HEK293T cells in triplicates in 96-well plates. Fluc and Rluc activities were measured 24 h later with the Dual-Luciferase Reporter Assay System (Promega, E1910) according to the manufacturer's instructions. The relative luciferase activity was calculated through dividing Fluc activity by individual Rluc activity and then normalizing to control of each assay. The sequences of PCR primers were listed in Supplementary Table 1.

Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
Source data are provided with this paper. RNA-sequencing and m 6 A-RIP sequencing raw data and processed expression matrix are uploaded to GEO DataSets under accession code GSE197564. The sequencing reads were mapped to the mouse mm10 genome. All other data analyzed or generated in this study are provided along with the article. Source data are provided with this paper.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/ licenses/by/4.0/.